1 /* 2 * QEMU PC System Emulator 3 * 4 * Copyright (c) 2003-2004 Fabrice Bellard 5 * 6 * Permission is hereby granted, free of charge, to any person obtaining a copy 7 * of this software and associated documentation files (the "Software"), to deal 8 * in the Software without restriction, including without limitation the rights 9 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell 10 * copies of the Software, and to permit persons to whom the Software is 11 * furnished to do so, subject to the following conditions: 12 * 13 * The above copyright notice and this permission notice shall be included in 14 * all copies or substantial portions of the Software. 15 * 16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR 17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, 18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL 19 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER 20 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, 21 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN 22 * THE SOFTWARE. 23 */ 24 #include "hw/hw.h" 25 #include "hw/i386/pc.h" 26 #include "hw/char/serial.h" 27 #include "hw/i386/apic.h" 28 #include "hw/i386/topology.h" 29 #include "sysemu/cpus.h" 30 #include "hw/block/fdc.h" 31 #include "hw/ide.h" 32 #include "hw/pci/pci.h" 33 #include "hw/pci/pci_bus.h" 34 #include "hw/nvram/fw_cfg.h" 35 #include "hw/timer/hpet.h" 36 #include "hw/smbios/smbios.h" 37 #include "hw/loader.h" 38 #include "elf.h" 39 #include "multiboot.h" 40 #include "hw/timer/mc146818rtc.h" 41 #include "hw/timer/i8254.h" 42 #include "hw/audio/pcspk.h" 43 #include "hw/pci/msi.h" 44 #include "hw/sysbus.h" 45 #include "sysemu/sysemu.h" 46 #include "sysemu/numa.h" 47 #include "sysemu/kvm.h" 48 #include "sysemu/qtest.h" 49 #include "kvm_i386.h" 50 #include "hw/xen/xen.h" 51 #include "sysemu/block-backend.h" 52 #include "hw/block/block.h" 53 #include "ui/qemu-spice.h" 54 #include "exec/memory.h" 55 #include "exec/address-spaces.h" 56 #include "sysemu/arch_init.h" 57 #include "qemu/bitmap.h" 58 #include "qemu/config-file.h" 59 #include "qemu/error-report.h" 60 #include "hw/acpi/acpi.h" 61 #include "hw/acpi/cpu_hotplug.h" 62 #include "hw/boards.h" 63 #include "hw/pci/pci_host.h" 64 #include "acpi-build.h" 65 #include "hw/mem/pc-dimm.h" 66 #include "qapi/visitor.h" 67 #include "qapi-visit.h" 68 69 /* debug PC/ISA interrupts */ 70 //#define DEBUG_IRQ 71 72 #ifdef DEBUG_IRQ 73 #define DPRINTF(fmt, ...) \ 74 do { printf("CPUIRQ: " fmt , ## __VA_ARGS__); } while (0) 75 #else 76 #define DPRINTF(fmt, ...) 77 #endif 78 79 /* Leave a chunk of memory at the top of RAM for the BIOS ACPI tables 80 * (128K) and other BIOS datastructures (less than 4K reported to be used at 81 * the moment, 32K should be enough for a while). */ 82 static unsigned acpi_data_size = 0x20000 + 0x8000; 83 void pc_set_legacy_acpi_data_size(void) 84 { 85 acpi_data_size = 0x10000; 86 } 87 88 #define BIOS_CFG_IOPORT 0x510 89 #define FW_CFG_ACPI_TABLES (FW_CFG_ARCH_LOCAL + 0) 90 #define FW_CFG_SMBIOS_ENTRIES (FW_CFG_ARCH_LOCAL + 1) 91 #define FW_CFG_IRQ0_OVERRIDE (FW_CFG_ARCH_LOCAL + 2) 92 #define FW_CFG_E820_TABLE (FW_CFG_ARCH_LOCAL + 3) 93 #define FW_CFG_HPET (FW_CFG_ARCH_LOCAL + 4) 94 95 #define E820_NR_ENTRIES 16 96 97 struct e820_entry { 98 uint64_t address; 99 uint64_t length; 100 uint32_t type; 101 } QEMU_PACKED __attribute((__aligned__(4))); 102 103 struct e820_table { 104 uint32_t count; 105 struct e820_entry entry[E820_NR_ENTRIES]; 106 } QEMU_PACKED __attribute((__aligned__(4))); 107 108 static struct e820_table e820_reserve; 109 static struct e820_entry *e820_table; 110 static unsigned e820_entries; 111 struct hpet_fw_config hpet_cfg = {.count = UINT8_MAX}; 112 113 void gsi_handler(void *opaque, int n, int level) 114 { 115 GSIState *s = opaque; 116 117 DPRINTF("pc: %s GSI %d\n", level ? "raising" : "lowering", n); 118 if (n < ISA_NUM_IRQS) { 119 qemu_set_irq(s->i8259_irq[n], level); 120 } 121 qemu_set_irq(s->ioapic_irq[n], level); 122 } 123 124 static void ioport80_write(void *opaque, hwaddr addr, uint64_t data, 125 unsigned size) 126 { 127 } 128 129 static uint64_t ioport80_read(void *opaque, hwaddr addr, unsigned size) 130 { 131 return 0xffffffffffffffffULL; 132 } 133 134 /* MSDOS compatibility mode FPU exception support */ 135 static qemu_irq ferr_irq; 136 137 void pc_register_ferr_irq(qemu_irq irq) 138 { 139 ferr_irq = irq; 140 } 141 142 /* XXX: add IGNNE support */ 143 void cpu_set_ferr(CPUX86State *s) 144 { 145 qemu_irq_raise(ferr_irq); 146 } 147 148 static void ioportF0_write(void *opaque, hwaddr addr, uint64_t data, 149 unsigned size) 150 { 151 qemu_irq_lower(ferr_irq); 152 } 153 154 static uint64_t ioportF0_read(void *opaque, hwaddr addr, unsigned size) 155 { 156 return 0xffffffffffffffffULL; 157 } 158 159 /* TSC handling */ 160 uint64_t cpu_get_tsc(CPUX86State *env) 161 { 162 return cpu_get_ticks(); 163 } 164 165 /* IRQ handling */ 166 int cpu_get_pic_interrupt(CPUX86State *env) 167 { 168 X86CPU *cpu = x86_env_get_cpu(env); 169 int intno; 170 171 intno = apic_get_interrupt(cpu->apic_state); 172 if (intno >= 0) { 173 return intno; 174 } 175 /* read the irq from the PIC */ 176 if (!apic_accept_pic_intr(cpu->apic_state)) { 177 return -1; 178 } 179 180 intno = pic_read_irq(isa_pic); 181 return intno; 182 } 183 184 static void pic_irq_request(void *opaque, int irq, int level) 185 { 186 CPUState *cs = first_cpu; 187 X86CPU *cpu = X86_CPU(cs); 188 189 DPRINTF("pic_irqs: %s irq %d\n", level? "raise" : "lower", irq); 190 if (cpu->apic_state) { 191 CPU_FOREACH(cs) { 192 cpu = X86_CPU(cs); 193 if (apic_accept_pic_intr(cpu->apic_state)) { 194 apic_deliver_pic_intr(cpu->apic_state, level); 195 } 196 } 197 } else { 198 if (level) { 199 cpu_interrupt(cs, CPU_INTERRUPT_HARD); 200 } else { 201 cpu_reset_interrupt(cs, CPU_INTERRUPT_HARD); 202 } 203 } 204 } 205 206 /* PC cmos mappings */ 207 208 #define REG_EQUIPMENT_BYTE 0x14 209 210 static int cmos_get_fd_drive_type(FDriveType fd0) 211 { 212 int val; 213 214 switch (fd0) { 215 case FDRIVE_DRV_144: 216 /* 1.44 Mb 3"5 drive */ 217 val = 4; 218 break; 219 case FDRIVE_DRV_288: 220 /* 2.88 Mb 3"5 drive */ 221 val = 5; 222 break; 223 case FDRIVE_DRV_120: 224 /* 1.2 Mb 5"5 drive */ 225 val = 2; 226 break; 227 case FDRIVE_DRV_NONE: 228 default: 229 val = 0; 230 break; 231 } 232 return val; 233 } 234 235 static void cmos_init_hd(ISADevice *s, int type_ofs, int info_ofs, 236 int16_t cylinders, int8_t heads, int8_t sectors) 237 { 238 rtc_set_memory(s, type_ofs, 47); 239 rtc_set_memory(s, info_ofs, cylinders); 240 rtc_set_memory(s, info_ofs + 1, cylinders >> 8); 241 rtc_set_memory(s, info_ofs + 2, heads); 242 rtc_set_memory(s, info_ofs + 3, 0xff); 243 rtc_set_memory(s, info_ofs + 4, 0xff); 244 rtc_set_memory(s, info_ofs + 5, 0xc0 | ((heads > 8) << 3)); 245 rtc_set_memory(s, info_ofs + 6, cylinders); 246 rtc_set_memory(s, info_ofs + 7, cylinders >> 8); 247 rtc_set_memory(s, info_ofs + 8, sectors); 248 } 249 250 /* convert boot_device letter to something recognizable by the bios */ 251 static int boot_device2nibble(char boot_device) 252 { 253 switch(boot_device) { 254 case 'a': 255 case 'b': 256 return 0x01; /* floppy boot */ 257 case 'c': 258 return 0x02; /* hard drive boot */ 259 case 'd': 260 return 0x03; /* CD-ROM boot */ 261 case 'n': 262 return 0x04; /* Network boot */ 263 } 264 return 0; 265 } 266 267 static void set_boot_dev(ISADevice *s, const char *boot_device, Error **errp) 268 { 269 #define PC_MAX_BOOT_DEVICES 3 270 int nbds, bds[3] = { 0, }; 271 int i; 272 273 nbds = strlen(boot_device); 274 if (nbds > PC_MAX_BOOT_DEVICES) { 275 error_setg(errp, "Too many boot devices for PC"); 276 return; 277 } 278 for (i = 0; i < nbds; i++) { 279 bds[i] = boot_device2nibble(boot_device[i]); 280 if (bds[i] == 0) { 281 error_setg(errp, "Invalid boot device for PC: '%c'", 282 boot_device[i]); 283 return; 284 } 285 } 286 rtc_set_memory(s, 0x3d, (bds[1] << 4) | bds[0]); 287 rtc_set_memory(s, 0x38, (bds[2] << 4) | (fd_bootchk ? 0x0 : 0x1)); 288 } 289 290 static void pc_boot_set(void *opaque, const char *boot_device, Error **errp) 291 { 292 set_boot_dev(opaque, boot_device, errp); 293 } 294 295 static void pc_cmos_init_floppy(ISADevice *rtc_state, ISADevice *floppy) 296 { 297 int val, nb, i; 298 FDriveType fd_type[2] = { FDRIVE_DRV_NONE, FDRIVE_DRV_NONE }; 299 300 /* floppy type */ 301 if (floppy) { 302 for (i = 0; i < 2; i++) { 303 fd_type[i] = isa_fdc_get_drive_type(floppy, i); 304 } 305 } 306 val = (cmos_get_fd_drive_type(fd_type[0]) << 4) | 307 cmos_get_fd_drive_type(fd_type[1]); 308 rtc_set_memory(rtc_state, 0x10, val); 309 310 val = rtc_get_memory(rtc_state, REG_EQUIPMENT_BYTE); 311 nb = 0; 312 if (fd_type[0] < FDRIVE_DRV_NONE) { 313 nb++; 314 } 315 if (fd_type[1] < FDRIVE_DRV_NONE) { 316 nb++; 317 } 318 switch (nb) { 319 case 0: 320 break; 321 case 1: 322 val |= 0x01; /* 1 drive, ready for boot */ 323 break; 324 case 2: 325 val |= 0x41; /* 2 drives, ready for boot */ 326 break; 327 } 328 rtc_set_memory(rtc_state, REG_EQUIPMENT_BYTE, val); 329 } 330 331 typedef struct pc_cmos_init_late_arg { 332 ISADevice *rtc_state; 333 BusState *idebus[2]; 334 } pc_cmos_init_late_arg; 335 336 typedef struct check_fdc_state { 337 ISADevice *floppy; 338 bool multiple; 339 } CheckFdcState; 340 341 static int check_fdc(Object *obj, void *opaque) 342 { 343 CheckFdcState *state = opaque; 344 Object *fdc; 345 uint32_t iobase; 346 Error *local_err = NULL; 347 348 fdc = object_dynamic_cast(obj, TYPE_ISA_FDC); 349 if (!fdc) { 350 return 0; 351 } 352 353 iobase = object_property_get_int(obj, "iobase", &local_err); 354 if (local_err || iobase != 0x3f0) { 355 error_free(local_err); 356 return 0; 357 } 358 359 if (state->floppy) { 360 state->multiple = true; 361 } else { 362 state->floppy = ISA_DEVICE(obj); 363 } 364 return 0; 365 } 366 367 static const char * const fdc_container_path[] = { 368 "/unattached", "/peripheral", "/peripheral-anon" 369 }; 370 371 static void pc_cmos_init_late(void *opaque) 372 { 373 pc_cmos_init_late_arg *arg = opaque; 374 ISADevice *s = arg->rtc_state; 375 int16_t cylinders; 376 int8_t heads, sectors; 377 int val; 378 int i, trans; 379 Object *container; 380 CheckFdcState state = { 0 }; 381 382 val = 0; 383 if (ide_get_geometry(arg->idebus[0], 0, 384 &cylinders, &heads, §ors) >= 0) { 385 cmos_init_hd(s, 0x19, 0x1b, cylinders, heads, sectors); 386 val |= 0xf0; 387 } 388 if (ide_get_geometry(arg->idebus[0], 1, 389 &cylinders, &heads, §ors) >= 0) { 390 cmos_init_hd(s, 0x1a, 0x24, cylinders, heads, sectors); 391 val |= 0x0f; 392 } 393 rtc_set_memory(s, 0x12, val); 394 395 val = 0; 396 for (i = 0; i < 4; i++) { 397 /* NOTE: ide_get_geometry() returns the physical 398 geometry. It is always such that: 1 <= sects <= 63, 1 399 <= heads <= 16, 1 <= cylinders <= 16383. The BIOS 400 geometry can be different if a translation is done. */ 401 if (ide_get_geometry(arg->idebus[i / 2], i % 2, 402 &cylinders, &heads, §ors) >= 0) { 403 trans = ide_get_bios_chs_trans(arg->idebus[i / 2], i % 2) - 1; 404 assert((trans & ~3) == 0); 405 val |= trans << (i * 2); 406 } 407 } 408 rtc_set_memory(s, 0x39, val); 409 410 /* 411 * Locate the FDC at IO address 0x3f0, and configure the CMOS registers 412 * accordingly. 413 */ 414 for (i = 0; i < ARRAY_SIZE(fdc_container_path); i++) { 415 container = container_get(qdev_get_machine(), fdc_container_path[i]); 416 object_child_foreach(container, check_fdc, &state); 417 } 418 419 if (state.multiple) { 420 error_report("warning: multiple floppy disk controllers with " 421 "iobase=0x3f0 have been found;\n" 422 "the one being picked for CMOS setup might not reflect " 423 "your intent"); 424 } 425 pc_cmos_init_floppy(s, state.floppy); 426 427 qemu_unregister_reset(pc_cmos_init_late, opaque); 428 } 429 430 void pc_cmos_init(PCMachineState *pcms, 431 BusState *idebus0, BusState *idebus1, 432 ISADevice *s) 433 { 434 int val; 435 static pc_cmos_init_late_arg arg; 436 Error *local_err = NULL; 437 438 /* various important CMOS locations needed by PC/Bochs bios */ 439 440 /* memory size */ 441 /* base memory (first MiB) */ 442 val = MIN(pcms->below_4g_mem_size / 1024, 640); 443 rtc_set_memory(s, 0x15, val); 444 rtc_set_memory(s, 0x16, val >> 8); 445 /* extended memory (next 64MiB) */ 446 if (pcms->below_4g_mem_size > 1024 * 1024) { 447 val = (pcms->below_4g_mem_size - 1024 * 1024) / 1024; 448 } else { 449 val = 0; 450 } 451 if (val > 65535) 452 val = 65535; 453 rtc_set_memory(s, 0x17, val); 454 rtc_set_memory(s, 0x18, val >> 8); 455 rtc_set_memory(s, 0x30, val); 456 rtc_set_memory(s, 0x31, val >> 8); 457 /* memory between 16MiB and 4GiB */ 458 if (pcms->below_4g_mem_size > 16 * 1024 * 1024) { 459 val = (pcms->below_4g_mem_size - 16 * 1024 * 1024) / 65536; 460 } else { 461 val = 0; 462 } 463 if (val > 65535) 464 val = 65535; 465 rtc_set_memory(s, 0x34, val); 466 rtc_set_memory(s, 0x35, val >> 8); 467 /* memory above 4GiB */ 468 val = pcms->above_4g_mem_size / 65536; 469 rtc_set_memory(s, 0x5b, val); 470 rtc_set_memory(s, 0x5c, val >> 8); 471 rtc_set_memory(s, 0x5d, val >> 16); 472 473 /* set the number of CPU */ 474 rtc_set_memory(s, 0x5f, smp_cpus - 1); 475 476 object_property_add_link(OBJECT(pcms), "rtc_state", 477 TYPE_ISA_DEVICE, 478 (Object **)&pcms->rtc, 479 object_property_allow_set_link, 480 OBJ_PROP_LINK_UNREF_ON_RELEASE, &error_abort); 481 object_property_set_link(OBJECT(pcms), OBJECT(s), 482 "rtc_state", &error_abort); 483 484 set_boot_dev(s, MACHINE(pcms)->boot_order, &local_err); 485 if (local_err) { 486 error_report_err(local_err); 487 exit(1); 488 } 489 490 val = 0; 491 val |= 0x02; /* FPU is there */ 492 val |= 0x04; /* PS/2 mouse installed */ 493 rtc_set_memory(s, REG_EQUIPMENT_BYTE, val); 494 495 /* hard drives and FDC */ 496 arg.rtc_state = s; 497 arg.idebus[0] = idebus0; 498 arg.idebus[1] = idebus1; 499 qemu_register_reset(pc_cmos_init_late, &arg); 500 } 501 502 #define TYPE_PORT92 "port92" 503 #define PORT92(obj) OBJECT_CHECK(Port92State, (obj), TYPE_PORT92) 504 505 /* port 92 stuff: could be split off */ 506 typedef struct Port92State { 507 ISADevice parent_obj; 508 509 MemoryRegion io; 510 uint8_t outport; 511 qemu_irq *a20_out; 512 } Port92State; 513 514 static void port92_write(void *opaque, hwaddr addr, uint64_t val, 515 unsigned size) 516 { 517 Port92State *s = opaque; 518 int oldval = s->outport; 519 520 DPRINTF("port92: write 0x%02" PRIx64 "\n", val); 521 s->outport = val; 522 qemu_set_irq(*s->a20_out, (val >> 1) & 1); 523 if ((val & 1) && !(oldval & 1)) { 524 qemu_system_reset_request(); 525 } 526 } 527 528 static uint64_t port92_read(void *opaque, hwaddr addr, 529 unsigned size) 530 { 531 Port92State *s = opaque; 532 uint32_t ret; 533 534 ret = s->outport; 535 DPRINTF("port92: read 0x%02x\n", ret); 536 return ret; 537 } 538 539 static void port92_init(ISADevice *dev, qemu_irq *a20_out) 540 { 541 Port92State *s = PORT92(dev); 542 543 s->a20_out = a20_out; 544 } 545 546 static const VMStateDescription vmstate_port92_isa = { 547 .name = "port92", 548 .version_id = 1, 549 .minimum_version_id = 1, 550 .fields = (VMStateField[]) { 551 VMSTATE_UINT8(outport, Port92State), 552 VMSTATE_END_OF_LIST() 553 } 554 }; 555 556 static void port92_reset(DeviceState *d) 557 { 558 Port92State *s = PORT92(d); 559 560 s->outport &= ~1; 561 } 562 563 static const MemoryRegionOps port92_ops = { 564 .read = port92_read, 565 .write = port92_write, 566 .impl = { 567 .min_access_size = 1, 568 .max_access_size = 1, 569 }, 570 .endianness = DEVICE_LITTLE_ENDIAN, 571 }; 572 573 static void port92_initfn(Object *obj) 574 { 575 Port92State *s = PORT92(obj); 576 577 memory_region_init_io(&s->io, OBJECT(s), &port92_ops, s, "port92", 1); 578 579 s->outport = 0; 580 } 581 582 static void port92_realizefn(DeviceState *dev, Error **errp) 583 { 584 ISADevice *isadev = ISA_DEVICE(dev); 585 Port92State *s = PORT92(dev); 586 587 isa_register_ioport(isadev, &s->io, 0x92); 588 } 589 590 static void port92_class_initfn(ObjectClass *klass, void *data) 591 { 592 DeviceClass *dc = DEVICE_CLASS(klass); 593 594 dc->realize = port92_realizefn; 595 dc->reset = port92_reset; 596 dc->vmsd = &vmstate_port92_isa; 597 /* 598 * Reason: unlike ordinary ISA devices, this one needs additional 599 * wiring: its A20 output line needs to be wired up by 600 * port92_init(). 601 */ 602 dc->cannot_instantiate_with_device_add_yet = true; 603 } 604 605 static const TypeInfo port92_info = { 606 .name = TYPE_PORT92, 607 .parent = TYPE_ISA_DEVICE, 608 .instance_size = sizeof(Port92State), 609 .instance_init = port92_initfn, 610 .class_init = port92_class_initfn, 611 }; 612 613 static void port92_register_types(void) 614 { 615 type_register_static(&port92_info); 616 } 617 618 type_init(port92_register_types) 619 620 static void handle_a20_line_change(void *opaque, int irq, int level) 621 { 622 X86CPU *cpu = opaque; 623 624 /* XXX: send to all CPUs ? */ 625 /* XXX: add logic to handle multiple A20 line sources */ 626 x86_cpu_set_a20(cpu, level); 627 } 628 629 int e820_add_entry(uint64_t address, uint64_t length, uint32_t type) 630 { 631 int index = le32_to_cpu(e820_reserve.count); 632 struct e820_entry *entry; 633 634 if (type != E820_RAM) { 635 /* old FW_CFG_E820_TABLE entry -- reservations only */ 636 if (index >= E820_NR_ENTRIES) { 637 return -EBUSY; 638 } 639 entry = &e820_reserve.entry[index++]; 640 641 entry->address = cpu_to_le64(address); 642 entry->length = cpu_to_le64(length); 643 entry->type = cpu_to_le32(type); 644 645 e820_reserve.count = cpu_to_le32(index); 646 } 647 648 /* new "etc/e820" file -- include ram too */ 649 e820_table = g_renew(struct e820_entry, e820_table, e820_entries + 1); 650 e820_table[e820_entries].address = cpu_to_le64(address); 651 e820_table[e820_entries].length = cpu_to_le64(length); 652 e820_table[e820_entries].type = cpu_to_le32(type); 653 e820_entries++; 654 655 return e820_entries; 656 } 657 658 int e820_get_num_entries(void) 659 { 660 return e820_entries; 661 } 662 663 bool e820_get_entry(int idx, uint32_t type, uint64_t *address, uint64_t *length) 664 { 665 if (idx < e820_entries && e820_table[idx].type == cpu_to_le32(type)) { 666 *address = le64_to_cpu(e820_table[idx].address); 667 *length = le64_to_cpu(e820_table[idx].length); 668 return true; 669 } 670 return false; 671 } 672 673 /* Enables contiguous-apic-ID mode, for compatibility */ 674 static bool compat_apic_id_mode; 675 676 void enable_compat_apic_id_mode(void) 677 { 678 compat_apic_id_mode = true; 679 } 680 681 /* Calculates initial APIC ID for a specific CPU index 682 * 683 * Currently we need to be able to calculate the APIC ID from the CPU index 684 * alone (without requiring a CPU object), as the QEMU<->Seabios interfaces have 685 * no concept of "CPU index", and the NUMA tables on fw_cfg need the APIC ID of 686 * all CPUs up to max_cpus. 687 */ 688 static uint32_t x86_cpu_apic_id_from_index(unsigned int cpu_index) 689 { 690 uint32_t correct_id; 691 static bool warned; 692 693 correct_id = x86_apicid_from_cpu_idx(smp_cores, smp_threads, cpu_index); 694 if (compat_apic_id_mode) { 695 if (cpu_index != correct_id && !warned && !qtest_enabled()) { 696 error_report("APIC IDs set in compatibility mode, " 697 "CPU topology won't match the configuration"); 698 warned = true; 699 } 700 return cpu_index; 701 } else { 702 return correct_id; 703 } 704 } 705 706 /* Calculates the limit to CPU APIC ID values 707 * 708 * This function returns the limit for the APIC ID value, so that all 709 * CPU APIC IDs are < pc_apic_id_limit(). 710 * 711 * This is used for FW_CFG_MAX_CPUS. See comments on bochs_bios_init(). 712 */ 713 static unsigned int pc_apic_id_limit(unsigned int max_cpus) 714 { 715 return x86_cpu_apic_id_from_index(max_cpus - 1) + 1; 716 } 717 718 static void pc_build_smbios(FWCfgState *fw_cfg) 719 { 720 uint8_t *smbios_tables, *smbios_anchor; 721 size_t smbios_tables_len, smbios_anchor_len; 722 struct smbios_phys_mem_area *mem_array; 723 unsigned i, array_count; 724 725 smbios_tables = smbios_get_table_legacy(&smbios_tables_len); 726 if (smbios_tables) { 727 fw_cfg_add_bytes(fw_cfg, FW_CFG_SMBIOS_ENTRIES, 728 smbios_tables, smbios_tables_len); 729 } 730 731 /* build the array of physical mem area from e820 table */ 732 mem_array = g_malloc0(sizeof(*mem_array) * e820_get_num_entries()); 733 for (i = 0, array_count = 0; i < e820_get_num_entries(); i++) { 734 uint64_t addr, len; 735 736 if (e820_get_entry(i, E820_RAM, &addr, &len)) { 737 mem_array[array_count].address = addr; 738 mem_array[array_count].length = len; 739 array_count++; 740 } 741 } 742 smbios_get_tables(mem_array, array_count, 743 &smbios_tables, &smbios_tables_len, 744 &smbios_anchor, &smbios_anchor_len); 745 g_free(mem_array); 746 747 if (smbios_anchor) { 748 fw_cfg_add_file(fw_cfg, "etc/smbios/smbios-tables", 749 smbios_tables, smbios_tables_len); 750 fw_cfg_add_file(fw_cfg, "etc/smbios/smbios-anchor", 751 smbios_anchor, smbios_anchor_len); 752 } 753 } 754 755 static FWCfgState *bochs_bios_init(void) 756 { 757 FWCfgState *fw_cfg; 758 uint64_t *numa_fw_cfg; 759 int i, j; 760 unsigned int apic_id_limit = pc_apic_id_limit(max_cpus); 761 762 fw_cfg = fw_cfg_init_io(BIOS_CFG_IOPORT); 763 /* FW_CFG_MAX_CPUS is a bit confusing/problematic on x86: 764 * 765 * SeaBIOS needs FW_CFG_MAX_CPUS for CPU hotplug, but the CPU hotplug 766 * QEMU<->SeaBIOS interface is not based on the "CPU index", but on the APIC 767 * ID of hotplugged CPUs[1]. This means that FW_CFG_MAX_CPUS is not the 768 * "maximum number of CPUs", but the "limit to the APIC ID values SeaBIOS 769 * may see". 770 * 771 * So, this means we must not use max_cpus, here, but the maximum possible 772 * APIC ID value, plus one. 773 * 774 * [1] The only kind of "CPU identifier" used between SeaBIOS and QEMU is 775 * the APIC ID, not the "CPU index" 776 */ 777 fw_cfg_add_i16(fw_cfg, FW_CFG_MAX_CPUS, (uint16_t)apic_id_limit); 778 fw_cfg_add_i64(fw_cfg, FW_CFG_RAM_SIZE, (uint64_t)ram_size); 779 fw_cfg_add_bytes(fw_cfg, FW_CFG_ACPI_TABLES, 780 acpi_tables, acpi_tables_len); 781 fw_cfg_add_i32(fw_cfg, FW_CFG_IRQ0_OVERRIDE, kvm_allows_irq0_override()); 782 783 pc_build_smbios(fw_cfg); 784 785 fw_cfg_add_bytes(fw_cfg, FW_CFG_E820_TABLE, 786 &e820_reserve, sizeof(e820_reserve)); 787 fw_cfg_add_file(fw_cfg, "etc/e820", e820_table, 788 sizeof(struct e820_entry) * e820_entries); 789 790 fw_cfg_add_bytes(fw_cfg, FW_CFG_HPET, &hpet_cfg, sizeof(hpet_cfg)); 791 /* allocate memory for the NUMA channel: one (64bit) word for the number 792 * of nodes, one word for each VCPU->node and one word for each node to 793 * hold the amount of memory. 794 */ 795 numa_fw_cfg = g_new0(uint64_t, 1 + apic_id_limit + nb_numa_nodes); 796 numa_fw_cfg[0] = cpu_to_le64(nb_numa_nodes); 797 for (i = 0; i < max_cpus; i++) { 798 unsigned int apic_id = x86_cpu_apic_id_from_index(i); 799 assert(apic_id < apic_id_limit); 800 for (j = 0; j < nb_numa_nodes; j++) { 801 if (test_bit(i, numa_info[j].node_cpu)) { 802 numa_fw_cfg[apic_id + 1] = cpu_to_le64(j); 803 break; 804 } 805 } 806 } 807 for (i = 0; i < nb_numa_nodes; i++) { 808 numa_fw_cfg[apic_id_limit + 1 + i] = cpu_to_le64(numa_info[i].node_mem); 809 } 810 fw_cfg_add_bytes(fw_cfg, FW_CFG_NUMA, numa_fw_cfg, 811 (1 + apic_id_limit + nb_numa_nodes) * 812 sizeof(*numa_fw_cfg)); 813 814 return fw_cfg; 815 } 816 817 static long get_file_size(FILE *f) 818 { 819 long where, size; 820 821 /* XXX: on Unix systems, using fstat() probably makes more sense */ 822 823 where = ftell(f); 824 fseek(f, 0, SEEK_END); 825 size = ftell(f); 826 fseek(f, where, SEEK_SET); 827 828 return size; 829 } 830 831 static void load_linux(PCMachineState *pcms, 832 FWCfgState *fw_cfg) 833 { 834 uint16_t protocol; 835 int setup_size, kernel_size, initrd_size = 0, cmdline_size; 836 uint32_t initrd_max; 837 uint8_t header[8192], *setup, *kernel, *initrd_data; 838 hwaddr real_addr, prot_addr, cmdline_addr, initrd_addr = 0; 839 FILE *f; 840 char *vmode; 841 MachineState *machine = MACHINE(pcms); 842 const char *kernel_filename = machine->kernel_filename; 843 const char *initrd_filename = machine->initrd_filename; 844 const char *kernel_cmdline = machine->kernel_cmdline; 845 846 /* Align to 16 bytes as a paranoia measure */ 847 cmdline_size = (strlen(kernel_cmdline)+16) & ~15; 848 849 /* load the kernel header */ 850 f = fopen(kernel_filename, "rb"); 851 if (!f || !(kernel_size = get_file_size(f)) || 852 fread(header, 1, MIN(ARRAY_SIZE(header), kernel_size), f) != 853 MIN(ARRAY_SIZE(header), kernel_size)) { 854 fprintf(stderr, "qemu: could not load kernel '%s': %s\n", 855 kernel_filename, strerror(errno)); 856 exit(1); 857 } 858 859 /* kernel protocol version */ 860 #if 0 861 fprintf(stderr, "header magic: %#x\n", ldl_p(header+0x202)); 862 #endif 863 if (ldl_p(header+0x202) == 0x53726448) { 864 protocol = lduw_p(header+0x206); 865 } else { 866 /* This looks like a multiboot kernel. If it is, let's stop 867 treating it like a Linux kernel. */ 868 if (load_multiboot(fw_cfg, f, kernel_filename, initrd_filename, 869 kernel_cmdline, kernel_size, header)) { 870 return; 871 } 872 protocol = 0; 873 } 874 875 if (protocol < 0x200 || !(header[0x211] & 0x01)) { 876 /* Low kernel */ 877 real_addr = 0x90000; 878 cmdline_addr = 0x9a000 - cmdline_size; 879 prot_addr = 0x10000; 880 } else if (protocol < 0x202) { 881 /* High but ancient kernel */ 882 real_addr = 0x90000; 883 cmdline_addr = 0x9a000 - cmdline_size; 884 prot_addr = 0x100000; 885 } else { 886 /* High and recent kernel */ 887 real_addr = 0x10000; 888 cmdline_addr = 0x20000; 889 prot_addr = 0x100000; 890 } 891 892 #if 0 893 fprintf(stderr, 894 "qemu: real_addr = 0x" TARGET_FMT_plx "\n" 895 "qemu: cmdline_addr = 0x" TARGET_FMT_plx "\n" 896 "qemu: prot_addr = 0x" TARGET_FMT_plx "\n", 897 real_addr, 898 cmdline_addr, 899 prot_addr); 900 #endif 901 902 /* highest address for loading the initrd */ 903 if (protocol >= 0x203) { 904 initrd_max = ldl_p(header+0x22c); 905 } else { 906 initrd_max = 0x37ffffff; 907 } 908 909 if (initrd_max >= pcms->below_4g_mem_size - acpi_data_size) { 910 initrd_max = pcms->below_4g_mem_size - acpi_data_size - 1; 911 } 912 913 fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_ADDR, cmdline_addr); 914 fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_SIZE, strlen(kernel_cmdline)+1); 915 fw_cfg_add_string(fw_cfg, FW_CFG_CMDLINE_DATA, kernel_cmdline); 916 917 if (protocol >= 0x202) { 918 stl_p(header+0x228, cmdline_addr); 919 } else { 920 stw_p(header+0x20, 0xA33F); 921 stw_p(header+0x22, cmdline_addr-real_addr); 922 } 923 924 /* handle vga= parameter */ 925 vmode = strstr(kernel_cmdline, "vga="); 926 if (vmode) { 927 unsigned int video_mode; 928 /* skip "vga=" */ 929 vmode += 4; 930 if (!strncmp(vmode, "normal", 6)) { 931 video_mode = 0xffff; 932 } else if (!strncmp(vmode, "ext", 3)) { 933 video_mode = 0xfffe; 934 } else if (!strncmp(vmode, "ask", 3)) { 935 video_mode = 0xfffd; 936 } else { 937 video_mode = strtol(vmode, NULL, 0); 938 } 939 stw_p(header+0x1fa, video_mode); 940 } 941 942 /* loader type */ 943 /* High nybble = B reserved for QEMU; low nybble is revision number. 944 If this code is substantially changed, you may want to consider 945 incrementing the revision. */ 946 if (protocol >= 0x200) { 947 header[0x210] = 0xB0; 948 } 949 /* heap */ 950 if (protocol >= 0x201) { 951 header[0x211] |= 0x80; /* CAN_USE_HEAP */ 952 stw_p(header+0x224, cmdline_addr-real_addr-0x200); 953 } 954 955 /* load initrd */ 956 if (initrd_filename) { 957 if (protocol < 0x200) { 958 fprintf(stderr, "qemu: linux kernel too old to load a ram disk\n"); 959 exit(1); 960 } 961 962 initrd_size = get_image_size(initrd_filename); 963 if (initrd_size < 0) { 964 fprintf(stderr, "qemu: error reading initrd %s: %s\n", 965 initrd_filename, strerror(errno)); 966 exit(1); 967 } 968 969 initrd_addr = (initrd_max-initrd_size) & ~4095; 970 971 initrd_data = g_malloc(initrd_size); 972 load_image(initrd_filename, initrd_data); 973 974 fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_ADDR, initrd_addr); 975 fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_SIZE, initrd_size); 976 fw_cfg_add_bytes(fw_cfg, FW_CFG_INITRD_DATA, initrd_data, initrd_size); 977 978 stl_p(header+0x218, initrd_addr); 979 stl_p(header+0x21c, initrd_size); 980 } 981 982 /* load kernel and setup */ 983 setup_size = header[0x1f1]; 984 if (setup_size == 0) { 985 setup_size = 4; 986 } 987 setup_size = (setup_size+1)*512; 988 kernel_size -= setup_size; 989 990 setup = g_malloc(setup_size); 991 kernel = g_malloc(kernel_size); 992 fseek(f, 0, SEEK_SET); 993 if (fread(setup, 1, setup_size, f) != setup_size) { 994 fprintf(stderr, "fread() failed\n"); 995 exit(1); 996 } 997 if (fread(kernel, 1, kernel_size, f) != kernel_size) { 998 fprintf(stderr, "fread() failed\n"); 999 exit(1); 1000 } 1001 fclose(f); 1002 memcpy(setup, header, MIN(sizeof(header), setup_size)); 1003 1004 fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ADDR, prot_addr); 1005 fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_SIZE, kernel_size); 1006 fw_cfg_add_bytes(fw_cfg, FW_CFG_KERNEL_DATA, kernel, kernel_size); 1007 1008 fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_ADDR, real_addr); 1009 fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_SIZE, setup_size); 1010 fw_cfg_add_bytes(fw_cfg, FW_CFG_SETUP_DATA, setup, setup_size); 1011 1012 option_rom[nb_option_roms].name = "linuxboot.bin"; 1013 option_rom[nb_option_roms].bootindex = 0; 1014 nb_option_roms++; 1015 } 1016 1017 #define NE2000_NB_MAX 6 1018 1019 static const int ne2000_io[NE2000_NB_MAX] = { 0x300, 0x320, 0x340, 0x360, 1020 0x280, 0x380 }; 1021 static const int ne2000_irq[NE2000_NB_MAX] = { 9, 10, 11, 3, 4, 5 }; 1022 1023 void pc_init_ne2k_isa(ISABus *bus, NICInfo *nd) 1024 { 1025 static int nb_ne2k = 0; 1026 1027 if (nb_ne2k == NE2000_NB_MAX) 1028 return; 1029 isa_ne2000_init(bus, ne2000_io[nb_ne2k], 1030 ne2000_irq[nb_ne2k], nd); 1031 nb_ne2k++; 1032 } 1033 1034 DeviceState *cpu_get_current_apic(void) 1035 { 1036 if (current_cpu) { 1037 X86CPU *cpu = X86_CPU(current_cpu); 1038 return cpu->apic_state; 1039 } else { 1040 return NULL; 1041 } 1042 } 1043 1044 void pc_acpi_smi_interrupt(void *opaque, int irq, int level) 1045 { 1046 X86CPU *cpu = opaque; 1047 1048 if (level) { 1049 cpu_interrupt(CPU(cpu), CPU_INTERRUPT_SMI); 1050 } 1051 } 1052 1053 static X86CPU *pc_new_cpu(const char *cpu_model, int64_t apic_id, 1054 Error **errp) 1055 { 1056 X86CPU *cpu = NULL; 1057 Error *local_err = NULL; 1058 1059 cpu = cpu_x86_create(cpu_model, &local_err); 1060 if (local_err != NULL) { 1061 goto out; 1062 } 1063 1064 object_property_set_int(OBJECT(cpu), apic_id, "apic-id", &local_err); 1065 object_property_set_bool(OBJECT(cpu), true, "realized", &local_err); 1066 1067 out: 1068 if (local_err) { 1069 error_propagate(errp, local_err); 1070 object_unref(OBJECT(cpu)); 1071 cpu = NULL; 1072 } 1073 return cpu; 1074 } 1075 1076 static const char *current_cpu_model; 1077 1078 void pc_hot_add_cpu(const int64_t id, Error **errp) 1079 { 1080 X86CPU *cpu; 1081 int64_t apic_id = x86_cpu_apic_id_from_index(id); 1082 Error *local_err = NULL; 1083 1084 if (id < 0) { 1085 error_setg(errp, "Invalid CPU id: %" PRIi64, id); 1086 return; 1087 } 1088 1089 if (cpu_exists(apic_id)) { 1090 error_setg(errp, "Unable to add CPU: %" PRIi64 1091 ", it already exists", id); 1092 return; 1093 } 1094 1095 if (id >= max_cpus) { 1096 error_setg(errp, "Unable to add CPU: %" PRIi64 1097 ", max allowed: %d", id, max_cpus - 1); 1098 return; 1099 } 1100 1101 if (apic_id >= ACPI_CPU_HOTPLUG_ID_LIMIT) { 1102 error_setg(errp, "Unable to add CPU: %" PRIi64 1103 ", resulting APIC ID (%" PRIi64 ") is too large", 1104 id, apic_id); 1105 return; 1106 } 1107 1108 cpu = pc_new_cpu(current_cpu_model, apic_id, &local_err); 1109 if (local_err) { 1110 error_propagate(errp, local_err); 1111 return; 1112 } 1113 object_unref(OBJECT(cpu)); 1114 } 1115 1116 void pc_cpus_init(const char *cpu_model) 1117 { 1118 int i; 1119 X86CPU *cpu = NULL; 1120 Error *error = NULL; 1121 unsigned long apic_id_limit; 1122 1123 /* init CPUs */ 1124 if (cpu_model == NULL) { 1125 #ifdef TARGET_X86_64 1126 cpu_model = "qemu64"; 1127 #else 1128 cpu_model = "qemu32"; 1129 #endif 1130 } 1131 current_cpu_model = cpu_model; 1132 1133 apic_id_limit = pc_apic_id_limit(max_cpus); 1134 if (apic_id_limit > ACPI_CPU_HOTPLUG_ID_LIMIT) { 1135 error_report("max_cpus is too large. APIC ID of last CPU is %lu", 1136 apic_id_limit - 1); 1137 exit(1); 1138 } 1139 1140 for (i = 0; i < smp_cpus; i++) { 1141 cpu = pc_new_cpu(cpu_model, x86_cpu_apic_id_from_index(i), 1142 &error); 1143 if (error) { 1144 error_report_err(error); 1145 exit(1); 1146 } 1147 object_unref(OBJECT(cpu)); 1148 } 1149 1150 /* tell smbios about cpuid version and features */ 1151 smbios_set_cpuid(cpu->env.cpuid_version, cpu->env.features[FEAT_1_EDX]); 1152 } 1153 1154 /* pci-info ROM file. Little endian format */ 1155 typedef struct PcRomPciInfo { 1156 uint64_t w32_min; 1157 uint64_t w32_max; 1158 uint64_t w64_min; 1159 uint64_t w64_max; 1160 } PcRomPciInfo; 1161 1162 typedef struct PcGuestInfoState { 1163 PcGuestInfo info; 1164 Notifier machine_done; 1165 } PcGuestInfoState; 1166 1167 static 1168 void pc_guest_info_machine_done(Notifier *notifier, void *data) 1169 { 1170 PcGuestInfoState *guest_info_state = container_of(notifier, 1171 PcGuestInfoState, 1172 machine_done); 1173 PCIBus *bus = find_i440fx(); 1174 1175 if (bus) { 1176 int extra_hosts = 0; 1177 1178 QLIST_FOREACH(bus, &bus->child, sibling) { 1179 /* look for expander root buses */ 1180 if (pci_bus_is_root(bus)) { 1181 extra_hosts++; 1182 } 1183 } 1184 if (extra_hosts && guest_info_state->info.fw_cfg) { 1185 uint64_t *val = g_malloc(sizeof(*val)); 1186 *val = cpu_to_le64(extra_hosts); 1187 fw_cfg_add_file(guest_info_state->info.fw_cfg, 1188 "etc/extra-pci-roots", val, sizeof(*val)); 1189 } 1190 } 1191 1192 acpi_setup(&guest_info_state->info); 1193 } 1194 1195 PcGuestInfo *pc_guest_info_init(PCMachineState *pcms) 1196 { 1197 PcGuestInfoState *guest_info_state = g_malloc0(sizeof *guest_info_state); 1198 PcGuestInfo *guest_info = &guest_info_state->info; 1199 int i, j; 1200 1201 guest_info->ram_size_below_4g = pcms->below_4g_mem_size; 1202 guest_info->ram_size = pcms->below_4g_mem_size + pcms->above_4g_mem_size; 1203 guest_info->apic_id_limit = pc_apic_id_limit(max_cpus); 1204 guest_info->apic_xrupt_override = kvm_allows_irq0_override(); 1205 guest_info->numa_nodes = nb_numa_nodes; 1206 guest_info->node_mem = g_malloc0(guest_info->numa_nodes * 1207 sizeof *guest_info->node_mem); 1208 for (i = 0; i < nb_numa_nodes; i++) { 1209 guest_info->node_mem[i] = numa_info[i].node_mem; 1210 } 1211 1212 guest_info->node_cpu = g_malloc0(guest_info->apic_id_limit * 1213 sizeof *guest_info->node_cpu); 1214 1215 for (i = 0; i < max_cpus; i++) { 1216 unsigned int apic_id = x86_cpu_apic_id_from_index(i); 1217 assert(apic_id < guest_info->apic_id_limit); 1218 for (j = 0; j < nb_numa_nodes; j++) { 1219 if (test_bit(i, numa_info[j].node_cpu)) { 1220 guest_info->node_cpu[apic_id] = j; 1221 break; 1222 } 1223 } 1224 } 1225 1226 guest_info_state->machine_done.notify = pc_guest_info_machine_done; 1227 qemu_add_machine_init_done_notifier(&guest_info_state->machine_done); 1228 return guest_info; 1229 } 1230 1231 /* setup pci memory address space mapping into system address space */ 1232 void pc_pci_as_mapping_init(Object *owner, MemoryRegion *system_memory, 1233 MemoryRegion *pci_address_space) 1234 { 1235 /* Set to lower priority than RAM */ 1236 memory_region_add_subregion_overlap(system_memory, 0x0, 1237 pci_address_space, -1); 1238 } 1239 1240 void pc_acpi_init(const char *default_dsdt) 1241 { 1242 char *filename; 1243 1244 if (acpi_tables != NULL) { 1245 /* manually set via -acpitable, leave it alone */ 1246 return; 1247 } 1248 1249 filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, default_dsdt); 1250 if (filename == NULL) { 1251 fprintf(stderr, "WARNING: failed to find %s\n", default_dsdt); 1252 } else { 1253 QemuOpts *opts = qemu_opts_create(qemu_find_opts("acpi"), NULL, 0, 1254 &error_abort); 1255 Error *err = NULL; 1256 1257 qemu_opt_set(opts, "file", filename, &error_abort); 1258 1259 acpi_table_add_builtin(opts, &err); 1260 if (err) { 1261 error_report("WARNING: failed to load %s: %s", filename, 1262 error_get_pretty(err)); 1263 error_free(err); 1264 } 1265 g_free(filename); 1266 } 1267 } 1268 1269 FWCfgState *xen_load_linux(PCMachineState *pcms, 1270 PcGuestInfo *guest_info) 1271 { 1272 int i; 1273 FWCfgState *fw_cfg; 1274 1275 assert(MACHINE(pcms)->kernel_filename != NULL); 1276 1277 fw_cfg = fw_cfg_init_io(BIOS_CFG_IOPORT); 1278 rom_set_fw(fw_cfg); 1279 1280 load_linux(pcms, fw_cfg); 1281 for (i = 0; i < nb_option_roms; i++) { 1282 assert(!strcmp(option_rom[i].name, "linuxboot.bin") || 1283 !strcmp(option_rom[i].name, "multiboot.bin")); 1284 rom_add_option(option_rom[i].name, option_rom[i].bootindex); 1285 } 1286 guest_info->fw_cfg = fw_cfg; 1287 return fw_cfg; 1288 } 1289 1290 FWCfgState *pc_memory_init(PCMachineState *pcms, 1291 MemoryRegion *system_memory, 1292 MemoryRegion *rom_memory, 1293 MemoryRegion **ram_memory, 1294 PcGuestInfo *guest_info) 1295 { 1296 int linux_boot, i; 1297 MemoryRegion *ram, *option_rom_mr; 1298 MemoryRegion *ram_below_4g, *ram_above_4g; 1299 FWCfgState *fw_cfg; 1300 MachineState *machine = MACHINE(pcms); 1301 1302 assert(machine->ram_size == pcms->below_4g_mem_size + 1303 pcms->above_4g_mem_size); 1304 1305 linux_boot = (machine->kernel_filename != NULL); 1306 1307 /* Allocate RAM. We allocate it as a single memory region and use 1308 * aliases to address portions of it, mostly for backwards compatibility 1309 * with older qemus that used qemu_ram_alloc(). 1310 */ 1311 ram = g_malloc(sizeof(*ram)); 1312 memory_region_allocate_system_memory(ram, NULL, "pc.ram", 1313 machine->ram_size); 1314 *ram_memory = ram; 1315 ram_below_4g = g_malloc(sizeof(*ram_below_4g)); 1316 memory_region_init_alias(ram_below_4g, NULL, "ram-below-4g", ram, 1317 0, pcms->below_4g_mem_size); 1318 memory_region_add_subregion(system_memory, 0, ram_below_4g); 1319 e820_add_entry(0, pcms->below_4g_mem_size, E820_RAM); 1320 if (pcms->above_4g_mem_size > 0) { 1321 ram_above_4g = g_malloc(sizeof(*ram_above_4g)); 1322 memory_region_init_alias(ram_above_4g, NULL, "ram-above-4g", ram, 1323 pcms->below_4g_mem_size, 1324 pcms->above_4g_mem_size); 1325 memory_region_add_subregion(system_memory, 0x100000000ULL, 1326 ram_above_4g); 1327 e820_add_entry(0x100000000ULL, pcms->above_4g_mem_size, E820_RAM); 1328 } 1329 1330 if (!guest_info->has_reserved_memory && 1331 (machine->ram_slots || 1332 (machine->maxram_size > machine->ram_size))) { 1333 MachineClass *mc = MACHINE_GET_CLASS(machine); 1334 1335 error_report("\"-memory 'slots|maxmem'\" is not supported by: %s", 1336 mc->name); 1337 exit(EXIT_FAILURE); 1338 } 1339 1340 /* initialize hotplug memory address space */ 1341 if (guest_info->has_reserved_memory && 1342 (machine->ram_size < machine->maxram_size)) { 1343 ram_addr_t hotplug_mem_size = 1344 machine->maxram_size - machine->ram_size; 1345 1346 if (machine->ram_slots > ACPI_MAX_RAM_SLOTS) { 1347 error_report("unsupported amount of memory slots: %"PRIu64, 1348 machine->ram_slots); 1349 exit(EXIT_FAILURE); 1350 } 1351 1352 if (QEMU_ALIGN_UP(machine->maxram_size, 1353 TARGET_PAGE_SIZE) != machine->maxram_size) { 1354 error_report("maximum memory size must by aligned to multiple of " 1355 "%d bytes", TARGET_PAGE_SIZE); 1356 exit(EXIT_FAILURE); 1357 } 1358 1359 pcms->hotplug_memory.base = 1360 ROUND_UP(0x100000000ULL + pcms->above_4g_mem_size, 1ULL << 30); 1361 1362 if (pcms->enforce_aligned_dimm) { 1363 /* size hotplug region assuming 1G page max alignment per slot */ 1364 hotplug_mem_size += (1ULL << 30) * machine->ram_slots; 1365 } 1366 1367 if ((pcms->hotplug_memory.base + hotplug_mem_size) < 1368 hotplug_mem_size) { 1369 error_report("unsupported amount of maximum memory: " RAM_ADDR_FMT, 1370 machine->maxram_size); 1371 exit(EXIT_FAILURE); 1372 } 1373 1374 memory_region_init(&pcms->hotplug_memory.mr, OBJECT(pcms), 1375 "hotplug-memory", hotplug_mem_size); 1376 memory_region_add_subregion(system_memory, pcms->hotplug_memory.base, 1377 &pcms->hotplug_memory.mr); 1378 } 1379 1380 /* Initialize PC system firmware */ 1381 pc_system_firmware_init(rom_memory, guest_info->isapc_ram_fw); 1382 1383 option_rom_mr = g_malloc(sizeof(*option_rom_mr)); 1384 memory_region_init_ram(option_rom_mr, NULL, "pc.rom", PC_ROM_SIZE, 1385 &error_fatal); 1386 vmstate_register_ram_global(option_rom_mr); 1387 memory_region_add_subregion_overlap(rom_memory, 1388 PC_ROM_MIN_VGA, 1389 option_rom_mr, 1390 1); 1391 1392 fw_cfg = bochs_bios_init(); 1393 rom_set_fw(fw_cfg); 1394 1395 if (guest_info->has_reserved_memory && pcms->hotplug_memory.base) { 1396 uint64_t *val = g_malloc(sizeof(*val)); 1397 PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms); 1398 uint64_t res_mem_end = pcms->hotplug_memory.base; 1399 1400 if (!pcmc->broken_reserved_end) { 1401 res_mem_end += memory_region_size(&pcms->hotplug_memory.mr); 1402 } 1403 *val = cpu_to_le64(ROUND_UP(res_mem_end, 0x1ULL << 30)); 1404 fw_cfg_add_file(fw_cfg, "etc/reserved-memory-end", val, sizeof(*val)); 1405 } 1406 1407 if (linux_boot) { 1408 load_linux(pcms, fw_cfg); 1409 } 1410 1411 for (i = 0; i < nb_option_roms; i++) { 1412 rom_add_option(option_rom[i].name, option_rom[i].bootindex); 1413 } 1414 guest_info->fw_cfg = fw_cfg; 1415 return fw_cfg; 1416 } 1417 1418 qemu_irq pc_allocate_cpu_irq(void) 1419 { 1420 return qemu_allocate_irq(pic_irq_request, NULL, 0); 1421 } 1422 1423 DeviceState *pc_vga_init(ISABus *isa_bus, PCIBus *pci_bus) 1424 { 1425 DeviceState *dev = NULL; 1426 1427 if (pci_bus) { 1428 PCIDevice *pcidev = pci_vga_init(pci_bus); 1429 dev = pcidev ? &pcidev->qdev : NULL; 1430 } else if (isa_bus) { 1431 ISADevice *isadev = isa_vga_init(isa_bus); 1432 dev = isadev ? DEVICE(isadev) : NULL; 1433 } 1434 return dev; 1435 } 1436 1437 static const MemoryRegionOps ioport80_io_ops = { 1438 .write = ioport80_write, 1439 .read = ioport80_read, 1440 .endianness = DEVICE_NATIVE_ENDIAN, 1441 .impl = { 1442 .min_access_size = 1, 1443 .max_access_size = 1, 1444 }, 1445 }; 1446 1447 static const MemoryRegionOps ioportF0_io_ops = { 1448 .write = ioportF0_write, 1449 .read = ioportF0_read, 1450 .endianness = DEVICE_NATIVE_ENDIAN, 1451 .impl = { 1452 .min_access_size = 1, 1453 .max_access_size = 1, 1454 }, 1455 }; 1456 1457 void pc_basic_device_init(ISABus *isa_bus, qemu_irq *gsi, 1458 ISADevice **rtc_state, 1459 bool create_fdctrl, 1460 bool no_vmport, 1461 uint32 hpet_irqs) 1462 { 1463 int i; 1464 DriveInfo *fd[MAX_FD]; 1465 DeviceState *hpet = NULL; 1466 int pit_isa_irq = 0; 1467 qemu_irq pit_alt_irq = NULL; 1468 qemu_irq rtc_irq = NULL; 1469 qemu_irq *a20_line; 1470 ISADevice *i8042, *port92, *vmmouse, *pit = NULL; 1471 MemoryRegion *ioport80_io = g_new(MemoryRegion, 1); 1472 MemoryRegion *ioportF0_io = g_new(MemoryRegion, 1); 1473 1474 memory_region_init_io(ioport80_io, NULL, &ioport80_io_ops, NULL, "ioport80", 1); 1475 memory_region_add_subregion(isa_bus->address_space_io, 0x80, ioport80_io); 1476 1477 memory_region_init_io(ioportF0_io, NULL, &ioportF0_io_ops, NULL, "ioportF0", 1); 1478 memory_region_add_subregion(isa_bus->address_space_io, 0xf0, ioportF0_io); 1479 1480 /* 1481 * Check if an HPET shall be created. 1482 * 1483 * Without KVM_CAP_PIT_STATE2, we cannot switch off the in-kernel PIT 1484 * when the HPET wants to take over. Thus we have to disable the latter. 1485 */ 1486 if (!no_hpet && (!kvm_irqchip_in_kernel() || kvm_has_pit_state2())) { 1487 /* In order to set property, here not using sysbus_try_create_simple */ 1488 hpet = qdev_try_create(NULL, TYPE_HPET); 1489 if (hpet) { 1490 /* For pc-piix-*, hpet's intcap is always IRQ2. For pc-q35-1.7 1491 * and earlier, use IRQ2 for compat. Otherwise, use IRQ16~23, 1492 * IRQ8 and IRQ2. 1493 */ 1494 uint8_t compat = object_property_get_int(OBJECT(hpet), 1495 HPET_INTCAP, NULL); 1496 if (!compat) { 1497 qdev_prop_set_uint32(hpet, HPET_INTCAP, hpet_irqs); 1498 } 1499 qdev_init_nofail(hpet); 1500 sysbus_mmio_map(SYS_BUS_DEVICE(hpet), 0, HPET_BASE); 1501 1502 for (i = 0; i < GSI_NUM_PINS; i++) { 1503 sysbus_connect_irq(SYS_BUS_DEVICE(hpet), i, gsi[i]); 1504 } 1505 pit_isa_irq = -1; 1506 pit_alt_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_PIT_INT); 1507 rtc_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_RTC_INT); 1508 } 1509 } 1510 *rtc_state = rtc_init(isa_bus, 2000, rtc_irq); 1511 1512 qemu_register_boot_set(pc_boot_set, *rtc_state); 1513 1514 if (!xen_enabled()) { 1515 if (kvm_irqchip_in_kernel()) { 1516 pit = kvm_pit_init(isa_bus, 0x40); 1517 } else { 1518 pit = pit_init(isa_bus, 0x40, pit_isa_irq, pit_alt_irq); 1519 } 1520 if (hpet) { 1521 /* connect PIT to output control line of the HPET */ 1522 qdev_connect_gpio_out(hpet, 0, qdev_get_gpio_in(DEVICE(pit), 0)); 1523 } 1524 pcspk_init(isa_bus, pit); 1525 } 1526 1527 serial_hds_isa_init(isa_bus, MAX_SERIAL_PORTS); 1528 parallel_hds_isa_init(isa_bus, MAX_PARALLEL_PORTS); 1529 1530 a20_line = qemu_allocate_irqs(handle_a20_line_change, first_cpu, 2); 1531 i8042 = isa_create_simple(isa_bus, "i8042"); 1532 i8042_setup_a20_line(i8042, &a20_line[0]); 1533 if (!no_vmport) { 1534 vmport_init(isa_bus); 1535 vmmouse = isa_try_create(isa_bus, "vmmouse"); 1536 } else { 1537 vmmouse = NULL; 1538 } 1539 if (vmmouse) { 1540 DeviceState *dev = DEVICE(vmmouse); 1541 qdev_prop_set_ptr(dev, "ps2_mouse", i8042); 1542 qdev_init_nofail(dev); 1543 } 1544 port92 = isa_create_simple(isa_bus, "port92"); 1545 port92_init(port92, &a20_line[1]); 1546 1547 DMA_init(0); 1548 1549 for(i = 0; i < MAX_FD; i++) { 1550 fd[i] = drive_get(IF_FLOPPY, 0, i); 1551 create_fdctrl |= !!fd[i]; 1552 } 1553 if (create_fdctrl) { 1554 fdctrl_init_isa(isa_bus, fd); 1555 } 1556 } 1557 1558 void pc_nic_init(ISABus *isa_bus, PCIBus *pci_bus) 1559 { 1560 int i; 1561 1562 for (i = 0; i < nb_nics; i++) { 1563 NICInfo *nd = &nd_table[i]; 1564 1565 if (!pci_bus || (nd->model && strcmp(nd->model, "ne2k_isa") == 0)) { 1566 pc_init_ne2k_isa(isa_bus, nd); 1567 } else { 1568 pci_nic_init_nofail(nd, pci_bus, "e1000", NULL); 1569 } 1570 } 1571 } 1572 1573 void pc_pci_device_init(PCIBus *pci_bus) 1574 { 1575 int max_bus; 1576 int bus; 1577 1578 max_bus = drive_get_max_bus(IF_SCSI); 1579 for (bus = 0; bus <= max_bus; bus++) { 1580 pci_create_simple(pci_bus, -1, "lsi53c895a"); 1581 } 1582 } 1583 1584 void ioapic_init_gsi(GSIState *gsi_state, const char *parent_name) 1585 { 1586 DeviceState *dev; 1587 SysBusDevice *d; 1588 unsigned int i; 1589 1590 if (kvm_irqchip_in_kernel()) { 1591 dev = qdev_create(NULL, "kvm-ioapic"); 1592 } else { 1593 dev = qdev_create(NULL, "ioapic"); 1594 } 1595 if (parent_name) { 1596 object_property_add_child(object_resolve_path(parent_name, NULL), 1597 "ioapic", OBJECT(dev), NULL); 1598 } 1599 qdev_init_nofail(dev); 1600 d = SYS_BUS_DEVICE(dev); 1601 sysbus_mmio_map(d, 0, IO_APIC_DEFAULT_ADDRESS); 1602 1603 for (i = 0; i < IOAPIC_NUM_PINS; i++) { 1604 gsi_state->ioapic_irq[i] = qdev_get_gpio_in(dev, i); 1605 } 1606 } 1607 1608 static void pc_dimm_plug(HotplugHandler *hotplug_dev, 1609 DeviceState *dev, Error **errp) 1610 { 1611 HotplugHandlerClass *hhc; 1612 Error *local_err = NULL; 1613 PCMachineState *pcms = PC_MACHINE(hotplug_dev); 1614 PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms); 1615 PCDIMMDevice *dimm = PC_DIMM(dev); 1616 PCDIMMDeviceClass *ddc = PC_DIMM_GET_CLASS(dimm); 1617 MemoryRegion *mr = ddc->get_memory_region(dimm); 1618 uint64_t align = TARGET_PAGE_SIZE; 1619 1620 if (memory_region_get_alignment(mr) && pcms->enforce_aligned_dimm) { 1621 align = memory_region_get_alignment(mr); 1622 } 1623 1624 if (!pcms->acpi_dev) { 1625 error_setg(&local_err, 1626 "memory hotplug is not enabled: missing acpi device"); 1627 goto out; 1628 } 1629 1630 pc_dimm_memory_plug(dev, &pcms->hotplug_memory, mr, align, 1631 pcmc->inter_dimm_gap, &local_err); 1632 if (local_err) { 1633 goto out; 1634 } 1635 1636 hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev); 1637 hhc->plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &error_abort); 1638 out: 1639 error_propagate(errp, local_err); 1640 } 1641 1642 static void pc_dimm_unplug_request(HotplugHandler *hotplug_dev, 1643 DeviceState *dev, Error **errp) 1644 { 1645 HotplugHandlerClass *hhc; 1646 Error *local_err = NULL; 1647 PCMachineState *pcms = PC_MACHINE(hotplug_dev); 1648 1649 if (!pcms->acpi_dev) { 1650 error_setg(&local_err, 1651 "memory hotplug is not enabled: missing acpi device"); 1652 goto out; 1653 } 1654 1655 hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev); 1656 hhc->unplug_request(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err); 1657 1658 out: 1659 error_propagate(errp, local_err); 1660 } 1661 1662 static void pc_dimm_unplug(HotplugHandler *hotplug_dev, 1663 DeviceState *dev, Error **errp) 1664 { 1665 PCMachineState *pcms = PC_MACHINE(hotplug_dev); 1666 PCDIMMDevice *dimm = PC_DIMM(dev); 1667 PCDIMMDeviceClass *ddc = PC_DIMM_GET_CLASS(dimm); 1668 MemoryRegion *mr = ddc->get_memory_region(dimm); 1669 HotplugHandlerClass *hhc; 1670 Error *local_err = NULL; 1671 1672 hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev); 1673 hhc->unplug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err); 1674 1675 if (local_err) { 1676 goto out; 1677 } 1678 1679 pc_dimm_memory_unplug(dev, &pcms->hotplug_memory, mr); 1680 object_unparent(OBJECT(dev)); 1681 1682 out: 1683 error_propagate(errp, local_err); 1684 } 1685 1686 static void pc_cpu_plug(HotplugHandler *hotplug_dev, 1687 DeviceState *dev, Error **errp) 1688 { 1689 HotplugHandlerClass *hhc; 1690 Error *local_err = NULL; 1691 PCMachineState *pcms = PC_MACHINE(hotplug_dev); 1692 1693 if (!dev->hotplugged) { 1694 goto out; 1695 } 1696 1697 if (!pcms->acpi_dev) { 1698 error_setg(&local_err, 1699 "cpu hotplug is not enabled: missing acpi device"); 1700 goto out; 1701 } 1702 1703 hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev); 1704 hhc->plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err); 1705 if (local_err) { 1706 goto out; 1707 } 1708 1709 /* increment the number of CPUs */ 1710 rtc_set_memory(pcms->rtc, 0x5f, rtc_get_memory(pcms->rtc, 0x5f) + 1); 1711 out: 1712 error_propagate(errp, local_err); 1713 } 1714 1715 static void pc_machine_device_plug_cb(HotplugHandler *hotplug_dev, 1716 DeviceState *dev, Error **errp) 1717 { 1718 if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) { 1719 pc_dimm_plug(hotplug_dev, dev, errp); 1720 } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) { 1721 pc_cpu_plug(hotplug_dev, dev, errp); 1722 } 1723 } 1724 1725 static void pc_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev, 1726 DeviceState *dev, Error **errp) 1727 { 1728 if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) { 1729 pc_dimm_unplug_request(hotplug_dev, dev, errp); 1730 } else { 1731 error_setg(errp, "acpi: device unplug request for not supported device" 1732 " type: %s", object_get_typename(OBJECT(dev))); 1733 } 1734 } 1735 1736 static void pc_machine_device_unplug_cb(HotplugHandler *hotplug_dev, 1737 DeviceState *dev, Error **errp) 1738 { 1739 if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) { 1740 pc_dimm_unplug(hotplug_dev, dev, errp); 1741 } else { 1742 error_setg(errp, "acpi: device unplug for not supported device" 1743 " type: %s", object_get_typename(OBJECT(dev))); 1744 } 1745 } 1746 1747 static HotplugHandler *pc_get_hotpug_handler(MachineState *machine, 1748 DeviceState *dev) 1749 { 1750 PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(machine); 1751 1752 if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) || 1753 object_dynamic_cast(OBJECT(dev), TYPE_CPU)) { 1754 return HOTPLUG_HANDLER(machine); 1755 } 1756 1757 return pcmc->get_hotplug_handler ? 1758 pcmc->get_hotplug_handler(machine, dev) : NULL; 1759 } 1760 1761 static void 1762 pc_machine_get_hotplug_memory_region_size(Object *obj, Visitor *v, void *opaque, 1763 const char *name, Error **errp) 1764 { 1765 PCMachineState *pcms = PC_MACHINE(obj); 1766 int64_t value = memory_region_size(&pcms->hotplug_memory.mr); 1767 1768 visit_type_int(v, &value, name, errp); 1769 } 1770 1771 static void pc_machine_get_max_ram_below_4g(Object *obj, Visitor *v, 1772 void *opaque, const char *name, 1773 Error **errp) 1774 { 1775 PCMachineState *pcms = PC_MACHINE(obj); 1776 uint64_t value = pcms->max_ram_below_4g; 1777 1778 visit_type_size(v, &value, name, errp); 1779 } 1780 1781 static void pc_machine_set_max_ram_below_4g(Object *obj, Visitor *v, 1782 void *opaque, const char *name, 1783 Error **errp) 1784 { 1785 PCMachineState *pcms = PC_MACHINE(obj); 1786 Error *error = NULL; 1787 uint64_t value; 1788 1789 visit_type_size(v, &value, name, &error); 1790 if (error) { 1791 error_propagate(errp, error); 1792 return; 1793 } 1794 if (value > (1ULL << 32)) { 1795 error_set(&error, ERROR_CLASS_GENERIC_ERROR, 1796 "Machine option 'max-ram-below-4g=%"PRIu64 1797 "' expects size less than or equal to 4G", value); 1798 error_propagate(errp, error); 1799 return; 1800 } 1801 1802 if (value < (1ULL << 20)) { 1803 error_report("Warning: small max_ram_below_4g(%"PRIu64 1804 ") less than 1M. BIOS may not work..", 1805 value); 1806 } 1807 1808 pcms->max_ram_below_4g = value; 1809 } 1810 1811 static void pc_machine_get_vmport(Object *obj, Visitor *v, void *opaque, 1812 const char *name, Error **errp) 1813 { 1814 PCMachineState *pcms = PC_MACHINE(obj); 1815 OnOffAuto vmport = pcms->vmport; 1816 1817 visit_type_OnOffAuto(v, &vmport, name, errp); 1818 } 1819 1820 static void pc_machine_set_vmport(Object *obj, Visitor *v, void *opaque, 1821 const char *name, Error **errp) 1822 { 1823 PCMachineState *pcms = PC_MACHINE(obj); 1824 1825 visit_type_OnOffAuto(v, &pcms->vmport, name, errp); 1826 } 1827 1828 bool pc_machine_is_smm_enabled(PCMachineState *pcms) 1829 { 1830 bool smm_available = false; 1831 1832 if (pcms->smm == ON_OFF_AUTO_OFF) { 1833 return false; 1834 } 1835 1836 if (tcg_enabled() || qtest_enabled()) { 1837 smm_available = true; 1838 } else if (kvm_enabled()) { 1839 smm_available = kvm_has_smm(); 1840 } 1841 1842 if (smm_available) { 1843 return true; 1844 } 1845 1846 if (pcms->smm == ON_OFF_AUTO_ON) { 1847 error_report("System Management Mode not supported by this hypervisor."); 1848 exit(1); 1849 } 1850 return false; 1851 } 1852 1853 static void pc_machine_get_smm(Object *obj, Visitor *v, void *opaque, 1854 const char *name, Error **errp) 1855 { 1856 PCMachineState *pcms = PC_MACHINE(obj); 1857 OnOffAuto smm = pcms->smm; 1858 1859 visit_type_OnOffAuto(v, &smm, name, errp); 1860 } 1861 1862 static void pc_machine_set_smm(Object *obj, Visitor *v, void *opaque, 1863 const char *name, Error **errp) 1864 { 1865 PCMachineState *pcms = PC_MACHINE(obj); 1866 1867 visit_type_OnOffAuto(v, &pcms->smm, name, errp); 1868 } 1869 1870 static bool pc_machine_get_aligned_dimm(Object *obj, Error **errp) 1871 { 1872 PCMachineState *pcms = PC_MACHINE(obj); 1873 1874 return pcms->enforce_aligned_dimm; 1875 } 1876 1877 static void pc_machine_initfn(Object *obj) 1878 { 1879 PCMachineState *pcms = PC_MACHINE(obj); 1880 1881 object_property_add(obj, PC_MACHINE_MEMHP_REGION_SIZE, "int", 1882 pc_machine_get_hotplug_memory_region_size, 1883 NULL, NULL, NULL, &error_abort); 1884 1885 pcms->max_ram_below_4g = 1ULL << 32; /* 4G */ 1886 object_property_add(obj, PC_MACHINE_MAX_RAM_BELOW_4G, "size", 1887 pc_machine_get_max_ram_below_4g, 1888 pc_machine_set_max_ram_below_4g, 1889 NULL, NULL, &error_abort); 1890 object_property_set_description(obj, PC_MACHINE_MAX_RAM_BELOW_4G, 1891 "Maximum ram below the 4G boundary (32bit boundary)", 1892 &error_abort); 1893 1894 pcms->smm = ON_OFF_AUTO_AUTO; 1895 object_property_add(obj, PC_MACHINE_SMM, "OnOffAuto", 1896 pc_machine_get_smm, 1897 pc_machine_set_smm, 1898 NULL, NULL, &error_abort); 1899 object_property_set_description(obj, PC_MACHINE_SMM, 1900 "Enable SMM (pc & q35)", 1901 &error_abort); 1902 1903 pcms->vmport = ON_OFF_AUTO_AUTO; 1904 object_property_add(obj, PC_MACHINE_VMPORT, "OnOffAuto", 1905 pc_machine_get_vmport, 1906 pc_machine_set_vmport, 1907 NULL, NULL, &error_abort); 1908 object_property_set_description(obj, PC_MACHINE_VMPORT, 1909 "Enable vmport (pc & q35)", 1910 &error_abort); 1911 1912 pcms->enforce_aligned_dimm = true; 1913 object_property_add_bool(obj, PC_MACHINE_ENFORCE_ALIGNED_DIMM, 1914 pc_machine_get_aligned_dimm, 1915 NULL, &error_abort); 1916 } 1917 1918 static void pc_machine_reset(void) 1919 { 1920 CPUState *cs; 1921 X86CPU *cpu; 1922 1923 qemu_devices_reset(); 1924 1925 /* Reset APIC after devices have been reset to cancel 1926 * any changes that qemu_devices_reset() might have done. 1927 */ 1928 CPU_FOREACH(cs) { 1929 cpu = X86_CPU(cs); 1930 1931 if (cpu->apic_state) { 1932 device_reset(cpu->apic_state); 1933 } 1934 } 1935 } 1936 1937 static unsigned pc_cpu_index_to_socket_id(unsigned cpu_index) 1938 { 1939 X86CPUTopoInfo topo; 1940 x86_topo_ids_from_idx(smp_cores, smp_threads, cpu_index, 1941 &topo); 1942 return topo.pkg_id; 1943 } 1944 1945 static void pc_machine_class_init(ObjectClass *oc, void *data) 1946 { 1947 MachineClass *mc = MACHINE_CLASS(oc); 1948 PCMachineClass *pcmc = PC_MACHINE_CLASS(oc); 1949 HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc); 1950 1951 pcmc->inter_dimm_gap = true; 1952 pcmc->get_hotplug_handler = mc->get_hotplug_handler; 1953 mc->get_hotplug_handler = pc_get_hotpug_handler; 1954 mc->cpu_index_to_socket_id = pc_cpu_index_to_socket_id; 1955 mc->default_boot_order = "cad"; 1956 mc->hot_add_cpu = pc_hot_add_cpu; 1957 mc->max_cpus = 255; 1958 mc->reset = pc_machine_reset; 1959 hc->plug = pc_machine_device_plug_cb; 1960 hc->unplug_request = pc_machine_device_unplug_request_cb; 1961 hc->unplug = pc_machine_device_unplug_cb; 1962 } 1963 1964 static const TypeInfo pc_machine_info = { 1965 .name = TYPE_PC_MACHINE, 1966 .parent = TYPE_MACHINE, 1967 .abstract = true, 1968 .instance_size = sizeof(PCMachineState), 1969 .instance_init = pc_machine_initfn, 1970 .class_size = sizeof(PCMachineClass), 1971 .class_init = pc_machine_class_init, 1972 .interfaces = (InterfaceInfo[]) { 1973 { TYPE_HOTPLUG_HANDLER }, 1974 { } 1975 }, 1976 }; 1977 1978 static void pc_machine_register_types(void) 1979 { 1980 type_register_static(&pc_machine_info); 1981 } 1982 1983 type_init(pc_machine_register_types) 1984