[qemu.git] / kvm-all.c

/*
 * QEMU KVM support
 *
 * Copyright IBM, Corp. 2008
 *
 * Authors:
 *  Anthony Liguori   <[email protected]>
 *
 * 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>
#include <stdarg.h>

#include <linux/kvm.h>

#include "qemu-common.h"
#include "sysemu.h"
#include "kvm.h"

//#define DEBUG_KVM

#ifdef DEBUG_KVM
#define dprintf(fmt, ...) \
    do { fprintf(stderr, fmt, ## __VA_ARGS__); } while (0)
#else
#define dprintf(fmt, ...) \
    do { } while (0)
#endif

typedef struct kvm_userspace_memory_region KVMSlot;

int kvm_allowed = 0;

struct KVMState
{
    KVMSlot slots[32];
    int fd;
    int vmfd;
};

static KVMState *kvm_state;

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

    for (i = 0; i < ARRAY_SIZE(s->slots); i++) {
        if (s->slots[i].memory_size == 0)
            return &s->slots[i];
    }

    return NULL;
}

static KVMSlot *kvm_lookup_slot(KVMState *s, target_phys_addr_t start_addr)
{
    int i;

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

        if (start_addr >= mem->guest_phys_addr &&
            start_addr < (mem->guest_phys_addr + mem->memory_size))
            return mem;
    }

    return NULL;
}

int kvm_init_vcpu(CPUState *env)
{
    KVMState *s = kvm_state;
    long mmap_size;
    int ret;

    dprintf("kvm_init_vcpu\n");

    ret = kvm_vm_ioctl(s, KVM_CREATE_VCPU, env->cpu_index);
    if (ret < 0) {
        dprintf("kvm_create_vcpu failed\n");
        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) {
        dprintf("KVM_GET_VCPU_MMAP_SIZE failed\n");
        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;
        dprintf("mmap'ing vcpu state failed\n");
        goto err;
    }

    ret = kvm_arch_init_vcpu(env);

err:
    return ret;
}

int kvm_init(int smp_cpus)
{
    KVMState *s;
    int ret;
    int i;

    if (smp_cpus > 1)
        return -EINVAL;

    s = qemu_mallocz(sizeof(KVMState));
    if (s == NULL)
        return -ENOMEM;

    for (i = 0; i < ARRAY_SIZE(s->slots); i++)
        s->slots[i].slot = i;

    s->vmfd = -1;
    s->fd = open("/dev/kvm", O_RDWR);
    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);
    if (s->vmfd < 0)
        goto err;

    /* 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
     * just use a user allocated buffer so we can use phys_ram_base
     * unmodified.  Make sure we have a sufficiently modern version of KVM.
     */
    ret = kvm_ioctl(s, KVM_CHECK_EXTENSION, KVM_CAP_USER_MEMORY);
    if (ret <= 0) {
        if (ret == 0)
            ret = -EINVAL;
        fprintf(stderr, "kvm does not support KVM_CAP_USER_MEMORY\n");
        goto err;
    }

    ret = kvm_arch_init(s, smp_cpus);
    if (ret < 0)
        goto err;

    kvm_state = s;

    return 0;

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

    return ret;
}

static int kvm_handle_io(CPUState *env, uint16_t port, void *data,
                         int direction, int size, uint32_t count)
{
    int i;
    uint8_t *ptr = data;

    for (i = 0; i < count; i++) {
        if (direction == KVM_EXIT_IO_IN) {
            switch (size) {
            case 1:
                stb_p(ptr, cpu_inb(env, port));
                break;
            case 2:
                stw_p(ptr, cpu_inw(env, port));
                break;
            case 4:
                stl_p(ptr, cpu_inl(env, port));
                break;
            }
        } else {
            switch (size) {
            case 1:
                cpu_outb(env, port, ldub_p(ptr));
                break;
            case 2:
                cpu_outw(env, port, lduw_p(ptr));
                break;
            case 4:
                cpu_outl(env, port, ldl_p(ptr));
                break;
            }
        }

        ptr += size;
    }

    return 1;
}

int kvm_cpu_exec(CPUState *env)
{
    struct kvm_run *run = env->kvm_run;
    int ret;

    dprintf("kvm_cpu_exec()\n");

    do {
        kvm_arch_pre_run(env, run);

        if ((env->interrupt_request & CPU_INTERRUPT_EXIT)) {
            dprintf("interrupt exit requested\n");
            ret = 0;
            break;
        }

        ret = kvm_vcpu_ioctl(env, KVM_RUN, 0);
        kvm_arch_post_run(env, run);

        if (ret == -EINTR || ret == -EAGAIN) {
            dprintf("io window exit\n");
            ret = 0;
            break;
        }

        if (ret < 0) {
            dprintf("kvm run failed %s\n", strerror(-ret));
            abort();
        }

        ret = 0; /* exit loop */
        switch (run->exit_reason) {
        case KVM_EXIT_IO:
            dprintf("handle_io\n");
            ret = kvm_handle_io(env, run->io.port,
                                (uint8_t *)run + run->io.data_offset,
                                run->io.direction,
                                run->io.size,
                                run->io.count);
            break;
        case KVM_EXIT_MMIO:
            dprintf("handle_mmio\n");
            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:
            dprintf("irq_window_open\n");
            break;
        case KVM_EXIT_SHUTDOWN:
            dprintf("shutdown\n");
            qemu_system_reset_request();
            ret = 1;
            break;
        case KVM_EXIT_UNKNOWN:
            dprintf("kvm_exit_unknown\n");
            break;
        case KVM_EXIT_FAIL_ENTRY:
            dprintf("kvm_exit_fail_entry\n");
            break;
        case KVM_EXIT_EXCEPTION:
            dprintf("kvm_exit_exception\n");
            break;
        case KVM_EXIT_DEBUG:
            dprintf("kvm_exit_debug\n");
            break;
        default:
            dprintf("kvm_arch_handle_exit\n");
            ret = kvm_arch_handle_exit(env, run);
            break;
        }
    } while (ret > 0);

    if ((env->interrupt_request & CPU_INTERRUPT_EXIT)) {
        env->interrupt_request &= ~CPU_INTERRUPT_EXIT;
        env->exception_index = EXCP_INTERRUPT;
    }

    return ret;
}

void kvm_set_phys_mem(target_phys_addr_t start_addr,
                      ram_addr_t size,
                      ram_addr_t phys_offset)
{
    KVMState *s = kvm_state;
    ram_addr_t flags = phys_offset & ~TARGET_PAGE_MASK;
    KVMSlot *mem;

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

    mem = kvm_lookup_slot(s, start_addr);
    if (mem) {
        if (flags == IO_MEM_UNASSIGNED) {
            mem->memory_size = 0;
            mem->guest_phys_addr = start_addr;
            mem->userspace_addr = 0;
            mem->flags = 0;

            kvm_vm_ioctl(s, KVM_SET_USER_MEMORY_REGION, mem);
        } else if (start_addr >= mem->guest_phys_addr &&
                   (start_addr + size) <= (mem->guest_phys_addr + mem->memory_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->guest_phys_addr = start_addr;
    mem->userspace_addr = (unsigned long)(phys_ram_base + phys_offset);
    mem->flags = 0;

    kvm_vm_ioctl(s, KVM_SET_USER_MEMORY_REGION, mem);
    /* FIXME deal with errors */
}

int kvm_ioctl(KVMState *s, int type, ...)
{
    int ret;
    void *arg;
    va_list ap;

    va_start(ap, type);
    arg = va_arg(ap, void *);
    va_end(ap);

    ret = ioctl(s->fd, type, arg);
    if (ret == -1)
        ret = -errno;

    return ret;
}

int kvm_vm_ioctl(KVMState *s, int type, ...)
{
    int ret;
    void *arg;
    va_list ap;

    va_start(ap, type);
    arg = va_arg(ap, void *);
    va_end(ap);

    ret = ioctl(s->vmfd, type, arg);
    if (ret == -1)
        ret = -errno;

    return ret;
}

int kvm_vcpu_ioctl(CPUState *env, int type, ...)
{
    int ret;
    void *arg;
    va_list ap;

    va_start(ap, type);
    arg = va_arg(ap, void *);
    va_end(ap);

    ret = ioctl(env->kvm_fd, type, arg);
    if (ret == -1)
        ret = -errno;

    return ret;
}
Commit	Line	Data
05330448 AL	1	/*
	2	* QEMU KVM support
	3	*
	4	* Copyright IBM, Corp. 2008
	5	*
	6	* Authors:
	7	* Anthony Liguori <[email protected]>
	8	*
	9	* This work is licensed under the terms of the GNU GPL, version 2 or later.
	10	* See the COPYING file in the top-level directory.
	11	*
	12	*/
	13
	14	#include <sys/types.h>
	15	#include <sys/ioctl.h>
	16	#include <sys/mman.h>
984b5181	17	#include <stdarg.h>
05330448 AL	18
	19	#include <linux/kvm.h>
	20
	21	#include "qemu-common.h"
	22	#include "sysemu.h"
	23	#include "kvm.h"
	24
	25	//#define DEBUG_KVM
	26
	27	#ifdef DEBUG_KVM
	28	#define dprintf(fmt, ...) \
	29	do { fprintf(stderr, fmt, ## __VA_ARGS__); } while (0)
	30	#else
	31	#define dprintf(fmt, ...) \
	32	do { } while (0)
	33	#endif
	34
	35	typedef struct kvm_userspace_memory_region KVMSlot;
	36
	37	int kvm_allowed = 0;
	38
	39	struct KVMState
	40	{
	41	KVMSlot slots[32];
	42	int fd;
	43	int vmfd;
	44	};
	45
	46	static KVMState *kvm_state;
	47
	48	static KVMSlot kvm_alloc_slot(KVMState s)
	49	{
	50	int i;
	51
	52	for (i = 0; i < ARRAY_SIZE(s->slots); i++) {
	53	if (s->slots[i].memory_size == 0)
	54	return &s->slots[i];
	55	}
	56
	57	return NULL;
	58	}
	59
	60	static KVMSlot kvm_lookup_slot(KVMState s, target_phys_addr_t start_addr)
	61	{
	62	int i;
	63
	64	for (i = 0; i < ARRAY_SIZE(s->slots); i++) {
	65	KVMSlot *mem = &s->slots[i];
	66
	67	if (start_addr >= mem->guest_phys_addr &&
	68	start_addr < (mem->guest_phys_addr + mem->memory_size))
	69	return mem;
	70	}
	71
	72	return NULL;
	73	}
	74
	75	int kvm_init_vcpu(CPUState *env)
	76	{
	77	KVMState *s = kvm_state;
	78	long mmap_size;
	79	int ret;
	80
	81	dprintf("kvm_init_vcpu\n");
82
984b5181	83	ret = kvm_vm_ioctl(s, KVM_CREATE_VCPU, env->cpu_index);
05330448 AL	84	if (ret < 0) {
	85	dprintf("kvm_create_vcpu failed\n");
	86	goto err;
	87	}
	88
	89	env->kvm_fd = ret;
	90	env->kvm_state = s;
	91
	92	mmap_size = kvm_ioctl(s, KVM_GET_VCPU_MMAP_SIZE, 0);
	93	if (mmap_size < 0) {
	94	dprintf("KVM_GET_VCPU_MMAP_SIZE failed\n");
	95	goto err;
	96	}
	97
	98	env->kvm_run = mmap(NULL, mmap_size, PROT_READ \| PROT_WRITE, MAP_SHARED,
	99	env->kvm_fd, 0);
	100	if (env->kvm_run == MAP_FAILED) {
	101	ret = -errno;
	102	dprintf("mmap'ing vcpu state failed\n");
	103	goto err;
	104	}
	105
	106	ret = kvm_arch_init_vcpu(env);
	107
	108	err:
	109	return ret;
	110	}
	111
	112	int kvm_init(int smp_cpus)
	113	{
	114	KVMState *s;
	115	int ret;
	116	int i;
	117
	118	if (smp_cpus > 1)
	119	return -EINVAL;
	120
	121	s = qemu_mallocz(sizeof(KVMState));
	122	if (s == NULL)
	123	return -ENOMEM;
	124
	125	for (i = 0; i < ARRAY_SIZE(s->slots); i++)
	126	s->slots[i].slot = i;
	127
	128	s->vmfd = -1;
	129	s->fd = open("/dev/kvm", O_RDWR);
	130	if (s->fd == -1) {
	131	fprintf(stderr, "Could not access KVM kernel module: %m\n");
	132	ret = -errno;
	133	goto err;
	134	}
	135
	136	ret = kvm_ioctl(s, KVM_GET_API_VERSION, 0);
	137	if (ret < KVM_API_VERSION) {
	138	if (ret > 0)
	139	ret = -EINVAL;
	140	fprintf(stderr, "kvm version too old\n");
	141	goto err;
	142	}
	143
	144	if (ret > KVM_API_VERSION) {
	145	ret = -EINVAL;
	146	fprintf(stderr, "kvm version not supported\n");
	147	goto err;
148	}
149
150	s->vmfd = kvm_ioctl(s, KVM_CREATE_VM, 0);
151	if (s->vmfd < 0)
152	goto err;
153
154	/* initially, KVM allocated its own memory and we had to jump through
155	* hooks to make phys_ram_base point to this. Modern versions of KVM
156	* just use a user allocated buffer so we can use phys_ram_base
157	* unmodified. Make sure we have a sufficiently modern version of KVM.
158	*/
984b5181	159	ret = kvm_ioctl(s, KVM_CHECK_EXTENSION, KVM_CAP_USER_MEMORY);
05330448 AL	160	if (ret <= 0) {
	161	if (ret == 0)
	162	ret = -EINVAL;
	163	fprintf(stderr, "kvm does not support KVM_CAP_USER_MEMORY\n");
	164	goto err;
	165	}
	166
	167	ret = kvm_arch_init(s, smp_cpus);
	168	if (ret < 0)
	169	goto err;
	170
	171	kvm_state = s;
	172
	173	return 0;
	174
	175	err:
	176	if (s) {
	177	if (s->vmfd != -1)
	178	close(s->vmfd);
	179	if (s->fd != -1)
	180	close(s->fd);
	181	}
	182	qemu_free(s);
	183
	184	return ret;
	185	}
	186
	187	static int kvm_handle_io(CPUState env, uint16_t port, void data,
	188	int direction, int size, uint32_t count)
	189	{
	190	int i;
	191	uint8_t *ptr = data;
	192
	193	for (i = 0; i < count; i++) {
	194	if (direction == KVM_EXIT_IO_IN) {
	195	switch (size) {
	196	case 1:
	197	stb_p(ptr, cpu_inb(env, port));
	198	break;
	199	case 2:
	200	stw_p(ptr, cpu_inw(env, port));
	201	break;
	202	case 4:
	203	stl_p(ptr, cpu_inl(env, port));
	204	break;
	205	}
	206	} else {
	207	switch (size) {
	208	case 1:
	209	cpu_outb(env, port, ldub_p(ptr));
	210	break;
	211	case 2:
	212	cpu_outw(env, port, lduw_p(ptr));
	213	break;
	214	case 4:
	215	cpu_outl(env, port, ldl_p(ptr));
	216	break;
	217	}
	218	}
	219
	220	ptr += size;
	221	}
	222
	223	return 1;
224	}
225
226	int kvm_cpu_exec(CPUState *env)
227	{
228	struct kvm_run *run = env->kvm_run;
229	int ret;
230
231	dprintf("kvm_cpu_exec()\n");
232
233	do {
234	kvm_arch_pre_run(env, run);
235
236	if ((env->interrupt_request & CPU_INTERRUPT_EXIT)) {
237	dprintf("interrupt exit requested\n");
238	ret = 0;
239	break;
240	}
241
242	ret = kvm_vcpu_ioctl(env, KVM_RUN, 0);
243	kvm_arch_post_run(env, run);
244
245	if (ret == -EINTR \|\| ret == -EAGAIN) {
246	dprintf("io window exit\n");
247	ret = 0;
248	break;
249	}
250
251	if (ret < 0) {
252	dprintf("kvm run failed %s\n", strerror(-ret));
253	abort();
254	}
255
256	ret = 0; /* exit loop */
257	switch (run->exit_reason) {
258	case KVM_EXIT_IO:
259	dprintf("handle_io\n");
260	ret = kvm_handle_io(env, run->io.port,
261	(uint8_t *)run + run->io.data_offset,
262	run->io.direction,
263	run->io.size,
264	run->io.count);
265	break;
266	case KVM_EXIT_MMIO:
267	dprintf("handle_mmio\n");
268	cpu_physical_memory_rw(run->mmio.phys_addr,
269	run->mmio.data,
270	run->mmio.len,
271	run->mmio.is_write);
272	ret = 1;
273	break;
274	case KVM_EXIT_IRQ_WINDOW_OPEN:
275	dprintf("irq_window_open\n");
276	break;
277	case KVM_EXIT_SHUTDOWN:
278	dprintf("shutdown\n");
279	qemu_system_reset_request();
280	ret = 1;
281	break;
282	case KVM_EXIT_UNKNOWN:
283	dprintf("kvm_exit_unknown\n");
284	break;
285	case KVM_EXIT_FAIL_ENTRY:
286	dprintf("kvm_exit_fail_entry\n");
287	break;
288	case KVM_EXIT_EXCEPTION:
289	dprintf("kvm_exit_exception\n");
290	break;
291	case KVM_EXIT_DEBUG:
292	dprintf("kvm_exit_debug\n");
293	break;
294	default:
295	dprintf("kvm_arch_handle_exit\n");
296	ret = kvm_arch_handle_exit(env, run);
297	break;
298	}
299	} while (ret > 0);
300
becfc390 AL	301	if ((env->interrupt_request & CPU_INTERRUPT_EXIT)) {
	302	env->interrupt_request &= ~CPU_INTERRUPT_EXIT;
	303	env->exception_index = EXCP_INTERRUPT;
	304	}
	305
05330448 AL	306	return ret;
	307	}
	308
	309	void kvm_set_phys_mem(target_phys_addr_t start_addr,
	310	ram_addr_t size,
	311	ram_addr_t phys_offset)
	312	{
	313	KVMState *s = kvm_state;
	314	ram_addr_t flags = phys_offset & ~TARGET_PAGE_MASK;
	315	KVMSlot *mem;
	316
	317	/* KVM does not support read-only slots */
	318	phys_offset &= ~IO_MEM_ROM;
	319
	320	mem = kvm_lookup_slot(s, start_addr);
	321	if (mem) {
	322	if (flags == IO_MEM_UNASSIGNED) {
	323	mem->memory_size = 0;
	324	mem->guest_phys_addr = start_addr;
	325	mem->userspace_addr = 0;
	326	mem->flags = 0;
	327
	328	kvm_vm_ioctl(s, KVM_SET_USER_MEMORY_REGION, mem);
	329	} else if (start_addr >= mem->guest_phys_addr &&
	330	(start_addr + size) <= (mem->guest_phys_addr + mem->memory_size))
	331	return;
	332	}
	333
	334	/* KVM does not need to know about this memory */
	335	if (flags >= IO_MEM_UNASSIGNED)
	336	return;
	337
	338	mem = kvm_alloc_slot(s);
	339	mem->memory_size = size;
	340	mem->guest_phys_addr = start_addr;
	341	mem->userspace_addr = (unsigned long)(phys_ram_base + phys_offset);
	342	mem->flags = 0;
	343
	344	kvm_vm_ioctl(s, KVM_SET_USER_MEMORY_REGION, mem);
	345	/* FIXME deal with errors */
	346	}
	347
984b5181	348	int kvm_ioctl(KVMState *s, int type, ...)
05330448 AL	349	{
05330448 AL	350	int ret;
984b5181 AL	351	void *arg;
984b5181 AL	352	va_list ap;
05330448	353
984b5181 AL	354	va_start(ap, type);
	355	arg = va_arg(ap, void *);
	356	va_end(ap);
	357
	358	ret = ioctl(s->fd, type, arg);
05330448 AL	359	if (ret == -1)
	360	ret = -errno;
	361
	362	return ret;
	363	}
	364
984b5181	365	int kvm_vm_ioctl(KVMState *s, int type, ...)
05330448 AL	366	{
05330448 AL	367	int ret;
984b5181 AL	368	void *arg;
	369	va_list ap;
	370
	371	va_start(ap, type);
	372	arg = va_arg(ap, void *);
	373	va_end(ap);
05330448	374
984b5181	375	ret = ioctl(s->vmfd, type, arg);
05330448 AL	376	if (ret == -1)
	377	ret = -errno;
	378
	379	return ret;
	380	}
	381
984b5181	382	int kvm_vcpu_ioctl(CPUState *env, int type, ...)
05330448 AL	383	{
05330448 AL	384	int ret;
984b5181 AL	385	void *arg;
	386	va_list ap;
	387
	388	va_start(ap, type);
	389	arg = va_arg(ap, void *);
	390	va_end(ap);
05330448	391
984b5181	392	ret = ioctl(env->kvm_fd, type, arg);
05330448 AL	393	if (ret == -1)
	394	ret = -errno;
	395
	396	return ret;
	397	}