[thirdparty/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

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;
}

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 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);
}

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