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 "qapi/error.h"
  33#include "hw/sysbus.h"
  34#include "hw/arm/arm.h"
  35#include "hw/arm/primecell.h"
  36#include "hw/arm/virt.h"
  37#include "hw/devices.h"
  38#include "net/net.h"
  39#include "sysemu/block-backend.h"
  40#include "sysemu/device_tree.h"
  41#include "sysemu/numa.h"
  42#include "sysemu/sysemu.h"
  43#include "sysemu/kvm.h"
  44#include "hw/compat.h"
  45#include "hw/loader.h"
  46#include "exec/address-spaces.h"
  47#include "qemu/bitops.h"
  48#include "qemu/error-report.h"
  49#include "hw/pci-host/gpex.h"
  50#include "hw/arm/sysbus-fdt.h"
  51#include "hw/platform-bus.h"
  52#include "hw/arm/fdt.h"
  53#include "hw/intc/arm_gic.h"
  54#include "hw/intc/arm_gicv3_common.h"
  55#include "kvm_arm.h"
  56#include "hw/smbios/smbios.h"
  57#include "qapi/visitor.h"
  58#include "standard-headers/linux/input.h"
  59
  60#define DEFINE_VIRT_MACHINE_LATEST(major, minor, latest) \
  61    static void virt_##major##_##minor##_class_init(ObjectClass *oc, \
  62                                                    void *data) \
  63    { \
  64        MachineClass *mc = MACHINE_CLASS(oc); \
  65        virt_machine_##major##_##minor##_options(mc); \
  66        mc->desc = "QEMU " # major "." # minor " ARM Virtual Machine"; \
  67        if (latest) { \
  68            mc->alias = "virt"; \
  69        } \
  70    } \
  71    static const TypeInfo machvirt_##major##_##minor##_info = { \
  72        .name = MACHINE_TYPE_NAME("virt-" # major "." # minor), \
  73        .parent = TYPE_VIRT_MACHINE, \
  74        .instance_init = virt_##major##_##minor##_instance_init, \
  75        .class_init = virt_##major##_##minor##_class_init, \
  76    }; \
  77    static void machvirt_machine_##major##_##minor##_init(void) \
  78    { \
  79        type_register_static(&machvirt_##major##_##minor##_info); \
  80    } \
  81    type_init(machvirt_machine_##major##_##minor##_init);
  82
  83#define DEFINE_VIRT_MACHINE_AS_LATEST(major, minor) \
  84    DEFINE_VIRT_MACHINE_LATEST(major, minor, true)
  85#define DEFINE_VIRT_MACHINE(major, minor) \
  86    DEFINE_VIRT_MACHINE_LATEST(major, minor, false)
  87
  88
  89/* Number of external interrupt lines to configure the GIC with */
  90#define NUM_IRQS 256
  91
  92#define PLATFORM_BUS_NUM_IRQS 64
  93
  94static ARMPlatformBusSystemParams platform_bus_params;
  95
  96/* RAM limit in GB. Since VIRT_MEM starts at the 1GB mark, this means
  97 * RAM can go up to the 256GB mark, leaving 256GB of the physical
  98 * address space unallocated and free for future use between 256G and 512G.
  99 * If we need to provide more RAM to VMs in the future then we need to:
 100 *  * allocate a second bank of RAM starting at 2TB and working up
 101 *  * fix the DT and ACPI table generation code in QEMU to correctly
 102 *    report two split lumps of RAM to the guest
 103 *  * fix KVM in the host kernel to allow guests with >40 bit address spaces
 104 * (We don't want to fill all the way up to 512GB with RAM because
 105 * we might want it for non-RAM purposes later. Conversely it seems
 106 * reasonable to assume that anybody configuring a VM with a quarter
 107 * of a terabyte of RAM will be doing it on a host with more than a
 108 * terabyte of physical address space.)
 109 */
 110#define RAMLIMIT_GB 255
 111#define RAMLIMIT_BYTES (RAMLIMIT_GB * 1024ULL * 1024 * 1024)
 112
 113/* Addresses and sizes of our components.
 114 * 0..128MB is space for a flash device so we can run bootrom code such as UEFI.
 115 * 128MB..256MB is used for miscellaneous device I/O.
 116 * 256MB..1GB is reserved for possible future PCI support (ie where the
 117 * PCI memory window will go if we add a PCI host controller).
 118 * 1GB and up is RAM (which may happily spill over into the
 119 * high memory region beyond 4GB).
 120 * This represents a compromise between how much RAM can be given to
 121 * a 32 bit VM and leaving space for expansion and in particular for PCI.
 122 * Note that devices should generally be placed at multiples of 0x10000,
 123 * to accommodate guests using 64K pages.
 124 */
 125static const MemMapEntry a15memmap[] = {
 126    /* Space up to 0x8000000 is reserved for a boot ROM */
 127    [VIRT_FLASH] =              {          0, 0x08000000 },
 128    [VIRT_CPUPERIPHS] =         { 0x08000000, 0x00020000 },
 129    /* GIC distributor and CPU interfaces sit inside the CPU peripheral space */
 130    [VIRT_GIC_DIST] =           { 0x08000000, 0x00010000 },
 131    [VIRT_GIC_CPU] =            { 0x08010000, 0x00010000 },
 132    [VIRT_GIC_V2M] =            { 0x08020000, 0x00001000 },
 133    /* The space in between here is reserved for GICv3 CPU/vCPU/HYP */
 134    [VIRT_GIC_ITS] =            { 0x08080000, 0x00020000 },
 135    /* This redistributor space allows up to 2*64kB*123 CPUs */
 136    [VIRT_GIC_REDIST] =         { 0x080A0000, 0x00F60000 },
 137    [VIRT_UART] =               { 0x09000000, 0x00001000 },
 138    [VIRT_RTC] =                { 0x09010000, 0x00001000 },
 139    [VIRT_FW_CFG] =             { 0x09020000, 0x00000018 },
 140    [VIRT_GPIO] =               { 0x09030000, 0x00001000 },
 141    [VIRT_SECURE_UART] =        { 0x09040000, 0x00001000 },
 142    [VIRT_MMIO] =               { 0x0a000000, 0x00000200 },
 143    /* ...repeating for a total of NUM_VIRTIO_TRANSPORTS, each of that size */
 144    [VIRT_PLATFORM_BUS] =       { 0x0c000000, 0x02000000 },
 145    [VIRT_SECURE_MEM] =         { 0x0e000000, 0x01000000 },
 146    [VIRT_PCIE_MMIO] =          { 0x10000000, 0x2eff0000 },
 147    [VIRT_PCIE_PIO] =           { 0x3eff0000, 0x00010000 },
 148    [VIRT_PCIE_ECAM] =          { 0x3f000000, 0x01000000 },
 149    [VIRT_MEM] =                { 0x40000000, RAMLIMIT_BYTES },
 150    /* Second PCIe window, 512GB wide at the 512GB boundary */
 151    [VIRT_PCIE_MMIO_HIGH] =   { 0x8000000000ULL, 0x8000000000ULL },
 152};
 153
 154static const int a15irqmap[] = {
 155    [VIRT_UART] = 1,
 156    [VIRT_RTC] = 2,
 157    [VIRT_PCIE] = 3, /* ... to 6 */
 158    [VIRT_GPIO] = 7,
 159    [VIRT_SECURE_UART] = 8,
 160    [VIRT_MMIO] = 16, /* ...to 16 + NUM_VIRTIO_TRANSPORTS - 1 */
 161    [VIRT_GIC_V2M] = 48, /* ...to 48 + NUM_GICV2M_SPIS - 1 */
 162    [VIRT_PLATFORM_BUS] = 112, /* ...to 112 + PLATFORM_BUS_NUM_IRQS -1 */
 163};
 164
 165static const char *valid_cpus[] = {
 166    ARM_CPU_TYPE_NAME("cortex-a15"),
 167    ARM_CPU_TYPE_NAME("cortex-a53"),
 168    ARM_CPU_TYPE_NAME("cortex-a57"),
 169    ARM_CPU_TYPE_NAME("host"),
 170};
 171
 172static bool cpu_type_valid(const char *cpu)
 173{
 174    int i;
 175
 176    for (i = 0; i < ARRAY_SIZE(valid_cpus); i++) {
 177        if (strcmp(cpu, valid_cpus[i]) == 0) {
 178            return true;
 179        }
 180    }
 181    return false;
 182}
 183
 184static void create_fdt(VirtMachineState *vms)
 185{
 186    void *fdt = create_device_tree(&vms->fdt_size);
 187
 188    if (!fdt) {
 189        error_report("create_device_tree() failed");
 190        exit(1);
 191    }
 192
 193    vms->fdt = fdt;
 194
 195    /* Header */
 196    qemu_fdt_setprop_string(fdt, "/", "compatible", "linux,dummy-virt");
 197    qemu_fdt_setprop_cell(fdt, "/", "#address-cells", 0x2);
 198    qemu_fdt_setprop_cell(fdt, "/", "#size-cells", 0x2);
 199
 200    /*
 201     * /chosen and /memory nodes must exist for load_dtb
 202     * to fill in necessary properties later
 203     */
 204    qemu_fdt_add_subnode(fdt, "/chosen");
 205    qemu_fdt_add_subnode(fdt, "/memory");
 206    qemu_fdt_setprop_string(fdt, "/memory", "device_type", "memory");
 207
 208    /* Clock node, for the benefit of the UART. The kernel device tree
 209     * binding documentation claims the PL011 node clock properties are
 210     * optional but in practice if you omit them the kernel refuses to
 211     * probe for the device.
 212     */
 213    vms->clock_phandle = qemu_fdt_alloc_phandle(fdt);
 214    qemu_fdt_add_subnode(fdt, "/apb-pclk");
 215    qemu_fdt_setprop_string(fdt, "/apb-pclk", "compatible", "fixed-clock");
 216    qemu_fdt_setprop_cell(fdt, "/apb-pclk", "#clock-cells", 0x0);
 217    qemu_fdt_setprop_cell(fdt, "/apb-pclk", "clock-frequency", 24000000);
 218    qemu_fdt_setprop_string(fdt, "/apb-pclk", "clock-output-names",
 219                                "clk24mhz");
 220    qemu_fdt_setprop_cell(fdt, "/apb-pclk", "phandle", vms->clock_phandle);
 221
 222    if (have_numa_distance) {
 223        int size = nb_numa_nodes * nb_numa_nodes * 3 * sizeof(uint32_t);
 224        uint32_t *matrix = g_malloc0(size);
 225        int idx, i, j;
 226
 227        for (i = 0; i < nb_numa_nodes; i++) {
 228            for (j = 0; j < nb_numa_nodes; j++) {
 229                idx = (i * nb_numa_nodes + j) * 3;
 230                matrix[idx + 0] = cpu_to_be32(i);
 231                matrix[idx + 1] = cpu_to_be32(j);
 232                matrix[idx + 2] = cpu_to_be32(numa_info[i].distance[j]);
 233            }
 234        }
 235
 236        qemu_fdt_add_subnode(fdt, "/distance-map");
 237        qemu_fdt_setprop_string(fdt, "/distance-map", "compatible",
 238                                "numa-distance-map-v1");
 239        qemu_fdt_setprop(fdt, "/distance-map", "distance-matrix",
 240                         matrix, size);
 241        g_free(matrix);
 242    }
 243}
 244
 245static void fdt_add_psci_node(const VirtMachineState *vms)
 246{
 247    uint32_t cpu_suspend_fn;
 248    uint32_t cpu_off_fn;
 249    uint32_t cpu_on_fn;
 250    uint32_t migrate_fn;
 251    void *fdt = vms->fdt;
 252    ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(0));
 253    const char *psci_method;
 254
 255    switch (vms->psci_conduit) {
 256    case QEMU_PSCI_CONDUIT_DISABLED:
 257        return;
 258    case QEMU_PSCI_CONDUIT_HVC:
 259        psci_method = "hvc";
 260        break;
 261    case QEMU_PSCI_CONDUIT_SMC:
 262        psci_method = "smc";
 263        break;
 264    default:
 265        g_assert_not_reached();
 266    }
 267
 268    qemu_fdt_add_subnode(fdt, "/psci");
 269    if (armcpu->psci_version == 2) {
 270        const char comp[] = "arm,psci-0.2\0arm,psci";
 271        qemu_fdt_setprop(fdt, "/psci", "compatible", comp, sizeof(comp));
 272
 273        cpu_off_fn = QEMU_PSCI_0_2_FN_CPU_OFF;
 274        if (arm_feature(&armcpu->env, ARM_FEATURE_AARCH64)) {
 275            cpu_suspend_fn = QEMU_PSCI_0_2_FN64_CPU_SUSPEND;
 276            cpu_on_fn = QEMU_PSCI_0_2_FN64_CPU_ON;
 277            migrate_fn = QEMU_PSCI_0_2_FN64_MIGRATE;
 278        } else {
 279            cpu_suspend_fn = QEMU_PSCI_0_2_FN_CPU_SUSPEND;
 280            cpu_on_fn = QEMU_PSCI_0_2_FN_CPU_ON;
 281            migrate_fn = QEMU_PSCI_0_2_FN_MIGRATE;
 282        }
 283    } else {
 284        qemu_fdt_setprop_string(fdt, "/psci", "compatible", "arm,psci");
 285
 286        cpu_suspend_fn = QEMU_PSCI_0_1_FN_CPU_SUSPEND;
 287        cpu_off_fn = QEMU_PSCI_0_1_FN_CPU_OFF;
 288        cpu_on_fn = QEMU_PSCI_0_1_FN_CPU_ON;
 289        migrate_fn = QEMU_PSCI_0_1_FN_MIGRATE;
 290    }
 291
 292    /* We adopt the PSCI spec's nomenclature, and use 'conduit' to refer
 293     * to the instruction that should be used to invoke PSCI functions.
 294     * However, the device tree binding uses 'method' instead, so that is
 295     * what we should use here.
 296     */
 297    qemu_fdt_setprop_string(fdt, "/psci", "method", psci_method);
 298
 299    qemu_fdt_setprop_cell(fdt, "/psci", "cpu_suspend", cpu_suspend_fn);
 300    qemu_fdt_setprop_cell(fdt, "/psci", "cpu_off", cpu_off_fn);
 301    qemu_fdt_setprop_cell(fdt, "/psci", "cpu_on", cpu_on_fn);
 302    qemu_fdt_setprop_cell(fdt, "/psci", "migrate", migrate_fn);
 303}
 304
 305static void fdt_add_timer_nodes(const VirtMachineState *vms)
 306{
 307    /* On real hardware these interrupts are level-triggered.
 308     * On KVM they were edge-triggered before host kernel version 4.4,
 309     * and level-triggered afterwards.
 310     * On emulated QEMU they are level-triggered.
 311     *
 312     * Getting the DTB info about them wrong is awkward for some
 313     * guest kernels:
 314     *  pre-4.8 ignore the DT and leave the interrupt configured
 315     *   with whatever the GIC reset value (or the bootloader) left it at
 316     *  4.8 before rc6 honour the incorrect data by programming it back
 317     *   into the GIC, causing problems
 318     *  4.8rc6 and later ignore the DT and always write "level triggered"
 319     *   into the GIC
 320     *
 321     * For backwards-compatibility, virt-2.8 and earlier will continue
 322     * to say these are edge-triggered, but later machines will report
 323     * the correct information.
 324     */
 325    ARMCPU *armcpu;
 326    VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
 327    uint32_t irqflags = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
 328
 329    if (vmc->claim_edge_triggered_timers) {
 330        irqflags = GIC_FDT_IRQ_FLAGS_EDGE_LO_HI;
 331    }
 332
 333    if (vms->gic_version == 2) {
 334        irqflags = deposit32(irqflags, GIC_FDT_IRQ_PPI_CPU_START,
 335                             GIC_FDT_IRQ_PPI_CPU_WIDTH,
 336                             (1 << vms->smp_cpus) - 1);
 337    }
 338
 339    qemu_fdt_add_subnode(vms->fdt, "/timer");
 340
 341    armcpu = ARM_CPU(qemu_get_cpu(0));
 342    if (arm_feature(&armcpu->env, ARM_FEATURE_V8)) {
 343        const char compat[] = "arm,armv8-timer\0arm,armv7-timer";
 344        qemu_fdt_setprop(vms->fdt, "/timer", "compatible",
 345                         compat, sizeof(compat));
 346    } else {
 347        qemu_fdt_setprop_string(vms->fdt, "/timer", "compatible",
 348                                "arm,armv7-timer");
 349    }
 350    qemu_fdt_setprop(vms->fdt, "/timer", "always-on", NULL, 0);
 351    qemu_fdt_setprop_cells(vms->fdt, "/timer", "interrupts",
 352                       GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_S_EL1_IRQ, irqflags,
 353                       GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_NS_EL1_IRQ, irqflags,
 354                       GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_VIRT_IRQ, irqflags,
 355                       GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_NS_EL2_IRQ, irqflags);
 356}
 357
 358static void fdt_add_cpu_nodes(const VirtMachineState *vms)
 359{
 360    int cpu;
 361    int addr_cells = 1;
 362    const MachineState *ms = MACHINE(vms);
 363
 364    /*
 365     * From Documentation/devicetree/bindings/arm/cpus.txt
 366     *  On ARM v8 64-bit systems value should be set to 2,
 367     *  that corresponds to the MPIDR_EL1 register size.
 368     *  If MPIDR_EL1[63:32] value is equal to 0 on all CPUs
 369     *  in the system, #address-cells can be set to 1, since
 370     *  MPIDR_EL1[63:32] bits are not used for CPUs
 371     *  identification.
 372     *
 373     *  Here we actually don't know whether our system is 32- or 64-bit one.
 374     *  The simplest way to go is to examine affinity IDs of all our CPUs. If
 375     *  at least one of them has Aff3 populated, we set #address-cells to 2.
 376     */
 377    for (cpu = 0; cpu < vms->smp_cpus; cpu++) {
 378        ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(cpu));
 379
 380        if (armcpu->mp_affinity & ARM_AFF3_MASK) {
 381            addr_cells = 2;
 382            break;
 383        }
 384    }
 385
 386    qemu_fdt_add_subnode(vms->fdt, "/cpus");
 387    qemu_fdt_setprop_cell(vms->fdt, "/cpus", "#address-cells", addr_cells);
 388    qemu_fdt_setprop_cell(vms->fdt, "/cpus", "#size-cells", 0x0);
 389
 390    for (cpu = vms->smp_cpus - 1; cpu >= 0; cpu--) {
 391        char *nodename = g_strdup_printf("/cpus/cpu@%d", cpu);
 392        ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(cpu));
 393        CPUState *cs = CPU(armcpu);
 394
 395        qemu_fdt_add_subnode(vms->fdt, nodename);
 396        qemu_fdt_setprop_string(vms->fdt, nodename, "device_type", "cpu");
 397        qemu_fdt_setprop_string(vms->fdt, nodename, "compatible",
 398                                    armcpu->dtb_compatible);
 399
 400        if (vms->psci_conduit != QEMU_PSCI_CONDUIT_DISABLED
 401            && vms->smp_cpus > 1) {
 402            qemu_fdt_setprop_string(vms->fdt, nodename,
 403                                        "enable-method", "psci");
 404        }
 405
 406        if (addr_cells == 2) {
 407            qemu_fdt_setprop_u64(vms->fdt, nodename, "reg",
 408                                 armcpu->mp_affinity);
 409        } else {
 410            qemu_fdt_setprop_cell(vms->fdt, nodename, "reg",
 411                                  armcpu->mp_affinity);
 412        }
 413
 414        if (ms->possible_cpus->cpus[cs->cpu_index].props.has_node_id) {
 415            qemu_fdt_setprop_cell(vms->fdt, nodename, "numa-node-id",
 416                ms->possible_cpus->cpus[cs->cpu_index].props.node_id);
 417        }
 418
 419        g_free(nodename);
 420    }
 421}
 422
 423static void fdt_add_its_gic_node(VirtMachineState *vms)
 424{
 425    vms->msi_phandle = qemu_fdt_alloc_phandle(vms->fdt);
 426    qemu_fdt_add_subnode(vms->fdt, "/intc/its");
 427    qemu_fdt_setprop_string(vms->fdt, "/intc/its", "compatible",
 428                            "arm,gic-v3-its");
 429    qemu_fdt_setprop(vms->fdt, "/intc/its", "msi-controller", NULL, 0);
 430    qemu_fdt_setprop_sized_cells(vms->fdt, "/intc/its", "reg",
 431                                 2, vms->memmap[VIRT_GIC_ITS].base,
 432                                 2, vms->memmap[VIRT_GIC_ITS].size);
 433    qemu_fdt_setprop_cell(vms->fdt, "/intc/its", "phandle", vms->msi_phandle);
 434}
 435
 436static void fdt_add_v2m_gic_node(VirtMachineState *vms)
 437{
 438    vms->msi_phandle = qemu_fdt_alloc_phandle(vms->fdt);
 439    qemu_fdt_add_subnode(vms->fdt, "/intc/v2m");
 440    qemu_fdt_setprop_string(vms->fdt, "/intc/v2m", "compatible",
 441                            "arm,gic-v2m-frame");
 442    qemu_fdt_setprop(vms->fdt, "/intc/v2m", "msi-controller", NULL, 0);
 443    qemu_fdt_setprop_sized_cells(vms->fdt, "/intc/v2m", "reg",
 444                                 2, vms->memmap[VIRT_GIC_V2M].base,
 445                                 2, vms->memmap[VIRT_GIC_V2M].size);
 446    qemu_fdt_setprop_cell(vms->fdt, "/intc/v2m", "phandle", vms->msi_phandle);
 447}
 448
 449static void fdt_add_gic_node(VirtMachineState *vms)
 450{
 451    vms->gic_phandle = qemu_fdt_alloc_phandle(vms->fdt);
 452    qemu_fdt_setprop_cell(vms->fdt, "/", "interrupt-parent", vms->gic_phandle);
 453
 454    qemu_fdt_add_subnode(vms->fdt, "/intc");
 455    qemu_fdt_setprop_cell(vms->fdt, "/intc", "#interrupt-cells", 3);
 456    qemu_fdt_setprop(vms->fdt, "/intc", "interrupt-controller", NULL, 0);
 457    qemu_fdt_setprop_cell(vms->fdt, "/intc", "#address-cells", 0x2);
 458    qemu_fdt_setprop_cell(vms->fdt, "/intc", "#size-cells", 0x2);
 459    qemu_fdt_setprop(vms->fdt, "/intc", "ranges", NULL, 0);
 460    if (vms->gic_version == 3) {
 461        qemu_fdt_setprop_string(vms->fdt, "/intc", "compatible",
 462                                "arm,gic-v3");
 463        qemu_fdt_setprop_sized_cells(vms->fdt, "/intc", "reg",
 464                                     2, vms->memmap[VIRT_GIC_DIST].base,
 465                                     2, vms->memmap[VIRT_GIC_DIST].size,
 466                                     2, vms->memmap[VIRT_GIC_REDIST].base,
 467                                     2, vms->memmap[VIRT_GIC_REDIST].size);
 468        if (vms->virt) {
 469            qemu_fdt_setprop_cells(vms->fdt, "/intc", "interrupts",
 470                                   GIC_FDT_IRQ_TYPE_PPI, ARCH_GICV3_MAINT_IRQ,
 471                                   GIC_FDT_IRQ_FLAGS_LEVEL_HI);
 472        }
 473    } else {
 474        /* 'cortex-a15-gic' means 'GIC v2' */
 475        qemu_fdt_setprop_string(vms->fdt, "/intc", "compatible",
 476                                "arm,cortex-a15-gic");
 477        qemu_fdt_setprop_sized_cells(vms->fdt, "/intc", "reg",
 478                                      2, vms->memmap[VIRT_GIC_DIST].base,
 479                                      2, vms->memmap[VIRT_GIC_DIST].size,
 480                                      2, vms->memmap[VIRT_GIC_CPU].base,
 481                                      2, vms->memmap[VIRT_GIC_CPU].size);
 482    }
 483
 484    qemu_fdt_setprop_cell(vms->fdt, "/intc", "phandle", vms->gic_phandle);
 485}
 486
 487static void fdt_add_pmu_nodes(const VirtMachineState *vms)
 488{
 489    CPUState *cpu;
 490    ARMCPU *armcpu;
 491    uint32_t irqflags = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
 492
 493    CPU_FOREACH(cpu) {
 494        armcpu = ARM_CPU(cpu);
 495        if (!arm_feature(&armcpu->env, ARM_FEATURE_PMU)) {
 496            return;
 497        }
 498        if (kvm_enabled()) {
 499            if (kvm_irqchip_in_kernel()) {
 500                kvm_arm_pmu_set_irq(cpu, PPI(VIRTUAL_PMU_IRQ));
 501            }
 502            kvm_arm_pmu_init(cpu);
 503        }
 504    }
 505
 506    if (vms->gic_version == 2) {
 507        irqflags = deposit32(irqflags, GIC_FDT_IRQ_PPI_CPU_START,
 508                             GIC_FDT_IRQ_PPI_CPU_WIDTH,
 509                             (1 << vms->smp_cpus) - 1);
 510    }
 511
 512    armcpu = ARM_CPU(qemu_get_cpu(0));
 513    qemu_fdt_add_subnode(vms->fdt, "/pmu");
 514    if (arm_feature(&armcpu->env, ARM_FEATURE_V8)) {
 515        const char compat[] = "arm,armv8-pmuv3";
 516        qemu_fdt_setprop(vms->fdt, "/pmu", "compatible",
 517                         compat, sizeof(compat));
 518        qemu_fdt_setprop_cells(vms->fdt, "/pmu", "interrupts",
 519                               GIC_FDT_IRQ_TYPE_PPI, VIRTUAL_PMU_IRQ, irqflags);
 520    }
 521}
 522
 523static void create_its(VirtMachineState *vms, DeviceState *gicdev)
 524{
 525    const char *itsclass = its_class_name();
 526    DeviceState *dev;
 527
 528    if (!itsclass) {
 529        /* Do nothing if not supported */
 530        return;
 531    }
 532
 533    dev = qdev_create(NULL, itsclass);
 534
 535    object_property_set_link(OBJECT(dev), OBJECT(gicdev), "parent-gicv3",
 536                             &error_abort);
 537    qdev_init_nofail(dev);
 538    sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_GIC_ITS].base);
 539
 540    fdt_add_its_gic_node(vms);
 541}
 542
 543static void create_v2m(VirtMachineState *vms, qemu_irq *pic)
 544{
 545    int i;
 546    int irq = vms->irqmap[VIRT_GIC_V2M];
 547    DeviceState *dev;
 548
 549    dev = qdev_create(NULL, "arm-gicv2m");
 550    sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_GIC_V2M].base);
 551    qdev_prop_set_uint32(dev, "base-spi", irq);
 552    qdev_prop_set_uint32(dev, "num-spi", NUM_GICV2M_SPIS);
 553    qdev_init_nofail(dev);
 554
 555    for (i = 0; i < NUM_GICV2M_SPIS; i++) {
 556        sysbus_connect_irq(SYS_BUS_DEVICE(dev), i, pic[irq + i]);
 557    }
 558
 559    fdt_add_v2m_gic_node(vms);
 560}
 561
 562static void create_gic(VirtMachineState *vms, qemu_irq *pic)
 563{
 564    /* We create a standalone GIC */
 565    DeviceState *gicdev;
 566    SysBusDevice *gicbusdev;
 567    const char *gictype;
 568    int type = vms->gic_version, i;
 569
 570    gictype = (type == 3) ? gicv3_class_name() : gic_class_name();
 571
 572    gicdev = qdev_create(NULL, gictype);
 573    qdev_prop_set_uint32(gicdev, "revision", type);
 574    qdev_prop_set_uint32(gicdev, "num-cpu", smp_cpus);
 575    /* Note that the num-irq property counts both internal and external
 576     * interrupts; there are always 32 of the former (mandated by GIC spec).
 577     */
 578    qdev_prop_set_uint32(gicdev, "num-irq", NUM_IRQS + 32);
 579    if (!kvm_irqchip_in_kernel()) {
 580        qdev_prop_set_bit(gicdev, "has-security-extensions", vms->secure);
 581    }
 582    qdev_init_nofail(gicdev);
 583    gicbusdev = SYS_BUS_DEVICE(gicdev);
 584    sysbus_mmio_map(gicbusdev, 0, vms->memmap[VIRT_GIC_DIST].base);
 585    if (type == 3) {
 586        sysbus_mmio_map(gicbusdev, 1, vms->memmap[VIRT_GIC_REDIST].base);
 587    } else {
 588        sysbus_mmio_map(gicbusdev, 1, vms->memmap[VIRT_GIC_CPU].base);
 589    }
 590
 591    /* Wire the outputs from each CPU's generic timer and the GICv3
 592     * maintenance interrupt signal to the appropriate GIC PPI inputs,
 593     * and the GIC's IRQ/FIQ/VIRQ/VFIQ interrupt outputs to the CPU's inputs.
 594     */
 595    for (i = 0; i < smp_cpus; i++) {
 596        DeviceState *cpudev = DEVICE(qemu_get_cpu(i));
 597        int ppibase = NUM_IRQS + i * GIC_INTERNAL + GIC_NR_SGIS;
 598        int irq;
 599        /* Mapping from the output timer irq lines from the CPU to the
 600         * GIC PPI inputs we use for the virt board.
 601         */
 602        const int timer_irq[] = {
 603            [GTIMER_PHYS] = ARCH_TIMER_NS_EL1_IRQ,
 604            [GTIMER_VIRT] = ARCH_TIMER_VIRT_IRQ,
 605            [GTIMER_HYP]  = ARCH_TIMER_NS_EL2_IRQ,
 606            [GTIMER_SEC]  = ARCH_TIMER_S_EL1_IRQ,
 607        };
 608
 609        for (irq = 0; irq < ARRAY_SIZE(timer_irq); irq++) {
 610            qdev_connect_gpio_out(cpudev, irq,
 611                                  qdev_get_gpio_in(gicdev,
 612                                                   ppibase + timer_irq[irq]));
 613        }
 614
 615        qdev_connect_gpio_out_named(cpudev, "gicv3-maintenance-interrupt", 0,
 616                                    qdev_get_gpio_in(gicdev, ppibase
 617                                                     + ARCH_GICV3_MAINT_IRQ));
 618        qdev_connect_gpio_out_named(cpudev, "pmu-interrupt", 0,
 619                                    qdev_get_gpio_in(gicdev, ppibase
 620                                                     + VIRTUAL_PMU_IRQ));
 621
 622        sysbus_connect_irq(gicbusdev, i, qdev_get_gpio_in(cpudev, ARM_CPU_IRQ));
 623        sysbus_connect_irq(gicbusdev, i + smp_cpus,
 624                           qdev_get_gpio_in(cpudev, ARM_CPU_FIQ));
 625        sysbus_connect_irq(gicbusdev, i + 2 * smp_cpus,
 626                           qdev_get_gpio_in(cpudev, ARM_CPU_VIRQ));
 627        sysbus_connect_irq(gicbusdev, i + 3 * smp_cpus,
 628                           qdev_get_gpio_in(cpudev, ARM_CPU_VFIQ));
 629    }
 630
 631    for (i = 0; i < NUM_IRQS; i++) {
 632        pic[i] = qdev_get_gpio_in(gicdev, i);
 633    }
 634
 635    fdt_add_gic_node(vms);
 636
 637    if (type == 3 && vms->its) {
 638        create_its(vms, gicdev);
 639    } else if (type == 2) {
 640        create_v2m(vms, pic);
 641    }
 642}
 643
 644static void create_uart(const VirtMachineState *vms, qemu_irq *pic, int uart,
 645                        MemoryRegion *mem, Chardev *chr)
 646{
 647    char *nodename;
 648    hwaddr base = vms->memmap[uart].base;
 649    hwaddr size = vms->memmap[uart].size;
 650    int irq = vms->irqmap[uart];
 651    const char compat[] = "arm,pl011\0arm,primecell";
 652    const char clocknames[] = "uartclk\0apb_pclk";
 653    DeviceState *dev = qdev_create(NULL, "pl011");
 654    SysBusDevice *s = SYS_BUS_DEVICE(dev);
 655
 656    qdev_prop_set_chr(dev, "chardev", chr);
 657    qdev_init_nofail(dev);
 658    memory_region_add_subregion(mem, base,
 659                                sysbus_mmio_get_region(s, 0));
 660    sysbus_connect_irq(s, 0, pic[irq]);
 661
 662    nodename = g_strdup_printf("/pl011@%" PRIx64, base);
 663    qemu_fdt_add_subnode(vms->fdt, nodename);
 664    /* Note that we can't use setprop_string because of the embedded NUL */
 665    qemu_fdt_setprop(vms->fdt, nodename, "compatible",
 666                         compat, sizeof(compat));
 667    qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
 668                                     2, base, 2, size);
 669    qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupts",
 670                               GIC_FDT_IRQ_TYPE_SPI, irq,
 671                               GIC_FDT_IRQ_FLAGS_LEVEL_HI);
 672    qemu_fdt_setprop_cells(vms->fdt, nodename, "clocks",
 673                               vms->clock_phandle, vms->clock_phandle);
 674    qemu_fdt_setprop(vms->fdt, nodename, "clock-names",
 675                         clocknames, sizeof(clocknames));
 676
 677    if (uart == VIRT_UART) {
 678        qemu_fdt_setprop_string(vms->fdt, "/chosen", "stdout-path", nodename);
 679    } else {
 680        /* Mark as not usable by the normal world */
 681        qemu_fdt_setprop_string(vms->fdt, nodename, "status", "disabled");
 682        qemu_fdt_setprop_string(vms->fdt, nodename, "secure-status", "okay");
 683    }
 684
 685    g_free(nodename);
 686}
 687
 688static void create_rtc(const VirtMachineState *vms, qemu_irq *pic)
 689{
 690    char *nodename;
 691    hwaddr base = vms->memmap[VIRT_RTC].base;
 692    hwaddr size = vms->memmap[VIRT_RTC].size;
 693    int irq = vms->irqmap[VIRT_RTC];
 694    const char compat[] = "arm,pl031\0arm,primecell";
 695
 696    sysbus_create_simple("pl031", base, pic[irq]);
 697
 698    nodename = g_strdup_printf("/pl031@%" PRIx64, base);
 699    qemu_fdt_add_subnode(vms->fdt, nodename);
 700    qemu_fdt_setprop(vms->fdt, nodename, "compatible", compat, sizeof(compat));
 701    qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
 702                                 2, base, 2, size);
 703    qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupts",
 704                           GIC_FDT_IRQ_TYPE_SPI, irq,
 705                           GIC_FDT_IRQ_FLAGS_LEVEL_HI);
 706    qemu_fdt_setprop_cell(vms->fdt, nodename, "clocks", vms->clock_phandle);
 707    qemu_fdt_setprop_string(vms->fdt, nodename, "clock-names", "apb_pclk");
 708    g_free(nodename);
 709}
 710
 711static DeviceState *gpio_key_dev;
 712static void virt_powerdown_req(Notifier *n, void *opaque)
 713{
 714    /* use gpio Pin 3 for power button event */
 715    qemu_set_irq(qdev_get_gpio_in(gpio_key_dev, 0), 1);
 716}
 717
 718static Notifier virt_system_powerdown_notifier = {
 719    .notify = virt_powerdown_req
 720};
 721
 722static void create_gpio(const VirtMachineState *vms, qemu_irq *pic)
 723{
 724    char *nodename;
 725    DeviceState *pl061_dev;
 726    hwaddr base = vms->memmap[VIRT_GPIO].base;
 727    hwaddr size = vms->memmap[VIRT_GPIO].size;
 728    int irq = vms->irqmap[VIRT_GPIO];
 729    const char compat[] = "arm,pl061\0arm,primecell";
 730
 731    pl061_dev = sysbus_create_simple("pl061", base, pic[irq]);
 732
 733    uint32_t phandle = qemu_fdt_alloc_phandle(vms->fdt);
 734    nodename = g_strdup_printf("/pl061@%" PRIx64, base);
 735    qemu_fdt_add_subnode(vms->fdt, nodename);
 736    qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
 737                                 2, base, 2, size);
 738    qemu_fdt_setprop(vms->fdt, nodename, "compatible", compat, sizeof(compat));
 739    qemu_fdt_setprop_cell(vms->fdt, nodename, "#gpio-cells", 2);
 740    qemu_fdt_setprop(vms->fdt, nodename, "gpio-controller", NULL, 0);
 741    qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupts",
 742                           GIC_FDT_IRQ_TYPE_SPI, irq,
 743                           GIC_FDT_IRQ_FLAGS_LEVEL_HI);
 744    qemu_fdt_setprop_cell(vms->fdt, nodename, "clocks", vms->clock_phandle);
 745    qemu_fdt_setprop_string(vms->fdt, nodename, "clock-names", "apb_pclk");
 746    qemu_fdt_setprop_cell(vms->fdt, nodename, "phandle", phandle);
 747
 748    gpio_key_dev = sysbus_create_simple("gpio-key", -1,
 749                                        qdev_get_gpio_in(pl061_dev, 3));
 750    qemu_fdt_add_subnode(vms->fdt, "/gpio-keys");
 751    qemu_fdt_setprop_string(vms->fdt, "/gpio-keys", "compatible", "gpio-keys");
 752    qemu_fdt_setprop_cell(vms->fdt, "/gpio-keys", "#size-cells", 0);
 753    qemu_fdt_setprop_cell(vms->fdt, "/gpio-keys", "#address-cells", 1);
 754
 755    qemu_fdt_add_subnode(vms->fdt, "/gpio-keys/poweroff");
 756    qemu_fdt_setprop_string(vms->fdt, "/gpio-keys/poweroff",
 757                            "label", "GPIO Key Poweroff");
 758    qemu_fdt_setprop_cell(vms->fdt, "/gpio-keys/poweroff", "linux,code",
 759                          KEY_POWER);
 760    qemu_fdt_setprop_cells(vms->fdt, "/gpio-keys/poweroff",
 761                           "gpios", phandle, 3, 0);
 762
 763    /* connect powerdown request */
 764    qemu_register_powerdown_notifier(&virt_system_powerdown_notifier);
 765
 766    g_free(nodename);
 767}
 768
 769static void create_virtio_devices(const VirtMachineState *vms, qemu_irq *pic)
 770{
 771    int i;
 772    hwaddr size = vms->memmap[VIRT_MMIO].size;
 773
 774    /* We create the transports in forwards order. Since qbus_realize()
 775     * prepends (not appends) new child buses, the incrementing loop below will
 776     * create a list of virtio-mmio buses with decreasing base addresses.
 777     *
 778     * When a -device option is processed from the command line,
 779     * qbus_find_recursive() picks the next free virtio-mmio bus in forwards
 780     * order. The upshot is that -device options in increasing command line
 781     * order are mapped to virtio-mmio buses with decreasing base addresses.
 782     *
 783     * When this code was originally written, that arrangement ensured that the
 784     * guest Linux kernel would give the lowest "name" (/dev/vda, eth0, etc) to
 785     * the first -device on the command line. (The end-to-end order is a
 786     * function of this loop, qbus_realize(), qbus_find_recursive(), and the
 787     * guest kernel's name-to-address assignment strategy.)
 788     *
 789     * Meanwhile, the kernel's traversal seems to have been reversed; see eg.
 790     * the message, if not necessarily the code, of commit 70161ff336.
 791     * Therefore the loop now establishes the inverse of the original intent.
 792     *
 793     * Unfortunately, we can't counteract the kernel change by reversing the
 794     * loop; it would break existing command lines.
 795     *
 796     * In any case, the kernel makes no guarantee about the stability of
 797     * enumeration order of virtio devices (as demonstrated by it changing
 798     * between kernel versions). For reliable and stable identification
 799     * of disks users must use UUIDs or similar mechanisms.
 800     */
 801    for (i = 0; i < NUM_VIRTIO_TRANSPORTS; i++) {
 802        int irq = vms->irqmap[VIRT_MMIO] + i;
 803        hwaddr base = vms->memmap[VIRT_MMIO].base + i * size;
 804
 805        sysbus_create_simple("virtio-mmio", base, pic[irq]);
 806    }
 807
 808    /* We add dtb nodes in reverse order so that they appear in the finished
 809     * device tree lowest address first.
 810     *
 811     * Note that this mapping is independent of the loop above. The previous
 812     * loop influences virtio device to virtio transport assignment, whereas
 813     * this loop controls how virtio transports are laid out in the dtb.
 814     */
 815    for (i = NUM_VIRTIO_TRANSPORTS - 1; i >= 0; i--) {
 816        char *nodename;
 817        int irq = vms->irqmap[VIRT_MMIO] + i;
 818        hwaddr base = vms->memmap[VIRT_MMIO].base + i * size;
 819
 820        nodename = g_strdup_printf("/virtio_mmio@%" PRIx64, base);
 821        qemu_fdt_add_subnode(vms->fdt, nodename);
 822        qemu_fdt_setprop_string(vms->fdt, nodename,
 823                                "compatible", "virtio,mmio");
 824        qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
 825                                     2, base, 2, size);
 826        qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupts",
 827                               GIC_FDT_IRQ_TYPE_SPI, irq,
 828                               GIC_FDT_IRQ_FLAGS_EDGE_LO_HI);
 829        qemu_fdt_setprop(vms->fdt, nodename, "dma-coherent", NULL, 0);
 830        g_free(nodename);
 831    }
 832}
 833
 834static void create_one_flash(const char *name, hwaddr flashbase,
 835                             hwaddr flashsize, const char *file,
 836                             MemoryRegion *sysmem)
 837{
 838    /* Create and map a single flash device. We use the same
 839     * parameters as the flash devices on the Versatile Express board.
 840     */
 841    DriveInfo *dinfo = drive_get_next(IF_PFLASH);
 842    DeviceState *dev = qdev_create(NULL, "cfi.pflash01");
 843    SysBusDevice *sbd = SYS_BUS_DEVICE(dev);
 844    const uint64_t sectorlength = 256 * 1024;
 845
 846    if (dinfo) {
 847        qdev_prop_set_drive(dev, "drive", blk_by_legacy_dinfo(dinfo),
 848                            &error_abort);
 849    }
 850
 851    qdev_prop_set_uint32(dev, "num-blocks", flashsize / sectorlength);
 852    qdev_prop_set_uint64(dev, "sector-length", sectorlength);
 853    qdev_prop_set_uint8(dev, "width", 4);
 854    qdev_prop_set_uint8(dev, "device-width", 2);
 855    qdev_prop_set_bit(dev, "big-endian", false);
 856    qdev_prop_set_uint16(dev, "id0", 0x89);
 857    qdev_prop_set_uint16(dev, "id1", 0x18);
 858    qdev_prop_set_uint16(dev, "id2", 0x00);
 859    qdev_prop_set_uint16(dev, "id3", 0x00);
 860    qdev_prop_set_string(dev, "name", name);
 861    qdev_init_nofail(dev);
 862
 863    memory_region_add_subregion(sysmem, flashbase,
 864                                sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 0));
 865
 866    if (file) {
 867        char *fn;
 868        int image_size;
 869
 870        if (drive_get(IF_PFLASH, 0, 0)) {
 871            error_report("The contents of the first flash device may be "
 872                         "specified with -bios or with -drive if=pflash... "
 873                         "but you cannot use both options at once");
 874            exit(1);
 875        }
 876        fn = qemu_find_file(QEMU_FILE_TYPE_BIOS, file);
 877        if (!fn) {
 878            error_report("Could not find ROM image '%s'", file);
 879            exit(1);
 880        }
 881        image_size = load_image_mr(fn, sysbus_mmio_get_region(sbd, 0));
 882        g_free(fn);
 883        if (image_size < 0) {
 884            error_report("Could not load ROM image '%s'", file);
 885            exit(1);
 886        }
 887    }
 888}
 889
 890static void create_flash(const VirtMachineState *vms,
 891                         MemoryRegion *sysmem,
 892                         MemoryRegion *secure_sysmem)
 893{
 894    /* Create two flash devices to fill the VIRT_FLASH space in the memmap.
 895     * Any file passed via -bios goes in the first of these.
 896     * sysmem is the system memory space. secure_sysmem is the secure view
 897     * of the system, and the first flash device should be made visible only
 898     * there. The second flash device is visible to both secure and nonsecure.
 899     * If sysmem == secure_sysmem this means there is no separate Secure
 900     * address space and both flash devices are generally visible.
 901     */
 902    hwaddr flashsize = vms->memmap[VIRT_FLASH].size / 2;
 903    hwaddr flashbase = vms->memmap[VIRT_FLASH].base;
 904    char *nodename;
 905
 906    create_one_flash("virt.flash0", flashbase, flashsize,
 907                     bios_name, secure_sysmem);
 908    create_one_flash("virt.flash1", flashbase + flashsize, flashsize,
 909                     NULL, sysmem);
 910
 911    if (sysmem == secure_sysmem) {
 912        /* Report both flash devices as a single node in the DT */
 913        nodename = g_strdup_printf("/flash@%" PRIx64, flashbase);
 914        qemu_fdt_add_subnode(vms->fdt, nodename);
 915        qemu_fdt_setprop_string(vms->fdt, nodename, "compatible", "cfi-flash");
 916        qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
 917                                     2, flashbase, 2, flashsize,
 918                                     2, flashbase + flashsize, 2, flashsize);
 919        qemu_fdt_setprop_cell(vms->fdt, nodename, "bank-width", 4);
 920        g_free(nodename);
 921    } else {
 922        /* Report the devices as separate nodes so we can mark one as
 923         * only visible to the secure world.
 924         */
 925        nodename = g_strdup_printf("/secflash@%" PRIx64, flashbase);
 926        qemu_fdt_add_subnode(vms->fdt, nodename);
 927        qemu_fdt_setprop_string(vms->fdt, nodename, "compatible", "cfi-flash");
 928        qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
 929                                     2, flashbase, 2, flashsize);
 930        qemu_fdt_setprop_cell(vms->fdt, nodename, "bank-width", 4);
 931        qemu_fdt_setprop_string(vms->fdt, nodename, "status", "disabled");
 932        qemu_fdt_setprop_string(vms->fdt, nodename, "secure-status", "okay");
 933        g_free(nodename);
 934
 935        nodename = g_strdup_printf("/flash@%" PRIx64, flashbase);
 936        qemu_fdt_add_subnode(vms->fdt, nodename);
 937        qemu_fdt_setprop_string(vms->fdt, nodename, "compatible", "cfi-flash");
 938        qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
 939                                     2, flashbase + flashsize, 2, flashsize);
 940        qemu_fdt_setprop_cell(vms->fdt, nodename, "bank-width", 4);
 941        g_free(nodename);
 942    }
 943}
 944
 945static FWCfgState *create_fw_cfg(const VirtMachineState *vms, AddressSpace *as)
 946{
 947    hwaddr base = vms->memmap[VIRT_FW_CFG].base;
 948    hwaddr size = vms->memmap[VIRT_FW_CFG].size;
 949    FWCfgState *fw_cfg;
 950    char *nodename;
 951
 952    fw_cfg = fw_cfg_init_mem_wide(base + 8, base, 8, base + 16, as);
 953    fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, (uint16_t)smp_cpus);
 954
 955    nodename = g_strdup_printf("/fw-cfg@%" PRIx64, base);
 956    qemu_fdt_add_subnode(vms->fdt, nodename);
 957    qemu_fdt_setprop_string(vms->fdt, nodename,
 958                            "compatible", "qemu,fw-cfg-mmio");
 959    qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
 960                                 2, base, 2, size);
 961    qemu_fdt_setprop(vms->fdt, nodename, "dma-coherent", NULL, 0);
 962    g_free(nodename);
 963    return fw_cfg;
 964}
 965
 966static void create_pcie_irq_map(const VirtMachineState *vms,
 967                                uint32_t gic_phandle,
 968                                int first_irq, const char *nodename)
 969{
 970    int devfn, pin;
 971    uint32_t full_irq_map[4 * 4 * 10] = { 0 };
 972    uint32_t *irq_map = full_irq_map;
 973
 974    for (devfn = 0; devfn <= 0x18; devfn += 0x8) {
 975        for (pin = 0; pin < 4; pin++) {
 976            int irq_type = GIC_FDT_IRQ_TYPE_SPI;
 977            int irq_nr = first_irq + ((pin + PCI_SLOT(devfn)) % PCI_NUM_PINS);
 978            int irq_level = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
 979            int i;
 980
 981            uint32_t map[] = {
 982                devfn << 8, 0, 0,                           /* devfn */
 983                pin + 1,                                    /* PCI pin */
 984                gic_phandle, 0, 0, irq_type, irq_nr, irq_level }; /* GIC irq */
 985
 986            /* Convert map to big endian */
 987            for (i = 0; i < 10; i++) {
 988                irq_map[i] = cpu_to_be32(map[i]);
 989            }
 990            irq_map += 10;
 991        }
 992    }
 993
 994    qemu_fdt_setprop(vms->fdt, nodename, "interrupt-map",
 995                     full_irq_map, sizeof(full_irq_map));
 996
 997    qemu_fdt_setprop_cells(vms->fdt, nodename, "interrupt-map-mask",
 998                           0x1800, 0, 0, /* devfn (PCI_SLOT(3)) */
 999                           0x7           /* PCI irq */);
1000}
1001
1002static void create_pcie(const VirtMachineState *vms, qemu_irq *pic)
1003{
1004    hwaddr base_mmio = vms->memmap[VIRT_PCIE_MMIO].base;
1005    hwaddr size_mmio = vms->memmap[VIRT_PCIE_MMIO].size;
1006    hwaddr base_mmio_high = vms->memmap[VIRT_PCIE_MMIO_HIGH].base;
1007    hwaddr size_mmio_high = vms->memmap[VIRT_PCIE_MMIO_HIGH].size;
1008    hwaddr base_pio = vms->memmap[VIRT_PCIE_PIO].base;
1009    hwaddr size_pio = vms->memmap[VIRT_PCIE_PIO].size;
1010    hwaddr base_ecam = vms->memmap[VIRT_PCIE_ECAM].base;
1011    hwaddr size_ecam = vms->memmap[VIRT_PCIE_ECAM].size;
1012    hwaddr base = base_mmio;
1013    int nr_pcie_buses = size_ecam / PCIE_MMCFG_SIZE_MIN;
1014    int irq = vms->irqmap[VIRT_PCIE];
1015    MemoryRegion *mmio_alias;
1016    MemoryRegion *mmio_reg;
1017    MemoryRegion *ecam_alias;
1018    MemoryRegion *ecam_reg;
1019    DeviceState *dev;
1020    char *nodename;
1021    int i;
1022    PCIHostState *pci;
1023
1024    dev = qdev_create(NULL, TYPE_GPEX_HOST);
1025    qdev_init_nofail(dev);
1026
1027    /* Map only the first size_ecam bytes of ECAM space */
1028    ecam_alias = g_new0(MemoryRegion, 1);
1029    ecam_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 0);
1030    memory_region_init_alias(ecam_alias, OBJECT(dev), "pcie-ecam",
1031                             ecam_reg, 0, size_ecam);
1032    memory_region_add_subregion(get_system_memory(), base_ecam, ecam_alias);
1033
1034    /* Map the MMIO window into system address space so as to expose
1035     * the section of PCI MMIO space which starts at the same base address
1036     * (ie 1:1 mapping for that part of PCI MMIO space visible through
1037     * the window).
1038     */
1039    mmio_alias = g_new0(MemoryRegion, 1);
1040    mmio_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 1);
1041    memory_region_init_alias(mmio_alias, OBJECT(dev), "pcie-mmio",
1042                             mmio_reg, base_mmio, size_mmio);
1043    memory_region_add_subregion(get_system_memory(), base_mmio, mmio_alias);
1044
1045    if (vms->highmem) {
1046        /* Map high MMIO space */
1047        MemoryRegion *high_mmio_alias = g_new0(MemoryRegion, 1);
1048
1049        memory_region_init_alias(high_mmio_alias, OBJECT(dev), "pcie-mmio-high",
1050                                 mmio_reg, base_mmio_high, size_mmio_high);
1051        memory_region_add_subregion(get_system_memory(), base_mmio_high,
1052                                    high_mmio_alias);
1053    }
1054
1055    /* Map IO port space */
1056    sysbus_mmio_map(SYS_BUS_DEVICE(dev), 2, base_pio);
1057
1058    for (i = 0; i < GPEX_NUM_IRQS; i++) {
1059        sysbus_connect_irq(SYS_BUS_DEVICE(dev), i, pic[irq + i]);
1060        gpex_set_irq_num(GPEX_HOST(dev), i, irq + i);
1061    }
1062
1063    pci = PCI_HOST_BRIDGE(dev);
1064    if (pci->bus) {
1065        for (i = 0; i < nb_nics; i++) {
1066            NICInfo *nd = &nd_table[i];
1067
1068            if (!nd->model) {
1069                nd->model = g_strdup("virtio");
1070            }
1071
1072            pci_nic_init_nofail(nd, pci->bus, nd->model, NULL);
1073        }
1074    }
1075
1076    nodename = g_strdup_printf("/pcie@%" PRIx64, base);
1077    qemu_fdt_add_subnode(vms->fdt, nodename);
1078    qemu_fdt_setprop_string(vms->fdt, nodename,
1079                            "compatible", "pci-host-ecam-generic");
1080    qemu_fdt_setprop_string(vms->fdt, nodename, "device_type", "pci");
1081    qemu_fdt_setprop_cell(vms->fdt, nodename, "#address-cells", 3);
1082    qemu_fdt_setprop_cell(vms->fdt, nodename, "#size-cells", 2);
1083    qemu_fdt_setprop_cells(vms->fdt, nodename, "bus-range", 0,
1084                           nr_pcie_buses - 1);
1085    qemu_fdt_setprop(vms->fdt, nodename, "dma-coherent", NULL, 0);
1086
1087    if (vms->msi_phandle) {
1088        qemu_fdt_setprop_cells(vms->fdt, nodename, "msi-parent",
1089                               vms->msi_phandle);
1090    }
1091
1092    qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg",
1093                                 2, base_ecam, 2, size_ecam);
1094
1095    if (vms->highmem) {
1096        qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "ranges",
1097                                     1, FDT_PCI_RANGE_IOPORT, 2, 0,
1098                                     2, base_pio, 2, size_pio,
1099                                     1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
1100                                     2, base_mmio, 2, size_mmio,
1101                                     1, FDT_PCI_RANGE_MMIO_64BIT,
1102                                     2, base_mmio_high,
1103                                     2, base_mmio_high, 2, size_mmio_high);
1104    } else {
1105        qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "ranges",
1106                                     1, FDT_PCI_RANGE_IOPORT, 2, 0,
1107                                     2, base_pio, 2, size_pio,
1108                                     1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
1109                                     2, base_mmio, 2, size_mmio);
1110    }
1111
1112    qemu_fdt_setprop_cell(vms->fdt, nodename, "#interrupt-cells", 1);
1113    create_pcie_irq_map(vms, vms->gic_phandle, irq, nodename);
1114
1115    g_free(nodename);
1116}
1117
1118static void create_platform_bus(VirtMachineState *vms, qemu_irq *pic)
1119{
1120    DeviceState *dev;
1121    SysBusDevice *s;
1122    int i;
1123    ARMPlatformBusFDTParams *fdt_params = g_new(ARMPlatformBusFDTParams, 1);
1124    MemoryRegion *sysmem = get_system_memory();
1125
1126    platform_bus_params.platform_bus_base = vms->memmap[VIRT_PLATFORM_BUS].base;
1127    platform_bus_params.platform_bus_size = vms->memmap[VIRT_PLATFORM_BUS].size;
1128    platform_bus_params.platform_bus_first_irq = vms->irqmap[VIRT_PLATFORM_BUS];
1129    platform_bus_params.platform_bus_num_irqs = PLATFORM_BUS_NUM_IRQS;
1130
1131    fdt_params->system_params = &platform_bus_params;
1132    fdt_params->binfo = &vms->bootinfo;
1133    fdt_params->intc = "/intc";
1134    /*
1135     * register a machine init done notifier that creates the device tree
1136     * nodes of the platform bus and its children dynamic sysbus devices
1137     */
1138    arm_register_platform_bus_fdt_creator(fdt_params);
1139
1140    dev = qdev_create(NULL, TYPE_PLATFORM_BUS_DEVICE);
1141    dev->id = TYPE_PLATFORM_BUS_DEVICE;
1142    qdev_prop_set_uint32(dev, "num_irqs",
1143        platform_bus_params.platform_bus_num_irqs);
1144    qdev_prop_set_uint32(dev, "mmio_size",
1145        platform_bus_params.platform_bus_size);
1146    qdev_init_nofail(dev);
1147    s = SYS_BUS_DEVICE(dev);
1148
1149    for (i = 0; i < platform_bus_params.platform_bus_num_irqs; i++) {
1150        int irqn = platform_bus_params.platform_bus_first_irq + i;
1151        sysbus_connect_irq(s, i, pic[irqn]);
1152    }
1153
1154    memory_region_add_subregion(sysmem,
1155                                platform_bus_params.platform_bus_base,
1156                                sysbus_mmio_get_region(s, 0));
1157}
1158
1159static void create_secure_ram(VirtMachineState *vms,
1160                              MemoryRegion *secure_sysmem)
1161{
1162    MemoryRegion *secram = g_new(MemoryRegion, 1);
1163    char *nodename;
1164    hwaddr base = vms->memmap[VIRT_SECURE_MEM].base;
1165    hwaddr size = vms->memmap[VIRT_SECURE_MEM].size;
1166
1167    memory_region_init_ram(secram, NULL, "virt.secure-ram", size,
1168                           &error_fatal);
1169    memory_region_add_subregion(secure_sysmem, base, secram);
1170
1171    nodename = g_strdup_printf("/secram@%" PRIx64, base);
1172    qemu_fdt_add_subnode(vms->fdt, nodename);
1173    qemu_fdt_setprop_string(vms->fdt, nodename, "device_type", "memory");
1174    qemu_fdt_setprop_sized_cells(vms->fdt, nodename, "reg", 2, base, 2, size);
1175    qemu_fdt_setprop_string(vms->fdt, nodename, "status", "disabled");
1176    qemu_fdt_setprop_string(vms->fdt, nodename, "secure-status", "okay");
1177
1178    g_free(nodename);
1179}
1180
1181static void *machvirt_dtb(const struct arm_boot_info *binfo, int *fdt_size)
1182{
1183    const VirtMachineState *board = container_of(binfo, VirtMachineState,
1184                                                 bootinfo);
1185
1186    *fdt_size = board->fdt_size;
1187    return board->fdt;
1188}
1189
1190static void virt_build_smbios(VirtMachineState *vms)
1191{
1192    uint8_t *smbios_tables, *smbios_anchor;
1193    size_t smbios_tables_len, smbios_anchor_len;
1194    const char *product = "QEMU Virtual Machine";
1195
1196    if (!vms->fw_cfg) {
1197        return;
1198    }
1199
1200    if (kvm_enabled()) {
1201        product = "KVM Virtual Machine";
1202    }
1203
1204    smbios_set_defaults("QEMU", product,
1205                        "1.0", false, true, SMBIOS_ENTRY_POINT_30);
1206
1207    smbios_get_tables(NULL, 0, &smbios_tables, &smbios_tables_len,
1208                      &smbios_anchor, &smbios_anchor_len);
1209
1210    if (smbios_anchor) {
1211        fw_cfg_add_file(vms->fw_cfg, "etc/smbios/smbios-tables",
1212                        smbios_tables, smbios_tables_len);
1213        fw_cfg_add_file(vms->fw_cfg, "etc/smbios/smbios-anchor",
1214                        smbios_anchor, smbios_anchor_len);
1215    }
1216}
1217
1218static
1219void virt_machine_done(Notifier *notifier, void *data)
1220{
1221    VirtMachineState *vms = container_of(notifier, VirtMachineState,
1222                                         machine_done);
1223
1224    virt_acpi_setup(vms);
1225    virt_build_smbios(vms);
1226}
1227
1228static uint64_t virt_cpu_mp_affinity(VirtMachineState *vms, int idx)
1229{
1230    uint8_t clustersz = ARM_DEFAULT_CPUS_PER_CLUSTER;
1231    VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
1232
1233    if (!vmc->disallow_affinity_adjustment) {
1234        /* Adjust MPIDR like 64-bit KVM hosts, which incorporate the
1235         * GIC's target-list limitations. 32-bit KVM hosts currently
1236         * always create clusters of 4 CPUs, but that is expected to
1237         * change when they gain support for gicv3. When KVM is enabled
1238         * it will override the changes we make here, therefore our
1239         * purposes are to make TCG consistent (with 64-bit KVM hosts)
1240         * and to improve SGI efficiency.
1241         */
1242        if (vms->gic_version == 3) {
1243            clustersz = GICV3_TARGETLIST_BITS;
1244        } else {
1245            clustersz = GIC_TARGETLIST_BITS;
1246        }
1247    }
1248    return arm_cpu_mp_affinity(idx, clustersz);
1249}
1250
1251static void machvirt_init(MachineState *machine)
1252{
1253    VirtMachineState *vms = VIRT_MACHINE(machine);
1254    VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(machine);
1255    MachineClass *mc = MACHINE_GET_CLASS(machine);
1256    const CPUArchIdList *possible_cpus;
1257    qemu_irq pic[NUM_IRQS];
1258    MemoryRegion *sysmem = get_system_memory();
1259    MemoryRegion *secure_sysmem = NULL;
1260    int n, virt_max_cpus;
1261    MemoryRegion *ram = g_new(MemoryRegion, 1);
1262    bool firmware_loaded = bios_name || drive_get(IF_PFLASH, 0, 0);
1263
1264    /* We can probe only here because during property set
1265     * KVM is not available yet
1266     */
1267    if (!vms->gic_version) {
1268        if (!kvm_enabled()) {
1269            error_report("gic-version=host requires KVM");
1270            exit(1);
1271        }
1272
1273        vms->gic_version = kvm_arm_vgic_probe();
1274        if (!vms->gic_version) {
1275            error_report("Unable to determine GIC version supported by host");
1276            exit(1);
1277        }
1278    }
1279
1280    if (!cpu_type_valid(machine->cpu_type)) {
1281        error_report("mach-virt: CPU type %s not supported", machine->cpu_type);
1282        exit(1);
1283    }
1284
1285    /* If we have an EL3 boot ROM then the assumption is that it will
1286     * implement PSCI itself, so disable QEMU's internal implementation
1287     * so it doesn't get in the way. Instead of starting secondary
1288     * CPUs in PSCI powerdown state we will start them all running and
1289     * let the boot ROM sort them out.
1290     * The usual case is that we do use QEMU's PSCI implementation;
1291     * if the guest has EL2 then we will use SMC as the conduit,
1292     * and otherwise we will use HVC (for backwards compatibility and
1293     * because if we're using KVM then we must use HVC).
1294     */
1295    if (vms->secure && firmware_loaded) {
1296        vms->psci_conduit = QEMU_PSCI_CONDUIT_DISABLED;
1297    } else if (vms->virt) {
1298        vms->psci_conduit = QEMU_PSCI_CONDUIT_SMC;
1299    } else {
1300        vms->psci_conduit = QEMU_PSCI_CONDUIT_HVC;
1301    }
1302
1303    /* The maximum number of CPUs depends on the GIC version, or on how
1304     * many redistributors we can fit into the memory map.
1305     */
1306    if (vms->gic_version == 3) {
1307        virt_max_cpus = vms->memmap[VIRT_GIC_REDIST].size / 0x20000;
1308    } else {
1309        virt_max_cpus = GIC_NCPU;
1310    }
1311
1312    if (max_cpus > virt_max_cpus) {
1313        error_report("Number of SMP CPUs requested (%d) exceeds max CPUs "
1314                     "supported by machine 'mach-virt' (%d)",
1315                     max_cpus, virt_max_cpus);
1316        exit(1);
1317    }
1318
1319    vms->smp_cpus = smp_cpus;
1320
1321    if (machine->ram_size > vms->memmap[VIRT_MEM].size) {
1322        error_report("mach-virt: cannot model more than %dGB RAM", RAMLIMIT_GB);
1323        exit(1);
1324    }
1325
1326    if (vms->virt && kvm_enabled()) {
1327        error_report("mach-virt: KVM does not support providing "
1328                     "Virtualization extensions to the guest CPU");
1329        exit(1);
1330    }
1331
1332    if (vms->secure) {
1333        if (kvm_enabled()) {
1334            error_report("mach-virt: KVM does not support Security extensions");
1335            exit(1);
1336        }
1337
1338        /* The Secure view of the world is the same as the NonSecure,
1339         * but with a few extra devices. Create it as a container region
1340         * containing the system memory at low priority; any secure-only
1341         * devices go in at higher priority and take precedence.
1342         */
1343        secure_sysmem = g_new(MemoryRegion, 1);
1344        memory_region_init(secure_sysmem, OBJECT(machine), "secure-memory",
1345                           UINT64_MAX);
1346        memory_region_add_subregion_overlap(secure_sysmem, 0, sysmem, -1);
1347    }
1348
1349    create_fdt(vms);
1350
1351    possible_cpus = mc->possible_cpu_arch_ids(machine);
1352    for (n = 0; n < possible_cpus->len; n++) {
1353        Object *cpuobj;
1354        CPUState *cs;
1355
1356        if (n >= smp_cpus) {
1357            break;
1358        }
1359
1360        cpuobj = object_new(machine->cpu_type);
1361        object_property_set_int(cpuobj, possible_cpus->cpus[n].arch_id,
1362                                "mp-affinity", NULL);
1363
1364        cs = CPU(cpuobj);
1365        cs->cpu_index = n;
1366
1367        numa_cpu_pre_plug(&possible_cpus->cpus[cs->cpu_index], DEVICE(cpuobj),
1368                          &error_fatal);
1369
1370        if (!vms->secure) {
1371            object_property_set_bool(cpuobj, false, "has_el3", NULL);
1372        }
1373
1374        if (!vms->virt && object_property_find(cpuobj, "has_el2", NULL)) {
1375            object_property_set_bool(cpuobj, false, "has_el2", NULL);
1376        }
1377
1378        if (vms->psci_conduit != QEMU_PSCI_CONDUIT_DISABLED) {
1379            object_property_set_int(cpuobj, vms->psci_conduit,
1380                                    "psci-conduit", NULL);
1381
1382            /* Secondary CPUs start in PSCI powered-down state */
1383            if (n > 0) {
1384                object_property_set_bool(cpuobj, true,
1385                                         "start-powered-off", NULL);
1386            }
1387        }
1388
1389        if (vmc->no_pmu && object_property_find(cpuobj, "pmu", NULL)) {
1390            object_property_set_bool(cpuobj, false, "pmu", NULL);
1391        }
1392
1393        if (object_property_find(cpuobj, "reset-cbar", NULL)) {
1394            object_property_set_int(cpuobj, vms->memmap[VIRT_CPUPERIPHS].base,
1395                                    "reset-cbar", &error_abort);
1396        }
1397
1398        object_property_set_link(cpuobj, OBJECT(sysmem), "memory",
1399                                 &error_abort);
1400        if (vms->secure) {
1401            object_property_set_link(cpuobj, OBJECT(secure_sysmem),
1402                                     "secure-memory", &error_abort);
1403        }
1404
1405        object_property_set_bool(cpuobj, true, "realized", NULL);
1406        object_unref(cpuobj);
1407    }
1408    fdt_add_timer_nodes(vms);
1409    fdt_add_cpu_nodes(vms);
1410    fdt_add_psci_node(vms);
1411
1412    memory_region_allocate_system_memory(ram, NULL, "mach-virt.ram",
1413                                         machine->ram_size);
1414    memory_region_add_subregion(sysmem, vms->memmap[VIRT_MEM].base, ram);
1415
1416    create_flash(vms, sysmem, secure_sysmem ? secure_sysmem : sysmem);
1417
1418    create_gic(vms, pic);
1419
1420    fdt_add_pmu_nodes(vms);
1421
1422    create_uart(vms, pic, VIRT_UART, sysmem, serial_hds[0]);
1423
1424    if (vms->secure) {
1425        create_secure_ram(vms, secure_sysmem);
1426        create_uart(vms, pic, VIRT_SECURE_UART, secure_sysmem, serial_hds[1]);
1427    }
1428
1429    create_rtc(vms, pic);
1430
1431    create_pcie(vms, pic);
1432
1433    create_gpio(vms, pic);
1434
1435    /* Create mmio transports, so the user can create virtio backends
1436     * (which will be automatically plugged in to the transports). If
1437     * no backend is created the transport will just sit harmlessly idle.
1438     */
1439    create_virtio_devices(vms, pic);
1440
1441    vms->fw_cfg = create_fw_cfg(vms, &address_space_memory);
1442    rom_set_fw(vms->fw_cfg);
1443
1444    vms->machine_done.notify = virt_machine_done;
1445    qemu_add_machine_init_done_notifier(&vms->machine_done);
1446
1447    vms->bootinfo.ram_size = machine->ram_size;
1448    vms->bootinfo.kernel_filename = machine->kernel_filename;
1449    vms->bootinfo.kernel_cmdline = machine->kernel_cmdline;
1450    vms->bootinfo.initrd_filename = machine->initrd_filename;
1451    vms->bootinfo.nb_cpus = smp_cpus;
1452    vms->bootinfo.board_id = -1;
1453    vms->bootinfo.loader_start = vms->memmap[VIRT_MEM].base;
1454    vms->bootinfo.get_dtb = machvirt_dtb;
1455    vms->bootinfo.firmware_loaded = firmware_loaded;
1456    arm_load_kernel(ARM_CPU(first_cpu), &vms->bootinfo);
1457
1458    /*
1459     * arm_load_kernel machine init done notifier registration must
1460     * happen before the platform_bus_create call. In this latter,
1461     * another notifier is registered which adds platform bus nodes.
1462     * Notifiers are executed in registration reverse order.
1463     */
1464    create_platform_bus(vms, pic);
1465}
1466
1467static bool virt_get_secure(Object *obj, Error **errp)
1468{
1469    VirtMachineState *vms = VIRT_MACHINE(obj);
1470
1471    return vms->secure;
1472}
1473
1474static void virt_set_secure(Object *obj, bool value, Error **errp)
1475{
1476    VirtMachineState *vms = VIRT_MACHINE(obj);
1477
1478    vms->secure = value;
1479}
1480
1481static bool virt_get_virt(Object *obj, Error **errp)
1482{
1483    VirtMachineState *vms = VIRT_MACHINE(obj);
1484
1485    return vms->virt;
1486}
1487
1488static void virt_set_virt(Object *obj, bool value, Error **errp)
1489{
1490    VirtMachineState *vms = VIRT_MACHINE(obj);
1491
1492    vms->virt = value;
1493}
1494
1495static bool virt_get_highmem(Object *obj, Error **errp)
1496{
1497    VirtMachineState *vms = VIRT_MACHINE(obj);
1498
1499    return vms->highmem;
1500}
1501
1502static void virt_set_highmem(Object *obj, bool value, Error **errp)
1503{
1504    VirtMachineState *vms = VIRT_MACHINE(obj);
1505
1506    vms->highmem = value;
1507}
1508
1509static bool virt_get_its(Object *obj, Error **errp)
1510{
1511    VirtMachineState *vms = VIRT_MACHINE(obj);
1512
1513    return vms->its;
1514}
1515
1516static void virt_set_its(Object *obj, bool value, Error **errp)
1517{
1518    VirtMachineState *vms = VIRT_MACHINE(obj);
1519
1520    vms->its = value;
1521}
1522
1523static char *virt_get_gic_version(Object *obj, Error **errp)
1524{
1525    VirtMachineState *vms = VIRT_MACHINE(obj);
1526    const char *val = vms->gic_version == 3 ? "3" : "2";
1527
1528    return g_strdup(val);
1529}
1530
1531static void virt_set_gic_version(Object *obj, const char *value, Error **errp)
1532{
1533    VirtMachineState *vms = VIRT_MACHINE(obj);
1534
1535    if (!strcmp(value, "3")) {
1536        vms->gic_version = 3;
1537    } else if (!strcmp(value, "2")) {
1538        vms->gic_version = 2;
1539    } else if (!strcmp(value, "host")) {
1540        vms->gic_version = 0; /* Will probe later */
1541    } else {
1542        error_setg(errp, "Invalid gic-version value");
1543        error_append_hint(errp, "Valid values are 3, 2, host.\n");
1544    }
1545}
1546
1547static CpuInstanceProperties
1548virt_cpu_index_to_props(MachineState *ms, unsigned cpu_index)
1549{
1550    MachineClass *mc = MACHINE_GET_CLASS(ms);
1551    const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(ms);
1552
1553    assert(cpu_index < possible_cpus->len);
1554    return possible_cpus->cpus[cpu_index].props;
1555}
1556
1557static int64_t virt_get_default_cpu_node_id(const MachineState *ms, int idx)
1558{
1559    return idx % nb_numa_nodes;
1560}
1561
1562static const CPUArchIdList *virt_possible_cpu_arch_ids(MachineState *ms)
1563{
1564    int n;
1565    VirtMachineState *vms = VIRT_MACHINE(ms);
1566
1567    if (ms->possible_cpus) {
1568        assert(ms->possible_cpus->len == max_cpus);
1569        return ms->possible_cpus;
1570    }
1571
1572    ms->possible_cpus = g_malloc0(sizeof(CPUArchIdList) +
1573                                  sizeof(CPUArchId) * max_cpus);
1574    ms->possible_cpus->len = max_cpus;
1575    for (n = 0; n < ms->possible_cpus->len; n++) {
1576        ms->possible_cpus->cpus[n].arch_id =
1577            virt_cpu_mp_affinity(vms, n);
1578        ms->possible_cpus->cpus[n].props.has_thread_id = true;
1579        ms->possible_cpus->cpus[n].props.thread_id = n;
1580    }
1581    return ms->possible_cpus;
1582}
1583
1584static void virt_machine_class_init(ObjectClass *oc, void *data)
1585{
1586    MachineClass *mc = MACHINE_CLASS(oc);
1587
1588    mc->init = machvirt_init;
1589    /* Start max_cpus at the maximum QEMU supports. We'll further restrict
1590     * it later in machvirt_init, where we have more information about the
1591     * configuration of the particular instance.
1592     */
1593    mc->max_cpus = 255;
1594    mc->has_dynamic_sysbus = true;
1595    mc->block_default_type = IF_VIRTIO;
1596    mc->no_cdrom = 1;
1597    mc->pci_allow_0_address = true;
1598    /* We know we will never create a pre-ARMv7 CPU which needs 1K pages */
1599    mc->minimum_page_bits = 12;
1600    mc->possible_cpu_arch_ids = virt_possible_cpu_arch_ids;
1601    mc->cpu_index_to_instance_props = virt_cpu_index_to_props;
1602    mc->default_cpu_type = ARM_CPU_TYPE_NAME("cortex-a15");
1603    mc->get_default_cpu_node_id = virt_get_default_cpu_node_id;
1604}
1605
1606static const TypeInfo virt_machine_info = {
1607    .name          = TYPE_VIRT_MACHINE,
1608    .parent        = TYPE_MACHINE,
1609    .abstract      = true,
1610    .instance_size = sizeof(VirtMachineState),
1611    .class_size    = sizeof(VirtMachineClass),
1612    .class_init    = virt_machine_class_init,
1613};
1614
1615static void machvirt_machine_init(void)
1616{
1617    type_register_static(&virt_machine_info);
1618}
1619type_init(machvirt_machine_init);
1620
1621static void virt_2_11_instance_init(Object *obj)
1622{
1623    VirtMachineState *vms = VIRT_MACHINE(obj);
1624    VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
1625
1626    /* EL3 is disabled by default on virt: this makes us consistent
1627     * between KVM and TCG for this board, and it also allows us to
1628     * boot UEFI blobs which assume no TrustZone support.
1629     */
1630    vms->secure = false;
1631    object_property_add_bool(obj, "secure", virt_get_secure,
1632                             virt_set_secure, NULL);
1633    object_property_set_description(obj, "secure",
1634                                    "Set on/off to enable/disable the ARM "
1635                                    "Security Extensions (TrustZone)",
1636                                    NULL);
1637
1638    /* EL2 is also disabled by default, for similar reasons */
1639    vms->virt = false;
1640    object_property_add_bool(obj, "virtualization", virt_get_virt,
1641                             virt_set_virt, NULL);
1642    object_property_set_description(obj, "virtualization",
1643                                    "Set on/off to enable/disable emulating a "
1644                                    "guest CPU which implements the ARM "
1645                                    "Virtualization Extensions",
1646                                    NULL);
1647
1648    /* High memory is enabled by default */
1649    vms->highmem = true;
1650    object_property_add_bool(obj, "highmem", virt_get_highmem,
1651                             virt_set_highmem, NULL);
1652    object_property_set_description(obj, "highmem",
1653                                    "Set on/off to enable/disable using "
1654                                    "physical address space above 32 bits",
1655                                    NULL);
1656    /* Default GIC type is v2 */
1657    vms->gic_version = 2;
1658    object_property_add_str(obj, "gic-version", virt_get_gic_version,
1659                        virt_set_gic_version, NULL);
1660    object_property_set_description(obj, "gic-version",
1661                                    "Set GIC version. "
1662                                    "Valid values are 2, 3 and host", NULL);
1663
1664    if (vmc->no_its) {
1665        vms->its = false;
1666    } else {
1667        /* Default allows ITS instantiation */
1668        vms->its = true;
1669        object_property_add_bool(obj, "its", virt_get_its,
1670                                 virt_set_its, NULL);
1671        object_property_set_description(obj, "its",
1672                                        "Set on/off to enable/disable "
1673                                        "ITS instantiation",
1674                                        NULL);
1675    }
1676
1677    vms->memmap = a15memmap;
1678    vms->irqmap = a15irqmap;
1679}
1680
1681static void virt_machine_2_11_options(MachineClass *mc)
1682{
1683}
1684DEFINE_VIRT_MACHINE_AS_LATEST(2, 11)
1685
1686#define VIRT_COMPAT_2_10 \
1687    HW_COMPAT_2_10
1688
1689static void virt_2_10_instance_init(Object *obj)
1690{
1691    virt_2_11_instance_init(obj);
1692}
1693
1694static void virt_machine_2_10_options(MachineClass *mc)
1695{
1696    virt_machine_2_11_options(mc);
1697    SET_MACHINE_COMPAT(mc, VIRT_COMPAT_2_10);
1698}
1699DEFINE_VIRT_MACHINE(2, 10)
1700
1701#define VIRT_COMPAT_2_9 \
1702    HW_COMPAT_2_9
1703
1704static void virt_2_9_instance_init(Object *obj)
1705{
1706    virt_2_10_instance_init(obj);
1707}
1708
1709static void virt_machine_2_9_options(MachineClass *mc)
1710{
1711    virt_machine_2_10_options(mc);
1712    SET_MACHINE_COMPAT(mc, VIRT_COMPAT_2_9);
1713}
1714DEFINE_VIRT_MACHINE(2, 9)
1715
1716#define VIRT_COMPAT_2_8 \
1717    HW_COMPAT_2_8
1718
1719static void virt_2_8_instance_init(Object *obj)
1720{
1721    virt_2_9_instance_init(obj);
1722}
1723
1724static void virt_machine_2_8_options(MachineClass *mc)
1725{
1726    VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
1727
1728    virt_machine_2_9_options(mc);
1729    SET_MACHINE_COMPAT(mc, VIRT_COMPAT_2_8);
1730    /* For 2.8 and earlier we falsely claimed in the DT that
1731     * our timers were edge-triggered, not level-triggered.
1732     */
1733    vmc->claim_edge_triggered_timers = true;
1734}
1735DEFINE_VIRT_MACHINE(2, 8)
1736
1737#define VIRT_COMPAT_2_7 \
1738    HW_COMPAT_2_7
1739
1740static void virt_2_7_instance_init(Object *obj)
1741{
1742    virt_2_8_instance_init(obj);
1743}
1744
1745static void virt_machine_2_7_options(MachineClass *mc)
1746{
1747    VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
1748
1749    virt_machine_2_8_options(mc);
1750    SET_MACHINE_COMPAT(mc, VIRT_COMPAT_2_7);
1751    /* ITS was introduced with 2.8 */
1752    vmc->no_its = true;
1753    /* Stick with 1K pages for migration compatibility */
1754    mc->minimum_page_bits = 0;
1755}
1756DEFINE_VIRT_MACHINE(2, 7)
1757
1758#define VIRT_COMPAT_2_6 \
1759    HW_COMPAT_2_6
1760
1761static void virt_2_6_instance_init(Object *obj)
1762{
1763    virt_2_7_instance_init(obj);
1764}
1765
1766static void virt_machine_2_6_options(MachineClass *mc)
1767{
1768    VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
1769
1770    virt_machine_2_7_options(mc);
1771    SET_MACHINE_COMPAT(mc, VIRT_COMPAT_2_6);
1772    vmc->disallow_affinity_adjustment = true;
1773    /* Disable PMU for 2.6 as PMU support was first introduced in 2.7 */
1774    vmc->no_pmu = true;
1775}
1776DEFINE_VIRT_MACHINE(2, 6)
1777