cachepc-qemu

Fork of AMDESE/qemu with changes for cachepc side-channel attack
git clone https://git.sinitax.com/sinitax/cachepc-qemu
Log | Files | Refs | Submodules | LICENSE | sfeed.txt

x86.c (45459B)


      1/*
      2 * Copyright (c) 2003-2004 Fabrice Bellard
      3 * Copyright (c) 2019 Red Hat, Inc.
      4 *
      5 * Permission is hereby granted, free of charge, to any person obtaining a copy
      6 * of this software and associated documentation files (the "Software"), to deal
      7 * in the Software without restriction, including without limitation the rights
      8 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
      9 * copies of the Software, and to permit persons to whom the Software is
     10 * furnished to do so, subject to the following conditions:
     11 *
     12 * The above copyright notice and this permission notice shall be included in
     13 * all copies or substantial portions of the Software.
     14 *
     15 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
     16 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
     17 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
     18 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
     19 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
     20 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
     21 * THE SOFTWARE.
     22 */
     23#include "qemu/osdep.h"
     24#include "qemu/error-report.h"
     25#include "qemu/option.h"
     26#include "qemu/cutils.h"
     27#include "qemu/units.h"
     28#include "qemu-common.h"
     29#include "qemu/datadir.h"
     30#include "qapi/error.h"
     31#include "qapi/qmp/qerror.h"
     32#include "qapi/qapi-visit-common.h"
     33#include "qapi/clone-visitor.h"
     34#include "qapi/qapi-visit-machine.h"
     35#include "qapi/visitor.h"
     36#include "sysemu/qtest.h"
     37#include "sysemu/whpx.h"
     38#include "sysemu/numa.h"
     39#include "sysemu/replay.h"
     40#include "sysemu/sysemu.h"
     41#include "sysemu/cpu-timers.h"
     42#include "trace.h"
     43
     44#include "hw/i386/x86.h"
     45#include "target/i386/cpu.h"
     46#include "hw/i386/topology.h"
     47#include "hw/i386/fw_cfg.h"
     48#include "hw/intc/i8259.h"
     49#include "hw/rtc/mc146818rtc.h"
     50#include "target/i386/sev_i386.h"
     51
     52#include "hw/acpi/cpu_hotplug.h"
     53#include "hw/irq.h"
     54#include "hw/nmi.h"
     55#include "hw/loader.h"
     56#include "multiboot.h"
     57#include "elf.h"
     58#include "standard-headers/asm-x86/bootparam.h"
     59#include CONFIG_DEVICES
     60#include "kvm/kvm_i386.h"
     61
     62/* Physical Address of PVH entry point read from kernel ELF NOTE */
     63static size_t pvh_start_addr;
     64
     65inline void init_topo_info(X86CPUTopoInfo *topo_info,
     66                           const X86MachineState *x86ms)
     67{
     68    MachineState *ms = MACHINE(x86ms);
     69
     70    topo_info->dies_per_pkg = ms->smp.dies;
     71    topo_info->cores_per_die = ms->smp.cores;
     72    topo_info->threads_per_core = ms->smp.threads;
     73}
     74
     75/*
     76 * Calculates initial APIC ID for a specific CPU index
     77 *
     78 * Currently we need to be able to calculate the APIC ID from the CPU index
     79 * alone (without requiring a CPU object), as the QEMU<->Seabios interfaces have
     80 * no concept of "CPU index", and the NUMA tables on fw_cfg need the APIC ID of
     81 * all CPUs up to max_cpus.
     82 */
     83uint32_t x86_cpu_apic_id_from_index(X86MachineState *x86ms,
     84                                    unsigned int cpu_index)
     85{
     86    X86MachineClass *x86mc = X86_MACHINE_GET_CLASS(x86ms);
     87    X86CPUTopoInfo topo_info;
     88    uint32_t correct_id;
     89    static bool warned;
     90
     91    init_topo_info(&topo_info, x86ms);
     92
     93    correct_id = x86_apicid_from_cpu_idx(&topo_info, cpu_index);
     94    if (x86mc->compat_apic_id_mode) {
     95        if (cpu_index != correct_id && !warned && !qtest_enabled()) {
     96            error_report("APIC IDs set in compatibility mode, "
     97                         "CPU topology won't match the configuration");
     98            warned = true;
     99        }
    100        return cpu_index;
    101    } else {
    102        return correct_id;
    103    }
    104}
    105
    106
    107void x86_cpu_new(X86MachineState *x86ms, int64_t apic_id, Error **errp)
    108{
    109    Object *cpu = object_new(MACHINE(x86ms)->cpu_type);
    110
    111    if (!object_property_set_uint(cpu, "apic-id", apic_id, errp)) {
    112        goto out;
    113    }
    114    qdev_realize(DEVICE(cpu), NULL, errp);
    115
    116out:
    117    object_unref(cpu);
    118}
    119
    120void x86_cpus_init(X86MachineState *x86ms, int default_cpu_version)
    121{
    122    int i;
    123    const CPUArchIdList *possible_cpus;
    124    MachineState *ms = MACHINE(x86ms);
    125    MachineClass *mc = MACHINE_GET_CLASS(x86ms);
    126
    127    x86_cpu_set_default_version(default_cpu_version);
    128
    129    /*
    130     * Calculates the limit to CPU APIC ID values
    131     *
    132     * Limit for the APIC ID value, so that all
    133     * CPU APIC IDs are < x86ms->apic_id_limit.
    134     *
    135     * This is used for FW_CFG_MAX_CPUS. See comments on fw_cfg_arch_create().
    136     */
    137    x86ms->apic_id_limit = x86_cpu_apic_id_from_index(x86ms,
    138                                                      ms->smp.max_cpus - 1) + 1;
    139    possible_cpus = mc->possible_cpu_arch_ids(ms);
    140    for (i = 0; i < ms->smp.cpus; i++) {
    141        x86_cpu_new(x86ms, possible_cpus->cpus[i].arch_id, &error_fatal);
    142    }
    143}
    144
    145void x86_rtc_set_cpus_count(ISADevice *rtc, uint16_t cpus_count)
    146{
    147    if (cpus_count > 0xff) {
    148        /*
    149         * If the number of CPUs can't be represented in 8 bits, the
    150         * BIOS must use "FW_CFG_NB_CPUS". Set RTC field to 0 just
    151         * to make old BIOSes fail more predictably.
    152         */
    153        rtc_set_memory(rtc, 0x5f, 0);
    154    } else {
    155        rtc_set_memory(rtc, 0x5f, cpus_count - 1);
    156    }
    157}
    158
    159static int x86_apic_cmp(const void *a, const void *b)
    160{
    161   CPUArchId *apic_a = (CPUArchId *)a;
    162   CPUArchId *apic_b = (CPUArchId *)b;
    163
    164   return apic_a->arch_id - apic_b->arch_id;
    165}
    166
    167/*
    168 * returns pointer to CPUArchId descriptor that matches CPU's apic_id
    169 * in ms->possible_cpus->cpus, if ms->possible_cpus->cpus has no
    170 * entry corresponding to CPU's apic_id returns NULL.
    171 */
    172CPUArchId *x86_find_cpu_slot(MachineState *ms, uint32_t id, int *idx)
    173{
    174    CPUArchId apic_id, *found_cpu;
    175
    176    apic_id.arch_id = id;
    177    found_cpu = bsearch(&apic_id, ms->possible_cpus->cpus,
    178        ms->possible_cpus->len, sizeof(*ms->possible_cpus->cpus),
    179        x86_apic_cmp);
    180    if (found_cpu && idx) {
    181        *idx = found_cpu - ms->possible_cpus->cpus;
    182    }
    183    return found_cpu;
    184}
    185
    186void x86_cpu_plug(HotplugHandler *hotplug_dev,
    187                  DeviceState *dev, Error **errp)
    188{
    189    CPUArchId *found_cpu;
    190    Error *local_err = NULL;
    191    X86CPU *cpu = X86_CPU(dev);
    192    X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
    193
    194    if (x86ms->acpi_dev) {
    195        hotplug_handler_plug(x86ms->acpi_dev, dev, &local_err);
    196        if (local_err) {
    197            goto out;
    198        }
    199    }
    200
    201    /* increment the number of CPUs */
    202    x86ms->boot_cpus++;
    203    if (x86ms->rtc) {
    204        x86_rtc_set_cpus_count(x86ms->rtc, x86ms->boot_cpus);
    205    }
    206    if (x86ms->fw_cfg) {
    207        fw_cfg_modify_i16(x86ms->fw_cfg, FW_CFG_NB_CPUS, x86ms->boot_cpus);
    208    }
    209
    210    found_cpu = x86_find_cpu_slot(MACHINE(x86ms), cpu->apic_id, NULL);
    211    found_cpu->cpu = OBJECT(dev);
    212out:
    213    error_propagate(errp, local_err);
    214}
    215
    216void x86_cpu_unplug_request_cb(HotplugHandler *hotplug_dev,
    217                               DeviceState *dev, Error **errp)
    218{
    219    int idx = -1;
    220    X86CPU *cpu = X86_CPU(dev);
    221    X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
    222
    223    if (!x86ms->acpi_dev) {
    224        error_setg(errp, "CPU hot unplug not supported without ACPI");
    225        return;
    226    }
    227
    228    x86_find_cpu_slot(MACHINE(x86ms), cpu->apic_id, &idx);
    229    assert(idx != -1);
    230    if (idx == 0) {
    231        error_setg(errp, "Boot CPU is unpluggable");
    232        return;
    233    }
    234
    235    hotplug_handler_unplug_request(x86ms->acpi_dev, dev,
    236                                   errp);
    237}
    238
    239void x86_cpu_unplug_cb(HotplugHandler *hotplug_dev,
    240                       DeviceState *dev, Error **errp)
    241{
    242    CPUArchId *found_cpu;
    243    Error *local_err = NULL;
    244    X86CPU *cpu = X86_CPU(dev);
    245    X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
    246
    247    hotplug_handler_unplug(x86ms->acpi_dev, dev, &local_err);
    248    if (local_err) {
    249        goto out;
    250    }
    251
    252    found_cpu = x86_find_cpu_slot(MACHINE(x86ms), cpu->apic_id, NULL);
    253    found_cpu->cpu = NULL;
    254    qdev_unrealize(dev);
    255
    256    /* decrement the number of CPUs */
    257    x86ms->boot_cpus--;
    258    /* Update the number of CPUs in CMOS */
    259    x86_rtc_set_cpus_count(x86ms->rtc, x86ms->boot_cpus);
    260    fw_cfg_modify_i16(x86ms->fw_cfg, FW_CFG_NB_CPUS, x86ms->boot_cpus);
    261 out:
    262    error_propagate(errp, local_err);
    263}
    264
    265void x86_cpu_pre_plug(HotplugHandler *hotplug_dev,
    266                      DeviceState *dev, Error **errp)
    267{
    268    int idx;
    269    CPUState *cs;
    270    CPUArchId *cpu_slot;
    271    X86CPUTopoIDs topo_ids;
    272    X86CPU *cpu = X86_CPU(dev);
    273    CPUX86State *env = &cpu->env;
    274    MachineState *ms = MACHINE(hotplug_dev);
    275    X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
    276    unsigned int smp_cores = ms->smp.cores;
    277    unsigned int smp_threads = ms->smp.threads;
    278    X86CPUTopoInfo topo_info;
    279
    280    if (!object_dynamic_cast(OBJECT(cpu), ms->cpu_type)) {
    281        error_setg(errp, "Invalid CPU type, expected cpu type: '%s'",
    282                   ms->cpu_type);
    283        return;
    284    }
    285
    286    if (x86ms->acpi_dev) {
    287        Error *local_err = NULL;
    288
    289        hotplug_handler_pre_plug(HOTPLUG_HANDLER(x86ms->acpi_dev), dev,
    290                                 &local_err);
    291        if (local_err) {
    292            error_propagate(errp, local_err);
    293            return;
    294        }
    295    }
    296
    297    init_topo_info(&topo_info, x86ms);
    298
    299    env->nr_dies = ms->smp.dies;
    300
    301    /*
    302     * If APIC ID is not set,
    303     * set it based on socket/die/core/thread properties.
    304     */
    305    if (cpu->apic_id == UNASSIGNED_APIC_ID) {
    306        int max_socket = (ms->smp.max_cpus - 1) /
    307                                smp_threads / smp_cores / ms->smp.dies;
    308
    309        /*
    310         * die-id was optional in QEMU 4.0 and older, so keep it optional
    311         * if there's only one die per socket.
    312         */
    313        if (cpu->die_id < 0 && ms->smp.dies == 1) {
    314            cpu->die_id = 0;
    315        }
    316
    317        if (cpu->socket_id < 0) {
    318            error_setg(errp, "CPU socket-id is not set");
    319            return;
    320        } else if (cpu->socket_id > max_socket) {
    321            error_setg(errp, "Invalid CPU socket-id: %u must be in range 0:%u",
    322                       cpu->socket_id, max_socket);
    323            return;
    324        }
    325        if (cpu->die_id < 0) {
    326            error_setg(errp, "CPU die-id is not set");
    327            return;
    328        } else if (cpu->die_id > ms->smp.dies - 1) {
    329            error_setg(errp, "Invalid CPU die-id: %u must be in range 0:%u",
    330                       cpu->die_id, ms->smp.dies - 1);
    331            return;
    332        }
    333        if (cpu->core_id < 0) {
    334            error_setg(errp, "CPU core-id is not set");
    335            return;
    336        } else if (cpu->core_id > (smp_cores - 1)) {
    337            error_setg(errp, "Invalid CPU core-id: %u must be in range 0:%u",
    338                       cpu->core_id, smp_cores - 1);
    339            return;
    340        }
    341        if (cpu->thread_id < 0) {
    342            error_setg(errp, "CPU thread-id is not set");
    343            return;
    344        } else if (cpu->thread_id > (smp_threads - 1)) {
    345            error_setg(errp, "Invalid CPU thread-id: %u must be in range 0:%u",
    346                       cpu->thread_id, smp_threads - 1);
    347            return;
    348        }
    349
    350        topo_ids.pkg_id = cpu->socket_id;
    351        topo_ids.die_id = cpu->die_id;
    352        topo_ids.core_id = cpu->core_id;
    353        topo_ids.smt_id = cpu->thread_id;
    354        cpu->apic_id = x86_apicid_from_topo_ids(&topo_info, &topo_ids);
    355    }
    356
    357    cpu_slot = x86_find_cpu_slot(MACHINE(x86ms), cpu->apic_id, &idx);
    358    if (!cpu_slot) {
    359        MachineState *ms = MACHINE(x86ms);
    360
    361        x86_topo_ids_from_apicid(cpu->apic_id, &topo_info, &topo_ids);
    362        error_setg(errp,
    363            "Invalid CPU [socket: %u, die: %u, core: %u, thread: %u] with"
    364            " APIC ID %" PRIu32 ", valid index range 0:%d",
    365            topo_ids.pkg_id, topo_ids.die_id, topo_ids.core_id, topo_ids.smt_id,
    366            cpu->apic_id, ms->possible_cpus->len - 1);
    367        return;
    368    }
    369
    370    if (cpu_slot->cpu) {
    371        error_setg(errp, "CPU[%d] with APIC ID %" PRIu32 " exists",
    372                   idx, cpu->apic_id);
    373        return;
    374    }
    375
    376    /* if 'address' properties socket-id/core-id/thread-id are not set, set them
    377     * so that machine_query_hotpluggable_cpus would show correct values
    378     */
    379    /* TODO: move socket_id/core_id/thread_id checks into x86_cpu_realizefn()
    380     * once -smp refactoring is complete and there will be CPU private
    381     * CPUState::nr_cores and CPUState::nr_threads fields instead of globals */
    382    x86_topo_ids_from_apicid(cpu->apic_id, &topo_info, &topo_ids);
    383    if (cpu->socket_id != -1 && cpu->socket_id != topo_ids.pkg_id) {
    384        error_setg(errp, "property socket-id: %u doesn't match set apic-id:"
    385            " 0x%x (socket-id: %u)", cpu->socket_id, cpu->apic_id,
    386            topo_ids.pkg_id);
    387        return;
    388    }
    389    cpu->socket_id = topo_ids.pkg_id;
    390
    391    if (cpu->die_id != -1 && cpu->die_id != topo_ids.die_id) {
    392        error_setg(errp, "property die-id: %u doesn't match set apic-id:"
    393            " 0x%x (die-id: %u)", cpu->die_id, cpu->apic_id, topo_ids.die_id);
    394        return;
    395    }
    396    cpu->die_id = topo_ids.die_id;
    397
    398    if (cpu->core_id != -1 && cpu->core_id != topo_ids.core_id) {
    399        error_setg(errp, "property core-id: %u doesn't match set apic-id:"
    400            " 0x%x (core-id: %u)", cpu->core_id, cpu->apic_id,
    401            topo_ids.core_id);
    402        return;
    403    }
    404    cpu->core_id = topo_ids.core_id;
    405
    406    if (cpu->thread_id != -1 && cpu->thread_id != topo_ids.smt_id) {
    407        error_setg(errp, "property thread-id: %u doesn't match set apic-id:"
    408            " 0x%x (thread-id: %u)", cpu->thread_id, cpu->apic_id,
    409            topo_ids.smt_id);
    410        return;
    411    }
    412    cpu->thread_id = topo_ids.smt_id;
    413
    414    if (hyperv_feat_enabled(cpu, HYPERV_FEAT_VPINDEX) &&
    415        !kvm_hv_vpindex_settable()) {
    416        error_setg(errp, "kernel doesn't allow setting HyperV VP_INDEX");
    417        return;
    418    }
    419
    420    cs = CPU(cpu);
    421    cs->cpu_index = idx;
    422
    423    numa_cpu_pre_plug(cpu_slot, dev, errp);
    424}
    425
    426CpuInstanceProperties
    427x86_cpu_index_to_props(MachineState *ms, unsigned cpu_index)
    428{
    429    MachineClass *mc = MACHINE_GET_CLASS(ms);
    430    const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(ms);
    431
    432    assert(cpu_index < possible_cpus->len);
    433    return possible_cpus->cpus[cpu_index].props;
    434}
    435
    436int64_t x86_get_default_cpu_node_id(const MachineState *ms, int idx)
    437{
    438   X86CPUTopoIDs topo_ids;
    439   X86MachineState *x86ms = X86_MACHINE(ms);
    440   X86CPUTopoInfo topo_info;
    441
    442   init_topo_info(&topo_info, x86ms);
    443
    444   assert(idx < ms->possible_cpus->len);
    445   x86_topo_ids_from_apicid(ms->possible_cpus->cpus[idx].arch_id,
    446                            &topo_info, &topo_ids);
    447   return topo_ids.pkg_id % ms->numa_state->num_nodes;
    448}
    449
    450const CPUArchIdList *x86_possible_cpu_arch_ids(MachineState *ms)
    451{
    452    X86MachineState *x86ms = X86_MACHINE(ms);
    453    unsigned int max_cpus = ms->smp.max_cpus;
    454    X86CPUTopoInfo topo_info;
    455    int i;
    456
    457    if (ms->possible_cpus) {
    458        /*
    459         * make sure that max_cpus hasn't changed since the first use, i.e.
    460         * -smp hasn't been parsed after it
    461         */
    462        assert(ms->possible_cpus->len == max_cpus);
    463        return ms->possible_cpus;
    464    }
    465
    466    ms->possible_cpus = g_malloc0(sizeof(CPUArchIdList) +
    467                                  sizeof(CPUArchId) * max_cpus);
    468    ms->possible_cpus->len = max_cpus;
    469
    470    init_topo_info(&topo_info, x86ms);
    471
    472    for (i = 0; i < ms->possible_cpus->len; i++) {
    473        X86CPUTopoIDs topo_ids;
    474
    475        ms->possible_cpus->cpus[i].type = ms->cpu_type;
    476        ms->possible_cpus->cpus[i].vcpus_count = 1;
    477        ms->possible_cpus->cpus[i].arch_id =
    478            x86_cpu_apic_id_from_index(x86ms, i);
    479        x86_topo_ids_from_apicid(ms->possible_cpus->cpus[i].arch_id,
    480                                 &topo_info, &topo_ids);
    481        ms->possible_cpus->cpus[i].props.has_socket_id = true;
    482        ms->possible_cpus->cpus[i].props.socket_id = topo_ids.pkg_id;
    483        if (ms->smp.dies > 1) {
    484            ms->possible_cpus->cpus[i].props.has_die_id = true;
    485            ms->possible_cpus->cpus[i].props.die_id = topo_ids.die_id;
    486        }
    487        ms->possible_cpus->cpus[i].props.has_core_id = true;
    488        ms->possible_cpus->cpus[i].props.core_id = topo_ids.core_id;
    489        ms->possible_cpus->cpus[i].props.has_thread_id = true;
    490        ms->possible_cpus->cpus[i].props.thread_id = topo_ids.smt_id;
    491    }
    492    return ms->possible_cpus;
    493}
    494
    495static void x86_nmi(NMIState *n, int cpu_index, Error **errp)
    496{
    497    /* cpu index isn't used */
    498    CPUState *cs;
    499
    500    CPU_FOREACH(cs) {
    501        X86CPU *cpu = X86_CPU(cs);
    502
    503        if (!cpu->apic_state) {
    504            cpu_interrupt(cs, CPU_INTERRUPT_NMI);
    505        } else {
    506            apic_deliver_nmi(cpu->apic_state);
    507        }
    508    }
    509}
    510
    511static long get_file_size(FILE *f)
    512{
    513    long where, size;
    514
    515    /* XXX: on Unix systems, using fstat() probably makes more sense */
    516
    517    where = ftell(f);
    518    fseek(f, 0, SEEK_END);
    519    size = ftell(f);
    520    fseek(f, where, SEEK_SET);
    521
    522    return size;
    523}
    524
    525/* TSC handling */
    526uint64_t cpu_get_tsc(CPUX86State *env)
    527{
    528    return cpus_get_elapsed_ticks();
    529}
    530
    531/* IRQ handling */
    532static void pic_irq_request(void *opaque, int irq, int level)
    533{
    534    CPUState *cs = first_cpu;
    535    X86CPU *cpu = X86_CPU(cs);
    536
    537    trace_x86_pic_interrupt(irq, level);
    538    if (cpu->apic_state && !kvm_irqchip_in_kernel() &&
    539        !whpx_apic_in_platform()) {
    540        CPU_FOREACH(cs) {
    541            cpu = X86_CPU(cs);
    542            if (apic_accept_pic_intr(cpu->apic_state)) {
    543                apic_deliver_pic_intr(cpu->apic_state, level);
    544            }
    545        }
    546    } else {
    547        if (level) {
    548            cpu_interrupt(cs, CPU_INTERRUPT_HARD);
    549        } else {
    550            cpu_reset_interrupt(cs, CPU_INTERRUPT_HARD);
    551        }
    552    }
    553}
    554
    555qemu_irq x86_allocate_cpu_irq(void)
    556{
    557    return qemu_allocate_irq(pic_irq_request, NULL, 0);
    558}
    559
    560int cpu_get_pic_interrupt(CPUX86State *env)
    561{
    562    X86CPU *cpu = env_archcpu(env);
    563    int intno;
    564
    565    if (!kvm_irqchip_in_kernel() && !whpx_apic_in_platform()) {
    566        intno = apic_get_interrupt(cpu->apic_state);
    567        if (intno >= 0) {
    568            return intno;
    569        }
    570        /* read the irq from the PIC */
    571        if (!apic_accept_pic_intr(cpu->apic_state)) {
    572            return -1;
    573        }
    574    }
    575
    576    intno = pic_read_irq(isa_pic);
    577    return intno;
    578}
    579
    580DeviceState *cpu_get_current_apic(void)
    581{
    582    if (current_cpu) {
    583        X86CPU *cpu = X86_CPU(current_cpu);
    584        return cpu->apic_state;
    585    } else {
    586        return NULL;
    587    }
    588}
    589
    590void gsi_handler(void *opaque, int n, int level)
    591{
    592    GSIState *s = opaque;
    593
    594    trace_x86_gsi_interrupt(n, level);
    595    switch (n) {
    596    case 0 ... ISA_NUM_IRQS - 1:
    597        if (s->i8259_irq[n]) {
    598            /* Under KVM, Kernel will forward to both PIC and IOAPIC */
    599            qemu_set_irq(s->i8259_irq[n], level);
    600        }
    601        /* fall through */
    602    case ISA_NUM_IRQS ... IOAPIC_NUM_PINS - 1:
    603        qemu_set_irq(s->ioapic_irq[n], level);
    604        break;
    605    case IO_APIC_SECONDARY_IRQBASE
    606        ... IO_APIC_SECONDARY_IRQBASE + IOAPIC_NUM_PINS - 1:
    607        qemu_set_irq(s->ioapic2_irq[n - IO_APIC_SECONDARY_IRQBASE], level);
    608        break;
    609    }
    610}
    611
    612void ioapic_init_gsi(GSIState *gsi_state, const char *parent_name)
    613{
    614    DeviceState *dev;
    615    SysBusDevice *d;
    616    unsigned int i;
    617
    618    assert(parent_name);
    619    if (kvm_ioapic_in_kernel()) {
    620        dev = qdev_new(TYPE_KVM_IOAPIC);
    621    } else {
    622        dev = qdev_new(TYPE_IOAPIC);
    623    }
    624    object_property_add_child(object_resolve_path(parent_name, NULL),
    625                              "ioapic", OBJECT(dev));
    626    d = SYS_BUS_DEVICE(dev);
    627    sysbus_realize_and_unref(d, &error_fatal);
    628    sysbus_mmio_map(d, 0, IO_APIC_DEFAULT_ADDRESS);
    629
    630    for (i = 0; i < IOAPIC_NUM_PINS; i++) {
    631        gsi_state->ioapic_irq[i] = qdev_get_gpio_in(dev, i);
    632    }
    633}
    634
    635DeviceState *ioapic_init_secondary(GSIState *gsi_state)
    636{
    637    DeviceState *dev;
    638    SysBusDevice *d;
    639    unsigned int i;
    640
    641    dev = qdev_new(TYPE_IOAPIC);
    642    d = SYS_BUS_DEVICE(dev);
    643    sysbus_realize_and_unref(d, &error_fatal);
    644    sysbus_mmio_map(d, 0, IO_APIC_SECONDARY_ADDRESS);
    645
    646    for (i = 0; i < IOAPIC_NUM_PINS; i++) {
    647        gsi_state->ioapic2_irq[i] = qdev_get_gpio_in(dev, i);
    648    }
    649    return dev;
    650}
    651
    652struct setup_data {
    653    uint64_t next;
    654    uint32_t type;
    655    uint32_t len;
    656    uint8_t data[];
    657} __attribute__((packed));
    658
    659
    660/*
    661 * The entry point into the kernel for PVH boot is different from
    662 * the native entry point.  The PVH entry is defined by the x86/HVM
    663 * direct boot ABI and is available in an ELFNOTE in the kernel binary.
    664 *
    665 * This function is passed to load_elf() when it is called from
    666 * load_elfboot() which then additionally checks for an ELF Note of
    667 * type XEN_ELFNOTE_PHYS32_ENTRY and passes it to this function to
    668 * parse the PVH entry address from the ELF Note.
    669 *
    670 * Due to trickery in elf_opts.h, load_elf() is actually available as
    671 * load_elf32() or load_elf64() and this routine needs to be able
    672 * to deal with being called as 32 or 64 bit.
    673 *
    674 * The address of the PVH entry point is saved to the 'pvh_start_addr'
    675 * global variable.  (although the entry point is 32-bit, the kernel
    676 * binary can be either 32-bit or 64-bit).
    677 */
    678static uint64_t read_pvh_start_addr(void *arg1, void *arg2, bool is64)
    679{
    680    size_t *elf_note_data_addr;
    681
    682    /* Check if ELF Note header passed in is valid */
    683    if (arg1 == NULL) {
    684        return 0;
    685    }
    686
    687    if (is64) {
    688        struct elf64_note *nhdr64 = (struct elf64_note *)arg1;
    689        uint64_t nhdr_size64 = sizeof(struct elf64_note);
    690        uint64_t phdr_align = *(uint64_t *)arg2;
    691        uint64_t nhdr_namesz = nhdr64->n_namesz;
    692
    693        elf_note_data_addr =
    694            ((void *)nhdr64) + nhdr_size64 +
    695            QEMU_ALIGN_UP(nhdr_namesz, phdr_align);
    696
    697        pvh_start_addr = *elf_note_data_addr;
    698    } else {
    699        struct elf32_note *nhdr32 = (struct elf32_note *)arg1;
    700        uint32_t nhdr_size32 = sizeof(struct elf32_note);
    701        uint32_t phdr_align = *(uint32_t *)arg2;
    702        uint32_t nhdr_namesz = nhdr32->n_namesz;
    703
    704        elf_note_data_addr =
    705            ((void *)nhdr32) + nhdr_size32 +
    706            QEMU_ALIGN_UP(nhdr_namesz, phdr_align);
    707
    708        pvh_start_addr = *(uint32_t *)elf_note_data_addr;
    709    }
    710
    711    return pvh_start_addr;
    712}
    713
    714static bool load_elfboot(const char *kernel_filename,
    715                         int kernel_file_size,
    716                         uint8_t *header,
    717                         size_t pvh_xen_start_addr,
    718                         FWCfgState *fw_cfg)
    719{
    720    uint32_t flags = 0;
    721    uint32_t mh_load_addr = 0;
    722    uint32_t elf_kernel_size = 0;
    723    uint64_t elf_entry;
    724    uint64_t elf_low, elf_high;
    725    int kernel_size;
    726
    727    if (ldl_p(header) != 0x464c457f) {
    728        return false; /* no elfboot */
    729    }
    730
    731    bool elf_is64 = header[EI_CLASS] == ELFCLASS64;
    732    flags = elf_is64 ?
    733        ((Elf64_Ehdr *)header)->e_flags : ((Elf32_Ehdr *)header)->e_flags;
    734
    735    if (flags & 0x00010004) { /* LOAD_ELF_HEADER_HAS_ADDR */
    736        error_report("elfboot unsupported flags = %x", flags);
    737        exit(1);
    738    }
    739
    740    uint64_t elf_note_type = XEN_ELFNOTE_PHYS32_ENTRY;
    741    kernel_size = load_elf(kernel_filename, read_pvh_start_addr,
    742                           NULL, &elf_note_type, &elf_entry,
    743                           &elf_low, &elf_high, NULL, 0, I386_ELF_MACHINE,
    744                           0, 0);
    745
    746    if (kernel_size < 0) {
    747        error_report("Error while loading elf kernel");
    748        exit(1);
    749    }
    750    mh_load_addr = elf_low;
    751    elf_kernel_size = elf_high - elf_low;
    752
    753    if (pvh_start_addr == 0) {
    754        error_report("Error loading uncompressed kernel without PVH ELF Note");
    755        exit(1);
    756    }
    757    fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ENTRY, pvh_start_addr);
    758    fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ADDR, mh_load_addr);
    759    fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_SIZE, elf_kernel_size);
    760
    761    return true;
    762}
    763
    764void x86_load_linux(X86MachineState *x86ms,
    765                    FWCfgState *fw_cfg,
    766                    int acpi_data_size,
    767                    bool pvh_enabled,
    768                    bool linuxboot_dma_enabled)
    769{
    770    uint16_t protocol;
    771    int setup_size, kernel_size, cmdline_size;
    772    int dtb_size, setup_data_offset;
    773    uint32_t initrd_max;
    774    uint8_t header[8192], *setup, *kernel;
    775    hwaddr real_addr, prot_addr, cmdline_addr, initrd_addr = 0;
    776    FILE *f;
    777    char *vmode;
    778    MachineState *machine = MACHINE(x86ms);
    779    struct setup_data *setup_data;
    780    const char *kernel_filename = machine->kernel_filename;
    781    const char *initrd_filename = machine->initrd_filename;
    782    const char *dtb_filename = machine->dtb;
    783    const char *kernel_cmdline = machine->kernel_cmdline;
    784    SevKernelLoaderContext sev_load_ctx = {};
    785
    786    /* Align to 16 bytes as a paranoia measure */
    787    cmdline_size = (strlen(kernel_cmdline) + 16) & ~15;
    788
    789    /* load the kernel header */
    790    f = fopen(kernel_filename, "rb");
    791    if (!f) {
    792        fprintf(stderr, "qemu: could not open kernel file '%s': %s\n",
    793                kernel_filename, strerror(errno));
    794        exit(1);
    795    }
    796
    797    kernel_size = get_file_size(f);
    798    if (!kernel_size ||
    799        fread(header, 1, MIN(ARRAY_SIZE(header), kernel_size), f) !=
    800        MIN(ARRAY_SIZE(header), kernel_size)) {
    801        fprintf(stderr, "qemu: could not load kernel '%s': %s\n",
    802                kernel_filename, strerror(errno));
    803        exit(1);
    804    }
    805
    806    /* kernel protocol version */
    807    if (ldl_p(header + 0x202) == 0x53726448) {
    808        protocol = lduw_p(header + 0x206);
    809    } else {
    810        /*
    811         * This could be a multiboot kernel. If it is, let's stop treating it
    812         * like a Linux kernel.
    813         * Note: some multiboot images could be in the ELF format (the same of
    814         * PVH), so we try multiboot first since we check the multiboot magic
    815         * header before to load it.
    816         */
    817        if (load_multiboot(fw_cfg, f, kernel_filename, initrd_filename,
    818                           kernel_cmdline, kernel_size, header)) {
    819            return;
    820        }
    821        /*
    822         * Check if the file is an uncompressed kernel file (ELF) and load it,
    823         * saving the PVH entry point used by the x86/HVM direct boot ABI.
    824         * If load_elfboot() is successful, populate the fw_cfg info.
    825         */
    826        if (pvh_enabled &&
    827            load_elfboot(kernel_filename, kernel_size,
    828                         header, pvh_start_addr, fw_cfg)) {
    829            fclose(f);
    830
    831            fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_SIZE,
    832                strlen(kernel_cmdline) + 1);
    833            fw_cfg_add_string(fw_cfg, FW_CFG_CMDLINE_DATA, kernel_cmdline);
    834
    835            fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_SIZE, sizeof(header));
    836            fw_cfg_add_bytes(fw_cfg, FW_CFG_SETUP_DATA,
    837                             header, sizeof(header));
    838
    839            /* load initrd */
    840            if (initrd_filename) {
    841                GMappedFile *mapped_file;
    842                gsize initrd_size;
    843                gchar *initrd_data;
    844                GError *gerr = NULL;
    845
    846                mapped_file = g_mapped_file_new(initrd_filename, false, &gerr);
    847                if (!mapped_file) {
    848                    fprintf(stderr, "qemu: error reading initrd %s: %s\n",
    849                            initrd_filename, gerr->message);
    850                    exit(1);
    851                }
    852                x86ms->initrd_mapped_file = mapped_file;
    853
    854                initrd_data = g_mapped_file_get_contents(mapped_file);
    855                initrd_size = g_mapped_file_get_length(mapped_file);
    856                initrd_max = x86ms->below_4g_mem_size - acpi_data_size - 1;
    857                if (initrd_size >= initrd_max) {
    858                    fprintf(stderr, "qemu: initrd is too large, cannot support."
    859                            "(max: %"PRIu32", need %"PRId64")\n",
    860                            initrd_max, (uint64_t)initrd_size);
    861                    exit(1);
    862                }
    863
    864                initrd_addr = (initrd_max - initrd_size) & ~4095;
    865
    866                fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_ADDR, initrd_addr);
    867                fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_SIZE, initrd_size);
    868                fw_cfg_add_bytes(fw_cfg, FW_CFG_INITRD_DATA, initrd_data,
    869                                 initrd_size);
    870            }
    871
    872            option_rom[nb_option_roms].bootindex = 0;
    873            option_rom[nb_option_roms].name = "pvh.bin";
    874            nb_option_roms++;
    875
    876            return;
    877        }
    878        protocol = 0;
    879    }
    880
    881    if (protocol < 0x200 || !(header[0x211] & 0x01)) {
    882        /* Low kernel */
    883        real_addr    = 0x90000;
    884        cmdline_addr = 0x9a000 - cmdline_size;
    885        prot_addr    = 0x10000;
    886    } else if (protocol < 0x202) {
    887        /* High but ancient kernel */
    888        real_addr    = 0x90000;
    889        cmdline_addr = 0x9a000 - cmdline_size;
    890        prot_addr    = 0x100000;
    891    } else {
    892        /* High and recent kernel */
    893        real_addr    = 0x10000;
    894        cmdline_addr = 0x20000;
    895        prot_addr    = 0x100000;
    896    }
    897
    898    /* highest address for loading the initrd */
    899    if (protocol >= 0x20c &&
    900        lduw_p(header + 0x236) & XLF_CAN_BE_LOADED_ABOVE_4G) {
    901        /*
    902         * Linux has supported initrd up to 4 GB for a very long time (2007,
    903         * long before XLF_CAN_BE_LOADED_ABOVE_4G which was added in 2013),
    904         * though it only sets initrd_max to 2 GB to "work around bootloader
    905         * bugs". Luckily, QEMU firmware(which does something like bootloader)
    906         * has supported this.
    907         *
    908         * It's believed that if XLF_CAN_BE_LOADED_ABOVE_4G is set, initrd can
    909         * be loaded into any address.
    910         *
    911         * In addition, initrd_max is uint32_t simply because QEMU doesn't
    912         * support the 64-bit boot protocol (specifically the ext_ramdisk_image
    913         * field).
    914         *
    915         * Therefore here just limit initrd_max to UINT32_MAX simply as well.
    916         */
    917        initrd_max = UINT32_MAX;
    918    } else if (protocol >= 0x203) {
    919        initrd_max = ldl_p(header + 0x22c);
    920    } else {
    921        initrd_max = 0x37ffffff;
    922    }
    923
    924    if (initrd_max >= x86ms->below_4g_mem_size - acpi_data_size) {
    925        initrd_max = x86ms->below_4g_mem_size - acpi_data_size - 1;
    926    }
    927
    928    fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_ADDR, cmdline_addr);
    929    fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_SIZE, strlen(kernel_cmdline) + 1);
    930    fw_cfg_add_string(fw_cfg, FW_CFG_CMDLINE_DATA, kernel_cmdline);
    931    sev_load_ctx.cmdline_data = (char *)kernel_cmdline;
    932    sev_load_ctx.cmdline_size = strlen(kernel_cmdline) + 1;
    933
    934    if (protocol >= 0x202) {
    935        stl_p(header + 0x228, cmdline_addr);
    936    } else {
    937        stw_p(header + 0x20, 0xA33F);
    938        stw_p(header + 0x22, cmdline_addr - real_addr);
    939    }
    940
    941    /* handle vga= parameter */
    942    vmode = strstr(kernel_cmdline, "vga=");
    943    if (vmode) {
    944        unsigned int video_mode;
    945        const char *end;
    946        int ret;
    947        /* skip "vga=" */
    948        vmode += 4;
    949        if (!strncmp(vmode, "normal", 6)) {
    950            video_mode = 0xffff;
    951        } else if (!strncmp(vmode, "ext", 3)) {
    952            video_mode = 0xfffe;
    953        } else if (!strncmp(vmode, "ask", 3)) {
    954            video_mode = 0xfffd;
    955        } else {
    956            ret = qemu_strtoui(vmode, &end, 0, &video_mode);
    957            if (ret != 0 || (*end && *end != ' ')) {
    958                fprintf(stderr, "qemu: invalid 'vga=' kernel parameter.\n");
    959                exit(1);
    960            }
    961        }
    962        stw_p(header + 0x1fa, video_mode);
    963    }
    964
    965    /* loader type */
    966    /*
    967     * High nybble = B reserved for QEMU; low nybble is revision number.
    968     * If this code is substantially changed, you may want to consider
    969     * incrementing the revision.
    970     */
    971    if (protocol >= 0x200) {
    972        header[0x210] = 0xB0;
    973    }
    974    /* heap */
    975    if (protocol >= 0x201) {
    976        header[0x211] |= 0x80; /* CAN_USE_HEAP */
    977        stw_p(header + 0x224, cmdline_addr - real_addr - 0x200);
    978    }
    979
    980    /* load initrd */
    981    if (initrd_filename) {
    982        GMappedFile *mapped_file;
    983        gsize initrd_size;
    984        gchar *initrd_data;
    985        GError *gerr = NULL;
    986
    987        if (protocol < 0x200) {
    988            fprintf(stderr, "qemu: linux kernel too old to load a ram disk\n");
    989            exit(1);
    990        }
    991
    992        mapped_file = g_mapped_file_new(initrd_filename, false, &gerr);
    993        if (!mapped_file) {
    994            fprintf(stderr, "qemu: error reading initrd %s: %s\n",
    995                    initrd_filename, gerr->message);
    996            exit(1);
    997        }
    998        x86ms->initrd_mapped_file = mapped_file;
    999
   1000        initrd_data = g_mapped_file_get_contents(mapped_file);
   1001        initrd_size = g_mapped_file_get_length(mapped_file);
   1002        if (initrd_size >= initrd_max) {
   1003            fprintf(stderr, "qemu: initrd is too large, cannot support."
   1004                    "(max: %"PRIu32", need %"PRId64")\n",
   1005                    initrd_max, (uint64_t)initrd_size);
   1006            exit(1);
   1007        }
   1008
   1009        initrd_addr = (initrd_max - initrd_size) & ~4095;
   1010
   1011        fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_ADDR, initrd_addr);
   1012        fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_SIZE, initrd_size);
   1013        fw_cfg_add_bytes(fw_cfg, FW_CFG_INITRD_DATA, initrd_data, initrd_size);
   1014        sev_load_ctx.initrd_data = initrd_data;
   1015        sev_load_ctx.initrd_size = initrd_size;
   1016
   1017        stl_p(header + 0x218, initrd_addr);
   1018        stl_p(header + 0x21c, initrd_size);
   1019    }
   1020
   1021    /* load kernel and setup */
   1022    setup_size = header[0x1f1];
   1023    if (setup_size == 0) {
   1024        setup_size = 4;
   1025    }
   1026    setup_size = (setup_size + 1) * 512;
   1027    if (setup_size > kernel_size) {
   1028        fprintf(stderr, "qemu: invalid kernel header\n");
   1029        exit(1);
   1030    }
   1031    kernel_size -= setup_size;
   1032
   1033    setup  = g_malloc(setup_size);
   1034    kernel = g_malloc(kernel_size);
   1035    fseek(f, 0, SEEK_SET);
   1036    if (fread(setup, 1, setup_size, f) != setup_size) {
   1037        fprintf(stderr, "fread() failed\n");
   1038        exit(1);
   1039    }
   1040    if (fread(kernel, 1, kernel_size, f) != kernel_size) {
   1041        fprintf(stderr, "fread() failed\n");
   1042        exit(1);
   1043    }
   1044    fclose(f);
   1045
   1046    /* append dtb to kernel */
   1047    if (dtb_filename) {
   1048        if (protocol < 0x209) {
   1049            fprintf(stderr, "qemu: Linux kernel too old to load a dtb\n");
   1050            exit(1);
   1051        }
   1052
   1053        dtb_size = get_image_size(dtb_filename);
   1054        if (dtb_size <= 0) {
   1055            fprintf(stderr, "qemu: error reading dtb %s: %s\n",
   1056                    dtb_filename, strerror(errno));
   1057            exit(1);
   1058        }
   1059
   1060        setup_data_offset = QEMU_ALIGN_UP(kernel_size, 16);
   1061        kernel_size = setup_data_offset + sizeof(struct setup_data) + dtb_size;
   1062        kernel = g_realloc(kernel, kernel_size);
   1063
   1064        stq_p(header + 0x250, prot_addr + setup_data_offset);
   1065
   1066        setup_data = (struct setup_data *)(kernel + setup_data_offset);
   1067        setup_data->next = 0;
   1068        setup_data->type = cpu_to_le32(SETUP_DTB);
   1069        setup_data->len = cpu_to_le32(dtb_size);
   1070
   1071        load_image_size(dtb_filename, setup_data->data, dtb_size);
   1072    }
   1073
   1074    /*
   1075     * If we're starting an encrypted VM, it will be OVMF based, which uses the
   1076     * efi stub for booting and doesn't require any values to be placed in the
   1077     * kernel header.  We therefore don't update the header so the hash of the
   1078     * kernel on the other side of the fw_cfg interface matches the hash of the
   1079     * file the user passed in.
   1080     */
   1081    if (!sev_enabled()) {
   1082        memcpy(setup, header, MIN(sizeof(header), setup_size));
   1083    }
   1084
   1085    fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ADDR, prot_addr);
   1086    fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_SIZE, kernel_size);
   1087    fw_cfg_add_bytes(fw_cfg, FW_CFG_KERNEL_DATA, kernel, kernel_size);
   1088    sev_load_ctx.kernel_data = (char *)kernel;
   1089    sev_load_ctx.kernel_size = kernel_size;
   1090
   1091    fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_ADDR, real_addr);
   1092    fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_SIZE, setup_size);
   1093    fw_cfg_add_bytes(fw_cfg, FW_CFG_SETUP_DATA, setup, setup_size);
   1094    sev_load_ctx.setup_data = (char *)setup;
   1095    sev_load_ctx.setup_size = setup_size;
   1096
   1097    if (sev_enabled()) {
   1098        sev_add_kernel_loader_hashes(&sev_load_ctx, &error_fatal);
   1099    }
   1100
   1101    option_rom[nb_option_roms].bootindex = 0;
   1102    option_rom[nb_option_roms].name = "linuxboot.bin";
   1103    if (linuxboot_dma_enabled && fw_cfg_dma_enabled(fw_cfg)) {
   1104        option_rom[nb_option_roms].name = "linuxboot_dma.bin";
   1105    }
   1106    nb_option_roms++;
   1107}
   1108
   1109void x86_bios_rom_init(MachineState *ms, const char *default_firmware,
   1110                       MemoryRegion *rom_memory, bool isapc_ram_fw)
   1111{
   1112    const char *bios_name;
   1113    char *filename;
   1114    MemoryRegion *bios, *isa_bios;
   1115    int bios_size, isa_bios_size;
   1116    int ret;
   1117
   1118    /* BIOS load */
   1119    bios_name = ms->firmware ?: default_firmware;
   1120    filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, bios_name);
   1121    if (filename) {
   1122        bios_size = get_image_size(filename);
   1123    } else {
   1124        bios_size = -1;
   1125    }
   1126    if (bios_size <= 0 ||
   1127        (bios_size % 65536) != 0) {
   1128        goto bios_error;
   1129    }
   1130    bios = g_malloc(sizeof(*bios));
   1131    memory_region_init_ram(bios, NULL, "pc.bios", bios_size, &error_fatal);
   1132    if (!isapc_ram_fw) {
   1133        memory_region_set_readonly(bios, true);
   1134    }
   1135    ret = rom_add_file_fixed(bios_name, (uint32_t)(-bios_size), -1);
   1136    if (ret != 0) {
   1137    bios_error:
   1138        fprintf(stderr, "qemu: could not load PC BIOS '%s'\n", bios_name);
   1139        exit(1);
   1140    }
   1141    g_free(filename);
   1142
   1143    /* map the last 128KB of the BIOS in ISA space */
   1144    isa_bios_size = MIN(bios_size, 128 * KiB);
   1145    isa_bios = g_malloc(sizeof(*isa_bios));
   1146    memory_region_init_alias(isa_bios, NULL, "isa-bios", bios,
   1147                             bios_size - isa_bios_size, isa_bios_size);
   1148    memory_region_add_subregion_overlap(rom_memory,
   1149                                        0x100000 - isa_bios_size,
   1150                                        isa_bios,
   1151                                        1);
   1152    if (!isapc_ram_fw) {
   1153        memory_region_set_readonly(isa_bios, true);
   1154    }
   1155
   1156    /* map all the bios at the top of memory */
   1157    memory_region_add_subregion(rom_memory,
   1158                                (uint32_t)(-bios_size),
   1159                                bios);
   1160}
   1161
   1162bool x86_machine_is_smm_enabled(const X86MachineState *x86ms)
   1163{
   1164    bool smm_available = false;
   1165
   1166    if (x86ms->smm == ON_OFF_AUTO_OFF) {
   1167        return false;
   1168    }
   1169
   1170    if (tcg_enabled() || qtest_enabled()) {
   1171        smm_available = true;
   1172    } else if (kvm_enabled()) {
   1173        smm_available = kvm_has_smm();
   1174    }
   1175
   1176    if (smm_available) {
   1177        return true;
   1178    }
   1179
   1180    if (x86ms->smm == ON_OFF_AUTO_ON) {
   1181        error_report("System Management Mode not supported by this hypervisor.");
   1182        exit(1);
   1183    }
   1184    return false;
   1185}
   1186
   1187static void x86_machine_get_smm(Object *obj, Visitor *v, const char *name,
   1188                               void *opaque, Error **errp)
   1189{
   1190    X86MachineState *x86ms = X86_MACHINE(obj);
   1191    OnOffAuto smm = x86ms->smm;
   1192
   1193    visit_type_OnOffAuto(v, name, &smm, errp);
   1194}
   1195
   1196static void x86_machine_set_smm(Object *obj, Visitor *v, const char *name,
   1197                               void *opaque, Error **errp)
   1198{
   1199    X86MachineState *x86ms = X86_MACHINE(obj);
   1200
   1201    visit_type_OnOffAuto(v, name, &x86ms->smm, errp);
   1202}
   1203
   1204bool x86_machine_is_acpi_enabled(const X86MachineState *x86ms)
   1205{
   1206    if (x86ms->acpi == ON_OFF_AUTO_OFF) {
   1207        return false;
   1208    }
   1209    return true;
   1210}
   1211
   1212static void x86_machine_get_acpi(Object *obj, Visitor *v, const char *name,
   1213                                 void *opaque, Error **errp)
   1214{
   1215    X86MachineState *x86ms = X86_MACHINE(obj);
   1216    OnOffAuto acpi = x86ms->acpi;
   1217
   1218    visit_type_OnOffAuto(v, name, &acpi, errp);
   1219}
   1220
   1221static void x86_machine_set_acpi(Object *obj, Visitor *v, const char *name,
   1222                                 void *opaque, Error **errp)
   1223{
   1224    X86MachineState *x86ms = X86_MACHINE(obj);
   1225
   1226    visit_type_OnOffAuto(v, name, &x86ms->acpi, errp);
   1227}
   1228
   1229static char *x86_machine_get_oem_id(Object *obj, Error **errp)
   1230{
   1231    X86MachineState *x86ms = X86_MACHINE(obj);
   1232
   1233    return g_strdup(x86ms->oem_id);
   1234}
   1235
   1236static void x86_machine_set_oem_id(Object *obj, const char *value, Error **errp)
   1237{
   1238    X86MachineState *x86ms = X86_MACHINE(obj);
   1239    size_t len = strlen(value);
   1240
   1241    if (len > 6) {
   1242        error_setg(errp,
   1243                   "User specified "X86_MACHINE_OEM_ID" value is bigger than "
   1244                   "6 bytes in size");
   1245        return;
   1246    }
   1247
   1248    strncpy(x86ms->oem_id, value, 6);
   1249}
   1250
   1251static char *x86_machine_get_oem_table_id(Object *obj, Error **errp)
   1252{
   1253    X86MachineState *x86ms = X86_MACHINE(obj);
   1254
   1255    return g_strdup(x86ms->oem_table_id);
   1256}
   1257
   1258static void x86_machine_set_oem_table_id(Object *obj, const char *value,
   1259                                         Error **errp)
   1260{
   1261    X86MachineState *x86ms = X86_MACHINE(obj);
   1262    size_t len = strlen(value);
   1263
   1264    if (len > 8) {
   1265        error_setg(errp,
   1266                   "User specified "X86_MACHINE_OEM_TABLE_ID
   1267                   " value is bigger than "
   1268                   "8 bytes in size");
   1269        return;
   1270    }
   1271    strncpy(x86ms->oem_table_id, value, 8);
   1272}
   1273
   1274static void x86_machine_get_bus_lock_ratelimit(Object *obj, Visitor *v,
   1275                                const char *name, void *opaque, Error **errp)
   1276{
   1277    X86MachineState *x86ms = X86_MACHINE(obj);
   1278    uint64_t bus_lock_ratelimit = x86ms->bus_lock_ratelimit;
   1279
   1280    visit_type_uint64(v, name, &bus_lock_ratelimit, errp);
   1281}
   1282
   1283static void x86_machine_set_bus_lock_ratelimit(Object *obj, Visitor *v,
   1284                               const char *name, void *opaque, Error **errp)
   1285{
   1286    X86MachineState *x86ms = X86_MACHINE(obj);
   1287
   1288    visit_type_uint64(v, name, &x86ms->bus_lock_ratelimit, errp);
   1289}
   1290
   1291static void machine_get_sgx_epc(Object *obj, Visitor *v, const char *name,
   1292                                void *opaque, Error **errp)
   1293{
   1294    X86MachineState *x86ms = X86_MACHINE(obj);
   1295    SgxEPCList *list = x86ms->sgx_epc_list;
   1296
   1297    visit_type_SgxEPCList(v, name, &list, errp);
   1298}
   1299
   1300static void machine_set_sgx_epc(Object *obj, Visitor *v, const char *name,
   1301                                void *opaque, Error **errp)
   1302{
   1303    X86MachineState *x86ms = X86_MACHINE(obj);
   1304    SgxEPCList *list;
   1305
   1306    list = x86ms->sgx_epc_list;
   1307    visit_type_SgxEPCList(v, name, &x86ms->sgx_epc_list, errp);
   1308
   1309    qapi_free_SgxEPCList(list);
   1310}
   1311
   1312static void x86_machine_initfn(Object *obj)
   1313{
   1314    X86MachineState *x86ms = X86_MACHINE(obj);
   1315
   1316    x86ms->smm = ON_OFF_AUTO_AUTO;
   1317    x86ms->acpi = ON_OFF_AUTO_AUTO;
   1318    x86ms->pci_irq_mask = ACPI_BUILD_PCI_IRQS;
   1319    x86ms->oem_id = g_strndup(ACPI_BUILD_APPNAME6, 6);
   1320    x86ms->oem_table_id = g_strndup(ACPI_BUILD_APPNAME8, 8);
   1321    x86ms->bus_lock_ratelimit = 0;
   1322}
   1323
   1324static void x86_machine_class_init(ObjectClass *oc, void *data)
   1325{
   1326    MachineClass *mc = MACHINE_CLASS(oc);
   1327    X86MachineClass *x86mc = X86_MACHINE_CLASS(oc);
   1328    NMIClass *nc = NMI_CLASS(oc);
   1329
   1330    mc->cpu_index_to_instance_props = x86_cpu_index_to_props;
   1331    mc->get_default_cpu_node_id = x86_get_default_cpu_node_id;
   1332    mc->possible_cpu_arch_ids = x86_possible_cpu_arch_ids;
   1333    x86mc->compat_apic_id_mode = false;
   1334    x86mc->save_tsc_khz = true;
   1335    nc->nmi_monitor_handler = x86_nmi;
   1336
   1337    object_class_property_add(oc, X86_MACHINE_SMM, "OnOffAuto",
   1338        x86_machine_get_smm, x86_machine_set_smm,
   1339        NULL, NULL);
   1340    object_class_property_set_description(oc, X86_MACHINE_SMM,
   1341        "Enable SMM");
   1342
   1343    object_class_property_add(oc, X86_MACHINE_ACPI, "OnOffAuto",
   1344        x86_machine_get_acpi, x86_machine_set_acpi,
   1345        NULL, NULL);
   1346    object_class_property_set_description(oc, X86_MACHINE_ACPI,
   1347        "Enable ACPI");
   1348
   1349    object_class_property_add_str(oc, X86_MACHINE_OEM_ID,
   1350                                  x86_machine_get_oem_id,
   1351                                  x86_machine_set_oem_id);
   1352    object_class_property_set_description(oc, X86_MACHINE_OEM_ID,
   1353                                          "Override the default value of field OEMID "
   1354                                          "in ACPI table header."
   1355                                          "The string may be up to 6 bytes in size");
   1356
   1357
   1358    object_class_property_add_str(oc, X86_MACHINE_OEM_TABLE_ID,
   1359                                  x86_machine_get_oem_table_id,
   1360                                  x86_machine_set_oem_table_id);
   1361    object_class_property_set_description(oc, X86_MACHINE_OEM_TABLE_ID,
   1362                                          "Override the default value of field OEM Table ID "
   1363                                          "in ACPI table header."
   1364                                          "The string may be up to 8 bytes in size");
   1365
   1366    object_class_property_add(oc, X86_MACHINE_BUS_LOCK_RATELIMIT, "uint64_t",
   1367                                x86_machine_get_bus_lock_ratelimit,
   1368                                x86_machine_set_bus_lock_ratelimit, NULL, NULL);
   1369    object_class_property_set_description(oc, X86_MACHINE_BUS_LOCK_RATELIMIT,
   1370            "Set the ratelimit for the bus locks acquired in VMs");
   1371
   1372    object_class_property_add(oc, "sgx-epc", "SgxEPC",
   1373        machine_get_sgx_epc, machine_set_sgx_epc,
   1374        NULL, NULL);
   1375    object_class_property_set_description(oc, "sgx-epc",
   1376        "SGX EPC device");
   1377}
   1378
   1379static const TypeInfo x86_machine_info = {
   1380    .name = TYPE_X86_MACHINE,
   1381    .parent = TYPE_MACHINE,
   1382    .abstract = true,
   1383    .instance_size = sizeof(X86MachineState),
   1384    .instance_init = x86_machine_initfn,
   1385    .class_size = sizeof(X86MachineClass),
   1386    .class_init = x86_machine_class_init,
   1387    .interfaces = (InterfaceInfo[]) {
   1388         { TYPE_NMI },
   1389         { }
   1390    },
   1391};
   1392
   1393static void x86_machine_register_types(void)
   1394{
   1395    type_register_static(&x86_machine_info);
   1396}
   1397
   1398type_init(x86_machine_register_types)