qemu/hw/arm/virt.c
<<
>>
Prefs
   1/*
   2 * ARM mach-virt emulation
   3 *
   4 * Copyright (c) 2013 Linaro Limited
   5 *
   6 * This program is free software; you can redistribute it and/or modify it
   7 * under the terms and conditions of the GNU General Public License,
   8 * version 2 or later, as published by the Free Software Foundation.
   9 *
  10 * This program is distributed in the hope it will be useful, but WITHOUT
  11 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
  12 * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
  13 * more details.
  14 *
  15 * You should have received a copy of the GNU General Public License along with
  16 * this program.  If not, see <http://www.gnu.org/licenses/>.
  17 *
  18 * Emulate a virtual board which works by passing Linux all the information
  19 * it needs about what devices are present via the device tree.
  20 * There are some restrictions about what we can do here:
  21 *  + we can only present devices whose Linux drivers will work based
  22 *    purely on the device tree with no platform data at all
  23 *  + we want to present a very stripped-down minimalist platform,
  24 *    both because this reduces the security attack surface from the guest
  25 *    and also because it reduces our exposure to being broken when
  26 *    the kernel updates its device tree bindings and requires further
  27 *    information in a device binding that we aren't providing.
  28 * This is essentially the same approach kvmtool uses.
  29 */
  30
  31#include "qemu/osdep.h"
  32#include "qemu/datadir.h"
  33#include "qemu/units.h"
  34#include "qemu/option.h"
  35#include "monitor/qdev.h"
  36#include "qapi/error.h"
  37#include "hw/sysbus.h"
  38#include "hw/arm/boot.h"
  39#include "hw/arm/primecell.h"
  40#include "hw/arm/virt.h"
  41#include "hw/block/flash.h"
  42#include "hw/vfio/vfio-calxeda-xgmac.h"
  43#include "hw/vfio/vfio-amd-xgbe.h"
  44#include "hw/display/ramfb.h"
  45#include "net/net.h"
  46#include "sysemu/device_tree.h"
  47#include "sysemu/numa.h"
  48#include "sysemu/runstate.h"
  49#include "sysemu/tpm.h"
  50#include "sysemu/kvm.h"
  51#include "sysemu/hvf.h"
  52#include "hw/loader.h"
  53#include "qapi/error.h"
  54#include "qemu/bitops.h"
  55#include "qemu/error-report.h"
  56#include "qemu/module.h"
  57#include "hw/pci-host/gpex.h"
  58#include "hw/virtio/virtio-pci.h"
  59#include "hw/core/sysbus-fdt.h"
  60#include "hw/platform-bus.h"
  61#include "hw/qdev-properties.h"
  62#include "hw/arm/fdt.h"
  63#include "hw/intc/arm_gic.h"
  64#include "hw/intc/arm_gicv3_common.h"
  65#include "hw/irq.h"
  66#include "kvm_arm.h"
  67#include "hw/firmware/smbios.h"
  68#include "qapi/visitor.h"
  69#include "qapi/qapi-visit-common.h"
  70#include "standard-headers/linux/input.h"
  71#include "hw/arm/smmuv3.h"
  72#include "hw/acpi/acpi.h"
  73#include "target/arm/internals.h"
  74#include "hw/mem/memory-device.h"
  75#include "hw/mem/pc-dimm.h"
  76#include "hw/mem/nvdimm.h"
  77#include "hw/acpi/generic_event_device.h"
  78#include "hw/virtio/virtio-mem-pci.h"
  79#include "hw/virtio/virtio-iommu.h"
  80#include "hw/char/pl011.h"
  81#include "qemu/guest-random.h"
  82
  83#define DEFINE_VIRT_MACHINE_LATEST(major, minor, latest) \
  84    static void virt_##major##_##minor##_class_init(ObjectClass *oc, \
  85                                                    void *data) \
  86    { \
  87        MachineClass *mc = MACHINE_CLASS(oc); \
  88        virt_machine_##major##_##minor##_options(mc); \
  89        mc->desc = "QEMU " # major "." # minor " ARM Virtual Machine"; \
  90        if (latest) { \
  91            mc->alias = "virt"; \
  92        } \
  93    } \
  94    static const TypeInfo machvirt_##major##_##minor##_info = { \
  95        .name = MACHINE_TYPE_NAME("virt-" # major "." # minor), \
  96        .parent = TYPE_VIRT_MACHINE, \
  97        .class_init = virt_##major##_##minor##_class_init, \
  98    }; \
  99    static void machvirt_machine_##major##_##minor##_init(void) \
 100    { \
 101        type_register_static(&machvirt_##major##_##minor##_info); \
 102    } \
 103    type_init(machvirt_machine_##major##_##minor##_init);
 104
 105#define DEFINE_VIRT_MACHINE_AS_LATEST(major, minor) \
 106    DEFINE_VIRT_MACHINE_LATEST(major, minor, true)
 107#define DEFINE_VIRT_MACHINE(major, minor) \
 108    DEFINE_VIRT_MACHINE_LATEST(major, minor, false)
 109
 110
 111/* Number of external interrupt lines to configure the GIC with */
 112#define NUM_IRQS 256
 113
 114#define PLATFORM_BUS_NUM_IRQS 64
 115
 116/* Legacy RAM limit in GB (< version 4.0) */
 117#define LEGACY_RAMLIMIT_GB 255
 118#define LEGACY_RAMLIMIT_BYTES (LEGACY_RAMLIMIT_GB * GiB)
 119
 120/* Addresses and sizes of our components.
 121 * 0..128MB is space for a flash device so we can run bootrom code such as UEFI.
 122 * 128MB..256MB is used for miscellaneous device I/O.
 123 * 256MB..1GB is reserved for possible future PCI support (ie where the
 124 * PCI memory window will go if we add a PCI host controller).
 125 * 1GB and up is RAM (which may happily spill over into the
 126 * high memory region beyond 4GB).
 127 * This represents a compromise between how much RAM can be given to
 128 * a 32 bit VM and leaving space for expansion and in particular for PCI.
 129 * Note that devices should generally be placed at multiples of 0x10000,
 130 * to accommodate guests using 64K pages.
 131 */
 132static const MemMapEntry base_memmap[] = {
 133    /* Space up to 0x8000000 is reserved for a boot ROM */
 134    [VIRT_FLASH] =              {          0, 0x08000000 },
 135    [VIRT_CPUPERIPHS] =         { 0x08000000, 0x00020000 },
 136    /* GIC distributor and CPU interfaces sit inside the CPU peripheral space */
 137    [VIRT_GIC_DIST] =           { 0x08000000, 0x00010000 },
 138    [VIRT_GIC_CPU] =            { 0x08010000, 0x00010000 },
 139    [VIRT_GIC_V2M] =            { 0x08020000, 0x00001000 },
 140    [VIRT_GIC_HYP] =            { 0x08030000, 0x00010000 },
 141    [VIRT_GIC_VCPU] =           { 0x08040000, 0x00010000 },
 142    /* The space in between here is reserved for GICv3 CPU/vCPU/HYP */
 143    [VIRT_GIC_ITS] =            { 0x08080000, 0x00020000 },
 144    /* This redistributor space allows up to 2*64kB*123 CPUs */
 145    [VIRT_GIC_REDIST] =         { 0x080A0000, 0x00F60000 },
 146    [VIRT_UART] =               { 0x09000000, 0x00001000 },
 147    [VIRT_RTC] =                { 0x09010000, 0x00001000 },
 148    [VIRT_FW_CFG] =             { 0x09020000, 0x00000018 },
 149    [VIRT_GPIO] =               { 0x09030000, 0x00001000 },
 150    [VIRT_SECURE_UART] =        { 0x09040000, 0x00001000 },
 151    [VIRT_SMMU] =               { 0x09050000, 0x00020000 },
 152    [VIRT_PCDIMM_ACPI] =        { 0x09070000, MEMORY_HOTPLUG_IO_LEN },
 153    [VIRT_ACPI_GED] =           { 0x09080000, ACPI_GED_EVT_SEL_LEN },
 154    [VIRT_NVDIMM_ACPI] =        { 0x09090000, NVDIMM_ACPI_IO_LEN},
 155    [VIRT_PVTIME] =             { 0x090a0000, 0x00010000 },
 156    [VIRT_SECURE_GPIO] =        { 0x090b0000, 0x00001000 },
 157    [VIRT_MMIO] =               { 0x0a000000, 0x00000200 },
 158    /* ...repeating for a total of NUM_VIRTIO_TRANSPORTS, each of that size */
 159    [VIRT_PLATFORM_BUS] =       { 0x0c000000, 0x02000000 },
 160    [VIRT_SECURE_MEM] =         { 0x0e000000, 0x01000000 },
 161    [VIRT_PCIE_MMIO] =          { 0x10000000, 0x2eff0000 },
 162    [VIRT_PCIE_PIO] =           { 0x3eff0000, 0x00010000 },
 163    [VIRT_PCIE_ECAM] =          { 0x3f000000, 0x01000000 },
 164    /* Actual RAM size depends on initial RAM and device memory settings */
 165    [VIRT_MEM] =                { GiB, LEGACY_RAMLIMIT_BYTES },
 166};
 167
 168/*
 169 * Highmem IO Regions: This memory map is floating, located after the RAM.
 170 * Each MemMapEntry base (GPA) will be dynamically computed, depending on the
 171 * top of the RAM, so that its base get the same alignment as the size,
 172 * ie. a 512GiB entry will be aligned on a 512GiB boundary. If there is
 173 * less than 256GiB of RAM, the floating area starts at the 256GiB mark.
 174 * Note the extended_memmap is sized so that it eventually also includes the
 175 * base_memmap entries (VIRT_HIGH_GIC_REDIST2 index is greater than the last
 176 * index of base_memmap).
 177 */
 178static MemMapEntry extended_memmap[] = {
 179    /* Additional 64 MB redist region (can contain up to 512 redistributors) */
 180    [VIRT_HIGH_GIC_REDIST2] =   { 0x0, 64 * MiB },
 181    [VIRT_HIGH_PCIE_ECAM] =     { 0x0, 256 * MiB },
 182    /* Second PCIe window */
 183    [VIRT_HIGH_PCIE_MMIO] =     { 0x0, 512 * GiB },
 184};
 185
 186static const int a15irqmap[] = {
 187    [VIRT_UART] = 1,
 188    [VIRT_RTC] = 2,
 189    [VIRT_PCIE] = 3, /* ... to 6 */
 190    [VIRT_GPIO] = 7,
 191    [VIRT_SECURE_UART] = 8,
 192    [VIRT_ACPI_GED] = 9,
 193    [VIRT_MMIO] = 16, /* ...to 16 + NUM_VIRTIO_TRANSPORTS - 1 */
 194    [VIRT_GIC_V2M] = 48, /* ...to 48 + NUM_GICV2M_SPIS - 1 */
 195    [VIRT_SMMU] = 74,    /* ...to 74 + NUM_SMMU_IRQS - 1 */
 196    [VIRT_PLATFORM_BUS] = 112, /* ...to 112 + PLATFORM_BUS_NUM_IRQS -1 */
 197};
 198
 199static const char *valid_cpus[] = {
 200    ARM_CPU_TYPE_NAME("cortex-a7"),
 201    ARM_CPU_TYPE_NAME("cortex-a15"),
 202    ARM_CPU_TYPE_NAME("cortex-a53"),
 203    ARM_CPU_TYPE_NAME("cortex-a57"),
 204    ARM_CPU_TYPE_NAME("cortex-a72"),
 205    ARM_CPU_TYPE_NAME("cortex-a76"),
 206    ARM_CPU_TYPE_NAME("a64fx"),
 207    ARM_CPU_TYPE_NAME("neoverse-n1"),
 208    ARM_CPU_TYPE_NAME("host"),
 209    ARM_CPU_TYPE_NAME("max"),
 210};
 211
 212static bool cpu_type_valid(const char *cpu)
 213{
 214    int i;
 215
 216    for (i = 0; i < ARRAY_SIZE(valid_cpus); i++) {
 217        if (strcmp(cpu, valid_cpus[i]) == 0) {
 218            return true;
 219        }
 220    }
 221    return false;
 222}
 223
 224static void create_randomness(MachineState *ms, const char *node)
 225{
 226    struct {
 227        uint64_t kaslr;
 228        uint8_t rng[32];
 229    } seed;
 230
 231    if (qemu_guest_getrandom(&seed, sizeof(seed), NULL)) {
 232        return;
 233    }
 234    qemu_fdt_setprop_u64(ms->fdt, node, "kaslr-seed", seed.kaslr);
 235    qemu_fdt_setprop(ms->fdt, node, "rng-seed", seed.rng, sizeof(seed.rng));
 236}
 237
 238static void create_fdt(VirtMachineState *vms)
 239{
 240    MachineState *ms = MACHINE(vms);
 241    int nb_numa_nodes = ms->numa_state->num_nodes;
 242    void *fdt = create_device_tree(&vms->fdt_size);
 243
 244    if (!fdt) {
 245        error_report("create_device_tree() failed");
 246        exit(1);
 247    }
 248
 249    ms->fdt = fdt;
 250
 251    /* Header */
 252    qemu_fdt_setprop_string(fdt, "/", "compatible", "linux,dummy-virt");
 253    qemu_fdt_setprop_cell(fdt, "/", "#address-cells", 0x2);
 254    qemu_fdt_setprop_cell(fdt, "/", "#size-cells", 0x2);
 255
 256    /* /chosen must exist for load_dtb to fill in necessary properties later */
 257    qemu_fdt_add_subnode(fdt, "/chosen");
 258    if (vms->dtb_randomness) {
 259        create_randomness(ms, "/chosen");
 260    }
 261
 262    if (vms->secure) {
 263        qemu_fdt_add_subnode(fdt, "/secure-chosen");
 264        if (vms->dtb_randomness) {
 265            create_randomness(ms, "/secure-chosen");
 266        }
 267    }
 268
 269    /* Clock node, for the benefit of the UART. The kernel device tree
 270     * binding documentation claims the PL011 node clock properties are
 271     * optional but in practice if you omit them the kernel refuses to
 272     * probe for the device.
 273     */
 274    vms->clock_phandle = qemu_fdt_alloc_phandle(fdt);
 275    qemu_fdt_add_subnode(fdt, "/apb-pclk");
 276    qemu_fdt_setprop_string(fdt, "/apb-pclk", "compatible", "fixed-clock");
 277    qemu_fdt_setprop_cell(fdt, "/apb-pclk", "#clock-cells", 0x0);
 278    qemu_fdt_setprop_cell(fdt, "/apb-pclk", "clock-frequency", 24000000);
 279    qemu_fdt_setprop_string(fdt, "/apb-pclk", "clock-output-names",
 280                                "clk24mhz");
 281    qemu_fdt_setprop_cell(fdt, "/apb-pclk", "phandle", vms->clock_phandle);
 282
 283    if (nb_numa_nodes > 0 && ms->numa_state->have_numa_distance) {
 284        int size = nb_numa_nodes * nb_numa_nodes * 3 * sizeof(uint32_t);
 285        uint32_t *matrix = g_malloc0(size);
 286        int idx, i, j;
 287
 288        for (i = 0; i < nb_numa_nodes; i++) {
 289            for (j = 0; j < nb_numa_nodes; j++) {
 290                idx = (i * nb_numa_nodes + j) * 3;
 291                matrix[idx + 0] = cpu_to_be32(i);
 292                matrix[idx + 1] = cpu_to_be32(j);
 293                matrix[idx + 2] =
 294                    cpu_to_be32(ms->numa_state->nodes[i].distance[j]);
 295            }
 296        }
 297
 298        qemu_fdt_add_subnode(fdt, "/distance-map");
 299        qemu_fdt_setprop_string(fdt, "/distance-map", "compatible",
 300                                "numa-distance-map-v1");
 301        qemu_fdt_setprop(fdt, "/distance-map", "distance-matrix",
 302                         matrix, size);
 303        g_free(matrix);
 304    }
 305}
 306
 307static void fdt_add_timer_nodes(const VirtMachineState *vms)
 308{
 309    /* On real hardware these interrupts are level-triggered.
 310     * On KVM they were edge-triggered before host kernel version 4.4,
 311     * and level-triggered afterwards.
 312     * On emulated QEMU they are level-triggered.
 313     *
 314     * Getting the DTB info about them wrong is awkward for some
 315     * guest kernels:
 316     *  pre-4.8 ignore the DT and leave the interrupt configured
 317     *   with whatever the GIC reset value (or the bootloader) left it at
 318     *  4.8 before rc6 honour the incorrect data by programming it back
 319     *   into the GIC, causing problems
 320     *  4.8rc6 and later ignore the DT and always write "level triggered"
 321     *   into the GIC
 322     *
 323     * For backwards-compatibility, virt-2.8 and earlier will continue
 324     * to say these are edge-triggered, but later machines will report
 325     * the correct information.
 326     */
 327    ARMCPU *armcpu;
 328    VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
 329    uint32_t irqflags = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
 330    MachineState *ms = MACHINE(vms);
 331
 332    if (vmc->claim_edge_triggered_timers) {
 333        irqflags = GIC_FDT_IRQ_FLAGS_EDGE_LO_HI;
 334    }
 335
 336    if (vms->gic_version == VIRT_GIC_VERSION_2) {
 337        irqflags = deposit32(irqflags, GIC_FDT_IRQ_PPI_CPU_START,
 338                             GIC_FDT_IRQ_PPI_CPU_WIDTH,
 339                             (1 << MACHINE(vms)->smp.cpus) - 1);
 340    }
 341
 342    qemu_fdt_add_subnode(ms->fdt, "/timer");
 343
 344    armcpu = ARM_CPU(qemu_get_cpu(0));
 345    if (arm_feature(&armcpu->env, ARM_FEATURE_V8)) {
 346        const char compat[] = "arm,armv8-timer\0arm,armv7-timer";
 347        qemu_fdt_setprop(ms->fdt, "/timer", "compatible",
 348                         compat, sizeof(compat));
 349    } else {
 350        qemu_fdt_setprop_string(ms->fdt, "/timer", "compatible",
 351                                "arm,armv7-timer");
 352    }
 353    qemu_fdt_setprop(ms->fdt, "/timer", "always-on", NULL, 0);
 354    qemu_fdt_setprop_cells(ms->fdt, "/timer", "interrupts",
 355                       GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_S_EL1_IRQ, irqflags,
 356                       GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_NS_EL1_IRQ, irqflags,
 357                       GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_VIRT_IRQ, irqflags,
 358                       GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_NS_EL2_IRQ, irqflags);
 359}
 360
 361static void fdt_add_cpu_nodes(const VirtMachineState *vms)
 362{
 363    int cpu;
 364    int addr_cells = 1;
 365    const MachineState *ms = MACHINE(vms);
 366    const VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
 367    int smp_cpus = ms->smp.cpus;
 368
 369    /*
 370     * See Linux Documentation/devicetree/bindings/arm/cpus.yaml
 371     * On ARM v8 64-bit systems value should be set to 2,
 372     * that corresponds to the MPIDR_EL1 register size.
 373     * If MPIDR_EL1[63:32] value is equal to 0 on all CPUs
 374     * in the system, #address-cells can be set to 1, since
 375     * MPIDR_EL1[63:32] bits are not used for CPUs
 376     * identification.
 377     *
 378     * Here we actually don't know whether our system is 32- or 64-bit one.
 379     * The simplest way to go is to examine affinity IDs of all our CPUs. If
 380     * at least one of them has Aff3 populated, we set #address-cells to 2.
 381     */
 382    for (cpu = 0; cpu < smp_cpus; cpu++) {
 383        ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(cpu));
 384
 385        if (armcpu->mp_affinity & ARM_AFF3_MASK) {
 386            addr_cells = 2;
 387            break;
 388        }
 389    }
 390
 391    qemu_fdt_add_subnode(ms->fdt, "/cpus");
 392    qemu_fdt_setprop_cell(ms->fdt, "/cpus", "#address-cells", addr_cells);
 393    qemu_fdt_setprop_cell(ms->fdt, "/cpus", "#size-cells", 0x0);
 394
 395    for (cpu = smp_cpus - 1; cpu >= 0; cpu--) {
 396        char *nodename = g_strdup_printf("/cpus/cpu@%d", cpu);
 397        ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(cpu));
 398        CPUState *cs = CPU(armcpu);
 399
 400        qemu_fdt_add_subnode(ms->fdt, nodename);
 401        qemu_fdt_setprop_string(ms->fdt, nodename, "device_type", "cpu");
 402        qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
 403                                    armcpu->dtb_compatible);
 404
 405        if (vms->psci_conduit != QEMU_PSCI_CONDUIT_DISABLED && smp_cpus > 1) {
 406            qemu_fdt_setprop_string(ms->fdt, nodename,
 407                                        "enable-method", "psci");
 408        }
 409
 410        if (addr_cells == 2) {
 411            qemu_fdt_setprop_u64(ms->fdt, nodename, "reg",
 412                                 armcpu->mp_affinity);
 413        } else {
 414            qemu_fdt_setprop_cell(ms->fdt, nodename, "reg",
 415                                  armcpu->mp_affinity);
 416        }
 417
 418        if (ms->possible_cpus->cpus[cs->cpu_index].props.has_node_id) {
 419            qemu_fdt_setprop_cell(ms->fdt, nodename, "numa-node-id",
 420                ms->possible_cpus->cpus[cs->cpu_index].props.node_id);
 421        }
 422
 423        if (!vmc->no_cpu_topology) {
 424            qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle",
 425                                  qemu_fdt_alloc_phandle(ms->fdt));
 426        }
 427
 428        g_free(nodename);
 429    }
 430
 431    if (!vmc->no_cpu_topology) {
 432        /*
 433         * Add vCPU topology description through fdt node cpu-map.
 434         *
 435         * See Linux Documentation/devicetree/bindings/cpu/cpu-topology.txt
 436         * In a SMP system, the hierarchy of CPUs can be defined through
 437         * four entities that are used to describe the layout of CPUs in
 438         * the system: socket/cluster/core/thread.
 439         *
 440         * A socket node represents the boundary of system physical package
 441         * and its child nodes must be one or more cluster nodes. A system
 442         * can contain several layers of clustering within a single physical
 443         * package and cluster nodes can be contained in parent cluster nodes.
 444         *
 445         * Note: currently we only support one layer of clustering within
 446         * each physical package.
 447         */
 448        qemu_fdt_add_subnode(ms->fdt, "/cpus/cpu-map");
 449
 450        for (cpu = smp_cpus - 1; cpu >= 0; cpu--) {
 451            char *cpu_path = g_strdup_printf("/cpus/cpu@%d", cpu);
 452            char *map_path;
 453
 454            if (ms->smp.threads > 1) {
 455                map_path = g_strdup_printf(
 456                    "/cpus/cpu-map/socket%d/cluster%d/core%d/thread%d",
 457                    cpu / (ms->smp.clusters * ms->smp.cores * ms->smp.threads),
 458                    (cpu / (ms->smp.cores * ms->smp.threads)) % ms->smp.clusters,
 459                    (cpu / ms->smp.threads) % ms->smp.cores,
 460                    cpu % ms->smp.threads);
 461            } else {
 462                map_path = g_strdup_printf(
 463                    "/cpus/cpu-map/socket%d/cluster%d/core%d",
 464                    cpu / (ms->smp.clusters * ms->smp.cores),
 465                    (cpu / ms->smp.cores) % ms->smp.clusters,
 466                    cpu % ms->smp.cores);
 467            }
 468            qemu_fdt_add_path(ms->fdt, map_path);
 469            qemu_fdt_setprop_phandle(ms->fdt, map_path, "cpu", cpu_path);
 470
 471            g_free(map_path);
 472            g_free(cpu_path);
 473        }
 474    }
 475}
 476
 477static void fdt_add_its_gic_node(VirtMachineState *vms)
 478{
 479    char *nodename;
 480    MachineState *ms = MACHINE(vms);
 481
 482    vms->msi_phandle = qemu_fdt_alloc_phandle(ms->fdt);
 483    nodename = g_strdup_printf("/intc/its@%" PRIx64,
 484                               vms->memmap[VIRT_GIC_ITS].base);
 485    qemu_fdt_add_subnode(ms->fdt, nodename);
 486    qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
 487                            "arm,gic-v3-its");
 488    qemu_fdt_setprop(ms->fdt, nodename, "msi-controller", NULL, 0);
 489    qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
 490                                 2, vms->memmap[VIRT_GIC_ITS].base,
 491                                 2, vms->memmap[VIRT_GIC_ITS].size);
 492    qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", vms->msi_phandle);
 493    g_free(nodename);
 494}
 495
 496static void fdt_add_v2m_gic_node(VirtMachineState *vms)
 497{
 498    MachineState *ms = MACHINE(vms);
 499    char *nodename;
 500
 501    nodename = g_strdup_printf("/intc/v2m@%" PRIx64,
 502                               vms->memmap[VIRT_GIC_V2M].base);
 503    vms->msi_phandle = qemu_fdt_alloc_phandle(ms->fdt);
 504    qemu_fdt_add_subnode(ms->fdt, nodename);
 505    qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
 506                            "arm,gic-v2m-frame");
 507    qemu_fdt_setprop(ms->fdt, nodename, "msi-controller", NULL, 0);
 508    qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
 509                                 2, vms->memmap[VIRT_GIC_V2M].base,
 510                                 2, vms->memmap[VIRT_GIC_V2M].size);
 511    qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", vms->msi_phandle);
 512    g_free(nodename);
 513}
 514
 515static void fdt_add_gic_node(VirtMachineState *vms)
 516{
 517    MachineState *ms = MACHINE(vms);
 518    char *nodename;
 519
 520    vms->gic_phandle = qemu_fdt_alloc_phandle(ms->fdt);
 521    qemu_fdt_setprop_cell(ms->fdt, "/", "interrupt-parent", vms->gic_phandle);
 522
 523    nodename = g_strdup_printf("/intc@%" PRIx64,
 524                               vms->memmap[VIRT_GIC_DIST].base);
 525    qemu_fdt_add_subnode(ms->fdt, nodename);
 526    qemu_fdt_setprop_cell(ms->fdt, nodename, "#interrupt-cells", 3);
 527    qemu_fdt_setprop(ms->fdt, nodename, "interrupt-controller", NULL, 0);
 528    qemu_fdt_setprop_cell(ms->fdt, nodename, "#address-cells", 0x2);
 529    qemu_fdt_setprop_cell(ms->fdt, nodename, "#size-cells", 0x2);
 530    qemu_fdt_setprop(ms->fdt, nodename, "ranges", NULL, 0);
 531    if (vms->gic_version != VIRT_GIC_VERSION_2) {
 532        int nb_redist_regions = virt_gicv3_redist_region_count(vms);
 533
 534        qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
 535                                "arm,gic-v3");
 536
 537        qemu_fdt_setprop_cell(ms->fdt, nodename,
 538                              "#redistributor-regions", nb_redist_regions);
 539
 540        if (nb_redist_regions == 1) {
 541            qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
 542                                         2, vms->memmap[VIRT_GIC_DIST].base,
 543                                         2, vms->memmap[VIRT_GIC_DIST].size,
 544                                         2, vms->memmap[VIRT_GIC_REDIST].base,
 545                                         2, vms->memmap[VIRT_GIC_REDIST].size);
 546        } else {
 547            qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
 548                                 2, vms->memmap[VIRT_GIC_DIST].base,
 549                                 2, vms->memmap[VIRT_GIC_DIST].size,
 550                                 2, vms->memmap[VIRT_GIC_REDIST].base,
 551                                 2, vms->memmap[VIRT_GIC_REDIST].size,
 552                                 2, vms->memmap[VIRT_HIGH_GIC_REDIST2].base,
 553                                 2, vms->memmap[VIRT_HIGH_GIC_REDIST2].size);
 554        }
 555
 556        if (vms->virt) {
 557            qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
 558                                   GIC_FDT_IRQ_TYPE_PPI, ARCH_GIC_MAINT_IRQ,
 559                                   GIC_FDT_IRQ_FLAGS_LEVEL_HI);
 560        }
 561    } else {
 562        /* 'cortex-a15-gic' means 'GIC v2' */
 563        qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
 564                                "arm,cortex-a15-gic");
 565        if (!vms->virt) {
 566            qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
 567                                         2, vms->memmap[VIRT_GIC_DIST].base,
 568                                         2, vms->memmap[VIRT_GIC_DIST].size,
 569                                         2, vms->memmap[VIRT_GIC_CPU].base,
 570                                         2, vms->memmap[VIRT_GIC_CPU].size);
 571        } else {
 572            qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
 573                                         2, vms->memmap[VIRT_GIC_DIST].base,
 574                                         2, vms->memmap[VIRT_GIC_DIST].size,
 575                                         2, vms->memmap[VIRT_GIC_CPU].base,
 576                                         2, vms->memmap[VIRT_GIC_CPU].size,
 577                                         2, vms->memmap[VIRT_GIC_HYP].base,
 578                                         2, vms->memmap[VIRT_GIC_HYP].size,
 579                                         2, vms->memmap[VIRT_GIC_VCPU].base,
 580                                         2, vms->memmap[VIRT_GIC_VCPU].size);
 581            qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
 582                                   GIC_FDT_IRQ_TYPE_PPI, ARCH_GIC_MAINT_IRQ,
 583                                   GIC_FDT_IRQ_FLAGS_LEVEL_HI);
 584        }
 585    }
 586
 587    qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", vms->gic_phandle);
 588    g_free(nodename);
 589}
 590
 591static void fdt_add_pmu_nodes(const VirtMachineState *vms)
 592{
 593    ARMCPU *armcpu = ARM_CPU(first_cpu);
 594    uint32_t irqflags = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
 595    MachineState *ms = MACHINE(vms);
 596
 597    if (!arm_feature(&armcpu->env, ARM_FEATURE_PMU)) {
 598        assert(!object_property_get_bool(OBJECT(armcpu), "pmu", NULL));
 599        return;
 600    }
 601
 602    if (vms->gic_version == VIRT_GIC_VERSION_2) {
 603        irqflags = deposit32(irqflags, GIC_FDT_IRQ_PPI_CPU_START,
 604                             GIC_FDT_IRQ_PPI_CPU_WIDTH,
 605                             (1 << MACHINE(vms)->smp.cpus) - 1);
 606    }
 607
 608    qemu_fdt_add_subnode(ms->fdt, "/pmu");
 609    if (arm_feature(&armcpu->env, ARM_FEATURE_V8)) {
 610        const char compat[] = "arm,armv8-pmuv3";
 611        qemu_fdt_setprop(ms->fdt, "/pmu", "compatible",
 612                         compat, sizeof(compat));
 613        qemu_fdt_setprop_cells(ms->fdt, "/pmu", "interrupts",
 614                               GIC_FDT_IRQ_TYPE_PPI, VIRTUAL_PMU_IRQ, irqflags);
 615    }
 616}
 617
 618static inline DeviceState *create_acpi_ged(VirtMachineState *vms)
 619{
 620    DeviceState *dev;
 621    MachineState *ms = MACHINE(vms);
 622    int irq = vms->irqmap[VIRT_ACPI_GED];
 623    uint32_t event = ACPI_GED_PWR_DOWN_EVT;
 624
 625    if (ms->ram_slots) {
 626        event |= ACPI_GED_MEM_HOTPLUG_EVT;
 627    }
 628
 629    if (ms->nvdimms_state->is_enabled) {
 630        event |= ACPI_GED_NVDIMM_HOTPLUG_EVT;
 631    }
 632
 633    dev = qdev_new(TYPE_ACPI_GED);
 634    qdev_prop_set_uint32(dev, "ged-event", event);
 635
 636    sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_ACPI_GED].base);
 637    sysbus_mmio_map(SYS_BUS_DEVICE(dev), 1, vms->memmap[VIRT_PCDIMM_ACPI].base);
 638    sysbus_connect_irq(SYS_BUS_DEVICE(dev), 0, qdev_get_gpio_in(vms->gic, irq));
 639
 640    sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
 641
 642    return dev;
 643}
 644
 645static void create_its(VirtMachineState *vms)
 646{
 647    const char *itsclass = its_class_name();
 648    DeviceState *dev;
 649
 650    if (!strcmp(itsclass, "arm-gicv3-its")) {
 651        if (!vms->tcg_its) {
 652            itsclass = NULL;
 653        }
 654    }
 655
 656    if (!itsclass) {
 657        /* Do nothing if not supported */
 658        return;
 659    }
 660
 661    dev = qdev_new(itsclass);
 662
 663    object_property_set_link(OBJECT(dev), "parent-gicv3", OBJECT(vms->gic),
 664                             &error_abort);
 665    sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
 666    sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_GIC_ITS].base);
 667
 668    fdt_add_its_gic_node(vms);
 669    vms->msi_controller = VIRT_MSI_CTRL_ITS;
 670}
 671
 672static void create_v2m(VirtMachineState *vms)
 673{
 674    int i;
 675    int irq = vms->irqmap[VIRT_GIC_V2M];
 676    DeviceState *dev;
 677
 678    dev = qdev_new("arm-gicv2m");
 679    sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_GIC_V2M].base);
 680    qdev_prop_set_uint32(dev, "base-spi", irq);
 681    qdev_prop_set_uint32(dev, "num-spi", NUM_GICV2M_SPIS);
 682    sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
 683
 684    for (i = 0; i < NUM_GICV2M_SPIS; i++) {
 685        sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
 686                           qdev_get_gpio_in(vms->gic, irq + i));
 687    }
 688
 689    fdt_add_v2m_gic_node(vms);
 690    vms->msi_controller = VIRT_MSI_CTRL_GICV2M;
 691}
 692
 693static void create_gic(VirtMachineState *vms, MemoryRegion *mem)
 694{
 695    MachineState *ms = MACHINE(vms);
 696    /* We create a standalone GIC */
 697    SysBusDevice *gicbusdev;
 698    const char *gictype;
 699    int i;
 700    unsigned int smp_cpus = ms->smp.cpus;
 701    uint32_t nb_redist_regions = 0;
 702    int revision;
 703
 704    if (vms->gic_version == VIRT_GIC_VERSION_2) {
 705        gictype = gic_class_name();
 706    } else {
 707        gictype = gicv3_class_name();
 708    }
 709
 710    switch (vms->gic_version) {
 711    case VIRT_GIC_VERSION_2:
 712        revision = 2;
 713        break;
 714    case VIRT_GIC_VERSION_3:
 715        revision = 3;
 716        break;
 717    case VIRT_GIC_VERSION_4:
 718        revision = 4;
 719        break;
 720    default:
 721        g_assert_not_reached();
 722    }
 723    vms->gic = qdev_new(gictype);
 724    qdev_prop_set_uint32(vms->gic, "revision", revision);
 725    qdev_prop_set_uint32(vms->gic, "num-cpu", smp_cpus);
 726    /* Note that the num-irq property counts both internal and external
 727     * interrupts; there are always 32 of the former (mandated by GIC spec).
 728     */
 729    qdev_prop_set_uint32(vms->gic, "num-irq", NUM_IRQS + 32);
 730    if (!kvm_irqchip_in_kernel()) {
 731        qdev_prop_set_bit(vms->gic, "has-security-extensions", vms->secure);
 732    }
 733
 734    if (vms->gic_version != VIRT_GIC_VERSION_2) {
 735        uint32_t redist0_capacity = virt_redist_capacity(vms, VIRT_GIC_REDIST);
 736        uint32_t redist0_count = MIN(smp_cpus, redist0_capacity);
 737
 738        nb_redist_regions = virt_gicv3_redist_region_count(vms);
 739
 740        qdev_prop_set_uint32(vms->gic, "len-redist-region-count",
 741                             nb_redist_regions);
 742        qdev_prop_set_uint32(vms->gic, "redist-region-count[0]", redist0_count);
 743
 744        if (!kvm_irqchip_in_kernel()) {
 745            if (vms->tcg_its) {
 746                object_property_set_link(OBJECT(vms->gic), "sysmem",
 747                                         OBJECT(mem), &error_fatal);
 748                qdev_prop_set_bit(vms->gic, "has-lpi", true);
 749            }
 750        }
 751
 752        if (nb_redist_regions == 2) {
 753            uint32_t redist1_capacity =
 754                virt_redist_capacity(vms, VIRT_HIGH_GIC_REDIST2);
 755
 756            qdev_prop_set_uint32(vms->gic, "redist-region-count[1]",
 757                MIN(smp_cpus - redist0_count, redist1_capacity));
 758        }
 759    } else {
 760        if (!kvm_irqchip_in_kernel()) {
 761            qdev_prop_set_bit(vms->gic, "has-virtualization-extensions",
 762                              vms->virt);
 763        }
 764    }
 765    gicbusdev = SYS_BUS_DEVICE(vms->gic);
 766    sysbus_realize_and_unref(gicbusdev, &error_fatal);
 767    sysbus_mmio_map(gicbusdev, 0, vms->memmap[VIRT_GIC_DIST].base);
 768    if (vms->gic_version != VIRT_GIC_VERSION_2) {
 769        sysbus_mmio_map(gicbusdev, 1, vms->memmap[VIRT_GIC_REDIST].base);
 770        if (nb_redist_regions == 2) {
 771            sysbus_mmio_map(gicbusdev, 2,
 772                            vms->memmap[VIRT_HIGH_GIC_REDIST2].base);
 773        }
 774    } else {
 775        sysbus_mmio_map(gicbusdev, 1, vms->memmap[VIRT_GIC_CPU].base);
 776        if (vms->virt) {
 777            sysbus_mmio_map(gicbusdev, 2, vms->memmap[VIRT_GIC_HYP].base);
 778            sysbus_mmio_map(gicbusdev, 3, vms->memmap[VIRT_GIC_VCPU].base);
 779        }
 780    }
 781
 782    /* Wire the outputs from each CPU's generic timer and the GICv3
 783     * maintenance interrupt signal to the appropriate GIC PPI inputs,
 784     * and the GIC's IRQ/FIQ/VIRQ/VFIQ interrupt outputs to the CPU's inputs.
 785     */
 786    for (i = 0; i < smp_cpus; i++) {
 787        DeviceState *cpudev = DEVICE(qemu_get_cpu(i));
 788        int ppibase = NUM_IRQS + i * GIC_INTERNAL + GIC_NR_SGIS;
 789        int irq;
 790        /* Mapping from the output timer irq lines from the CPU to the
 791         * GIC PPI inputs we use for the virt board.
 792         */
 793        const int timer_irq[] = {
 794            [GTIMER_PHYS] = ARCH_TIMER_NS_EL1_IRQ,
 795            [GTIMER_VIRT] = ARCH_TIMER_VIRT_IRQ,
 796            [GTIMER_HYP]  = ARCH_TIMER_NS_EL2_IRQ,
 797            [GTIMER_SEC]  = ARCH_TIMER_S_EL1_IRQ,
 798        };
 799
 800        for (irq = 0; irq < ARRAY_SIZE(timer_irq); irq++) {
 801            qdev_connect_gpio_out(cpudev, irq,
 802                                  qdev_get_gpio_in(vms->gic,
 803                                                   ppibase + timer_irq[irq]));
 804        }
 805
 806        if (vms->gic_version != VIRT_GIC_VERSION_2) {
 807            qemu_irq irq = qdev_get_gpio_in(vms->gic,
 808                                            ppibase + ARCH_GIC_MAINT_IRQ);
 809            qdev_connect_gpio_out_named(cpudev, "gicv3-maintenance-interrupt",
 810                                        0, irq);
 811        } else if (vms->virt) {
 812            qemu_irq irq = qdev_get_gpio_in(vms->gic,
 813                                            ppibase + ARCH_GIC_MAINT_IRQ);
 814            sysbus_connect_irq(gicbusdev, i + 4 * smp_cpus, irq);
 815        }
 816
 817        qdev_connect_gpio_out_named(cpudev, "pmu-interrupt", 0,
 818                                    qdev_get_gpio_in(vms->gic, ppibase
 819                                                     + VIRTUAL_PMU_IRQ));
 820
 821        sysbus_connect_irq(gicbusdev, i, qdev_get_gpio_in(cpudev, ARM_CPU_IRQ));
 822        sysbus_connect_irq(gicbusdev, i + smp_cpus,
 823                           qdev_get_gpio_in(cpudev, ARM_CPU_FIQ));
 824        sysbus_connect_irq(gicbusdev, i + 2 * smp_cpus,
 825                           qdev_get_gpio_in(cpudev, ARM_CPU_VIRQ));
 826        sysbus_connect_irq(gicbusdev, i + 3 * smp_cpus,
 827                           qdev_get_gpio_in(cpudev, ARM_CPU_VFIQ));
 828    }
 829
 830    fdt_add_gic_node(vms);
 831
 832    if (vms->gic_version != VIRT_GIC_VERSION_2 && vms->its) {
 833        create_its(vms);
 834    } else if (vms->gic_version == VIRT_GIC_VERSION_2) {
 835        create_v2m(vms);
 836    }
 837}
 838
 839static void create_uart(const VirtMachineState *vms, int uart,
 840                        MemoryRegion *mem, Chardev *chr)
 841{
 842    char *nodename;
 843    hwaddr base = vms->memmap[uart].base;
 844    hwaddr size = vms->memmap[uart].size;
 845    int irq = vms->irqmap[uart];
 846    const char compat[] = "arm,pl011\0arm,primecell";
 847    const char clocknames[] = "uartclk\0apb_pclk";
 848    DeviceState *dev = qdev_new(TYPE_PL011);
 849    SysBusDevice *s = SYS_BUS_DEVICE(dev);
 850    MachineState *ms = MACHINE(vms);
 851
 852    qdev_prop_set_chr(dev, "chardev", chr);
 853    sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
 854    memory_region_add_subregion(mem, base,
 855                                sysbus_mmio_get_region(s, 0));
 856    sysbus_connect_irq(s, 0, qdev_get_gpio_in(vms->gic, irq));
 857
 858    nodename = g_strdup_printf("/pl011@%" PRIx64, base);
 859    qemu_fdt_add_subnode(ms->fdt, nodename);
 860    /* Note that we can't use setprop_string because of the embedded NUL */
 861    qemu_fdt_setprop(ms->fdt, nodename, "compatible",
 862                         compat, sizeof(compat));
 863    qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
 864                                     2, base, 2, size);
 865    qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
 866                               GIC_FDT_IRQ_TYPE_SPI, irq,
 867                               GIC_FDT_IRQ_FLAGS_LEVEL_HI);
 868    qemu_fdt_setprop_cells(ms->fdt, nodename, "clocks",
 869                               vms->clock_phandle, vms->clock_phandle);
 870    qemu_fdt_setprop(ms->fdt, nodename, "clock-names",
 871                         clocknames, sizeof(clocknames));
 872
 873    if (uart == VIRT_UART) {
 874        qemu_fdt_setprop_string(ms->fdt, "/chosen", "stdout-path", nodename);
 875    } else {
 876        /* Mark as not usable by the normal world */
 877        qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
 878        qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
 879
 880        qemu_fdt_setprop_string(ms->fdt, "/secure-chosen", "stdout-path",
 881                                nodename);
 882    }
 883
 884    g_free(nodename);
 885}
 886
 887static void create_rtc(const VirtMachineState *vms)
 888{
 889    char *nodename;
 890    hwaddr base = vms->memmap[VIRT_RTC].base;
 891    hwaddr size = vms->memmap[VIRT_RTC].size;
 892    int irq = vms->irqmap[VIRT_RTC];
 893    const char compat[] = "arm,pl031\0arm,primecell";
 894    MachineState *ms = MACHINE(vms);
 895
 896    sysbus_create_simple("pl031", base, qdev_get_gpio_in(vms->gic, irq));
 897
 898    nodename = g_strdup_printf("/pl031@%" PRIx64, base);
 899    qemu_fdt_add_subnode(ms->fdt, nodename);
 900    qemu_fdt_setprop(ms->fdt, nodename, "compatible", compat, sizeof(compat));
 901    qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
 902                                 2, base, 2, size);
 903    qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
 904                           GIC_FDT_IRQ_TYPE_SPI, irq,
 905                           GIC_FDT_IRQ_FLAGS_LEVEL_HI);
 906    qemu_fdt_setprop_cell(ms->fdt, nodename, "clocks", vms->clock_phandle);
 907    qemu_fdt_setprop_string(ms->fdt, nodename, "clock-names", "apb_pclk");
 908    g_free(nodename);
 909}
 910
 911static DeviceState *gpio_key_dev;
 912static void virt_powerdown_req(Notifier *n, void *opaque)
 913{
 914    VirtMachineState *s = container_of(n, VirtMachineState, powerdown_notifier);
 915
 916    if (s->acpi_dev) {
 917        acpi_send_event(s->acpi_dev, ACPI_POWER_DOWN_STATUS);
 918    } else {
 919        /* use gpio Pin 3 for power button event */
 920        qemu_set_irq(qdev_get_gpio_in(gpio_key_dev, 0), 1);
 921    }
 922}
 923
 924static void create_gpio_keys(char *fdt, DeviceState *pl061_dev,
 925                             uint32_t phandle)
 926{
 927    gpio_key_dev = sysbus_create_simple("gpio-key", -1,
 928                                        qdev_get_gpio_in(pl061_dev, 3));
 929
 930    qemu_fdt_add_subnode(fdt, "/gpio-keys");
 931    qemu_fdt_setprop_string(fdt, "/gpio-keys", "compatible", "gpio-keys");
 932
 933    qemu_fdt_add_subnode(fdt, "/gpio-keys/poweroff");
 934    qemu_fdt_setprop_string(fdt, "/gpio-keys/poweroff",
 935                            "label", "GPIO Key Poweroff");
 936    qemu_fdt_setprop_cell(fdt, "/gpio-keys/poweroff", "linux,code",
 937                          KEY_POWER);
 938    qemu_fdt_setprop_cells(fdt, "/gpio-keys/poweroff",
 939                           "gpios", phandle, 3, 0);
 940}
 941
 942#define SECURE_GPIO_POWEROFF 0
 943#define SECURE_GPIO_RESET    1
 944
 945static void create_secure_gpio_pwr(char *fdt, DeviceState *pl061_dev,
 946                                   uint32_t phandle)
 947{
 948    DeviceState *gpio_pwr_dev;
 949
 950    /* gpio-pwr */
 951    gpio_pwr_dev = sysbus_create_simple("gpio-pwr", -1, NULL);
 952
 953    /* connect secure pl061 to gpio-pwr */
 954    qdev_connect_gpio_out(pl061_dev, SECURE_GPIO_RESET,
 955                          qdev_get_gpio_in_named(gpio_pwr_dev, "reset", 0));
 956    qdev_connect_gpio_out(pl061_dev, SECURE_GPIO_POWEROFF,
 957                          qdev_get_gpio_in_named(gpio_pwr_dev, "shutdown", 0));
 958
 959    qemu_fdt_add_subnode(fdt, "/gpio-poweroff");
 960    qemu_fdt_setprop_string(fdt, "/gpio-poweroff", "compatible",
 961                            "gpio-poweroff");
 962    qemu_fdt_setprop_cells(fdt, "/gpio-poweroff",
 963                           "gpios", phandle, SECURE_GPIO_POWEROFF, 0);
 964    qemu_fdt_setprop_string(fdt, "/gpio-poweroff", "status", "disabled");
 965    qemu_fdt_setprop_string(fdt, "/gpio-poweroff", "secure-status",
 966                            "okay");
 967
 968    qemu_fdt_add_subnode(fdt, "/gpio-restart");
 969    qemu_fdt_setprop_string(fdt, "/gpio-restart", "compatible",
 970                            "gpio-restart");
 971    qemu_fdt_setprop_cells(fdt, "/gpio-restart",
 972                           "gpios", phandle, SECURE_GPIO_RESET, 0);
 973    qemu_fdt_setprop_string(fdt, "/gpio-restart", "status", "disabled");
 974    qemu_fdt_setprop_string(fdt, "/gpio-restart", "secure-status",
 975                            "okay");
 976}
 977
 978static void create_gpio_devices(const VirtMachineState *vms, int gpio,
 979                                MemoryRegion *mem)
 980{
 981    char *nodename;
 982    DeviceState *pl061_dev;
 983    hwaddr base = vms->memmap[gpio].base;
 984    hwaddr size = vms->memmap[gpio].size;
 985    int irq = vms->irqmap[gpio];
 986    const char compat[] = "arm,pl061\0arm,primecell";
 987    SysBusDevice *s;
 988    MachineState *ms = MACHINE(vms);
 989
 990    pl061_dev = qdev_new("pl061");
 991    /* Pull lines down to 0 if not driven by the PL061 */
 992    qdev_prop_set_uint32(pl061_dev, "pullups", 0);
 993    qdev_prop_set_uint32(pl061_dev, "pulldowns", 0xff);
 994    s = SYS_BUS_DEVICE(pl061_dev);
 995    sysbus_realize_and_unref(s, &error_fatal);
 996    memory_region_add_subregion(mem, base, sysbus_mmio_get_region(s, 0));
 997    sysbus_connect_irq(s, 0, qdev_get_gpio_in(vms->gic, irq));
 998
 999    uint32_t phandle = qemu_fdt_alloc_phandle(ms->fdt);
1000    nodename = g_strdup_printf("/pl061@%" PRIx64, base);
1001    qemu_fdt_add_subnode(ms->fdt, nodename);
1002    qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1003                                 2, base, 2, size);
1004    qemu_fdt_setprop(ms->fdt, nodename, "compatible", compat, sizeof(compat));
1005    qemu_fdt_setprop_cell(ms->fdt, nodename, "#gpio-cells", 2);
1006    qemu_fdt_setprop(ms->fdt, nodename, "gpio-controller", NULL, 0);
1007    qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
1008                           GIC_FDT_IRQ_TYPE_SPI, irq,
1009                           GIC_FDT_IRQ_FLAGS_LEVEL_HI);
1010    qemu_fdt_setprop_cell(ms->fdt, nodename, "clocks", vms->clock_phandle);
1011    qemu_fdt_setprop_string(ms->fdt, nodename, "clock-names", "apb_pclk");
1012    qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", phandle);
1013
1014    if (gpio != VIRT_GPIO) {
1015        /* Mark as not usable by the normal world */
1016        qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
1017        qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
1018    }
1019    g_free(nodename);
1020
1021    /* Child gpio devices */
1022    if (gpio == VIRT_GPIO) {
1023        create_gpio_keys(ms->fdt, pl061_dev, phandle);
1024    } else {
1025        create_secure_gpio_pwr(ms->fdt, pl061_dev, phandle);
1026    }
1027}
1028
1029static void create_virtio_devices(const VirtMachineState *vms)
1030{
1031    int i;
1032    hwaddr size = vms->memmap[VIRT_MMIO].size;
1033    MachineState *ms = MACHINE(vms);
1034
1035    /* We create the transports in forwards order. Since qbus_realize()
1036     * prepends (not appends) new child buses, the incrementing loop below will
1037     * create a list of virtio-mmio buses with decreasing base addresses.
1038     *
1039     * When a -device option is processed from the command line,
1040     * qbus_find_recursive() picks the next free virtio-mmio bus in forwards
1041     * order. The upshot is that -device options in increasing command line
1042     * order are mapped to virtio-mmio buses with decreasing base addresses.
1043     *
1044     * When this code was originally written, that arrangement ensured that the
1045     * guest Linux kernel would give the lowest "name" (/dev/vda, eth0, etc) to
1046     * the first -device on the command line. (The end-to-end order is a
1047     * function of this loop, qbus_realize(), qbus_find_recursive(), and the
1048     * guest kernel's name-to-address assignment strategy.)
1049     *
1050     * Meanwhile, the kernel's traversal seems to have been reversed; see eg.
1051     * the message, if not necessarily the code, of commit 70161ff336.
1052     * Therefore the loop now establishes the inverse of the original intent.
1053     *
1054     * Unfortunately, we can't counteract the kernel change by reversing the
1055     * loop; it would break existing command lines.
1056     *
1057     * In any case, the kernel makes no guarantee about the stability of
1058     * enumeration order of virtio devices (as demonstrated by it changing
1059     * between kernel versions). For reliable and stable identification
1060     * of disks users must use UUIDs or similar mechanisms.
1061     */
1062    for (i = 0; i < NUM_VIRTIO_TRANSPORTS; i++) {
1063        int irq = vms->irqmap[VIRT_MMIO] + i;
1064        hwaddr base = vms->memmap[VIRT_MMIO].base + i * size;
1065
1066        sysbus_create_simple("virtio-mmio", base,
1067                             qdev_get_gpio_in(vms->gic, irq));
1068    }
1069
1070    /* We add dtb nodes in reverse order so that they appear in the finished
1071     * device tree lowest address first.
1072     *
1073     * Note that this mapping is independent of the loop above. The previous
1074     * loop influences virtio device to virtio transport assignment, whereas
1075     * this loop controls how virtio transports are laid out in the dtb.
1076     */
1077    for (i = NUM_VIRTIO_TRANSPORTS - 1; i >= 0; i--) {
1078        char *nodename;
1079        int irq = vms->irqmap[VIRT_MMIO] + i;
1080        hwaddr base = vms->memmap[VIRT_MMIO].base + i * size;
1081
1082        nodename = g_strdup_printf("/virtio_mmio@%" PRIx64, base);
1083        qemu_fdt_add_subnode(ms->fdt, nodename);
1084        qemu_fdt_setprop_string(ms->fdt, nodename,
1085                                "compatible", "virtio,mmio");
1086        qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1087                                     2, base, 2, size);
1088        qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
1089                               GIC_FDT_IRQ_TYPE_SPI, irq,
1090                               GIC_FDT_IRQ_FLAGS_EDGE_LO_HI);
1091        qemu_fdt_setprop(ms->fdt, nodename, "dma-coherent", NULL, 0);
1092        g_free(nodename);
1093    }
1094}
1095
1096#define VIRT_FLASH_SECTOR_SIZE (256 * KiB)
1097
1098static PFlashCFI01 *virt_flash_create1(VirtMachineState *vms,
1099                                        const char *name,
1100                                        const char *alias_prop_name)
1101{
1102    /*
1103     * Create a single flash device.  We use the same parameters as
1104     * the flash devices on the Versatile Express board.
1105     */
1106    DeviceState *dev = qdev_new(TYPE_PFLASH_CFI01);
1107
1108    qdev_prop_set_uint64(dev, "sector-length", VIRT_FLASH_SECTOR_SIZE);
1109    qdev_prop_set_uint8(dev, "width", 4);
1110    qdev_prop_set_uint8(dev, "device-width", 2);
1111    qdev_prop_set_bit(dev, "big-endian", false);
1112    qdev_prop_set_uint16(dev, "id0", 0x89);
1113    qdev_prop_set_uint16(dev, "id1", 0x18);
1114    qdev_prop_set_uint16(dev, "id2", 0x00);
1115    qdev_prop_set_uint16(dev, "id3", 0x00);
1116    qdev_prop_set_string(dev, "name", name);
1117    object_property_add_child(OBJECT(vms), name, OBJECT(dev));
1118    object_property_add_alias(OBJECT(vms), alias_prop_name,
1119                              OBJECT(dev), "drive");
1120    return PFLASH_CFI01(dev);
1121}
1122
1123static void virt_flash_create(VirtMachineState *vms)
1124{
1125    vms->flash[0] = virt_flash_create1(vms, "virt.flash0", "pflash0");
1126    vms->flash[1] = virt_flash_create1(vms, "virt.flash1", "pflash1");
1127}
1128
1129static void virt_flash_map1(PFlashCFI01 *flash,
1130                            hwaddr base, hwaddr size,
1131                            MemoryRegion *sysmem)
1132{
1133    DeviceState *dev = DEVICE(flash);
1134
1135    assert(QEMU_IS_ALIGNED(size, VIRT_FLASH_SECTOR_SIZE));
1136    assert(size / VIRT_FLASH_SECTOR_SIZE <= UINT32_MAX);
1137    qdev_prop_set_uint32(dev, "num-blocks", size / VIRT_FLASH_SECTOR_SIZE);
1138    sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1139
1140    memory_region_add_subregion(sysmem, base,
1141                                sysbus_mmio_get_region(SYS_BUS_DEVICE(dev),
1142                                                       0));
1143}
1144
1145static void virt_flash_map(VirtMachineState *vms,
1146                           MemoryRegion *sysmem,
1147                           MemoryRegion *secure_sysmem)
1148{
1149    /*
1150     * Map two flash devices to fill the VIRT_FLASH space in the memmap.
1151     * sysmem is the system memory space. secure_sysmem is the secure view
1152     * of the system, and the first flash device should be made visible only
1153     * there. The second flash device is visible to both secure and nonsecure.
1154     * If sysmem == secure_sysmem this means there is no separate Secure
1155     * address space and both flash devices are generally visible.
1156     */
1157    hwaddr flashsize = vms->memmap[VIRT_FLASH].size / 2;
1158    hwaddr flashbase = vms->memmap[VIRT_FLASH].base;
1159
1160    virt_flash_map1(vms->flash[0], flashbase, flashsize,
1161                    secure_sysmem);
1162    virt_flash_map1(vms->flash[1], flashbase + flashsize, flashsize,
1163                    sysmem);
1164}
1165
1166static void virt_flash_fdt(VirtMachineState *vms,
1167                           MemoryRegion *sysmem,
1168                           MemoryRegion *secure_sysmem)
1169{
1170    hwaddr flashsize = vms->memmap[VIRT_FLASH].size / 2;
1171    hwaddr flashbase = vms->memmap[VIRT_FLASH].base;
1172    MachineState *ms = MACHINE(vms);
1173    char *nodename;
1174
1175    if (sysmem == secure_sysmem) {
1176        /* Report both flash devices as a single node in the DT */
1177        nodename = g_strdup_printf("/flash@%" PRIx64, flashbase);
1178        qemu_fdt_add_subnode(ms->fdt, nodename);
1179        qemu_fdt_setprop_string(ms->fdt, nodename, "compatible", "cfi-flash");
1180        qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1181                                     2, flashbase, 2, flashsize,
1182                                     2, flashbase + flashsize, 2, flashsize);
1183        qemu_fdt_setprop_cell(ms->fdt, nodename, "bank-width", 4);
1184        g_free(nodename);
1185    } else {
1186        /*
1187         * Report the devices as separate nodes so we can mark one as
1188         * only visible to the secure world.
1189         */
1190        nodename = g_strdup_printf("/secflash@%" PRIx64, flashbase);
1191        qemu_fdt_add_subnode(ms->fdt, nodename);
1192        qemu_fdt_setprop_string(ms->fdt, nodename, "compatible", "cfi-flash");
1193        qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1194                                     2, flashbase, 2, flashsize);
1195        qemu_fdt_setprop_cell(ms->fdt, nodename, "bank-width", 4);
1196        qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
1197        qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
1198        g_free(nodename);
1199
1200        nodename = g_strdup_printf("/flash@%" PRIx64, flashbase + flashsize);
1201        qemu_fdt_add_subnode(ms->fdt, nodename);
1202        qemu_fdt_setprop_string(ms->fdt, nodename, "compatible", "cfi-flash");
1203        qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1204                                     2, flashbase + flashsize, 2, flashsize);
1205        qemu_fdt_setprop_cell(ms->fdt, nodename, "bank-width", 4);
1206        g_free(nodename);
1207    }
1208}
1209
1210static bool virt_firmware_init(VirtMachineState *vms,
1211                               MemoryRegion *sysmem,
1212                               MemoryRegion *secure_sysmem)
1213{
1214    int i;
1215    const char *bios_name;
1216    BlockBackend *pflash_blk0;
1217
1218    /* Map legacy -drive if=pflash to machine properties */
1219    for (i = 0; i < ARRAY_SIZE(vms->flash); i++) {
1220        pflash_cfi01_legacy_drive(vms->flash[i],
1221                                  drive_get(IF_PFLASH, 0, i));
1222    }
1223
1224    virt_flash_map(vms, sysmem, secure_sysmem);
1225
1226    pflash_blk0 = pflash_cfi01_get_blk(vms->flash[0]);
1227
1228    bios_name = MACHINE(vms)->firmware;
1229    if (bios_name) {
1230        char *fname;
1231        MemoryRegion *mr;
1232        int image_size;
1233
1234        if (pflash_blk0) {
1235            error_report("The contents of the first flash device may be "
1236                         "specified with -bios or with -drive if=pflash... "
1237                         "but you cannot use both options at once");
1238            exit(1);
1239        }
1240
1241        /* Fall back to -bios */
1242
1243        fname = qemu_find_file(QEMU_FILE_TYPE_BIOS, bios_name);
1244        if (!fname) {
1245            error_report("Could not find ROM image '%s'", bios_name);
1246            exit(1);
1247        }
1248        mr = sysbus_mmio_get_region(SYS_BUS_DEVICE(vms->flash[0]), 0);
1249        image_size = load_image_mr(fname, mr);
1250        g_free(fname);
1251        if (image_size < 0) {
1252            error_report("Could not load ROM image '%s'", bios_name);
1253            exit(1);
1254        }
1255    }
1256
1257    return pflash_blk0 || bios_name;
1258}
1259
1260static FWCfgState *create_fw_cfg(const VirtMachineState *vms, AddressSpace *as)
1261{
1262    MachineState *ms = MACHINE(vms);
1263    hwaddr base = vms->memmap[VIRT_FW_CFG].base;
1264    hwaddr size = vms->memmap[VIRT_FW_CFG].size;
1265    FWCfgState *fw_cfg;
1266    char *nodename;
1267
1268    fw_cfg = fw_cfg_init_mem_wide(base + 8, base, 8, base + 16, as);
1269    fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, (uint16_t)ms->smp.cpus);
1270
1271    nodename = g_strdup_printf("/fw-cfg@%" PRIx64, base);
1272    qemu_fdt_add_subnode(ms->fdt, nodename);
1273    qemu_fdt_setprop_string(ms->fdt, nodename,
1274                            "compatible", "qemu,fw-cfg-mmio");
1275    qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1276                                 2, base, 2, size);
1277    qemu_fdt_setprop(ms->fdt, nodename, "dma-coherent", NULL, 0);
1278    g_free(nodename);
1279    return fw_cfg;
1280}
1281
1282static void create_pcie_irq_map(const MachineState *ms,
1283                                uint32_t gic_phandle,
1284                                int first_irq, const char *nodename)
1285{
1286    int devfn, pin;
1287    uint32_t full_irq_map[4 * 4 * 10] = { 0 };
1288    uint32_t *irq_map = full_irq_map;
1289
1290    for (devfn = 0; devfn <= 0x18; devfn += 0x8) {
1291        for (pin = 0; pin < 4; pin++) {
1292            int irq_type = GIC_FDT_IRQ_TYPE_SPI;
1293            int irq_nr = first_irq + ((pin + PCI_SLOT(devfn)) % PCI_NUM_PINS);
1294            int irq_level = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
1295            int i;
1296
1297            uint32_t map[] = {
1298                devfn << 8, 0, 0,                           /* devfn */
1299                pin + 1,                                    /* PCI pin */
1300                gic_phandle, 0, 0, irq_type, irq_nr, irq_level }; /* GIC irq */
1301
1302            /* Convert map to big endian */
1303            for (i = 0; i < 10; i++) {
1304                irq_map[i] = cpu_to_be32(map[i]);
1305            }
1306            irq_map += 10;
1307        }
1308    }
1309
1310    qemu_fdt_setprop(ms->fdt, nodename, "interrupt-map",
1311                     full_irq_map, sizeof(full_irq_map));
1312
1313    qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupt-map-mask",
1314                           cpu_to_be16(PCI_DEVFN(3, 0)), /* Slot 3 */
1315                           0, 0,
1316                           0x7           /* PCI irq */);
1317}
1318
1319static void create_smmu(const VirtMachineState *vms,
1320                        PCIBus *bus)
1321{
1322    char *node;
1323    const char compat[] = "arm,smmu-v3";
1324    int irq =  vms->irqmap[VIRT_SMMU];
1325    int i;
1326    hwaddr base = vms->memmap[VIRT_SMMU].base;
1327    hwaddr size = vms->memmap[VIRT_SMMU].size;
1328    const char irq_names[] = "eventq\0priq\0cmdq-sync\0gerror";
1329    DeviceState *dev;
1330    MachineState *ms = MACHINE(vms);
1331
1332    if (vms->iommu != VIRT_IOMMU_SMMUV3 || !vms->iommu_phandle) {
1333        return;
1334    }
1335
1336    dev = qdev_new("arm-smmuv3");
1337
1338    object_property_set_link(OBJECT(dev), "primary-bus", OBJECT(bus),
1339                             &error_abort);
1340    sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1341    sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, base);
1342    for (i = 0; i < NUM_SMMU_IRQS; i++) {
1343        sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
1344                           qdev_get_gpio_in(vms->gic, irq + i));
1345    }
1346
1347    node = g_strdup_printf("/smmuv3@%" PRIx64, base);
1348    qemu_fdt_add_subnode(ms->fdt, node);
1349    qemu_fdt_setprop(ms->fdt, node, "compatible", compat, sizeof(compat));
1350    qemu_fdt_setprop_sized_cells(ms->fdt, node, "reg", 2, base, 2, size);
1351
1352    qemu_fdt_setprop_cells(ms->fdt, node, "interrupts",
1353            GIC_FDT_IRQ_TYPE_SPI, irq    , GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1354            GIC_FDT_IRQ_TYPE_SPI, irq + 1, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1355            GIC_FDT_IRQ_TYPE_SPI, irq + 2, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1356            GIC_FDT_IRQ_TYPE_SPI, irq + 3, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI);
1357
1358    qemu_fdt_setprop(ms->fdt, node, "interrupt-names", irq_names,
1359                     sizeof(irq_names));
1360
1361    qemu_fdt_setprop_cell(ms->fdt, node, "clocks", vms->clock_phandle);
1362    qemu_fdt_setprop_string(ms->fdt, node, "clock-names", "apb_pclk");
1363    qemu_fdt_setprop(ms->fdt, node, "dma-coherent", NULL, 0);
1364
1365    qemu_fdt_setprop_cell(ms->fdt, node, "#iommu-cells", 1);
1366
1367    qemu_fdt_setprop_cell(ms->fdt, node, "phandle", vms->iommu_phandle);
1368    g_free(node);
1369}
1370
1371static void create_virtio_iommu_dt_bindings(VirtMachineState *vms)
1372{
1373    const char compat[] = "virtio,pci-iommu";
1374    uint16_t bdf = vms->virtio_iommu_bdf;
1375    MachineState *ms = MACHINE(vms);
1376    char *node;
1377
1378    vms->iommu_phandle = qemu_fdt_alloc_phandle(ms->fdt);
1379
1380    node = g_strdup_printf("%s/virtio_iommu@%d", vms->pciehb_nodename, bdf);
1381    qemu_fdt_add_subnode(ms->fdt, node);
1382    qemu_fdt_setprop(ms->fdt, node, "compatible", compat, sizeof(compat));
1383    qemu_fdt_setprop_sized_cells(ms->fdt, node, "reg",
1384                                 1, bdf << 8, 1, 0, 1, 0,
1385                                 1, 0, 1, 0);
1386
1387    qemu_fdt_setprop_cell(ms->fdt, node, "#iommu-cells", 1);
1388    qemu_fdt_setprop_cell(ms->fdt, node, "phandle", vms->iommu_phandle);
1389    g_free(node);
1390
1391    qemu_fdt_setprop_cells(ms->fdt, vms->pciehb_nodename, "iommu-map",
1392                           0x0, vms->iommu_phandle, 0x0, bdf,
1393                           bdf + 1, vms->iommu_phandle, bdf + 1, 0xffff - bdf);
1394}
1395
1396static void create_pcie(VirtMachineState *vms)
1397{
1398    hwaddr base_mmio = vms->memmap[VIRT_PCIE_MMIO].base;
1399    hwaddr size_mmio = vms->memmap[VIRT_PCIE_MMIO].size;
1400    hwaddr base_mmio_high = vms->memmap[VIRT_HIGH_PCIE_MMIO].base;
1401    hwaddr size_mmio_high = vms->memmap[VIRT_HIGH_PCIE_MMIO].size;
1402    hwaddr base_pio = vms->memmap[VIRT_PCIE_PIO].base;
1403    hwaddr size_pio = vms->memmap[VIRT_PCIE_PIO].size;
1404    hwaddr base_ecam, size_ecam;
1405    hwaddr base = base_mmio;
1406    int nr_pcie_buses;
1407    int irq = vms->irqmap[VIRT_PCIE];
1408    MemoryRegion *mmio_alias;
1409    MemoryRegion *mmio_reg;
1410    MemoryRegion *ecam_alias;
1411    MemoryRegion *ecam_reg;
1412    DeviceState *dev;
1413    char *nodename;
1414    int i, ecam_id;
1415    PCIHostState *pci;
1416    MachineState *ms = MACHINE(vms);
1417
1418    dev = qdev_new(TYPE_GPEX_HOST);
1419    sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1420
1421    ecam_id = VIRT_ECAM_ID(vms->highmem_ecam);
1422    base_ecam = vms->memmap[ecam_id].base;
1423    size_ecam = vms->memmap[ecam_id].size;
1424    nr_pcie_buses = size_ecam / PCIE_MMCFG_SIZE_MIN;
1425    /* Map only the first size_ecam bytes of ECAM space */
1426    ecam_alias = g_new0(MemoryRegion, 1);
1427    ecam_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 0);
1428    memory_region_init_alias(ecam_alias, OBJECT(dev), "pcie-ecam",
1429                             ecam_reg, 0, size_ecam);
1430    memory_region_add_subregion(get_system_memory(), base_ecam, ecam_alias);
1431
1432    /* Map the MMIO window into system address space so as to expose
1433     * the section of PCI MMIO space which starts at the same base address
1434     * (ie 1:1 mapping for that part of PCI MMIO space visible through
1435     * the window).
1436     */
1437    mmio_alias = g_new0(MemoryRegion, 1);
1438    mmio_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 1);
1439    memory_region_init_alias(mmio_alias, OBJECT(dev), "pcie-mmio",
1440                             mmio_reg, base_mmio, size_mmio);
1441    memory_region_add_subregion(get_system_memory(), base_mmio, mmio_alias);
1442
1443    if (vms->highmem_mmio) {
1444        /* Map high MMIO space */
1445        MemoryRegion *high_mmio_alias = g_new0(MemoryRegion, 1);
1446
1447        memory_region_init_alias(high_mmio_alias, OBJECT(dev), "pcie-mmio-high",
1448                                 mmio_reg, base_mmio_high, size_mmio_high);
1449        memory_region_add_subregion(get_system_memory(), base_mmio_high,
1450                                    high_mmio_alias);
1451    }
1452
1453    /* Map IO port space */
1454    sysbus_mmio_map(SYS_BUS_DEVICE(dev), 2, base_pio);
1455
1456    for (i = 0; i < GPEX_NUM_IRQS; i++) {
1457        sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
1458                           qdev_get_gpio_in(vms->gic, irq + i));
1459        gpex_set_irq_num(GPEX_HOST(dev), i, irq + i);
1460    }
1461
1462    pci = PCI_HOST_BRIDGE(dev);
1463    pci->bypass_iommu = vms->default_bus_bypass_iommu;
1464    vms->bus = pci->bus;
1465    if (vms->bus) {
1466        for (i = 0; i < nb_nics; i++) {
1467            NICInfo *nd = &nd_table[i];
1468
1469            if (!nd->model) {
1470                nd->model = g_strdup("virtio");
1471            }
1472
1473            pci_nic_init_nofail(nd, pci->bus, nd->model, NULL);
1474        }
1475    }
1476
1477    nodename = vms->pciehb_nodename = g_strdup_printf("/pcie@%" PRIx64, base);
1478    qemu_fdt_add_subnode(ms->fdt, nodename);
1479    qemu_fdt_setprop_string(ms->fdt, nodename,
1480                            "compatible", "pci-host-ecam-generic");
1481    qemu_fdt_setprop_string(ms->fdt, nodename, "device_type", "pci");
1482    qemu_fdt_setprop_cell(ms->fdt, nodename, "#address-cells", 3);
1483    qemu_fdt_setprop_cell(ms->fdt, nodename, "#size-cells", 2);
1484    qemu_fdt_setprop_cell(ms->fdt, nodename, "linux,pci-domain", 0);
1485    qemu_fdt_setprop_cells(ms->fdt, nodename, "bus-range", 0,
1486                           nr_pcie_buses - 1);
1487    qemu_fdt_setprop(ms->fdt, nodename, "dma-coherent", NULL, 0);
1488
1489    if (vms->msi_phandle) {
1490        qemu_fdt_setprop_cells(ms->fdt, nodename, "msi-parent",
1491                               vms->msi_phandle);
1492    }
1493
1494    qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1495                                 2, base_ecam, 2, size_ecam);
1496
1497    if (vms->highmem_mmio) {
1498        qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "ranges",
1499                                     1, FDT_PCI_RANGE_IOPORT, 2, 0,
1500                                     2, base_pio, 2, size_pio,
1501                                     1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
1502                                     2, base_mmio, 2, size_mmio,
1503                                     1, FDT_PCI_RANGE_MMIO_64BIT,
1504                                     2, base_mmio_high,
1505                                     2, base_mmio_high, 2, size_mmio_high);
1506    } else {
1507        qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "ranges",
1508                                     1, FDT_PCI_RANGE_IOPORT, 2, 0,
1509                                     2, base_pio, 2, size_pio,
1510                                     1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
1511                                     2, base_mmio, 2, size_mmio);
1512    }
1513
1514    qemu_fdt_setprop_cell(ms->fdt, nodename, "#interrupt-cells", 1);
1515    create_pcie_irq_map(ms, vms->gic_phandle, irq, nodename);
1516
1517    if (vms->iommu) {
1518        vms->iommu_phandle = qemu_fdt_alloc_phandle(ms->fdt);
1519
1520        switch (vms->iommu) {
1521        case VIRT_IOMMU_SMMUV3:
1522            create_smmu(vms, vms->bus);
1523            qemu_fdt_setprop_cells(ms->fdt, nodename, "iommu-map",
1524                                   0x0, vms->iommu_phandle, 0x0, 0x10000);
1525            break;
1526        default:
1527            g_assert_not_reached();
1528        }
1529    }
1530}
1531
1532static void create_platform_bus(VirtMachineState *vms)
1533{
1534    DeviceState *dev;
1535    SysBusDevice *s;
1536    int i;
1537    MemoryRegion *sysmem = get_system_memory();
1538
1539    dev = qdev_new(TYPE_PLATFORM_BUS_DEVICE);
1540    dev->id = g_strdup(TYPE_PLATFORM_BUS_DEVICE);
1541    qdev_prop_set_uint32(dev, "num_irqs", PLATFORM_BUS_NUM_IRQS);
1542    qdev_prop_set_uint32(dev, "mmio_size", vms->memmap[VIRT_PLATFORM_BUS].size);
1543    sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1544    vms->platform_bus_dev = dev;
1545
1546    s = SYS_BUS_DEVICE(dev);
1547    for (i = 0; i < PLATFORM_BUS_NUM_IRQS; i++) {
1548        int irq = vms->irqmap[VIRT_PLATFORM_BUS] + i;
1549        sysbus_connect_irq(s, i, qdev_get_gpio_in(vms->gic, irq));
1550    }
1551
1552    memory_region_add_subregion(sysmem,
1553                                vms->memmap[VIRT_PLATFORM_BUS].base,
1554                                sysbus_mmio_get_region(s, 0));
1555}
1556
1557static void create_tag_ram(MemoryRegion *tag_sysmem,
1558                           hwaddr base, hwaddr size,
1559                           const char *name)
1560{
1561    MemoryRegion *tagram = g_new(MemoryRegion, 1);
1562
1563    memory_region_init_ram(tagram, NULL, name, size / 32, &error_fatal);
1564    memory_region_add_subregion(tag_sysmem, base / 32, tagram);
1565}
1566
1567static void create_secure_ram(VirtMachineState *vms,
1568                              MemoryRegion *secure_sysmem,
1569                              MemoryRegion *secure_tag_sysmem)
1570{
1571    MemoryRegion *secram = g_new(MemoryRegion, 1);
1572    char *nodename;
1573    hwaddr base = vms->memmap[VIRT_SECURE_MEM].base;
1574    hwaddr size = vms->memmap[VIRT_SECURE_MEM].size;
1575    MachineState *ms = MACHINE(vms);
1576
1577    memory_region_init_ram(secram, NULL, "virt.secure-ram", size,
1578                           &error_fatal);
1579    memory_region_add_subregion(secure_sysmem, base, secram);
1580
1581    nodename = g_strdup_printf("/secram@%" PRIx64, base);
1582    qemu_fdt_add_subnode(ms->fdt, nodename);
1583    qemu_fdt_setprop_string(ms->fdt, nodename, "device_type", "memory");
1584    qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg", 2, base, 2, size);
1585    qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
1586    qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
1587
1588    if (secure_tag_sysmem) {
1589        create_tag_ram(secure_tag_sysmem, base, size, "mach-virt.secure-tag");
1590    }
1591
1592    g_free(nodename);
1593}
1594
1595static void *machvirt_dtb(const struct arm_boot_info *binfo, int *fdt_size)
1596{
1597    const VirtMachineState *board = container_of(binfo, VirtMachineState,
1598                                                 bootinfo);
1599    MachineState *ms = MACHINE(board);
1600
1601
1602    *fdt_size = board->fdt_size;
1603    return ms->fdt;
1604}
1605
1606static void virt_build_smbios(VirtMachineState *vms)
1607{
1608    MachineClass *mc = MACHINE_GET_CLASS(vms);
1609    VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
1610    uint8_t *smbios_tables, *smbios_anchor;
1611    size_t smbios_tables_len, smbios_anchor_len;
1612    const char *product = "QEMU Virtual Machine";
1613
1614    if (kvm_enabled()) {
1615        product = "KVM Virtual Machine";
1616    }
1617
1618    smbios_set_defaults("QEMU", product,
1619                        vmc->smbios_old_sys_ver ? "1.0" : mc->name, false,
1620                        true, SMBIOS_ENTRY_POINT_TYPE_64);
1621
1622    smbios_get_tables(MACHINE(vms), NULL, 0,
1623                      &smbios_tables, &smbios_tables_len,
1624                      &smbios_anchor, &smbios_anchor_len,
1625                      &error_fatal);
1626
1627    if (smbios_anchor) {
1628        fw_cfg_add_file(vms->fw_cfg, "etc/smbios/smbios-tables",
1629                        smbios_tables, smbios_tables_len);
1630        fw_cfg_add_file(vms->fw_cfg, "etc/smbios/smbios-anchor",
1631                        smbios_anchor, smbios_anchor_len);
1632    }
1633}
1634
1635static
1636void virt_machine_done(Notifier *notifier, void *data)
1637{
1638    VirtMachineState *vms = container_of(notifier, VirtMachineState,
1639                                         machine_done);
1640    MachineState *ms = MACHINE(vms);
1641    ARMCPU *cpu = ARM_CPU(first_cpu);
1642    struct arm_boot_info *info = &vms->bootinfo;
1643    AddressSpace *as = arm_boot_address_space(cpu, info);
1644
1645    /*
1646     * If the user provided a dtb, we assume the dynamic sysbus nodes
1647     * already are integrated there. This corresponds to a use case where
1648     * the dynamic sysbus nodes are complex and their generation is not yet
1649     * supported. In that case the user can take charge of the guest dt
1650     * while qemu takes charge of the qom stuff.
1651     */
1652    if (info->dtb_filename == NULL) {
1653        platform_bus_add_all_fdt_nodes(ms->fdt, "/intc",
1654                                       vms->memmap[VIRT_PLATFORM_BUS].base,
1655                                       vms->memmap[VIRT_PLATFORM_BUS].size,
1656                                       vms->irqmap[VIRT_PLATFORM_BUS]);
1657    }
1658    if (arm_load_dtb(info->dtb_start, info, info->dtb_limit, as, ms) < 0) {
1659        exit(1);
1660    }
1661
1662    fw_cfg_add_extra_pci_roots(vms->bus, vms->fw_cfg);
1663
1664    virt_acpi_setup(vms);
1665    virt_build_smbios(vms);
1666}
1667
1668static uint64_t virt_cpu_mp_affinity(VirtMachineState *vms, int idx)
1669{
1670    uint8_t clustersz = ARM_DEFAULT_CPUS_PER_CLUSTER;
1671    VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
1672
1673    if (!vmc->disallow_affinity_adjustment) {
1674        /* Adjust MPIDR like 64-bit KVM hosts, which incorporate the
1675         * GIC's target-list limitations. 32-bit KVM hosts currently
1676         * always create clusters of 4 CPUs, but that is expected to
1677         * change when they gain support for gicv3. When KVM is enabled
1678         * it will override the changes we make here, therefore our
1679         * purposes are to make TCG consistent (with 64-bit KVM hosts)
1680         * and to improve SGI efficiency.
1681         */
1682        if (vms->gic_version == VIRT_GIC_VERSION_2) {
1683            clustersz = GIC_TARGETLIST_BITS;
1684        } else {
1685            clustersz = GICV3_TARGETLIST_BITS;
1686        }
1687    }
1688    return arm_cpu_mp_affinity(idx, clustersz);
1689}
1690
1691static void virt_set_memmap(VirtMachineState *vms, int pa_bits)
1692{
1693    MachineState *ms = MACHINE(vms);
1694    hwaddr base, device_memory_base, device_memory_size, memtop;
1695    int i;
1696
1697    vms->memmap = extended_memmap;
1698
1699    for (i = 0; i < ARRAY_SIZE(base_memmap); i++) {
1700        vms->memmap[i] = base_memmap[i];
1701    }
1702
1703    if (ms->ram_slots > ACPI_MAX_RAM_SLOTS) {
1704        error_report("unsupported number of memory slots: %"PRIu64,
1705                     ms->ram_slots);
1706        exit(EXIT_FAILURE);
1707    }
1708
1709    /*
1710     * !highmem is exactly the same as limiting the PA space to 32bit,
1711     * irrespective of the underlying capabilities of the HW.
1712     */
1713    if (!vms->highmem) {
1714        pa_bits = 32;
1715    }
1716
1717    /*
1718     * We compute the base of the high IO region depending on the
1719     * amount of initial and device memory. The device memory start/size
1720     * is aligned on 1GiB. We never put the high IO region below 256GiB
1721     * so that if maxram_size is < 255GiB we keep the legacy memory map.
1722     * The device region size assumes 1GiB page max alignment per slot.
1723     */
1724    device_memory_base =
1725        ROUND_UP(vms->memmap[VIRT_MEM].base + ms->ram_size, GiB);
1726    device_memory_size = ms->maxram_size - ms->ram_size + ms->ram_slots * GiB;
1727
1728    /* Base address of the high IO region */
1729    memtop = base = device_memory_base + ROUND_UP(device_memory_size, GiB);
1730    if (memtop > BIT_ULL(pa_bits)) {
1731            error_report("Addressing limited to %d bits, but memory exceeds it by %llu bytes\n",
1732                         pa_bits, memtop - BIT_ULL(pa_bits));
1733        exit(EXIT_FAILURE);
1734    }
1735    if (base < device_memory_base) {
1736        error_report("maxmem/slots too huge");
1737        exit(EXIT_FAILURE);
1738    }
1739    if (base < vms->memmap[VIRT_MEM].base + LEGACY_RAMLIMIT_BYTES) {
1740        base = vms->memmap[VIRT_MEM].base + LEGACY_RAMLIMIT_BYTES;
1741    }
1742
1743    /* We know for sure that at least the memory fits in the PA space */
1744    vms->highest_gpa = memtop - 1;
1745
1746    for (i = VIRT_LOWMEMMAP_LAST; i < ARRAY_SIZE(extended_memmap); i++) {
1747        hwaddr size = extended_memmap[i].size;
1748        bool fits;
1749
1750        base = ROUND_UP(base, size);
1751        vms->memmap[i].base = base;
1752        vms->memmap[i].size = size;
1753
1754        /*
1755         * Check each device to see if they fit in the PA space,
1756         * moving highest_gpa as we go.
1757         *
1758         * For each device that doesn't fit, disable it.
1759         */
1760        fits = (base + size) <= BIT_ULL(pa_bits);
1761        if (fits) {
1762            vms->highest_gpa = base + size - 1;
1763        }
1764
1765        switch (i) {
1766        case VIRT_HIGH_GIC_REDIST2:
1767            vms->highmem_redists &= fits;
1768            break;
1769        case VIRT_HIGH_PCIE_ECAM:
1770            vms->highmem_ecam &= fits;
1771            break;
1772        case VIRT_HIGH_PCIE_MMIO:
1773            vms->highmem_mmio &= fits;
1774            break;
1775        }
1776
1777        base += size;
1778    }
1779
1780    if (device_memory_size > 0) {
1781        ms->device_memory = g_malloc0(sizeof(*ms->device_memory));
1782        ms->device_memory->base = device_memory_base;
1783        memory_region_init(&ms->device_memory->mr, OBJECT(vms),
1784                           "device-memory", device_memory_size);
1785    }
1786}
1787
1788/*
1789 * finalize_gic_version - Determines the final gic_version
1790 * according to the gic-version property
1791 *
1792 * Default GIC type is v2
1793 */
1794static void finalize_gic_version(VirtMachineState *vms)
1795{
1796    unsigned int max_cpus = MACHINE(vms)->smp.max_cpus;
1797
1798    if (kvm_enabled()) {
1799        int probe_bitmap;
1800
1801        if (!kvm_irqchip_in_kernel()) {
1802            switch (vms->gic_version) {
1803            case VIRT_GIC_VERSION_HOST:
1804                warn_report(
1805                    "gic-version=host not relevant with kernel-irqchip=off "
1806                     "as only userspace GICv2 is supported. Using v2 ...");
1807                return;
1808            case VIRT_GIC_VERSION_MAX:
1809            case VIRT_GIC_VERSION_NOSEL:
1810                vms->gic_version = VIRT_GIC_VERSION_2;
1811                return;
1812            case VIRT_GIC_VERSION_2:
1813                return;
1814            case VIRT_GIC_VERSION_3:
1815                error_report(
1816                    "gic-version=3 is not supported with kernel-irqchip=off");
1817                exit(1);
1818            case VIRT_GIC_VERSION_4:
1819                error_report(
1820                    "gic-version=4 is not supported with kernel-irqchip=off");
1821                exit(1);
1822            }
1823        }
1824
1825        probe_bitmap = kvm_arm_vgic_probe();
1826        if (!probe_bitmap) {
1827            error_report("Unable to determine GIC version supported by host");
1828            exit(1);
1829        }
1830
1831        switch (vms->gic_version) {
1832        case VIRT_GIC_VERSION_HOST:
1833        case VIRT_GIC_VERSION_MAX:
1834            if (probe_bitmap & KVM_ARM_VGIC_V3) {
1835                vms->gic_version = VIRT_GIC_VERSION_3;
1836            } else {
1837                vms->gic_version = VIRT_GIC_VERSION_2;
1838            }
1839            return;
1840        case VIRT_GIC_VERSION_NOSEL:
1841            if ((probe_bitmap & KVM_ARM_VGIC_V2) && max_cpus <= GIC_NCPU) {
1842                vms->gic_version = VIRT_GIC_VERSION_2;
1843            } else if (probe_bitmap & KVM_ARM_VGIC_V3) {
1844                /*
1845                 * in case the host does not support v2 in-kernel emulation or
1846                 * the end-user requested more than 8 VCPUs we now default
1847                 * to v3. In any case defaulting to v2 would be broken.
1848                 */
1849                vms->gic_version = VIRT_GIC_VERSION_3;
1850            } else if (max_cpus > GIC_NCPU) {
1851                error_report("host only supports in-kernel GICv2 emulation "
1852                             "but more than 8 vcpus are requested");
1853                exit(1);
1854            }
1855            break;
1856        case VIRT_GIC_VERSION_2:
1857        case VIRT_GIC_VERSION_3:
1858            break;
1859        case VIRT_GIC_VERSION_4:
1860            error_report("gic-version=4 is not supported with KVM");
1861            exit(1);
1862        }
1863
1864        /* Check chosen version is effectively supported by the host */
1865        if (vms->gic_version == VIRT_GIC_VERSION_2 &&
1866            !(probe_bitmap & KVM_ARM_VGIC_V2)) {
1867            error_report("host does not support in-kernel GICv2 emulation");
1868            exit(1);
1869        } else if (vms->gic_version == VIRT_GIC_VERSION_3 &&
1870                   !(probe_bitmap & KVM_ARM_VGIC_V3)) {
1871            error_report("host does not support in-kernel GICv3 emulation");
1872            exit(1);
1873        }
1874        return;
1875    }
1876
1877    /* TCG mode */
1878    switch (vms->gic_version) {
1879    case VIRT_GIC_VERSION_NOSEL:
1880        vms->gic_version = VIRT_GIC_VERSION_2;
1881        break;
1882    case VIRT_GIC_VERSION_MAX:
1883        if (module_object_class_by_name("arm-gicv3")) {
1884            /* CONFIG_ARM_GICV3_TCG was set */
1885            if (vms->virt) {
1886                /* GICv4 only makes sense if CPU has EL2 */
1887                vms->gic_version = VIRT_GIC_VERSION_4;
1888            } else {
1889                vms->gic_version = VIRT_GIC_VERSION_3;
1890            }
1891        } else {
1892            vms->gic_version = VIRT_GIC_VERSION_2;
1893        }
1894        break;
1895    case VIRT_GIC_VERSION_HOST:
1896        error_report("gic-version=host requires KVM");
1897        exit(1);
1898    case VIRT_GIC_VERSION_4:
1899        if (!vms->virt) {
1900            error_report("gic-version=4 requires virtualization enabled");
1901            exit(1);
1902        }
1903        break;
1904    case VIRT_GIC_VERSION_2:
1905    case VIRT_GIC_VERSION_3:
1906        break;
1907    }
1908}
1909
1910/*
1911 * virt_cpu_post_init() must be called after the CPUs have
1912 * been realized and the GIC has been created.
1913 */
1914static void virt_cpu_post_init(VirtMachineState *vms, MemoryRegion *sysmem)
1915{
1916    int max_cpus = MACHINE(vms)->smp.max_cpus;
1917    bool aarch64, pmu, steal_time;
1918    CPUState *cpu;
1919
1920    aarch64 = object_property_get_bool(OBJECT(first_cpu), "aarch64", NULL);
1921    pmu = object_property_get_bool(OBJECT(first_cpu), "pmu", NULL);
1922    steal_time = object_property_get_bool(OBJECT(first_cpu),
1923                                          "kvm-steal-time", NULL);
1924
1925    if (kvm_enabled()) {
1926        hwaddr pvtime_reg_base = vms->memmap[VIRT_PVTIME].base;
1927        hwaddr pvtime_reg_size = vms->memmap[VIRT_PVTIME].size;
1928
1929        if (steal_time) {
1930            MemoryRegion *pvtime = g_new(MemoryRegion, 1);
1931            hwaddr pvtime_size = max_cpus * PVTIME_SIZE_PER_CPU;
1932
1933            /* The memory region size must be a multiple of host page size. */
1934            pvtime_size = REAL_HOST_PAGE_ALIGN(pvtime_size);
1935
1936            if (pvtime_size > pvtime_reg_size) {
1937                error_report("pvtime requires a %" HWADDR_PRId
1938                             " byte memory region for %d CPUs,"
1939                             " but only %" HWADDR_PRId " has been reserved",
1940                             pvtime_size, max_cpus, pvtime_reg_size);
1941                exit(1);
1942            }
1943
1944            memory_region_init_ram(pvtime, NULL, "pvtime", pvtime_size, NULL);
1945            memory_region_add_subregion(sysmem, pvtime_reg_base, pvtime);
1946        }
1947
1948        CPU_FOREACH(cpu) {
1949            if (pmu) {
1950                assert(arm_feature(&ARM_CPU(cpu)->env, ARM_FEATURE_PMU));
1951                if (kvm_irqchip_in_kernel()) {
1952                    kvm_arm_pmu_set_irq(cpu, PPI(VIRTUAL_PMU_IRQ));
1953                }
1954                kvm_arm_pmu_init(cpu);
1955            }
1956            if (steal_time) {
1957                kvm_arm_pvtime_init(cpu, pvtime_reg_base +
1958                                         cpu->cpu_index * PVTIME_SIZE_PER_CPU);
1959            }
1960        }
1961    } else {
1962        if (aarch64 && vms->highmem) {
1963            int requested_pa_size = 64 - clz64(vms->highest_gpa);
1964            int pamax = arm_pamax(ARM_CPU(first_cpu));
1965
1966            if (pamax < requested_pa_size) {
1967                error_report("VCPU supports less PA bits (%d) than "
1968                             "requested by the memory map (%d)",
1969                             pamax, requested_pa_size);
1970                exit(1);
1971            }
1972        }
1973    }
1974}
1975
1976static void machvirt_init(MachineState *machine)
1977{
1978    VirtMachineState *vms = VIRT_MACHINE(machine);
1979    VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(machine);
1980    MachineClass *mc = MACHINE_GET_CLASS(machine);
1981    const CPUArchIdList *possible_cpus;
1982    MemoryRegion *sysmem = get_system_memory();
1983    MemoryRegion *secure_sysmem = NULL;
1984    MemoryRegion *tag_sysmem = NULL;
1985    MemoryRegion *secure_tag_sysmem = NULL;
1986    int n, virt_max_cpus;
1987    bool firmware_loaded;
1988    bool aarch64 = true;
1989    bool has_ged = !vmc->no_ged;
1990    unsigned int smp_cpus = machine->smp.cpus;
1991    unsigned int max_cpus = machine->smp.max_cpus;
1992
1993    if (!cpu_type_valid(machine->cpu_type)) {
1994        error_report("mach-virt: CPU type %s not supported", machine->cpu_type);
1995        exit(1);
1996    }
1997
1998    possible_cpus = mc->possible_cpu_arch_ids(machine);
1999
2000    /*
2001     * In accelerated mode, the memory map is computed earlier in kvm_type()
2002     * to create a VM with the right number of IPA bits.
2003     */
2004    if (!vms->memmap) {
2005        Object *cpuobj;
2006        ARMCPU *armcpu;
2007        int pa_bits;
2008
2009        /*
2010         * Instanciate a temporary CPU object to find out about what
2011         * we are about to deal with. Once this is done, get rid of
2012         * the object.
2013         */
2014        cpuobj = object_new(possible_cpus->cpus[0].type);
2015        armcpu = ARM_CPU(cpuobj);
2016
2017        pa_bits = arm_pamax(armcpu);
2018
2019        object_unref(cpuobj);
2020
2021        virt_set_memmap(vms, pa_bits);
2022    }
2023
2024    /* We can probe only here because during property set
2025     * KVM is not available yet
2026     */
2027    finalize_gic_version(vms);
2028
2029    if (vms->secure) {
2030        /*
2031         * The Secure view of the world is the same as the NonSecure,
2032         * but with a few extra devices. Create it as a container region
2033         * containing the system memory at low priority; any secure-only
2034         * devices go in at higher priority and take precedence.
2035         */
2036        secure_sysmem = g_new(MemoryRegion, 1);
2037        memory_region_init(secure_sysmem, OBJECT(machine), "secure-memory",
2038                           UINT64_MAX);
2039        memory_region_add_subregion_overlap(secure_sysmem, 0, sysmem, -1);
2040    }
2041
2042    firmware_loaded = virt_firmware_init(vms, sysmem,
2043                                         secure_sysmem ?: sysmem);
2044
2045    /* If we have an EL3 boot ROM then the assumption is that it will
2046     * implement PSCI itself, so disable QEMU's internal implementation
2047     * so it doesn't get in the way. Instead of starting secondary
2048     * CPUs in PSCI powerdown state we will start them all running and
2049     * let the boot ROM sort them out.
2050     * The usual case is that we do use QEMU's PSCI implementation;
2051     * if the guest has EL2 then we will use SMC as the conduit,
2052     * and otherwise we will use HVC (for backwards compatibility and
2053     * because if we're using KVM then we must use HVC).
2054     */
2055    if (vms->secure && firmware_loaded) {
2056        vms->psci_conduit = QEMU_PSCI_CONDUIT_DISABLED;
2057    } else if (vms->virt) {
2058        vms->psci_conduit = QEMU_PSCI_CONDUIT_SMC;
2059    } else {
2060        vms->psci_conduit = QEMU_PSCI_CONDUIT_HVC;
2061    }
2062
2063    /*
2064     * The maximum number of CPUs depends on the GIC version, or on how
2065     * many redistributors we can fit into the memory map (which in turn
2066     * depends on whether this is a GICv3 or v4).
2067     */
2068    if (vms->gic_version == VIRT_GIC_VERSION_2) {
2069        virt_max_cpus = GIC_NCPU;
2070    } else {
2071        virt_max_cpus = virt_redist_capacity(vms, VIRT_GIC_REDIST) +
2072            virt_redist_capacity(vms, VIRT_HIGH_GIC_REDIST2);
2073    }
2074
2075    if (max_cpus > virt_max_cpus) {
2076        error_report("Number of SMP CPUs requested (%d) exceeds max CPUs "
2077                     "supported by machine 'mach-virt' (%d)",
2078                     max_cpus, virt_max_cpus);
2079        exit(1);
2080    }
2081
2082    if (vms->secure && (kvm_enabled() || hvf_enabled())) {
2083        error_report("mach-virt: %s does not support providing "
2084                     "Security extensions (TrustZone) to the guest CPU",
2085                     kvm_enabled() ? "KVM" : "HVF");
2086        exit(1);
2087    }
2088
2089    if (vms->virt && (kvm_enabled() || hvf_enabled())) {
2090        error_report("mach-virt: %s does not support providing "
2091                     "Virtualization extensions to the guest CPU",
2092                     kvm_enabled() ? "KVM" : "HVF");
2093        exit(1);
2094    }
2095
2096    if (vms->mte && (kvm_enabled() || hvf_enabled())) {
2097        error_report("mach-virt: %s does not support providing "
2098                     "MTE to the guest CPU",
2099                     kvm_enabled() ? "KVM" : "HVF");
2100        exit(1);
2101    }
2102
2103    create_fdt(vms);
2104
2105    assert(possible_cpus->len == max_cpus);
2106    for (n = 0; n < possible_cpus->len; n++) {
2107        Object *cpuobj;
2108        CPUState *cs;
2109
2110        if (n >= smp_cpus) {
2111            break;
2112        }
2113
2114        cpuobj = object_new(possible_cpus->cpus[n].type);
2115        object_property_set_int(cpuobj, "mp-affinity",
2116                                possible_cpus->cpus[n].arch_id, NULL);
2117
2118        cs = CPU(cpuobj);
2119        cs->cpu_index = n;
2120
2121        numa_cpu_pre_plug(&possible_cpus->cpus[cs->cpu_index], DEVICE(cpuobj),
2122                          &error_fatal);
2123
2124        aarch64 &= object_property_get_bool(cpuobj, "aarch64", NULL);
2125
2126        if (!vms->secure) {
2127            object_property_set_bool(cpuobj, "has_el3", false, NULL);
2128        }
2129
2130        if (!vms->virt && object_property_find(cpuobj, "has_el2")) {
2131            object_property_set_bool(cpuobj, "has_el2", false, NULL);
2132        }
2133
2134        if (vmc->kvm_no_adjvtime &&
2135            object_property_find(cpuobj, "kvm-no-adjvtime")) {
2136            object_property_set_bool(cpuobj, "kvm-no-adjvtime", true, NULL);
2137        }
2138
2139        if (vmc->no_kvm_steal_time &&
2140            object_property_find(cpuobj, "kvm-steal-time")) {
2141            object_property_set_bool(cpuobj, "kvm-steal-time", false, NULL);
2142        }
2143
2144        if (vmc->no_pmu && object_property_find(cpuobj, "pmu")) {
2145            object_property_set_bool(cpuobj, "pmu", false, NULL);
2146        }
2147
2148        if (vmc->no_tcg_lpa2 && object_property_find(cpuobj, "lpa2")) {
2149            object_property_set_bool(cpuobj, "lpa2", false, NULL);
2150        }
2151
2152        if (object_property_find(cpuobj, "reset-cbar")) {
2153            object_property_set_int(cpuobj, "reset-cbar",
2154                                    vms->memmap[VIRT_CPUPERIPHS].base,
2155                                    &error_abort);
2156        }
2157
2158        object_property_set_link(cpuobj, "memory", OBJECT(sysmem),
2159                                 &error_abort);
2160        if (vms->secure) {
2161            object_property_set_link(cpuobj, "secure-memory",
2162                                     OBJECT(secure_sysmem), &error_abort);
2163        }
2164
2165        if (vms->mte) {
2166            /* Create the memory region only once, but link to all cpus. */
2167            if (!tag_sysmem) {
2168                /*
2169                 * The property exists only if MemTag is supported.
2170                 * If it is, we must allocate the ram to back that up.
2171                 */
2172                if (!object_property_find(cpuobj, "tag-memory")) {
2173                    error_report("MTE requested, but not supported "
2174                                 "by the guest CPU");
2175                    exit(1);
2176                }
2177
2178                tag_sysmem = g_new(MemoryRegion, 1);
2179                memory_region_init(tag_sysmem, OBJECT(machine),
2180                                   "tag-memory", UINT64_MAX / 32);
2181
2182                if (vms->secure) {
2183                    secure_tag_sysmem = g_new(MemoryRegion, 1);
2184                    memory_region_init(secure_tag_sysmem, OBJECT(machine),
2185                                       "secure-tag-memory", UINT64_MAX / 32);
2186
2187                    /* As with ram, secure-tag takes precedence over tag.  */
2188                    memory_region_add_subregion_overlap(secure_tag_sysmem, 0,
2189                                                        tag_sysmem, -1);
2190                }
2191            }
2192
2193            object_property_set_link(cpuobj, "tag-memory", OBJECT(tag_sysmem),
2194                                     &error_abort);
2195            if (vms->secure) {
2196                object_property_set_link(cpuobj, "secure-tag-memory",
2197                                         OBJECT(secure_tag_sysmem),
2198                                         &error_abort);
2199            }
2200        }
2201
2202        qdev_realize(DEVICE(cpuobj), NULL, &error_fatal);
2203        object_unref(cpuobj);
2204    }
2205    fdt_add_timer_nodes(vms);
2206    fdt_add_cpu_nodes(vms);
2207
2208    memory_region_add_subregion(sysmem, vms->memmap[VIRT_MEM].base,
2209                                machine->ram);
2210    if (machine->device_memory) {
2211        memory_region_add_subregion(sysmem, machine->device_memory->base,
2212                                    &machine->device_memory->mr);
2213    }
2214
2215    virt_flash_fdt(vms, sysmem, secure_sysmem ?: sysmem);
2216
2217    create_gic(vms, sysmem);
2218
2219    virt_cpu_post_init(vms, sysmem);
2220
2221    fdt_add_pmu_nodes(vms);
2222
2223    create_uart(vms, VIRT_UART, sysmem, serial_hd(0));
2224
2225    if (vms->secure) {
2226        create_secure_ram(vms, secure_sysmem, secure_tag_sysmem);
2227        create_uart(vms, VIRT_SECURE_UART, secure_sysmem, serial_hd(1));
2228    }
2229
2230    if (tag_sysmem) {
2231        create_tag_ram(tag_sysmem, vms->memmap[VIRT_MEM].base,
2232                       machine->ram_size, "mach-virt.tag");
2233    }
2234
2235    vms->highmem_ecam &= (!firmware_loaded || aarch64);
2236
2237    create_rtc(vms);
2238
2239    create_pcie(vms);
2240
2241    if (has_ged && aarch64 && firmware_loaded && virt_is_acpi_enabled(vms)) {
2242        vms->acpi_dev = create_acpi_ged(vms);
2243    } else {
2244        create_gpio_devices(vms, VIRT_GPIO, sysmem);
2245    }
2246
2247    if (vms->secure && !vmc->no_secure_gpio) {
2248        create_gpio_devices(vms, VIRT_SECURE_GPIO, secure_sysmem);
2249    }
2250
2251     /* connect powerdown request */
2252     vms->powerdown_notifier.notify = virt_powerdown_req;
2253     qemu_register_powerdown_notifier(&vms->powerdown_notifier);
2254
2255    /* Create mmio transports, so the user can create virtio backends
2256     * (which will be automatically plugged in to the transports). If
2257     * no backend is created the transport will just sit harmlessly idle.
2258     */
2259    create_virtio_devices(vms);
2260
2261    vms->fw_cfg = create_fw_cfg(vms, &address_space_memory);
2262    rom_set_fw(vms->fw_cfg);
2263
2264    create_platform_bus(vms);
2265
2266    if (machine->nvdimms_state->is_enabled) {
2267        const struct AcpiGenericAddress arm_virt_nvdimm_acpi_dsmio = {
2268            .space_id = AML_AS_SYSTEM_MEMORY,
2269            .address = vms->memmap[VIRT_NVDIMM_ACPI].base,
2270            .bit_width = NVDIMM_ACPI_IO_LEN << 3
2271        };
2272
2273        nvdimm_init_acpi_state(machine->nvdimms_state, sysmem,
2274                               arm_virt_nvdimm_acpi_dsmio,
2275                               vms->fw_cfg, OBJECT(vms));
2276    }
2277
2278    vms->bootinfo.ram_size = machine->ram_size;
2279    vms->bootinfo.board_id = -1;
2280    vms->bootinfo.loader_start = vms->memmap[VIRT_MEM].base;
2281    vms->bootinfo.get_dtb = machvirt_dtb;
2282    vms->bootinfo.skip_dtb_autoload = true;
2283    vms->bootinfo.firmware_loaded = firmware_loaded;
2284    vms->bootinfo.psci_conduit = vms->psci_conduit;
2285    arm_load_kernel(ARM_CPU(first_cpu), machine, &vms->bootinfo);
2286
2287    vms->machine_done.notify = virt_machine_done;
2288    qemu_add_machine_init_done_notifier(&vms->machine_done);
2289}
2290
2291static bool virt_get_secure(Object *obj, Error **errp)
2292{
2293    VirtMachineState *vms = VIRT_MACHINE(obj);
2294
2295    return vms->secure;
2296}
2297
2298static void virt_set_secure(Object *obj, bool value, Error **errp)
2299{
2300    VirtMachineState *vms = VIRT_MACHINE(obj);
2301
2302    vms->secure = value;
2303}
2304
2305static bool virt_get_virt(Object *obj, Error **errp)
2306{
2307    VirtMachineState *vms = VIRT_MACHINE(obj);
2308
2309    return vms->virt;
2310}
2311
2312static void virt_set_virt(Object *obj, bool value, Error **errp)
2313{
2314    VirtMachineState *vms = VIRT_MACHINE(obj);
2315
2316    vms->virt = value;
2317}
2318
2319static bool virt_get_highmem(Object *obj, Error **errp)
2320{
2321    VirtMachineState *vms = VIRT_MACHINE(obj);
2322
2323    return vms->highmem;
2324}
2325
2326static void virt_set_highmem(Object *obj, bool value, Error **errp)
2327{
2328    VirtMachineState *vms = VIRT_MACHINE(obj);
2329
2330    vms->highmem = value;
2331}
2332
2333static bool virt_get_its(Object *obj, Error **errp)
2334{
2335    VirtMachineState *vms = VIRT_MACHINE(obj);
2336
2337    return vms->its;
2338}
2339
2340static void virt_set_its(Object *obj, bool value, Error **errp)
2341{
2342    VirtMachineState *vms = VIRT_MACHINE(obj);
2343
2344    vms->its = value;
2345}
2346
2347static bool virt_get_dtb_randomness(Object *obj, Error **errp)
2348{
2349    VirtMachineState *vms = VIRT_MACHINE(obj);
2350
2351    return vms->dtb_randomness;
2352}
2353
2354static void virt_set_dtb_randomness(Object *obj, bool value, Error **errp)
2355{
2356    VirtMachineState *vms = VIRT_MACHINE(obj);
2357
2358    vms->dtb_randomness = value;
2359}
2360
2361static char *virt_get_oem_id(Object *obj, Error **errp)
2362{
2363    VirtMachineState *vms = VIRT_MACHINE(obj);
2364
2365    return g_strdup(vms->oem_id);
2366}
2367
2368static void virt_set_oem_id(Object *obj, const char *value, Error **errp)
2369{
2370    VirtMachineState *vms = VIRT_MACHINE(obj);
2371    size_t len = strlen(value);
2372
2373    if (len > 6) {
2374        error_setg(errp,
2375                   "User specified oem-id value is bigger than 6 bytes in size");
2376        return;
2377    }
2378
2379    strncpy(vms->oem_id, value, 6);
2380}
2381
2382static char *virt_get_oem_table_id(Object *obj, Error **errp)
2383{
2384    VirtMachineState *vms = VIRT_MACHINE(obj);
2385
2386    return g_strdup(vms->oem_table_id);
2387}
2388
2389static void virt_set_oem_table_id(Object *obj, const char *value,
2390                                  Error **errp)
2391{
2392    VirtMachineState *vms = VIRT_MACHINE(obj);
2393    size_t len = strlen(value);
2394
2395    if (len > 8) {
2396        error_setg(errp,
2397                   "User specified oem-table-id value is bigger than 8 bytes in size");
2398        return;
2399    }
2400    strncpy(vms->oem_table_id, value, 8);
2401}
2402
2403
2404bool virt_is_acpi_enabled(VirtMachineState *vms)
2405{
2406    if (vms->acpi == ON_OFF_AUTO_OFF) {
2407        return false;
2408    }
2409    return true;
2410}
2411
2412static void virt_get_acpi(Object *obj, Visitor *v, const char *name,
2413                          void *opaque, Error **errp)
2414{
2415    VirtMachineState *vms = VIRT_MACHINE(obj);
2416    OnOffAuto acpi = vms->acpi;
2417
2418    visit_type_OnOffAuto(v, name, &acpi, errp);
2419}
2420
2421static void virt_set_acpi(Object *obj, Visitor *v, const char *name,
2422                          void *opaque, Error **errp)
2423{
2424    VirtMachineState *vms = VIRT_MACHINE(obj);
2425
2426    visit_type_OnOffAuto(v, name, &vms->acpi, errp);
2427}
2428
2429static bool virt_get_ras(Object *obj, Error **errp)
2430{
2431    VirtMachineState *vms = VIRT_MACHINE(obj);
2432
2433    return vms->ras;
2434}
2435
2436static void virt_set_ras(Object *obj, bool value, Error **errp)
2437{
2438    VirtMachineState *vms = VIRT_MACHINE(obj);
2439
2440    vms->ras = value;
2441}
2442
2443static bool virt_get_mte(Object *obj, Error **errp)
2444{
2445    VirtMachineState *vms = VIRT_MACHINE(obj);
2446
2447    return vms->mte;
2448}
2449
2450static void virt_set_mte(Object *obj, bool value, Error **errp)
2451{
2452    VirtMachineState *vms = VIRT_MACHINE(obj);
2453
2454    vms->mte = value;
2455}
2456
2457static char *virt_get_gic_version(Object *obj, Error **errp)
2458{
2459    VirtMachineState *vms = VIRT_MACHINE(obj);
2460    const char *val;
2461
2462    switch (vms->gic_version) {
2463    case VIRT_GIC_VERSION_4:
2464        val = "4";
2465        break;
2466    case VIRT_GIC_VERSION_3:
2467        val = "3";
2468        break;
2469    default:
2470        val = "2";
2471        break;
2472    }
2473    return g_strdup(val);
2474}
2475
2476static void virt_set_gic_version(Object *obj, const char *value, Error **errp)
2477{
2478    VirtMachineState *vms = VIRT_MACHINE(obj);
2479
2480    if (!strcmp(value, "4")) {
2481        vms->gic_version = VIRT_GIC_VERSION_4;
2482    } else if (!strcmp(value, "3")) {
2483        vms->gic_version = VIRT_GIC_VERSION_3;
2484    } else if (!strcmp(value, "2")) {
2485        vms->gic_version = VIRT_GIC_VERSION_2;
2486    } else if (!strcmp(value, "host")) {
2487        vms->gic_version = VIRT_GIC_VERSION_HOST; /* Will probe later */
2488    } else if (!strcmp(value, "max")) {
2489        vms->gic_version = VIRT_GIC_VERSION_MAX; /* Will probe later */
2490    } else {
2491        error_setg(errp, "Invalid gic-version value");
2492        error_append_hint(errp, "Valid values are 3, 2, host, max.\n");
2493    }
2494}
2495
2496static char *virt_get_iommu(Object *obj, Error **errp)
2497{
2498    VirtMachineState *vms = VIRT_MACHINE(obj);
2499
2500    switch (vms->iommu) {
2501    case VIRT_IOMMU_NONE:
2502        return g_strdup("none");
2503    case VIRT_IOMMU_SMMUV3:
2504        return g_strdup("smmuv3");
2505    default:
2506        g_assert_not_reached();
2507    }
2508}
2509
2510static void virt_set_iommu(Object *obj, const char *value, Error **errp)
2511{
2512    VirtMachineState *vms = VIRT_MACHINE(obj);
2513
2514    if (!strcmp(value, "smmuv3")) {
2515        vms->iommu = VIRT_IOMMU_SMMUV3;
2516    } else if (!strcmp(value, "none")) {
2517        vms->iommu = VIRT_IOMMU_NONE;
2518    } else {
2519        error_setg(errp, "Invalid iommu value");
2520        error_append_hint(errp, "Valid values are none, smmuv3.\n");
2521    }
2522}
2523
2524static bool virt_get_default_bus_bypass_iommu(Object *obj, Error **errp)
2525{
2526    VirtMachineState *vms = VIRT_MACHINE(obj);
2527
2528    return vms->default_bus_bypass_iommu;
2529}
2530
2531static void virt_set_default_bus_bypass_iommu(Object *obj, bool value,
2532                                              Error **errp)
2533{
2534    VirtMachineState *vms = VIRT_MACHINE(obj);
2535
2536    vms->default_bus_bypass_iommu = value;
2537}
2538
2539static CpuInstanceProperties
2540virt_cpu_index_to_props(MachineState *ms, unsigned cpu_index)
2541{
2542    MachineClass *mc = MACHINE_GET_CLASS(ms);
2543    const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(ms);
2544
2545    assert(cpu_index < possible_cpus->len);
2546    return possible_cpus->cpus[cpu_index].props;
2547}
2548
2549static int64_t virt_get_default_cpu_node_id(const MachineState *ms, int idx)
2550{
2551    int64_t socket_id = ms->possible_cpus->cpus[idx].props.socket_id;
2552
2553    return socket_id % ms->numa_state->num_nodes;
2554}
2555
2556static const CPUArchIdList *virt_possible_cpu_arch_ids(MachineState *ms)
2557{
2558    int n;
2559    unsigned int max_cpus = ms->smp.max_cpus;
2560    VirtMachineState *vms = VIRT_MACHINE(ms);
2561    MachineClass *mc = MACHINE_GET_CLASS(vms);
2562
2563    if (ms->possible_cpus) {
2564        assert(ms->possible_cpus->len == max_cpus);
2565        return ms->possible_cpus;
2566    }
2567
2568    ms->possible_cpus = g_malloc0(sizeof(CPUArchIdList) +
2569                                  sizeof(CPUArchId) * max_cpus);
2570    ms->possible_cpus->len = max_cpus;
2571    for (n = 0; n < ms->possible_cpus->len; n++) {
2572        ms->possible_cpus->cpus[n].type = ms->cpu_type;
2573        ms->possible_cpus->cpus[n].arch_id =
2574            virt_cpu_mp_affinity(vms, n);
2575
2576        assert(!mc->smp_props.dies_supported);
2577        ms->possible_cpus->cpus[n].props.has_socket_id = true;
2578        ms->possible_cpus->cpus[n].props.socket_id =
2579            n / (ms->smp.clusters * ms->smp.cores * ms->smp.threads);
2580        ms->possible_cpus->cpus[n].props.has_cluster_id = true;
2581        ms->possible_cpus->cpus[n].props.cluster_id =
2582            (n / (ms->smp.cores * ms->smp.threads)) % ms->smp.clusters;
2583        ms->possible_cpus->cpus[n].props.has_core_id = true;
2584        ms->possible_cpus->cpus[n].props.core_id =
2585            (n / ms->smp.threads) % ms->smp.cores;
2586        ms->possible_cpus->cpus[n].props.has_thread_id = true;
2587        ms->possible_cpus->cpus[n].props.thread_id =
2588            n % ms->smp.threads;
2589    }
2590    return ms->possible_cpus;
2591}
2592
2593static void virt_memory_pre_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
2594                                 Error **errp)
2595{
2596    VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2597    const MachineState *ms = MACHINE(hotplug_dev);
2598    const bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2599
2600    if (!vms->acpi_dev) {
2601        error_setg(errp,
2602                   "memory hotplug is not enabled: missing acpi-ged device");
2603        return;
2604    }
2605
2606    if (vms->mte) {
2607        error_setg(errp, "memory hotplug is not enabled: MTE is enabled");
2608        return;
2609    }
2610
2611    if (is_nvdimm && !ms->nvdimms_state->is_enabled) {
2612        error_setg(errp, "nvdimm is not enabled: add 'nvdimm=on' to '-M'");
2613        return;
2614    }
2615
2616    pc_dimm_pre_plug(PC_DIMM(dev), MACHINE(hotplug_dev), NULL, errp);
2617}
2618
2619static void virt_memory_plug(HotplugHandler *hotplug_dev,
2620                             DeviceState *dev, Error **errp)
2621{
2622    VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2623    MachineState *ms = MACHINE(hotplug_dev);
2624    bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2625
2626    pc_dimm_plug(PC_DIMM(dev), MACHINE(vms));
2627
2628    if (is_nvdimm) {
2629        nvdimm_plug(ms->nvdimms_state);
2630    }
2631
2632    hotplug_handler_plug(HOTPLUG_HANDLER(vms->acpi_dev),
2633                         dev, &error_abort);
2634}
2635
2636static void virt_virtio_md_pci_pre_plug(HotplugHandler *hotplug_dev,
2637                                        DeviceState *dev, Error **errp)
2638{
2639    HotplugHandler *hotplug_dev2 = qdev_get_bus_hotplug_handler(dev);
2640    Error *local_err = NULL;
2641
2642    if (!hotplug_dev2 && dev->hotplugged) {
2643        /*
2644         * Without a bus hotplug handler, we cannot control the plug/unplug
2645         * order. We should never reach this point when hotplugging on ARM.
2646         * However, it's nice to add a safety net, similar to what we have
2647         * on x86.
2648         */
2649        error_setg(errp, "hotplug of virtio based memory devices not supported"
2650                   " on this bus.");
2651        return;
2652    }
2653    /*
2654     * First, see if we can plug this memory device at all. If that
2655     * succeeds, branch of to the actual hotplug handler.
2656     */
2657    memory_device_pre_plug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev), NULL,
2658                           &local_err);
2659    if (!local_err && hotplug_dev2) {
2660        hotplug_handler_pre_plug(hotplug_dev2, dev, &local_err);
2661    }
2662    error_propagate(errp, local_err);
2663}
2664
2665static void virt_virtio_md_pci_plug(HotplugHandler *hotplug_dev,
2666                                    DeviceState *dev, Error **errp)
2667{
2668    HotplugHandler *hotplug_dev2 = qdev_get_bus_hotplug_handler(dev);
2669    Error *local_err = NULL;
2670
2671    /*
2672     * Plug the memory device first and then branch off to the actual
2673     * hotplug handler. If that one fails, we can easily undo the memory
2674     * device bits.
2675     */
2676    memory_device_plug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev));
2677    if (hotplug_dev2) {
2678        hotplug_handler_plug(hotplug_dev2, dev, &local_err);
2679        if (local_err) {
2680            memory_device_unplug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev));
2681        }
2682    }
2683    error_propagate(errp, local_err);
2684}
2685
2686static void virt_virtio_md_pci_unplug_request(HotplugHandler *hotplug_dev,
2687                                              DeviceState *dev, Error **errp)
2688{
2689    /* We don't support hot unplug of virtio based memory devices */
2690    error_setg(errp, "virtio based memory devices cannot be unplugged.");
2691}
2692
2693
2694static void virt_machine_device_pre_plug_cb(HotplugHandler *hotplug_dev,
2695                                            DeviceState *dev, Error **errp)
2696{
2697    VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2698
2699    if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2700        virt_memory_pre_plug(hotplug_dev, dev, errp);
2701    } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI)) {
2702        virt_virtio_md_pci_pre_plug(hotplug_dev, dev, errp);
2703    } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2704        hwaddr db_start = 0, db_end = 0;
2705        char *resv_prop_str;
2706
2707        if (vms->iommu != VIRT_IOMMU_NONE) {
2708            error_setg(errp, "virt machine does not support multiple IOMMUs");
2709            return;
2710        }
2711
2712        switch (vms->msi_controller) {
2713        case VIRT_MSI_CTRL_NONE:
2714            return;
2715        case VIRT_MSI_CTRL_ITS:
2716            /* GITS_TRANSLATER page */
2717            db_start = base_memmap[VIRT_GIC_ITS].base + 0x10000;
2718            db_end = base_memmap[VIRT_GIC_ITS].base +
2719                     base_memmap[VIRT_GIC_ITS].size - 1;
2720            break;
2721        case VIRT_MSI_CTRL_GICV2M:
2722            /* MSI_SETSPI_NS page */
2723            db_start = base_memmap[VIRT_GIC_V2M].base;
2724            db_end = db_start + base_memmap[VIRT_GIC_V2M].size - 1;
2725            break;
2726        }
2727        resv_prop_str = g_strdup_printf("0x%"PRIx64":0x%"PRIx64":%u",
2728                                        db_start, db_end,
2729                                        VIRTIO_IOMMU_RESV_MEM_T_MSI);
2730
2731        object_property_set_uint(OBJECT(dev), "len-reserved-regions", 1, errp);
2732        object_property_set_str(OBJECT(dev), "reserved-regions[0]",
2733                                resv_prop_str, errp);
2734        g_free(resv_prop_str);
2735    }
2736}
2737
2738static void virt_machine_device_plug_cb(HotplugHandler *hotplug_dev,
2739                                        DeviceState *dev, Error **errp)
2740{
2741    VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2742
2743    if (vms->platform_bus_dev) {
2744        MachineClass *mc = MACHINE_GET_CLASS(vms);
2745
2746        if (device_is_dynamic_sysbus(mc, dev)) {
2747            platform_bus_link_device(PLATFORM_BUS_DEVICE(vms->platform_bus_dev),
2748                                     SYS_BUS_DEVICE(dev));
2749        }
2750    }
2751    if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2752        virt_memory_plug(hotplug_dev, dev, errp);
2753    }
2754
2755    if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI)) {
2756        virt_virtio_md_pci_plug(hotplug_dev, dev, errp);
2757    }
2758
2759    if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2760        PCIDevice *pdev = PCI_DEVICE(dev);
2761
2762        vms->iommu = VIRT_IOMMU_VIRTIO;
2763        vms->virtio_iommu_bdf = pci_get_bdf(pdev);
2764        create_virtio_iommu_dt_bindings(vms);
2765    }
2766}
2767
2768static void virt_dimm_unplug_request(HotplugHandler *hotplug_dev,
2769                                     DeviceState *dev, Error **errp)
2770{
2771    VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2772    Error *local_err = NULL;
2773
2774    if (!vms->acpi_dev) {
2775        error_setg(&local_err,
2776                   "memory hotplug is not enabled: missing acpi-ged device");
2777        goto out;
2778    }
2779
2780    if (object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM)) {
2781        error_setg(&local_err,
2782                   "nvdimm device hot unplug is not supported yet.");
2783        goto out;
2784    }
2785
2786    hotplug_handler_unplug_request(HOTPLUG_HANDLER(vms->acpi_dev), dev,
2787                                   &local_err);
2788out:
2789    error_propagate(errp, local_err);
2790}
2791
2792static void virt_dimm_unplug(HotplugHandler *hotplug_dev,
2793                             DeviceState *dev, Error **errp)
2794{
2795    VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2796    Error *local_err = NULL;
2797
2798    hotplug_handler_unplug(HOTPLUG_HANDLER(vms->acpi_dev), dev, &local_err);
2799    if (local_err) {
2800        goto out;
2801    }
2802
2803    pc_dimm_unplug(PC_DIMM(dev), MACHINE(vms));
2804    qdev_unrealize(dev);
2805
2806out:
2807    error_propagate(errp, local_err);
2808}
2809
2810static void virt_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
2811                                          DeviceState *dev, Error **errp)
2812{
2813    if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2814        virt_dimm_unplug_request(hotplug_dev, dev, errp);
2815    } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI)) {
2816        virt_virtio_md_pci_unplug_request(hotplug_dev, dev, errp);
2817    } else {
2818        error_setg(errp, "device unplug request for unsupported device"
2819                   " type: %s", object_get_typename(OBJECT(dev)));
2820    }
2821}
2822
2823static void virt_machine_device_unplug_cb(HotplugHandler *hotplug_dev,
2824                                          DeviceState *dev, Error **errp)
2825{
2826    if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2827        virt_dimm_unplug(hotplug_dev, dev, errp);
2828    } else {
2829        error_setg(errp, "virt: device unplug for unsupported device"
2830                   " type: %s", object_get_typename(OBJECT(dev)));
2831    }
2832}
2833
2834static HotplugHandler *virt_machine_get_hotplug_handler(MachineState *machine,
2835                                                        DeviceState *dev)
2836{
2837    MachineClass *mc = MACHINE_GET_CLASS(machine);
2838
2839    if (device_is_dynamic_sysbus(mc, dev) ||
2840        object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) ||
2841        object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI) ||
2842        object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2843        return HOTPLUG_HANDLER(machine);
2844    }
2845    return NULL;
2846}
2847
2848/*
2849 * for arm64 kvm_type [7-0] encodes the requested number of bits
2850 * in the IPA address space
2851 */
2852static int virt_kvm_type(MachineState *ms, const char *type_str)
2853{
2854    VirtMachineState *vms = VIRT_MACHINE(ms);
2855    int max_vm_pa_size, requested_pa_size;
2856    bool fixed_ipa;
2857
2858    max_vm_pa_size = kvm_arm_get_max_vm_ipa_size(ms, &fixed_ipa);
2859
2860    /* we freeze the memory map to compute the highest gpa */
2861    virt_set_memmap(vms, max_vm_pa_size);
2862
2863    requested_pa_size = 64 - clz64(vms->highest_gpa);
2864
2865    /*
2866     * KVM requires the IPA size to be at least 32 bits.
2867     */
2868    if (requested_pa_size < 32) {
2869        requested_pa_size = 32;
2870    }
2871
2872    if (requested_pa_size > max_vm_pa_size) {
2873        error_report("-m and ,maxmem option values "
2874                     "require an IPA range (%d bits) larger than "
2875                     "the one supported by the host (%d bits)",
2876                     requested_pa_size, max_vm_pa_size);
2877        exit(1);
2878    }
2879    /*
2880     * We return the requested PA log size, unless KVM only supports
2881     * the implicit legacy 40b IPA setting, in which case the kvm_type
2882     * must be 0.
2883     */
2884    return fixed_ipa ? 0 : requested_pa_size;
2885}
2886
2887static void virt_machine_class_init(ObjectClass *oc, void *data)
2888{
2889    MachineClass *mc = MACHINE_CLASS(oc);
2890    HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
2891
2892    mc->init = machvirt_init;
2893    /* Start with max_cpus set to 512, which is the maximum supported by KVM.
2894     * The value may be reduced later when we have more information about the
2895     * configuration of the particular instance.
2896     */
2897    mc->max_cpus = 512;
2898    machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_CALXEDA_XGMAC);
2899    machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_AMD_XGBE);
2900    machine_class_allow_dynamic_sysbus_dev(mc, TYPE_RAMFB_DEVICE);
2901    machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_PLATFORM);
2902#ifdef CONFIG_TPM
2903    machine_class_allow_dynamic_sysbus_dev(mc, TYPE_TPM_TIS_SYSBUS);
2904#endif
2905    mc->block_default_type = IF_VIRTIO;
2906    mc->no_cdrom = 1;
2907    mc->pci_allow_0_address = true;
2908    /* We know we will never create a pre-ARMv7 CPU which needs 1K pages */
2909    mc->minimum_page_bits = 12;
2910    mc->possible_cpu_arch_ids = virt_possible_cpu_arch_ids;
2911    mc->cpu_index_to_instance_props = virt_cpu_index_to_props;
2912    mc->default_cpu_type = ARM_CPU_TYPE_NAME("cortex-a15");
2913    mc->get_default_cpu_node_id = virt_get_default_cpu_node_id;
2914    mc->kvm_type = virt_kvm_type;
2915    assert(!mc->get_hotplug_handler);
2916    mc->get_hotplug_handler = virt_machine_get_hotplug_handler;
2917    hc->pre_plug = virt_machine_device_pre_plug_cb;
2918    hc->plug = virt_machine_device_plug_cb;
2919    hc->unplug_request = virt_machine_device_unplug_request_cb;
2920    hc->unplug = virt_machine_device_unplug_cb;
2921    mc->nvdimm_supported = true;
2922    mc->smp_props.clusters_supported = true;
2923    mc->auto_enable_numa_with_memhp = true;
2924    mc->auto_enable_numa_with_memdev = true;
2925    mc->default_ram_id = "mach-virt.ram";
2926
2927    object_class_property_add(oc, "acpi", "OnOffAuto",
2928        virt_get_acpi, virt_set_acpi,
2929        NULL, NULL);
2930    object_class_property_set_description(oc, "acpi",
2931        "Enable ACPI");
2932    object_class_property_add_bool(oc, "secure", virt_get_secure,
2933                                   virt_set_secure);
2934    object_class_property_set_description(oc, "secure",
2935                                                "Set on/off to enable/disable the ARM "
2936                                                "Security Extensions (TrustZone)");
2937
2938    object_class_property_add_bool(oc, "virtualization", virt_get_virt,
2939                                   virt_set_virt);
2940    object_class_property_set_description(oc, "virtualization",
2941                                          "Set on/off to enable/disable emulating a "
2942                                          "guest CPU which implements the ARM "
2943                                          "Virtualization Extensions");
2944
2945    object_class_property_add_bool(oc, "highmem", virt_get_highmem,
2946                                   virt_set_highmem);
2947    object_class_property_set_description(oc, "highmem",
2948                                          "Set on/off to enable/disable using "
2949                                          "physical address space above 32 bits");
2950
2951    object_class_property_add_str(oc, "gic-version", virt_get_gic_version,
2952                                  virt_set_gic_version);
2953    object_class_property_set_description(oc, "gic-version",
2954                                          "Set GIC version. "
2955                                          "Valid values are 2, 3, 4, host and max");
2956
2957    object_class_property_add_str(oc, "iommu", virt_get_iommu, virt_set_iommu);
2958    object_class_property_set_description(oc, "iommu",
2959                                          "Set the IOMMU type. "
2960                                          "Valid values are none and smmuv3");
2961
2962    object_class_property_add_bool(oc, "default-bus-bypass-iommu",
2963                                   virt_get_default_bus_bypass_iommu,
2964                                   virt_set_default_bus_bypass_iommu);
2965    object_class_property_set_description(oc, "default-bus-bypass-iommu",
2966                                          "Set on/off to enable/disable "
2967                                          "bypass_iommu for default root bus");
2968
2969    object_class_property_add_bool(oc, "ras", virt_get_ras,
2970                                   virt_set_ras);
2971    object_class_property_set_description(oc, "ras",
2972                                          "Set on/off to enable/disable reporting host memory errors "
2973                                          "to a KVM guest using ACPI and guest external abort exceptions");
2974
2975    object_class_property_add_bool(oc, "mte", virt_get_mte, virt_set_mte);
2976    object_class_property_set_description(oc, "mte",
2977                                          "Set on/off to enable/disable emulating a "
2978                                          "guest CPU which implements the ARM "
2979                                          "Memory Tagging Extension");
2980
2981    object_class_property_add_bool(oc, "its", virt_get_its,
2982                                   virt_set_its);
2983    object_class_property_set_description(oc, "its",
2984                                          "Set on/off to enable/disable "
2985                                          "ITS instantiation");
2986
2987    object_class_property_add_bool(oc, "dtb-randomness",
2988                                   virt_get_dtb_randomness,
2989                                   virt_set_dtb_randomness);
2990    object_class_property_set_description(oc, "dtb-randomness",
2991                                          "Set off to disable passing random or "
2992                                          "non-deterministic dtb nodes to guest");
2993
2994    object_class_property_add_bool(oc, "dtb-kaslr-seed",
2995                                   virt_get_dtb_randomness,
2996                                   virt_set_dtb_randomness);
2997    object_class_property_set_description(oc, "dtb-kaslr-seed",
2998                                          "Deprecated synonym of dtb-randomness");
2999
3000    object_class_property_add_str(oc, "x-oem-id",
3001                                  virt_get_oem_id,
3002                                  virt_set_oem_id);
3003    object_class_property_set_description(oc, "x-oem-id",
3004                                          "Override the default value of field OEMID "
3005                                          "in ACPI table header."
3006                                          "The string may be up to 6 bytes in size");
3007
3008
3009    object_class_property_add_str(oc, "x-oem-table-id",
3010                                  virt_get_oem_table_id,
3011                                  virt_set_oem_table_id);
3012    object_class_property_set_description(oc, "x-oem-table-id",
3013                                          "Override the default value of field OEM Table ID "
3014                                          "in ACPI table header."
3015                                          "The string may be up to 8 bytes in size");
3016
3017}
3018
3019static void virt_instance_init(Object *obj)
3020{
3021    VirtMachineState *vms = VIRT_MACHINE(obj);
3022    VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
3023
3024    /* EL3 is disabled by default on virt: this makes us consistent
3025     * between KVM and TCG for this board, and it also allows us to
3026     * boot UEFI blobs which assume no TrustZone support.
3027     */
3028    vms->secure = false;
3029
3030    /* EL2 is also disabled by default, for similar reasons */
3031    vms->virt = false;
3032
3033    /* High memory is enabled by default */
3034    vms->highmem = true;
3035    vms->gic_version = VIRT_GIC_VERSION_NOSEL;
3036
3037    vms->highmem_ecam = !vmc->no_highmem_ecam;
3038    vms->highmem_mmio = true;
3039    vms->highmem_redists = true;
3040
3041    if (vmc->no_its) {
3042        vms->its = false;
3043    } else {
3044        /* Default allows ITS instantiation */
3045        vms->its = true;
3046
3047        if (vmc->no_tcg_its) {
3048            vms->tcg_its = false;
3049        } else {
3050            vms->tcg_its = true;
3051        }
3052    }
3053
3054    /* Default disallows iommu instantiation */
3055    vms->iommu = VIRT_IOMMU_NONE;
3056
3057    /* The default root bus is attached to iommu by default */
3058    vms->default_bus_bypass_iommu = false;
3059
3060    /* Default disallows RAS instantiation */
3061    vms->ras = false;
3062
3063    /* MTE is disabled by default.  */
3064    vms->mte = false;
3065
3066    /* Supply kaslr-seed and rng-seed by default */
3067    vms->dtb_randomness = true;
3068
3069    vms->irqmap = a15irqmap;
3070
3071    virt_flash_create(vms);
3072
3073    vms->oem_id = g_strndup(ACPI_BUILD_APPNAME6, 6);
3074    vms->oem_table_id = g_strndup(ACPI_BUILD_APPNAME8, 8);
3075}
3076
3077static const TypeInfo virt_machine_info = {
3078    .name          = TYPE_VIRT_MACHINE,
3079    .parent        = TYPE_MACHINE,
3080    .abstract      = true,
3081    .instance_size = sizeof(VirtMachineState),
3082    .class_size    = sizeof(VirtMachineClass),
3083    .class_init    = virt_machine_class_init,
3084    .instance_init = virt_instance_init,
3085    .interfaces = (InterfaceInfo[]) {
3086         { TYPE_HOTPLUG_HANDLER },
3087         { }
3088    },
3089};
3090
3091static void machvirt_machine_init(void)
3092{
3093    type_register_static(&virt_machine_info);
3094}
3095type_init(machvirt_machine_init);
3096
3097static void virt_machine_7_1_options(MachineClass *mc)
3098{
3099}
3100DEFINE_VIRT_MACHINE_AS_LATEST(7, 1)
3101
3102static void virt_machine_7_0_options(MachineClass *mc)
3103{
3104    virt_machine_7_1_options(mc);
3105    compat_props_add(mc->compat_props, hw_compat_7_0, hw_compat_7_0_len);
3106}
3107DEFINE_VIRT_MACHINE(7, 0)
3108
3109static void virt_machine_6_2_options(MachineClass *mc)
3110{
3111    VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3112
3113    virt_machine_7_0_options(mc);
3114    compat_props_add(mc->compat_props, hw_compat_6_2, hw_compat_6_2_len);
3115    vmc->no_tcg_lpa2 = true;
3116}
3117DEFINE_VIRT_MACHINE(6, 2)
3118
3119static void virt_machine_6_1_options(MachineClass *mc)
3120{
3121    VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3122
3123    virt_machine_6_2_options(mc);
3124    compat_props_add(mc->compat_props, hw_compat_6_1, hw_compat_6_1_len);
3125    mc->smp_props.prefer_sockets = true;
3126    vmc->no_cpu_topology = true;
3127
3128    /* qemu ITS was introduced with 6.2 */
3129    vmc->no_tcg_its = true;
3130}
3131DEFINE_VIRT_MACHINE(6, 1)
3132
3133static void virt_machine_6_0_options(MachineClass *mc)
3134{
3135    virt_machine_6_1_options(mc);
3136    compat_props_add(mc->compat_props, hw_compat_6_0, hw_compat_6_0_len);
3137}
3138DEFINE_VIRT_MACHINE(6, 0)
3139
3140static void virt_machine_5_2_options(MachineClass *mc)
3141{
3142    VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3143
3144    virt_machine_6_0_options(mc);
3145    compat_props_add(mc->compat_props, hw_compat_5_2, hw_compat_5_2_len);
3146    vmc->no_secure_gpio = true;
3147}
3148DEFINE_VIRT_MACHINE(5, 2)
3149
3150static void virt_machine_5_1_options(MachineClass *mc)
3151{
3152    VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3153
3154    virt_machine_5_2_options(mc);
3155    compat_props_add(mc->compat_props, hw_compat_5_1, hw_compat_5_1_len);
3156    vmc->no_kvm_steal_time = true;
3157}
3158DEFINE_VIRT_MACHINE(5, 1)
3159
3160static void virt_machine_5_0_options(MachineClass *mc)
3161{
3162    VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3163
3164    virt_machine_5_1_options(mc);
3165    compat_props_add(mc->compat_props, hw_compat_5_0, hw_compat_5_0_len);
3166    mc->numa_mem_supported = true;
3167    vmc->acpi_expose_flash = true;
3168    mc->auto_enable_numa_with_memdev = false;
3169}
3170DEFINE_VIRT_MACHINE(5, 0)
3171
3172static void virt_machine_4_2_options(MachineClass *mc)
3173{
3174    VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3175
3176    virt_machine_5_0_options(mc);
3177    compat_props_add(mc->compat_props, hw_compat_4_2, hw_compat_4_2_len);
3178    vmc->kvm_no_adjvtime = true;
3179}
3180DEFINE_VIRT_MACHINE(4, 2)
3181
3182static void virt_machine_4_1_options(MachineClass *mc)
3183{
3184    VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3185
3186    virt_machine_4_2_options(mc);
3187    compat_props_add(mc->compat_props, hw_compat_4_1, hw_compat_4_1_len);
3188    vmc->no_ged = true;
3189    mc->auto_enable_numa_with_memhp = false;
3190}
3191DEFINE_VIRT_MACHINE(4, 1)
3192
3193static void virt_machine_4_0_options(MachineClass *mc)
3194{
3195    virt_machine_4_1_options(mc);
3196    compat_props_add(mc->compat_props, hw_compat_4_0, hw_compat_4_0_len);
3197}
3198DEFINE_VIRT_MACHINE(4, 0)
3199
3200static void virt_machine_3_1_options(MachineClass *mc)
3201{
3202    virt_machine_4_0_options(mc);
3203    compat_props_add(mc->compat_props, hw_compat_3_1, hw_compat_3_1_len);
3204}
3205DEFINE_VIRT_MACHINE(3, 1)
3206
3207static void virt_machine_3_0_options(MachineClass *mc)
3208{
3209    virt_machine_3_1_options(mc);
3210    compat_props_add(mc->compat_props, hw_compat_3_0, hw_compat_3_0_len);
3211}
3212DEFINE_VIRT_MACHINE(3, 0)
3213
3214static void virt_machine_2_12_options(MachineClass *mc)
3215{
3216    VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3217
3218    virt_machine_3_0_options(mc);
3219    compat_props_add(mc->compat_props, hw_compat_2_12, hw_compat_2_12_len);
3220    vmc->no_highmem_ecam = true;
3221    mc->max_cpus = 255;
3222}
3223DEFINE_VIRT_MACHINE(2, 12)
3224
3225static void virt_machine_2_11_options(MachineClass *mc)
3226{
3227    VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3228
3229    virt_machine_2_12_options(mc);
3230    compat_props_add(mc->compat_props, hw_compat_2_11, hw_compat_2_11_len);
3231    vmc->smbios_old_sys_ver = true;
3232}
3233DEFINE_VIRT_MACHINE(2, 11)
3234
3235static void virt_machine_2_10_options(MachineClass *mc)
3236{
3237    virt_machine_2_11_options(mc);
3238    compat_props_add(mc->compat_props, hw_compat_2_10, hw_compat_2_10_len);
3239    /* before 2.11 we never faulted accesses to bad addresses */
3240    mc->ignore_memory_transaction_failures = true;
3241}
3242DEFINE_VIRT_MACHINE(2, 10)
3243
3244static void virt_machine_2_9_options(MachineClass *mc)
3245{
3246    virt_machine_2_10_options(mc);
3247    compat_props_add(mc->compat_props, hw_compat_2_9, hw_compat_2_9_len);
3248}
3249DEFINE_VIRT_MACHINE(2, 9)
3250
3251static void virt_machine_2_8_options(MachineClass *mc)
3252{
3253    VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3254
3255    virt_machine_2_9_options(mc);
3256    compat_props_add(mc->compat_props, hw_compat_2_8, hw_compat_2_8_len);
3257    /* For 2.8 and earlier we falsely claimed in the DT that
3258     * our timers were edge-triggered, not level-triggered.
3259     */
3260    vmc->claim_edge_triggered_timers = true;
3261}
3262DEFINE_VIRT_MACHINE(2, 8)
3263
3264static void virt_machine_2_7_options(MachineClass *mc)
3265{
3266    VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3267
3268    virt_machine_2_8_options(mc);
3269    compat_props_add(mc->compat_props, hw_compat_2_7, hw_compat_2_7_len);
3270    /* ITS was introduced with 2.8 */
3271    vmc->no_its = true;
3272    /* Stick with 1K pages for migration compatibility */
3273    mc->minimum_page_bits = 0;
3274}
3275DEFINE_VIRT_MACHINE(2, 7)
3276
3277static void virt_machine_2_6_options(MachineClass *mc)
3278{
3279    VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3280
3281    virt_machine_2_7_options(mc);
3282    compat_props_add(mc->compat_props, hw_compat_2_6, hw_compat_2_6_len);
3283    vmc->disallow_affinity_adjustment = true;
3284    /* Disable PMU for 2.6 as PMU support was first introduced in 2.7 */
3285    vmc->no_pmu = true;
3286}
3287DEFINE_VIRT_MACHINE(2, 6)
3288