[mirror_qemu.git] / backends / hostmem.c

/*
 * QEMU Host Memory Backend
 *
 * Copyright (C) 2013-2014 Red Hat Inc
 *
 * Authors:
 *   Igor Mammedov <imammedo@redhat.com>
 *
 * 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 "qemu/osdep.h"
#include "sysemu/hostmem.h"
#include "hw/boards.h"
#include "qapi/error.h"
#include "qapi/qapi-builtin-visit.h"
#include "qapi/visitor.h"
#include "qemu/config-file.h"
#include "qom/object_interfaces.h"
#include "qemu/mmap-alloc.h"

#ifdef CONFIG_NUMA
#include <numaif.h>
QEMU_BUILD_BUG_ON(HOST_MEM_POLICY_DEFAULT != MPOL_DEFAULT);
QEMU_BUILD_BUG_ON(HOST_MEM_POLICY_PREFERRED != MPOL_PREFERRED);
QEMU_BUILD_BUG_ON(HOST_MEM_POLICY_BIND != MPOL_BIND);
QEMU_BUILD_BUG_ON(HOST_MEM_POLICY_INTERLEAVE != MPOL_INTERLEAVE);
#endif

char *
host_memory_backend_get_name(HostMemoryBackend *backend)
{
    if (!backend->use_canonical_path) {
        return object_get_canonical_path_component(OBJECT(backend));
    }

    return object_get_canonical_path(OBJECT(backend));
}

static void
host_memory_backend_get_size(Object *obj, Visitor *v, const char *name,
                             void *opaque, Error **errp)
{
    HostMemoryBackend *backend = MEMORY_BACKEND(obj);
    uint64_t value = backend->size;

    visit_type_size(v, name, &value, errp);
}

static void
host_memory_backend_set_size(Object *obj, Visitor *v, const char *name,
                             void *opaque, Error **errp)
{
    HostMemoryBackend *backend = MEMORY_BACKEND(obj);
    Error *local_err = NULL;
    uint64_t value;

    if (host_memory_backend_mr_inited(backend)) {
        error_setg(&local_err, "cannot change property value");
        goto out;
    }

    visit_type_size(v, name, &value, &local_err);
    if (local_err) {
        goto out;
    }
    if (!value) {
        error_setg(&local_err, "Property '%s.%s' doesn't take value '%"
                   PRIu64 "'", object_get_typename(obj), name, value);
        goto out;
    }
    backend->size = value;
out:
    error_propagate(errp, local_err);
}

static void
host_memory_backend_get_host_nodes(Object *obj, Visitor *v, const char *name,
                                   void *opaque, Error **errp)
{
    HostMemoryBackend *backend = MEMORY_BACKEND(obj);
    uint16List *host_nodes = NULL;
    uint16List **node = &host_nodes;
    unsigned long value;

    value = find_first_bit(backend->host_nodes, MAX_NODES);
    if (value == MAX_NODES) {
        return;
    }

    *node = g_malloc0(sizeof(**node));
    (*node)->value = value;
    node = &(*node)->next;

    do {
        value = find_next_bit(backend->host_nodes, MAX_NODES, value + 1);
        if (value == MAX_NODES) {
            break;
        }

        *node = g_malloc0(sizeof(**node));
        (*node)->value = value;
        node = &(*node)->next;
    } while (true);

    visit_type_uint16List(v, name, &host_nodes, errp);
}

static void
host_memory_backend_set_host_nodes(Object *obj, Visitor *v, const char *name,
                                   void *opaque, Error **errp)
{
#ifdef CONFIG_NUMA
    HostMemoryBackend *backend = MEMORY_BACKEND(obj);
    uint16List *l, *host_nodes = NULL;

    visit_type_uint16List(v, name, &host_nodes, errp);

    for (l = host_nodes; l; l = l->next) {
        if (l->value >= MAX_NODES) {
            error_setg(errp, "Invalid host-nodes value: %d", l->value);
            goto out;
        }
    }

    for (l = host_nodes; l; l = l->next) {
        bitmap_set(backend->host_nodes, l->value, 1);
    }

out:
    qapi_free_uint16List(host_nodes);
#else
    error_setg(errp, "NUMA node binding are not supported by this QEMU");
#endif
}

static int
host_memory_backend_get_policy(Object *obj, Error **errp G_GNUC_UNUSED)
{
    HostMemoryBackend *backend = MEMORY_BACKEND(obj);
    return backend->policy;
}

static void
host_memory_backend_set_policy(Object *obj, int policy, Error **errp)
{
    HostMemoryBackend *backend = MEMORY_BACKEND(obj);
    backend->policy = policy;

#ifndef CONFIG_NUMA
    if (policy != HOST_MEM_POLICY_DEFAULT) {
        error_setg(errp, "NUMA policies are not supported by this QEMU");
    }
#endif
}

static bool host_memory_backend_get_merge(Object *obj, Error **errp)
{
    HostMemoryBackend *backend = MEMORY_BACKEND(obj);

    return backend->merge;
}

static void host_memory_backend_set_merge(Object *obj, bool value, Error **errp)
{
    HostMemoryBackend *backend = MEMORY_BACKEND(obj);

    if (!host_memory_backend_mr_inited(backend)) {
        backend->merge = value;
        return;
    }

    if (value != backend->merge) {
        void *ptr = memory_region_get_ram_ptr(&backend->mr);
        uint64_t sz = memory_region_size(&backend->mr);

        qemu_madvise(ptr, sz,
                     value ? QEMU_MADV_MERGEABLE : QEMU_MADV_UNMERGEABLE);
        backend->merge = value;
    }
}

static bool host_memory_backend_get_dump(Object *obj, Error **errp)
{
    HostMemoryBackend *backend = MEMORY_BACKEND(obj);

    return backend->dump;
}

static void host_memory_backend_set_dump(Object *obj, bool value, Error **errp)
{
    HostMemoryBackend *backend = MEMORY_BACKEND(obj);

    if (!host_memory_backend_mr_inited(backend)) {
        backend->dump = value;
        return;
    }

    if (value != backend->dump) {
        void *ptr = memory_region_get_ram_ptr(&backend->mr);
        uint64_t sz = memory_region_size(&backend->mr);

        qemu_madvise(ptr, sz,
                     value ? QEMU_MADV_DODUMP : QEMU_MADV_DONTDUMP);
        backend->dump = value;
    }
}

static bool host_memory_backend_get_prealloc(Object *obj, Error **errp)
{
    HostMemoryBackend *backend = MEMORY_BACKEND(obj);

    return backend->prealloc || backend->force_prealloc;
}

static void host_memory_backend_set_prealloc(Object *obj, bool value,
                                             Error **errp)
{
    Error *local_err = NULL;
    HostMemoryBackend *backend = MEMORY_BACKEND(obj);

    if (backend->force_prealloc) {
        if (value) {
            error_setg(errp,
                       "remove -mem-prealloc to use the prealloc property");
            return;
        }
    }

    if (!host_memory_backend_mr_inited(backend)) {
        backend->prealloc = value;
        return;
    }

    if (value && !backend->prealloc) {
        int fd = memory_region_get_fd(&backend->mr);
        void *ptr = memory_region_get_ram_ptr(&backend->mr);
        uint64_t sz = memory_region_size(&backend->mr);

        os_mem_prealloc(fd, ptr, sz, smp_cpus, &local_err);
        if (local_err) {
            error_propagate(errp, local_err);
            return;
        }
        backend->prealloc = true;
    }
}

static void host_memory_backend_init(Object *obj)
{
    HostMemoryBackend *backend = MEMORY_BACKEND(obj);
    MachineState *machine = MACHINE(qdev_get_machine());

    backend->merge = machine_mem_merge(machine);
    backend->dump = machine_dump_guest_core(machine);
    backend->prealloc = mem_prealloc;
}

static void host_memory_backend_post_init(Object *obj)
{
    object_apply_compat_props(obj);
}

bool host_memory_backend_mr_inited(HostMemoryBackend *backend)
{
    /*
     * NOTE: We forbid zero-length memory backend, so here zero means
     * "we haven't inited the backend memory region yet".
     */
    return memory_region_size(&backend->mr) != 0;
}

MemoryRegion *host_memory_backend_get_memory(HostMemoryBackend *backend)
{
    return host_memory_backend_mr_inited(backend) ? &backend->mr : NULL;
}

void host_memory_backend_set_mapped(HostMemoryBackend *backend, bool mapped)
{
    backend->is_mapped = mapped;
}

bool host_memory_backend_is_mapped(HostMemoryBackend *backend)
{
    return backend->is_mapped;
}

#ifdef __linux__
size_t host_memory_backend_pagesize(HostMemoryBackend *memdev)
{
    Object *obj = OBJECT(memdev);
    char *path = object_property_get_str(obj, "mem-path", NULL);
    size_t pagesize = qemu_mempath_getpagesize(path);

    g_free(path);
    return pagesize;
}
#else
size_t host_memory_backend_pagesize(HostMemoryBackend *memdev)
{
    return getpagesize();
}
#endif

static void
host_memory_backend_memory_complete(UserCreatable *uc, Error **errp)
{
    HostMemoryBackend *backend = MEMORY_BACKEND(uc);
    HostMemoryBackendClass *bc = MEMORY_BACKEND_GET_CLASS(uc);
    Error *local_err = NULL;
    void *ptr;
    uint64_t sz;

    if (bc->alloc) {
        bc->alloc(backend, &local_err);
        if (local_err) {
            goto out;
        }

        ptr = memory_region_get_ram_ptr(&backend->mr);
        sz = memory_region_size(&backend->mr);

        if (backend->merge) {
            qemu_madvise(ptr, sz, QEMU_MADV_MERGEABLE);
        }
        if (!backend->dump) {
            qemu_madvise(ptr, sz, QEMU_MADV_DONTDUMP);
        }
#ifdef CONFIG_NUMA
        unsigned long lastbit = find_last_bit(backend->host_nodes, MAX_NODES);
        /* lastbit == MAX_NODES means maxnode = 0 */
        unsigned long maxnode = (lastbit + 1) % (MAX_NODES + 1);
        /* ensure policy won't be ignored in case memory is preallocated
         * before mbind(). note: MPOL_MF_STRICT is ignored on hugepages so
         * this doesn't catch hugepage case. */
        unsigned flags = MPOL_MF_STRICT | MPOL_MF_MOVE;

        /* check for invalid host-nodes and policies and give more verbose
         * error messages than mbind(). */
        if (maxnode && backend->policy == MPOL_DEFAULT) {
            error_setg(errp, "host-nodes must be empty for policy default,"
                       " or you should explicitly specify a policy other"
                       " than default");
            return;
        } else if (maxnode == 0 && backend->policy != MPOL_DEFAULT) {
            error_setg(errp, "host-nodes must be set for policy %s",
                       HostMemPolicy_str(backend->policy));
            return;
        }

        /* We can have up to MAX_NODES nodes, but we need to pass maxnode+1
         * as argument to mbind() due to an old Linux bug (feature?) which
         * cuts off the last specified node. This means backend->host_nodes
         * must have MAX_NODES+1 bits available.
         */
        assert(sizeof(backend->host_nodes) >=
               BITS_TO_LONGS(MAX_NODES + 1) * sizeof(unsigned long));
        assert(maxnode <= MAX_NODES);
        if (mbind(ptr, sz, backend->policy,
                  maxnode ? backend->host_nodes : NULL, maxnode + 1, flags)) {
            if (backend->policy != MPOL_DEFAULT || errno != ENOSYS) {
                error_setg_errno(errp, errno,
                                 "cannot bind memory to host NUMA nodes");
                return;
            }
        }
#endif
        /* Preallocate memory after the NUMA policy has been instantiated.
         * This is necessary to guarantee memory is allocated with
         * specified NUMA policy in place.
         */
        if (backend->prealloc) {
            os_mem_prealloc(memory_region_get_fd(&backend->mr), ptr, sz,
                            smp_cpus, &local_err);
            if (local_err) {
                goto out;
            }
        }
    }
out:
    error_propagate(errp, local_err);
}

static bool
host_memory_backend_can_be_deleted(UserCreatable *uc)
{
    if (host_memory_backend_is_mapped(MEMORY_BACKEND(uc))) {
        return false;
    } else {
        return true;
    }
}

static bool host_memory_backend_get_share(Object *o, Error **errp)
{
    HostMemoryBackend *backend = MEMORY_BACKEND(o);

    return backend->share;
}

static void host_memory_backend_set_share(Object *o, bool value, Error **errp)
{
    HostMemoryBackend *backend = MEMORY_BACKEND(o);

    if (host_memory_backend_mr_inited(backend)) {
        error_setg(errp, "cannot change property value");
        return;
    }
    backend->share = value;
}

static bool
host_memory_backend_get_use_canonical_path(Object *obj, Error **errp)
{
    HostMemoryBackend *backend = MEMORY_BACKEND(obj);

    return backend->use_canonical_path;
}

static void
host_memory_backend_set_use_canonical_path(Object *obj, bool value,
                                           Error **errp)
{
    HostMemoryBackend *backend = MEMORY_BACKEND(obj);

    backend->use_canonical_path = value;
}

static void
host_memory_backend_class_init(ObjectClass *oc, void *data)
{
    UserCreatableClass *ucc = USER_CREATABLE_CLASS(oc);

    ucc->complete = host_memory_backend_memory_complete;
    ucc->can_be_deleted = host_memory_backend_can_be_deleted;

    object_class_property_add_bool(oc, "merge",
        host_memory_backend_get_merge,
        host_memory_backend_set_merge, &error_abort);
    object_class_property_set_description(oc, "merge",
        "Mark memory as mergeable", &error_abort);
    object_class_property_add_bool(oc, "dump",
        host_memory_backend_get_dump,
        host_memory_backend_set_dump, &error_abort);
    object_class_property_set_description(oc, "dump",
        "Set to 'off' to exclude from core dump", &error_abort);
    object_class_property_add_bool(oc, "prealloc",
        host_memory_backend_get_prealloc,
        host_memory_backend_set_prealloc, &error_abort);
    object_class_property_set_description(oc, "prealloc",
        "Preallocate memory", &error_abort);
    object_class_property_add(oc, "size", "int",
        host_memory_backend_get_size,
        host_memory_backend_set_size,
        NULL, NULL, &error_abort);
    object_class_property_set_description(oc, "size",
        "Size of the memory region (ex: 500M)", &error_abort);
    object_class_property_add(oc, "host-nodes", "int",
        host_memory_backend_get_host_nodes,
        host_memory_backend_set_host_nodes,
        NULL, NULL, &error_abort);
    object_class_property_set_description(oc, "host-nodes",
        "Binds memory to the list of NUMA host nodes", &error_abort);
    object_class_property_add_enum(oc, "policy", "HostMemPolicy",
        &HostMemPolicy_lookup,
        host_memory_backend_get_policy,
        host_memory_backend_set_policy, &error_abort);
    object_class_property_set_description(oc, "policy",
        "Set the NUMA policy", &error_abort);
    object_class_property_add_bool(oc, "share",
        host_memory_backend_get_share, host_memory_backend_set_share,
        &error_abort);
    object_class_property_set_description(oc, "share",
        "Mark the memory as private to QEMU or shared", &error_abort);
    object_class_property_add_bool(oc, "x-use-canonical-path-for-ramblock-id",
        host_memory_backend_get_use_canonical_path,
        host_memory_backend_set_use_canonical_path, &error_abort);
}

static const TypeInfo host_memory_backend_info = {
    .name = TYPE_MEMORY_BACKEND,
    .parent = TYPE_OBJECT,
    .abstract = true,
    .class_size = sizeof(HostMemoryBackendClass),
    .class_init = host_memory_backend_class_init,
    .instance_size = sizeof(HostMemoryBackend),
    .instance_init = host_memory_backend_init,
    .instance_post_init = host_memory_backend_post_init,
    .interfaces = (InterfaceInfo[]) {
        { TYPE_USER_CREATABLE },
        { }
    }
};

static void register_types(void)
{
    type_register_static(&host_memory_backend_info);
}

type_init(register_types);
Commit	Line	Data
1f070489 IM	1	/*
	2	* QEMU Host Memory Backend
	3	*
	4	* Copyright (C) 2013-2014 Red Hat Inc
	5	*
	6	* Authors:
	7	* Igor Mammedov <imammedo@redhat.com>
	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	*/
9af23989	12
9c058332	13	#include "qemu/osdep.h"
1f070489	14	#include "sysemu/hostmem.h"
6b269967	15	#include "hw/boards.h"
da34e65c	16	#include "qapi/error.h"
eb815e24	17	#include "qapi/qapi-builtin-visit.h"
1f070489	18	#include "qapi/visitor.h"
1f070489 IM	19	#include "qemu/config-file.h"
1f070489 IM	20	#include "qom/object_interfaces.h"
2b108085	21	#include "qemu/mmap-alloc.h"
1f070489	22
4cf1b76b HT	23	#ifdef CONFIG_NUMA
	24	#include <numaif.h>
	25	QEMU_BUILD_BUG_ON(HOST_MEM_POLICY_DEFAULT != MPOL_DEFAULT);
	26	QEMU_BUILD_BUG_ON(HOST_MEM_POLICY_PREFERRED != MPOL_PREFERRED);
	27	QEMU_BUILD_BUG_ON(HOST_MEM_POLICY_BIND != MPOL_BIND);
	28	QEMU_BUILD_BUG_ON(HOST_MEM_POLICY_INTERLEAVE != MPOL_INTERLEAVE);
	29	#endif
	30
fa0cb34d MAL	31	char *
	32	host_memory_backend_get_name(HostMemoryBackend *backend)
	33	{
	34	if (!backend->use_canonical_path) {
	35	return object_get_canonical_path_component(OBJECT(backend));
	36	}
	37
	38	return object_get_canonical_path(OBJECT(backend));
	39	}
	40
1f070489	41	static void
d7bce999 EB	42	host_memory_backend_get_size(Object obj, Visitor v, const char *name,
d7bce999 EB	43	void opaque, Error *errp)
1f070489 IM	44	{
	45	HostMemoryBackend *backend = MEMORY_BACKEND(obj);
	46	uint64_t value = backend->size;
	47
51e72bc1	48	visit_type_size(v, name, &value, errp);
1f070489 IM	49	}
	50
	51	static void
d7bce999 EB	52	host_memory_backend_set_size(Object obj, Visitor v, const char *name,
d7bce999 EB	53	void opaque, Error *errp)
1f070489 IM	54	{
	55	HostMemoryBackend *backend = MEMORY_BACKEND(obj);
	56	Error *local_err = NULL;
	57	uint64_t value;
	58
6f4c60e4	59	if (host_memory_backend_mr_inited(backend)) {
1f070489 IM	60	error_setg(&local_err, "cannot change property value");
	61	goto out;
	62	}
	63
51e72bc1	64	visit_type_size(v, name, &value, &local_err);
1f070489 IM	65	if (local_err) {
	66	goto out;
	67	}
	68	if (!value) {
	69	error_setg(&local_err, "Property '%s.%s' doesn't take value '%"
	70	PRIu64 "'", object_get_typename(obj), name, value);
	71	goto out;
	72	}
	73	backend->size = value;
	74	out:
	75	error_propagate(errp, local_err);
	76	}
	77
4cf1b76b	78	static void
d7bce999 EB	79	host_memory_backend_get_host_nodes(Object obj, Visitor v, const char *name,
d7bce999 EB	80	void opaque, Error *errp)
4cf1b76b HT	81	{
	82	HostMemoryBackend *backend = MEMORY_BACKEND(obj);
	83	uint16List *host_nodes = NULL;
	84	uint16List **node = &host_nodes;
	85	unsigned long value;
	86
	87	value = find_first_bit(backend->host_nodes, MAX_NODES);
	88	if (value == MAX_NODES) {
658ae5a7	89	return;
4cf1b76b HT	90	}
4cf1b76b HT	91
658ae5a7 MA	92	node = g_malloc0(sizeof(*node));
	93	(*node)->value = value;
	94	node = &(*node)->next;
	95
4cf1b76b HT	96	do {
	97	value = find_next_bit(backend->host_nodes, MAX_NODES, value + 1);
	98	if (value == MAX_NODES) {
	99	break;
	100	}
	101
658ae5a7 MA	102	node = g_malloc0(sizeof(*node));
	103	(*node)->value = value;
	104	node = &(*node)->next;
4cf1b76b HT	105	} while (true);
4cf1b76b HT	106
51e72bc1	107	visit_type_uint16List(v, name, &host_nodes, errp);
4cf1b76b HT	108	}
	109
	110	static void
d7bce999 EB	111	host_memory_backend_set_host_nodes(Object obj, Visitor v, const char *name,
d7bce999 EB	112	void opaque, Error *errp)
4cf1b76b HT	113	{
	114	#ifdef CONFIG_NUMA
	115	HostMemoryBackend *backend = MEMORY_BACKEND(obj);
ffa144b3	116	uint16List l, host_nodes = NULL;
4cf1b76b	117
ffa144b3 EH	118	visit_type_uint16List(v, name, &host_nodes, errp);
	119
	120	for (l = host_nodes; l; l = l->next) {
	121	if (l->value >= MAX_NODES) {
	122	error_setg(errp, "Invalid host-nodes value: %d", l->value);
	123	goto out;
	124	}
	125	}
4cf1b76b	126
ffa144b3	127	for (l = host_nodes; l; l = l->next) {
4cf1b76b	128	bitmap_set(backend->host_nodes, l->value, 1);
4cf1b76b	129	}
ffa144b3 EH	130
	131	out:
	132	qapi_free_uint16List(host_nodes);
4cf1b76b HT	133	#else
	134	error_setg(errp, "NUMA node binding are not supported by this QEMU");
	135	#endif
	136	}
	137
a3590dac DB	138	static int
a3590dac DB	139	host_memory_backend_get_policy(Object obj, Error *errp G_GNUC_UNUSED)
4cf1b76b HT	140	{
4cf1b76b HT	141	HostMemoryBackend *backend = MEMORY_BACKEND(obj);
a3590dac	142	return backend->policy;
4cf1b76b HT	143	}
	144
	145	static void
a3590dac	146	host_memory_backend_set_policy(Object obj, int policy, Error *errp)
4cf1b76b HT	147	{
4cf1b76b HT	148	HostMemoryBackend *backend = MEMORY_BACKEND(obj);
4cf1b76b HT	149	backend->policy = policy;
	150
	151	#ifndef CONFIG_NUMA
	152	if (policy != HOST_MEM_POLICY_DEFAULT) {
	153	error_setg(errp, "NUMA policies are not supported by this QEMU");
	154	}
	155	#endif
	156	}
	157
605d0a94 PB	158	static bool host_memory_backend_get_merge(Object obj, Error *errp)
	159	{
	160	HostMemoryBackend *backend = MEMORY_BACKEND(obj);
	161
	162	return backend->merge;
	163	}
	164
	165	static void host_memory_backend_set_merge(Object obj, bool value, Error *errp)
	166	{
	167	HostMemoryBackend *backend = MEMORY_BACKEND(obj);
	168
6f4c60e4	169	if (!host_memory_backend_mr_inited(backend)) {
605d0a94 PB	170	backend->merge = value;
	171	return;
	172	}
	173
	174	if (value != backend->merge) {
	175	void *ptr = memory_region_get_ram_ptr(&backend->mr);
	176	uint64_t sz = memory_region_size(&backend->mr);
	177
	178	qemu_madvise(ptr, sz,
	179	value ? QEMU_MADV_MERGEABLE : QEMU_MADV_UNMERGEABLE);
	180	backend->merge = value;
	181	}
	182	}
	183
	184	static bool host_memory_backend_get_dump(Object obj, Error *errp)
	185	{
	186	HostMemoryBackend *backend = MEMORY_BACKEND(obj);
	187
	188	return backend->dump;
	189	}
	190
	191	static void host_memory_backend_set_dump(Object obj, bool value, Error *errp)
	192	{
	193	HostMemoryBackend *backend = MEMORY_BACKEND(obj);
	194
6f4c60e4	195	if (!host_memory_backend_mr_inited(backend)) {
605d0a94 PB	196	backend->dump = value;
	197	return;
	198	}
	199
	200	if (value != backend->dump) {
	201	void *ptr = memory_region_get_ram_ptr(&backend->mr);
	202	uint64_t sz = memory_region_size(&backend->mr);
	203
	204	qemu_madvise(ptr, sz,
	205	value ? QEMU_MADV_DODUMP : QEMU_MADV_DONTDUMP);
	206	backend->dump = value;
	207	}
	208	}
	209
a35ba7be PB	210	static bool host_memory_backend_get_prealloc(Object obj, Error *errp)
	211	{
	212	HostMemoryBackend *backend = MEMORY_BACKEND(obj);
	213
	214	return backend->prealloc \|\| backend->force_prealloc;
	215	}
	216
	217	static void host_memory_backend_set_prealloc(Object *obj, bool value,
	218	Error **errp)
	219	{
056b68af	220	Error *local_err = NULL;
a35ba7be PB	221	HostMemoryBackend *backend = MEMORY_BACKEND(obj);
	222
	223	if (backend->force_prealloc) {
	224	if (value) {
	225	error_setg(errp,
	226	"remove -mem-prealloc to use the prealloc property");
	227	return;
	228	}
	229	}
	230
6f4c60e4	231	if (!host_memory_backend_mr_inited(backend)) {
a35ba7be PB	232	backend->prealloc = value;
	233	return;
	234	}
	235
	236	if (value && !backend->prealloc) {
	237	int fd = memory_region_get_fd(&backend->mr);
	238	void *ptr = memory_region_get_ram_ptr(&backend->mr);
	239	uint64_t sz = memory_region_size(&backend->mr);
	240
1e356fc1	241	os_mem_prealloc(fd, ptr, sz, smp_cpus, &local_err);
056b68af IM	242	if (local_err) {
	243	error_propagate(errp, local_err);
	244	return;
	245	}
a35ba7be PB	246	backend->prealloc = true;
	247	}
	248	}
	249
58f4662c	250	static void host_memory_backend_init(Object *obj)
1f070489	251	{
605d0a94	252	HostMemoryBackend *backend = MEMORY_BACKEND(obj);
6b269967	253	MachineState *machine = MACHINE(qdev_get_machine());
605d0a94	254
6b269967 EH	255	backend->merge = machine_mem_merge(machine);
6b269967 EH	256	backend->dump = machine_dump_guest_core(machine);
a35ba7be	257	backend->prealloc = mem_prealloc;
1f070489 IM	258	}
1f070489 IM	259
fa0cb34d MAL	260	static void host_memory_backend_post_init(Object *obj)
	261	{
	262	object_apply_compat_props(obj);
	263	}
	264
4728b574 PX	265	bool host_memory_backend_mr_inited(HostMemoryBackend *backend)
	266	{
	267	/*
	268	* NOTE: We forbid zero-length memory backend, so here zero means
	269	* "we haven't inited the backend memory region yet".
	270	*/
	271	return memory_region_size(&backend->mr) != 0;
	272	}
	273
7943e97b	274	MemoryRegion host_memory_backend_get_memory(HostMemoryBackend backend)
1f070489	275	{
6f4c60e4	276	return host_memory_backend_mr_inited(backend) ? &backend->mr : NULL;
1f070489 IM	277	}
1f070489 IM	278
2aece63c XG	279	void host_memory_backend_set_mapped(HostMemoryBackend *backend, bool mapped)
	280	{
	281	backend->is_mapped = mapped;
	282	}
	283
	284	bool host_memory_backend_is_mapped(HostMemoryBackend *backend)
	285	{
	286	return backend->is_mapped;
	287	}
	288
2b108085 DG	289	#ifdef __linux__
	290	size_t host_memory_backend_pagesize(HostMemoryBackend *memdev)
	291	{
	292	Object *obj = OBJECT(memdev);
	293	char *path = object_property_get_str(obj, "mem-path", NULL);
	294	size_t pagesize = qemu_mempath_getpagesize(path);
	295
	296	g_free(path);
	297	return pagesize;
	298	}
	299	#else
	300	size_t host_memory_backend_pagesize(HostMemoryBackend *memdev)
	301	{
	302	return getpagesize();
	303	}
	304	#endif
	305
bd9262d9 HT	306	static void
	307	host_memory_backend_memory_complete(UserCreatable uc, Error *errp)
	308	{
	309	HostMemoryBackend *backend = MEMORY_BACKEND(uc);
	310	HostMemoryBackendClass *bc = MEMORY_BACKEND_GET_CLASS(uc);
605d0a94 PB	311	Error *local_err = NULL;
	312	void *ptr;
	313	uint64_t sz;
bd9262d9 HT	314
bd9262d9 HT	315	if (bc->alloc) {
605d0a94 PB	316	bc->alloc(backend, &local_err);
605d0a94 PB	317	if (local_err) {
056b68af	318	goto out;
605d0a94 PB	319	}
	320
	321	ptr = memory_region_get_ram_ptr(&backend->mr);
	322	sz = memory_region_size(&backend->mr);
	323
	324	if (backend->merge) {
	325	qemu_madvise(ptr, sz, QEMU_MADV_MERGEABLE);
	326	}
	327	if (!backend->dump) {
	328	qemu_madvise(ptr, sz, QEMU_MADV_DONTDUMP);
	329	}
4cf1b76b HT	330	#ifdef CONFIG_NUMA
	331	unsigned long lastbit = find_last_bit(backend->host_nodes, MAX_NODES);
	332	/* lastbit == MAX_NODES means maxnode = 0 */
	333	unsigned long maxnode = (lastbit + 1) % (MAX_NODES + 1);
	334	/* ensure policy won't be ignored in case memory is preallocated
	335	* before mbind(). note: MPOL_MF_STRICT is ignored on hugepages so
	336	* this doesn't catch hugepage case. */
288d3322	337	unsigned flags = MPOL_MF_STRICT \| MPOL_MF_MOVE;
4cf1b76b HT	338
	339	/* check for invalid host-nodes and policies and give more verbose
	340	* error messages than mbind(). */
	341	if (maxnode && backend->policy == MPOL_DEFAULT) {
	342	error_setg(errp, "host-nodes must be empty for policy default,"
	343	" or you should explicitly specify a policy other"
	344	" than default");
	345	return;
	346	} else if (maxnode == 0 && backend->policy != MPOL_DEFAULT) {
	347	error_setg(errp, "host-nodes must be set for policy %s",
977c736f	348	HostMemPolicy_str(backend->policy));
4cf1b76b HT	349	return;
	350	}
	351
	352	/* We can have up to MAX_NODES nodes, but we need to pass maxnode+1
	353	* as argument to mbind() due to an old Linux bug (feature?) which
	354	* cuts off the last specified node. This means backend->host_nodes
	355	* must have MAX_NODES+1 bits available.
	356	*/
	357	assert(sizeof(backend->host_nodes) >=
	358	BITS_TO_LONGS(MAX_NODES + 1) * sizeof(unsigned long));
	359	assert(maxnode <= MAX_NODES);
	360	if (mbind(ptr, sz, backend->policy,
	361	maxnode ? backend->host_nodes : NULL, maxnode + 1, flags)) {
a3567ba1 PF	362	if (backend->policy != MPOL_DEFAULT \|\| errno != ENOSYS) {
	363	error_setg_errno(errp, errno,
	364	"cannot bind memory to host NUMA nodes");
	365	return;
	366	}
4cf1b76b HT	367	}
	368	#endif
	369	/* Preallocate memory after the NUMA policy has been instantiated.
	370	* This is necessary to guarantee memory is allocated with
	371	* specified NUMA policy in place.
	372	*/
a35ba7be	373	if (backend->prealloc) {
056b68af	374	os_mem_prealloc(memory_region_get_fd(&backend->mr), ptr, sz,
1e356fc1	375	smp_cpus, &local_err);
056b68af IM	376	if (local_err) {
	377	goto out;
	378	}
a35ba7be	379	}
bd9262d9	380	}
056b68af IM	381	out:
056b68af IM	382	error_propagate(errp, local_err);
bd9262d9 HT	383	}
bd9262d9 HT	384
36bce5ca	385	static bool
3beacfb9	386	host_memory_backend_can_be_deleted(UserCreatable *uc)
36bce5ca	387	{
2aece63c	388	if (host_memory_backend_is_mapped(MEMORY_BACKEND(uc))) {
36bce5ca LM	389	return false;
	390	} else {
	391	return true;
	392	}
	393	}
	394
06329cce MA	395	static bool host_memory_backend_get_share(Object o, Error *errp)
	396	{
	397	HostMemoryBackend *backend = MEMORY_BACKEND(o);
	398
	399	return backend->share;
	400	}
	401
	402	static void host_memory_backend_set_share(Object o, bool value, Error *errp)
	403	{
	404	HostMemoryBackend *backend = MEMORY_BACKEND(o);
	405
	406	if (host_memory_backend_mr_inited(backend)) {
	407	error_setg(errp, "cannot change property value");
	408	return;
	409	}
	410	backend->share = value;
	411	}
	412
fa0cb34d MAL	413	static bool
	414	host_memory_backend_get_use_canonical_path(Object obj, Error *errp)
	415	{
	416	HostMemoryBackend *backend = MEMORY_BACKEND(obj);
	417
	418	return backend->use_canonical_path;
	419	}
	420
	421	static void
	422	host_memory_backend_set_use_canonical_path(Object *obj, bool value,
	423	Error **errp)
	424	{
	425	HostMemoryBackend *backend = MEMORY_BACKEND(obj);
	426
	427	backend->use_canonical_path = value;
	428	}
	429
bd9262d9 HT	430	static void
	431	host_memory_backend_class_init(ObjectClass oc, void data)
	432	{
	433	UserCreatableClass *ucc = USER_CREATABLE_CLASS(oc);
	434
	435	ucc->complete = host_memory_backend_memory_complete;
36bce5ca	436	ucc->can_be_deleted = host_memory_backend_can_be_deleted;
e62834ca EH	437
	438	object_class_property_add_bool(oc, "merge",
	439	host_memory_backend_get_merge,
	440	host_memory_backend_set_merge, &error_abort);
033bfc5e MAL	441	object_class_property_set_description(oc, "merge",
033bfc5e MAL	442	"Mark memory as mergeable", &error_abort);
e62834ca EH	443	object_class_property_add_bool(oc, "dump",
	444	host_memory_backend_get_dump,
	445	host_memory_backend_set_dump, &error_abort);
033bfc5e MAL	446	object_class_property_set_description(oc, "dump",
033bfc5e MAL	447	"Set to 'off' to exclude from core dump", &error_abort);
e62834ca EH	448	object_class_property_add_bool(oc, "prealloc",
	449	host_memory_backend_get_prealloc,
	450	host_memory_backend_set_prealloc, &error_abort);
033bfc5e MAL	451	object_class_property_set_description(oc, "prealloc",
033bfc5e MAL	452	"Preallocate memory", &error_abort);
e62834ca EH	453	object_class_property_add(oc, "size", "int",
	454	host_memory_backend_get_size,
	455	host_memory_backend_set_size,
	456	NULL, NULL, &error_abort);
033bfc5e MAL	457	object_class_property_set_description(oc, "size",
033bfc5e MAL	458	"Size of the memory region (ex: 500M)", &error_abort);
e62834ca EH	459	object_class_property_add(oc, "host-nodes", "int",
	460	host_memory_backend_get_host_nodes,
	461	host_memory_backend_set_host_nodes,
	462	NULL, NULL, &error_abort);
033bfc5e MAL	463	object_class_property_set_description(oc, "host-nodes",
033bfc5e MAL	464	"Binds memory to the list of NUMA host nodes", &error_abort);
e62834ca	465	object_class_property_add_enum(oc, "policy", "HostMemPolicy",
f7abe0ec	466	&HostMemPolicy_lookup,
e62834ca EH	467	host_memory_backend_get_policy,
e62834ca EH	468	host_memory_backend_set_policy, &error_abort);
033bfc5e MAL	469	object_class_property_set_description(oc, "policy",
033bfc5e MAL	470	"Set the NUMA policy", &error_abort);
06329cce MA	471	object_class_property_add_bool(oc, "share",
	472	host_memory_backend_get_share, host_memory_backend_set_share,
	473	&error_abort);
033bfc5e MAL	474	object_class_property_set_description(oc, "share",
033bfc5e MAL	475	"Mark the memory as private to QEMU or shared", &error_abort);
fa0cb34d MAL	476	object_class_property_add_bool(oc, "x-use-canonical-path-for-ramblock-id",
	477	host_memory_backend_get_use_canonical_path,
	478	host_memory_backend_set_use_canonical_path, &error_abort);
e1ff3c67 IM	479	}
e1ff3c67 IM	480
58f4662c	481	static const TypeInfo host_memory_backend_info = {
1f070489 IM	482	.name = TYPE_MEMORY_BACKEND,
	483	.parent = TYPE_OBJECT,
	484	.abstract = true,
	485	.class_size = sizeof(HostMemoryBackendClass),
bd9262d9	486	.class_init = host_memory_backend_class_init,
1f070489	487	.instance_size = sizeof(HostMemoryBackend),
58f4662c	488	.instance_init = host_memory_backend_init,
fa0cb34d	489	.instance_post_init = host_memory_backend_post_init,
1f070489 IM	490	.interfaces = (InterfaceInfo[]) {
	491	{ TYPE_USER_CREATABLE },
	492	{ }
	493	}
	494	};
	495
	496	static void register_types(void)
	497	{
58f4662c	498	type_register_static(&host_memory_backend_info);
1f070489 IM	499	}
	500
	501	type_init(register_types);