kvm-all.c 31.8 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;
A
aliguori 已提交
74 75 76 77 78 79 80 81 82
};

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
83 84 85
        /* KVM private memory slots */
        if (i >= 8 && i < 12)
            continue;
A
aliguori 已提交
86 87 88 89
        if (s->slots[i].memory_size == 0)
            return &s->slots[i];
    }

90 91 92 93 94
    fprintf(stderr, "%s: no free slot available\n", __func__);
    abort();
}

static KVMSlot *kvm_lookup_matching_slot(KVMState *s,
A
Anthony Liguori 已提交
95 96
                                         target_phys_addr_t start_addr,
                                         target_phys_addr_t end_addr)
97 98 99 100 101 102 103 104 105 106 107 108
{
    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
109 110 111
    return NULL;
}

112 113 114 115
/*
 * Find overlapping slot with lowest start address
 */
static KVMSlot *kvm_lookup_overlapping_slot(KVMState *s,
A
Anthony Liguori 已提交
116 117
                                            target_phys_addr_t start_addr,
                                            target_phys_addr_t end_addr)
A
aliguori 已提交
118
{
119
    KVMSlot *found = NULL;
A
aliguori 已提交
120 121 122 123 124
    int i;

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

125 126 127 128 129 130 131 132 133
        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
134 135
    }

136
    return found;
A
aliguori 已提交
137 138
}

139 140 141 142 143 144 145
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
146
    mem.userspace_addr = (unsigned long)qemu_get_ram_ptr(slot->phys_offset);
147
    mem.flags = slot->flags;
148 149 150
    if (s->migration_log) {
        mem.flags |= KVM_MEM_LOG_DIRTY_PAGES;
    }
151 152 153
    return kvm_vm_ioctl(s, KVM_SET_USER_MEMORY_REGION, &mem);
}

J
Jan Kiszka 已提交
154 155 156 157
static void kvm_reset_vcpu(void *opaque)
{
    CPUState *env = opaque;

J
Jan Kiszka 已提交
158
    kvm_arch_reset_vcpu(env);
J
Jan Kiszka 已提交
159
}
160

161 162 163 164 165 166 167 168 169 170 171
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
172 173 174 175 176 177
int kvm_init_vcpu(CPUState *env)
{
    KVMState *s = kvm_state;
    long mmap_size;
    int ret;

178
    DPRINTF("kvm_init_vcpu\n");
A
aliguori 已提交
179

180
    ret = kvm_vm_ioctl(s, KVM_CREATE_VCPU, env->cpu_index);
A
aliguori 已提交
181
    if (ret < 0) {
182
        DPRINTF("kvm_create_vcpu failed\n");
A
aliguori 已提交
183 184 185 186 187 188 189 190
        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) {
191
        DPRINTF("KVM_GET_VCPU_MMAP_SIZE failed\n");
A
aliguori 已提交
192 193 194 195 196 197 198
        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;
199
        DPRINTF("mmap'ing vcpu state failed\n");
A
aliguori 已提交
200 201 202
        goto err;
    }

203 204 205 206 207 208
#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
209
    ret = kvm_arch_init_vcpu(env);
J
Jan Kiszka 已提交
210
    if (ret == 0) {
211
        qemu_register_reset(kvm_reset_vcpu, env);
J
Jan Kiszka 已提交
212
        kvm_arch_reset_vcpu(env);
J
Jan Kiszka 已提交
213
    }
A
aliguori 已提交
214 215 216 217
err:
    return ret;
}

218 219 220
/*
 * dirty pages logging control
 */
A
Anthony Liguori 已提交
221 222
static int kvm_dirty_pages_log_change(target_phys_addr_t phys_addr,
                                      ram_addr_t size, int flags, int mask)
223 224
{
    KVMState *s = kvm_state;
225
    KVMSlot *mem = kvm_lookup_matching_slot(s, phys_addr, phys_addr + size);
226 227
    int old_flags;

228
    if (mem == NULL)  {
229 230
            fprintf(stderr, "BUG: %s: invalid parameters " TARGET_FMT_plx "-"
                    TARGET_FMT_plx "\n", __func__, phys_addr,
A
Anthony Liguori 已提交
231
                    (target_phys_addr_t)(phys_addr + size - 1));
232 233 234
            return -EINVAL;
    }

235
    old_flags = mem->flags;
236

237
    flags = (mem->flags & ~mask) | flags;
238 239
    mem->flags = flags;

240 241 242 243 244 245 246 247
    /* 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;
    }

248 249 250
    return kvm_set_user_memory_region(s, mem);
}

A
Anthony Liguori 已提交
251
int kvm_log_start(target_phys_addr_t phys_addr, ram_addr_t size)
252
{
253
        return kvm_dirty_pages_log_change(phys_addr, size,
254 255 256 257
                                          KVM_MEM_LOG_DIRTY_PAGES,
                                          KVM_MEM_LOG_DIRTY_PAGES);
}

A
Anthony Liguori 已提交
258
int kvm_log_stop(target_phys_addr_t phys_addr, ram_addr_t size)
259
{
260
        return kvm_dirty_pages_log_change(phys_addr, size,
261 262 263 264
                                          0,
                                          KVM_MEM_LOG_DIRTY_PAGES);
}

265
static int kvm_set_migration_log(int enable)
266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286
{
    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];

        if (!!(mem->flags & KVM_MEM_LOG_DIRTY_PAGES) == enable) {
            continue;
        }
        err = kvm_set_user_memory_region(s, mem);
        if (err) {
            return err;
        }
    }
    return 0;
}

287 288 289 290 291
/* 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
292
{
293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317
    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
318 319
}

320 321
#define ALIGN(x, y)  (((x)+(y)-1) & ~((y)-1))

322 323 324 325 326
/**
 * 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.
 *
327
 * @start_add: start of logged region.
328 329
 * @end_addr: end of logged region.
 */
330 331
static int kvm_physical_sync_dirty_bitmap(target_phys_addr_t start_addr,
					  target_phys_addr_t end_addr)
332 333
{
    KVMState *s = kvm_state;
334 335 336 337
    unsigned long size, allocated_size = 0;
    KVMDirtyLog d;
    KVMSlot *mem;
    int ret = 0;
338

339 340 341 342 343 344
    d.dirty_bitmap = NULL;
    while (start_addr < end_addr) {
        mem = kvm_lookup_overlapping_slot(s, start_addr, end_addr);
        if (mem == NULL) {
            break;
        }
345

346
        size = ALIGN(((mem->memory_size) >> TARGET_PAGE_BITS), HOST_LONG_BITS) / 8;
347 348 349 350 351 352 353
        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);
354

355
        d.slot = mem->slot;
356

357
        if (kvm_vm_ioctl(s, KVM_GET_DIRTY_LOG, &d) == -1) {
358
            DPRINTF("ioctl failed %d\n", errno);
359 360 361
            ret = -1;
            break;
        }
362

363 364 365
        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;
366 367
    }
    qemu_free(d.dirty_bitmap);
368 369

    return ret;
370 371
}

A
Anthony Liguori 已提交
372
int kvm_coalesce_mmio_region(target_phys_addr_t start, ram_addr_t size)
A
aliguori 已提交
373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390
{
    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
391
int kvm_uncoalesce_mmio_region(target_phys_addr_t start, ram_addr_t size)
A
aliguori 已提交
392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409
{
    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;
}

410 411 412 413 414 415 416 417 418 419 420 421
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;
}

422 423 424
static void kvm_set_phys_mem(target_phys_addr_t start_addr,
			     ram_addr_t size,
			     ram_addr_t phys_offset)
425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559
{
    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();
    }
}

560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586
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
587 588
int kvm_init(int smp_cpus)
{
589 590 591
    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
592 593 594 595
    KVMState *s;
    int ret;
    int i;

596 597
    if (smp_cpus > 1) {
        fprintf(stderr, "No SMP KVM support, use '-smp 1'\n");
A
aliguori 已提交
598
        return -EINVAL;
599
    }
A
aliguori 已提交
600 601 602

    s = qemu_mallocz(sizeof(KVMState));

603
#ifdef KVM_CAP_SET_GUEST_DEBUG
B
Blue Swirl 已提交
604
    QTAILQ_INIT(&s->kvm_sw_breakpoints);
605
#endif
A
aliguori 已提交
606 607 608 609
    for (i = 0; i < ARRAY_SIZE(s->slots); i++)
        s->slots[i].slot = i;

    s->vmfd = -1;
K
Kevin Wolf 已提交
610
    s->fd = qemu_open("/dev/kvm", O_RDWR);
A
aliguori 已提交
611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631
    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);
632 633 634 635 636
    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
637
        goto err;
638
    }
A
aliguori 已提交
639 640 641

    /* 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
642
     * just use a user allocated buffer so we can use regular pages
A
aliguori 已提交
643 644
     * unmodified.  Make sure we have a sufficiently modern version of KVM.
     */
645 646
    if (!kvm_check_extension(s, KVM_CAP_USER_MEMORY)) {
        ret = -EINVAL;
647 648
        fprintf(stderr, "kvm does not support KVM_CAP_USER_MEMORY\n%s",
                upgrade_note);
A
aliguori 已提交
649 650 651
        goto err;
    }

652 653 654
    /* 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. */
655 656
    if (!kvm_check_extension(s, KVM_CAP_DESTROY_MEMORY_REGION_WORKS)) {
        ret = -EINVAL;
657 658

        fprintf(stderr,
659 660
                "KVM kernel module broken (DESTROY_MEMORY_REGION).\n%s",
                upgrade_note);
661 662 663
        goto err;
    }

664
    s->coalesced_mmio = 0;
A
aliguori 已提交
665
#ifdef KVM_CAP_COALESCED_MMIO
666
    s->coalesced_mmio = kvm_check_extension(s, KVM_CAP_COALESCED_MMIO);
667
    s->coalesced_mmio_ring = NULL;
A
aliguori 已提交
668 669
#endif

670 671 672 673 674 675 676 677
    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

678 679 680 681 682
    s->vcpu_events = 0;
#ifdef KVM_CAP_VCPU_EVENTS
    s->vcpu_events = kvm_check_extension(s, KVM_CAP_VCPU_EVENTS);
#endif

683 684 685 686 687 688
    s->robust_singlestep = 0;
#ifdef KVM_CAP_X86_ROBUST_SINGLESTEP
    s->robust_singlestep =
        kvm_check_extension(s, KVM_CAP_X86_ROBUST_SINGLESTEP);
#endif

689 690 691 692 693
    s->debugregs = 0;
#ifdef KVM_CAP_DEBUGREGS
    s->debugregs = kvm_check_extension(s, KVM_CAP_DEBUGREGS);
#endif

A
aliguori 已提交
694 695 696 697 698
    ret = kvm_arch_init(s, smp_cpus);
    if (ret < 0)
        goto err;

    kvm_state = s;
699
    cpu_register_phys_memory_client(&kvm_cpu_phys_memory_client);
A
aliguori 已提交
700 701 702 703 704 705 706 707 708 709 710 711 712 713 714

    return 0;

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

    return ret;
}

715 716
static int kvm_handle_io(uint16_t port, void *data, int direction, int size,
                         uint32_t count)
A
aliguori 已提交
717 718 719 720 721 722 723 724
{
    int i;
    uint8_t *ptr = data;

    for (i = 0; i < count; i++) {
        if (direction == KVM_EXIT_IO_IN) {
            switch (size) {
            case 1:
725
                stb_p(ptr, cpu_inb(port));
A
aliguori 已提交
726 727
                break;
            case 2:
728
                stw_p(ptr, cpu_inw(port));
A
aliguori 已提交
729 730
                break;
            case 4:
731
                stl_p(ptr, cpu_inl(port));
A
aliguori 已提交
732 733 734 735 736
                break;
            }
        } else {
            switch (size) {
            case 1:
737
                cpu_outb(port, ldub_p(ptr));
A
aliguori 已提交
738 739
                break;
            case 2:
740
                cpu_outw(port, lduw_p(ptr));
A
aliguori 已提交
741 742
                break;
            case 4:
743
                cpu_outl(port, ldl_p(ptr));
A
aliguori 已提交
744 745 746 747 748 749 750 751 752 753
                break;
            }
        }

        ptr += size;
    }

    return 1;
}

M
Marcelo Tosatti 已提交
754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779
#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");
    }
    /* FIXME: Should trigger a qmp message to let management know
     * something went wrong.
     */
    vm_stop(0);
}
#endif

780
void kvm_flush_coalesced_mmio_buffer(void)
A
aliguori 已提交
781 782 783
{
#ifdef KVM_CAP_COALESCED_MMIO
    KVMState *s = kvm_state;
784 785
    if (s->coalesced_mmio_ring) {
        struct kvm_coalesced_mmio_ring *ring = s->coalesced_mmio_ring;
A
aliguori 已提交
786 787 788 789 790 791
        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);
792
            smp_wmb();
A
aliguori 已提交
793 794 795 796 797 798
            ring->first = (ring->first + 1) % KVM_COALESCED_MMIO_MAX;
        }
    }
#endif
}

799
static void do_kvm_cpu_synchronize_state(void *_env)
800
{
801 802
    CPUState *env = _env;

J
Jan Kiszka 已提交
803
    if (!env->kvm_vcpu_dirty) {
804
        kvm_arch_get_registers(env);
J
Jan Kiszka 已提交
805
        env->kvm_vcpu_dirty = 1;
806 807 808
    }
}

809 810 811 812 813 814
void kvm_cpu_synchronize_state(CPUState *env)
{
    if (!env->kvm_vcpu_dirty)
        run_on_cpu(env, do_kvm_cpu_synchronize_state, env);
}

815 816 817 818 819 820 821 822 823 824 825 826
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
827 828 829 830 831
int kvm_cpu_exec(CPUState *env)
{
    struct kvm_run *run = env->kvm_run;
    int ret;

832
    DPRINTF("kvm_cpu_exec()\n");
A
aliguori 已提交
833 834

    do {
835
#ifndef CONFIG_IOTHREAD
836
        if (env->exit_request) {
837
            DPRINTF("interrupt exit requested\n");
A
aliguori 已提交
838 839 840
            ret = 0;
            break;
        }
841
#endif
A
aliguori 已提交
842

J
Jan Kiszka 已提交
843
        if (env->kvm_vcpu_dirty) {
844
            kvm_arch_put_registers(env, KVM_PUT_RUNTIME_STATE);
J
Jan Kiszka 已提交
845
            env->kvm_vcpu_dirty = 0;
846 847
        }

848
        kvm_arch_pre_run(env, run);
849
        cpu_single_env = NULL;
850
        qemu_mutex_unlock_iothread();
A
aliguori 已提交
851
        ret = kvm_vcpu_ioctl(env, KVM_RUN, 0);
852
        qemu_mutex_lock_iothread();
853
        cpu_single_env = env;
A
aliguori 已提交
854 855 856
        kvm_arch_post_run(env, run);

        if (ret == -EINTR || ret == -EAGAIN) {
857
            cpu_exit(env);
858
            DPRINTF("io window exit\n");
A
aliguori 已提交
859 860 861 862 863
            ret = 0;
            break;
        }

        if (ret < 0) {
864
            DPRINTF("kvm run failed %s\n", strerror(-ret));
A
aliguori 已提交
865 866 867
            abort();
        }

868
        kvm_flush_coalesced_mmio_buffer();
A
aliguori 已提交
869

A
aliguori 已提交
870 871 872
        ret = 0; /* exit loop */
        switch (run->exit_reason) {
        case KVM_EXIT_IO:
873
            DPRINTF("handle_io\n");
874
            ret = kvm_handle_io(run->io.port,
A
aliguori 已提交
875 876 877 878 879 880
                                (uint8_t *)run + run->io.data_offset,
                                run->io.direction,
                                run->io.size,
                                run->io.count);
            break;
        case KVM_EXIT_MMIO:
881
            DPRINTF("handle_mmio\n");
A
aliguori 已提交
882 883 884 885 886 887 888
            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:
889
            DPRINTF("irq_window_open\n");
A
aliguori 已提交
890 891
            break;
        case KVM_EXIT_SHUTDOWN:
892
            DPRINTF("shutdown\n");
A
aliguori 已提交
893 894 895 896
            qemu_system_reset_request();
            ret = 1;
            break;
        case KVM_EXIT_UNKNOWN:
897
            DPRINTF("kvm_exit_unknown\n");
A
aliguori 已提交
898 899
            break;
        case KVM_EXIT_FAIL_ENTRY:
900
            DPRINTF("kvm_exit_fail_entry\n");
A
aliguori 已提交
901 902
            break;
        case KVM_EXIT_EXCEPTION:
903
            DPRINTF("kvm_exit_exception\n");
A
aliguori 已提交
904
            break;
M
Marcelo Tosatti 已提交
905 906 907 908 909
#ifdef KVM_CAP_INTERNAL_ERROR_DATA
        case KVM_EXIT_INTERNAL_ERROR:
            kvm_handle_internal_error(env, run);
            break;
#endif
A
aliguori 已提交
910
        case KVM_EXIT_DEBUG:
911
            DPRINTF("kvm_exit_debug\n");
912 913 914 915 916 917 918 919 920 921
#ifdef KVM_CAP_SET_GUEST_DEBUG
            if (kvm_arch_debug(&run->debug.arch)) {
                gdb_set_stop_cpu(env);
                vm_stop(EXCP_DEBUG);
                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
922 923
            break;
        default:
924
            DPRINTF("kvm_arch_handle_exit\n");
A
aliguori 已提交
925 926 927 928 929
            ret = kvm_arch_handle_exit(env, run);
            break;
        }
    } while (ret > 0);

930 931
    if (env->exit_request) {
        env->exit_request = 0;
A
aliguori 已提交
932 933 934
        env->exception_index = EXCP_INTERRUPT;
    }

A
aliguori 已提交
935 936 937
    return ret;
}

938
int kvm_ioctl(KVMState *s, int type, ...)
A
aliguori 已提交
939 940
{
    int ret;
941 942
    void *arg;
    va_list ap;
A
aliguori 已提交
943

944 945 946 947 948
    va_start(ap, type);
    arg = va_arg(ap, void *);
    va_end(ap);

    ret = ioctl(s->fd, type, arg);
A
aliguori 已提交
949 950 951 952 953 954
    if (ret == -1)
        ret = -errno;

    return ret;
}

955
int kvm_vm_ioctl(KVMState *s, int type, ...)
A
aliguori 已提交
956 957
{
    int ret;
958 959 960 961 962 963
    void *arg;
    va_list ap;

    va_start(ap, type);
    arg = va_arg(ap, void *);
    va_end(ap);
A
aliguori 已提交
964

965
    ret = ioctl(s->vmfd, type, arg);
A
aliguori 已提交
966 967 968 969 970 971
    if (ret == -1)
        ret = -errno;

    return ret;
}

972
int kvm_vcpu_ioctl(CPUState *env, int type, ...)
A
aliguori 已提交
973 974
{
    int ret;
975 976 977 978 979 980
    void *arg;
    va_list ap;

    va_start(ap, type);
    arg = va_arg(ap, void *);
    va_end(ap);
A
aliguori 已提交
981

982
    ret = ioctl(env->kvm_fd, type, arg);
A
aliguori 已提交
983 984 985 986 987
    if (ret == -1)
        ret = -errno;

    return ret;
}
A
aliguori 已提交
988 989 990

int kvm_has_sync_mmu(void)
{
A
aurel32 已提交
991
#ifdef KVM_CAP_SYNC_MMU
A
aliguori 已提交
992 993
    KVMState *s = kvm_state;

994 995
    return kvm_check_extension(s, KVM_CAP_SYNC_MMU);
#else
A
aliguori 已提交
996
    return 0;
997
#endif
A
aliguori 已提交
998
}
999

1000 1001 1002 1003 1004
int kvm_has_vcpu_events(void)
{
    return kvm_state->vcpu_events;
}

1005 1006 1007 1008 1009
int kvm_has_robust_singlestep(void)
{
    return kvm_state->robust_singlestep;
}

1010 1011 1012 1013 1014
int kvm_has_debugregs(void)
{
    return kvm_state->debugregs;
}

1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032
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
    }
}

1033
#ifdef KVM_CAP_SET_GUEST_DEBUG
L
Luiz Capitulino 已提交
1034 1035
static void on_vcpu(CPUState *env, void (*func)(void *data), void *data)
{
G
Glauber Costa 已提交
1036
#ifdef CONFIG_IOTHREAD
1037 1038
    if (env != cpu_single_env) {
        abort();
L
Luiz Capitulino 已提交
1039
    }
G
Glauber Costa 已提交
1040
#endif
1041
    func(data);
L
Luiz Capitulino 已提交
1042 1043
}

1044 1045 1046 1047 1048
struct kvm_sw_breakpoint *kvm_find_sw_breakpoint(CPUState *env,
                                                 target_ulong pc)
{
    struct kvm_sw_breakpoint *bp;

B
Blue Swirl 已提交
1049
    QTAILQ_FOREACH(bp, &env->kvm_state->kvm_sw_breakpoints, entry) {
1050 1051 1052 1053 1054 1055 1056 1057
        if (bp->pc == pc)
            return bp;
    }
    return NULL;
}

int kvm_sw_breakpoints_active(CPUState *env)
{
B
Blue Swirl 已提交
1058
    return !QTAILQ_EMPTY(&env->kvm_state->kvm_sw_breakpoints);
1059 1060
}

G
Glauber Costa 已提交
1061 1062 1063 1064 1065 1066 1067 1068 1069
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
1070 1071 1072
    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
1073 1074
}

1075 1076
int kvm_update_guest_debug(CPUState *env, unsigned long reinject_trap)
{
G
Glauber Costa 已提交
1077
    struct kvm_set_guest_debug_data data;
1078

1079
    data.dbg.control = reinject_trap;
1080

1081 1082 1083
    if (env->singlestep_enabled) {
        data.dbg.control |= KVM_GUESTDBG_ENABLE | KVM_GUESTDBG_SINGLESTEP;
    }
G
Glauber Costa 已提交
1084 1085
    kvm_arch_update_guest_debug(env, &data.dbg);
    data.env = env;
1086

G
Glauber Costa 已提交
1087 1088
    on_vcpu(env, kvm_invoke_set_guest_debug, &data);
    return data.err;
1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116
}

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
1117
        QTAILQ_INSERT_HEAD(&current_env->kvm_state->kvm_sw_breakpoints,
1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153
                          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
1154
        QTAILQ_REMOVE(&current_env->kvm_state->kvm_sw_breakpoints, bp, entry);
1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175
        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
1176
    QTAILQ_FOREACH_SAFE(bp, &s->kvm_sw_breakpoints, entry, next) {
1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213
        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 */
1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231

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;
}
1232 1233 1234

int kvm_set_ioeventfd_pio_word(int fd, uint16_t addr, uint16_t val, bool assign)
{
1235
#ifdef KVM_IOEVENTFD
1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251
    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;
1252 1253
#else
    return -ENOSYS;
1254
#endif
1255
}