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 "qemu/osdep.h" 25 #include "hw/hw.h" 26 #include "hw/i386/pc.h" 27 #include "hw/char/serial.h" 28 #include "hw/i386/apic.h" 29 #include "hw/i386/topology.h" 30 #include "sysemu/cpus.h" 31 #include "hw/block/fdc.h" 32 #include "hw/ide.h" 33 #include "hw/pci/pci.h" 34 #include "hw/pci/pci_bus.h" 35 #include "hw/nvram/fw_cfg.h" 36 #include "hw/timer/hpet.h" 37 #include "hw/smbios/smbios.h" 38 #include "hw/loader.h" 39 #include "elf.h" 40 #include "multiboot.h" 41 #include "hw/timer/mc146818rtc.h" 42 #include "hw/timer/i8254.h" 43 #include "hw/audio/pcspk.h" 44 #include "hw/pci/msi.h" 45 #include "hw/sysbus.h" 46 #include "sysemu/sysemu.h" 47 #include "sysemu/numa.h" 48 #include "sysemu/kvm.h" 49 #include "sysemu/qtest.h" 50 #include "kvm_i386.h" 51 #include "hw/xen/xen.h" 52 #include "sysemu/block-backend.h" 53 #include "hw/block/block.h" 54 #include "ui/qemu-spice.h" 55 #include "exec/memory.h" 56 #include "exec/address-spaces.h" 57 #include "sysemu/arch_init.h" 58 #include "qemu/bitmap.h" 59 #include "qemu/config-file.h" 60 #include "qemu/error-report.h" 61 #include "hw/acpi/acpi.h" 62 #include "hw/acpi/cpu_hotplug.h" 63 #include "hw/boards.h" 64 #include "hw/pci/pci_host.h" 65 #include "acpi-build.h" 66 #include "hw/mem/pc-dimm.h" 67 #include "qapi/visitor.h" 68 #include "qapi-visit.h" 69 #include "qom/cpu.h" 70 #include "hw/nmi.h" 71 72 /* debug PC/ISA interrupts */ 73 //#define DEBUG_IRQ 74 75 #ifdef DEBUG_IRQ 76 #define DPRINTF(fmt, ...) \ 77 do { printf("CPUIRQ: " fmt , ## __VA_ARGS__); } while (0) 78 #else 79 #define DPRINTF(fmt, ...) 80 #endif 81 82 #define FW_CFG_ACPI_TABLES (FW_CFG_ARCH_LOCAL + 0) 83 #define FW_CFG_SMBIOS_ENTRIES (FW_CFG_ARCH_LOCAL + 1) 84 #define FW_CFG_IRQ0_OVERRIDE (FW_CFG_ARCH_LOCAL + 2) 85 #define FW_CFG_E820_TABLE (FW_CFG_ARCH_LOCAL + 3) 86 #define FW_CFG_HPET (FW_CFG_ARCH_LOCAL + 4) 87 88 #define E820_NR_ENTRIES 16 89 90 struct e820_entry { 91 uint64_t address; 92 uint64_t length; 93 uint32_t type; 94 } QEMU_PACKED __attribute((__aligned__(4))); 95 96 struct e820_table { 97 uint32_t count; 98 struct e820_entry entry[E820_NR_ENTRIES]; 99 } QEMU_PACKED __attribute((__aligned__(4))); 100 101 static struct e820_table e820_reserve; 102 static struct e820_entry *e820_table; 103 static unsigned e820_entries; 104 struct hpet_fw_config hpet_cfg = {.count = UINT8_MAX}; 105 106 void gsi_handler(void *opaque, int n, int level) 107 { 108 GSIState *s = opaque; 109 110 DPRINTF("pc: %s GSI %d\n", level ? "raising" : "lowering", n); 111 if (n < ISA_NUM_IRQS) { 112 qemu_set_irq(s->i8259_irq[n], level); 113 } 114 qemu_set_irq(s->ioapic_irq[n], level); 115 } 116 117 static void ioport80_write(void *opaque, hwaddr addr, uint64_t data, 118 unsigned size) 119 { 120 } 121 122 static uint64_t ioport80_read(void *opaque, hwaddr addr, unsigned size) 123 { 124 return 0xffffffffffffffffULL; 125 } 126 127 /* MSDOS compatibility mode FPU exception support */ 128 static qemu_irq ferr_irq; 129 130 void pc_register_ferr_irq(qemu_irq irq) 131 { 132 ferr_irq = irq; 133 } 134 135 /* XXX: add IGNNE support */ 136 void cpu_set_ferr(CPUX86State *s) 137 { 138 qemu_irq_raise(ferr_irq); 139 } 140 141 static void ioportF0_write(void *opaque, hwaddr addr, uint64_t data, 142 unsigned size) 143 { 144 qemu_irq_lower(ferr_irq); 145 } 146 147 static uint64_t ioportF0_read(void *opaque, hwaddr addr, unsigned size) 148 { 149 return 0xffffffffffffffffULL; 150 } 151 152 /* TSC handling */ 153 uint64_t cpu_get_tsc(CPUX86State *env) 154 { 155 return cpu_get_ticks(); 156 } 157 158 /* IRQ handling */ 159 int cpu_get_pic_interrupt(CPUX86State *env) 160 { 161 X86CPU *cpu = x86_env_get_cpu(env); 162 int intno; 163 164 intno = apic_get_interrupt(cpu->apic_state); 165 if (intno >= 0) { 166 return intno; 167 } 168 /* read the irq from the PIC */ 169 if (!apic_accept_pic_intr(cpu->apic_state)) { 170 return -1; 171 } 172 173 intno = pic_read_irq(isa_pic); 174 return intno; 175 } 176 177 static void pic_irq_request(void *opaque, int irq, int level) 178 { 179 CPUState *cs = first_cpu; 180 X86CPU *cpu = X86_CPU(cs); 181 182 DPRINTF("pic_irqs: %s irq %d\n", level? "raise" : "lower", irq); 183 if (cpu->apic_state) { 184 CPU_FOREACH(cs) { 185 cpu = X86_CPU(cs); 186 if (apic_accept_pic_intr(cpu->apic_state)) { 187 apic_deliver_pic_intr(cpu->apic_state, level); 188 } 189 } 190 } else { 191 if (level) { 192 cpu_interrupt(cs, CPU_INTERRUPT_HARD); 193 } else { 194 cpu_reset_interrupt(cs, CPU_INTERRUPT_HARD); 195 } 196 } 197 } 198 199 /* PC cmos mappings */ 200 201 #define REG_EQUIPMENT_BYTE 0x14 202 203 int cmos_get_fd_drive_type(FloppyDriveType fd0) 204 { 205 int val; 206 207 switch (fd0) { 208 case FLOPPY_DRIVE_TYPE_144: 209 /* 1.44 Mb 3"5 drive */ 210 val = 4; 211 break; 212 case FLOPPY_DRIVE_TYPE_288: 213 /* 2.88 Mb 3"5 drive */ 214 val = 5; 215 break; 216 case FLOPPY_DRIVE_TYPE_120: 217 /* 1.2 Mb 5"5 drive */ 218 val = 2; 219 break; 220 case FLOPPY_DRIVE_TYPE_NONE: 221 default: 222 val = 0; 223 break; 224 } 225 return val; 226 } 227 228 static void cmos_init_hd(ISADevice *s, int type_ofs, int info_ofs, 229 int16_t cylinders, int8_t heads, int8_t sectors) 230 { 231 rtc_set_memory(s, type_ofs, 47); 232 rtc_set_memory(s, info_ofs, cylinders); 233 rtc_set_memory(s, info_ofs + 1, cylinders >> 8); 234 rtc_set_memory(s, info_ofs + 2, heads); 235 rtc_set_memory(s, info_ofs + 3, 0xff); 236 rtc_set_memory(s, info_ofs + 4, 0xff); 237 rtc_set_memory(s, info_ofs + 5, 0xc0 | ((heads > 8) << 3)); 238 rtc_set_memory(s, info_ofs + 6, cylinders); 239 rtc_set_memory(s, info_ofs + 7, cylinders >> 8); 240 rtc_set_memory(s, info_ofs + 8, sectors); 241 } 242 243 /* convert boot_device letter to something recognizable by the bios */ 244 static int boot_device2nibble(char boot_device) 245 { 246 switch(boot_device) { 247 case 'a': 248 case 'b': 249 return 0x01; /* floppy boot */ 250 case 'c': 251 return 0x02; /* hard drive boot */ 252 case 'd': 253 return 0x03; /* CD-ROM boot */ 254 case 'n': 255 return 0x04; /* Network boot */ 256 } 257 return 0; 258 } 259 260 static void set_boot_dev(ISADevice *s, const char *boot_device, Error **errp) 261 { 262 #define PC_MAX_BOOT_DEVICES 3 263 int nbds, bds[3] = { 0, }; 264 int i; 265 266 nbds = strlen(boot_device); 267 if (nbds > PC_MAX_BOOT_DEVICES) { 268 error_setg(errp, "Too many boot devices for PC"); 269 return; 270 } 271 for (i = 0; i < nbds; i++) { 272 bds[i] = boot_device2nibble(boot_device[i]); 273 if (bds[i] == 0) { 274 error_setg(errp, "Invalid boot device for PC: '%c'", 275 boot_device[i]); 276 return; 277 } 278 } 279 rtc_set_memory(s, 0x3d, (bds[1] << 4) | bds[0]); 280 rtc_set_memory(s, 0x38, (bds[2] << 4) | (fd_bootchk ? 0x0 : 0x1)); 281 } 282 283 static void pc_boot_set(void *opaque, const char *boot_device, Error **errp) 284 { 285 set_boot_dev(opaque, boot_device, errp); 286 } 287 288 static void pc_cmos_init_floppy(ISADevice *rtc_state, ISADevice *floppy) 289 { 290 int val, nb, i; 291 FloppyDriveType fd_type[2] = { FLOPPY_DRIVE_TYPE_NONE, 292 FLOPPY_DRIVE_TYPE_NONE }; 293 294 /* floppy type */ 295 if (floppy) { 296 for (i = 0; i < 2; i++) { 297 fd_type[i] = isa_fdc_get_drive_type(floppy, i); 298 } 299 } 300 val = (cmos_get_fd_drive_type(fd_type[0]) << 4) | 301 cmos_get_fd_drive_type(fd_type[1]); 302 rtc_set_memory(rtc_state, 0x10, val); 303 304 val = rtc_get_memory(rtc_state, REG_EQUIPMENT_BYTE); 305 nb = 0; 306 if (fd_type[0] != FLOPPY_DRIVE_TYPE_NONE) { 307 nb++; 308 } 309 if (fd_type[1] != FLOPPY_DRIVE_TYPE_NONE) { 310 nb++; 311 } 312 switch (nb) { 313 case 0: 314 break; 315 case 1: 316 val |= 0x01; /* 1 drive, ready for boot */ 317 break; 318 case 2: 319 val |= 0x41; /* 2 drives, ready for boot */ 320 break; 321 } 322 rtc_set_memory(rtc_state, REG_EQUIPMENT_BYTE, val); 323 } 324 325 typedef struct pc_cmos_init_late_arg { 326 ISADevice *rtc_state; 327 BusState *idebus[2]; 328 } pc_cmos_init_late_arg; 329 330 typedef struct check_fdc_state { 331 ISADevice *floppy; 332 bool multiple; 333 } CheckFdcState; 334 335 static int check_fdc(Object *obj, void *opaque) 336 { 337 CheckFdcState *state = opaque; 338 Object *fdc; 339 uint32_t iobase; 340 Error *local_err = NULL; 341 342 fdc = object_dynamic_cast(obj, TYPE_ISA_FDC); 343 if (!fdc) { 344 return 0; 345 } 346 347 iobase = object_property_get_int(obj, "iobase", &local_err); 348 if (local_err || iobase != 0x3f0) { 349 error_free(local_err); 350 return 0; 351 } 352 353 if (state->floppy) { 354 state->multiple = true; 355 } else { 356 state->floppy = ISA_DEVICE(obj); 357 } 358 return 0; 359 } 360 361 static const char * const fdc_container_path[] = { 362 "/unattached", "/peripheral", "/peripheral-anon" 363 }; 364 365 /* 366 * Locate the FDC at IO address 0x3f0, in order to configure the CMOS registers 367 * and ACPI objects. 368 */ 369 ISADevice *pc_find_fdc0(void) 370 { 371 int i; 372 Object *container; 373 CheckFdcState state = { 0 }; 374 375 for (i = 0; i < ARRAY_SIZE(fdc_container_path); i++) { 376 container = container_get(qdev_get_machine(), fdc_container_path[i]); 377 object_child_foreach(container, check_fdc, &state); 378 } 379 380 if (state.multiple) { 381 error_report("warning: multiple floppy disk controllers with " 382 "iobase=0x3f0 have been found"); 383 error_printf("the one being picked for CMOS setup might not reflect " 384 "your intent"); 385 } 386 387 return state.floppy; 388 } 389 390 static void pc_cmos_init_late(void *opaque) 391 { 392 pc_cmos_init_late_arg *arg = opaque; 393 ISADevice *s = arg->rtc_state; 394 int16_t cylinders; 395 int8_t heads, sectors; 396 int val; 397 int i, trans; 398 399 val = 0; 400 if (ide_get_geometry(arg->idebus[0], 0, 401 &cylinders, &heads, §ors) >= 0) { 402 cmos_init_hd(s, 0x19, 0x1b, cylinders, heads, sectors); 403 val |= 0xf0; 404 } 405 if (ide_get_geometry(arg->idebus[0], 1, 406 &cylinders, &heads, §ors) >= 0) { 407 cmos_init_hd(s, 0x1a, 0x24, cylinders, heads, sectors); 408 val |= 0x0f; 409 } 410 rtc_set_memory(s, 0x12, val); 411 412 val = 0; 413 for (i = 0; i < 4; i++) { 414 /* NOTE: ide_get_geometry() returns the physical 415 geometry. It is always such that: 1 <= sects <= 63, 1 416 <= heads <= 16, 1 <= cylinders <= 16383. The BIOS 417 geometry can be different if a translation is done. */ 418 if (ide_get_geometry(arg->idebus[i / 2], i % 2, 419 &cylinders, &heads, §ors) >= 0) { 420 trans = ide_get_bios_chs_trans(arg->idebus[i / 2], i % 2) - 1; 421 assert((trans & ~3) == 0); 422 val |= trans << (i * 2); 423 } 424 } 425 rtc_set_memory(s, 0x39, val); 426 427 pc_cmos_init_floppy(s, pc_find_fdc0()); 428 429 qemu_unregister_reset(pc_cmos_init_late, opaque); 430 } 431 432 void pc_cmos_init(PCMachineState *pcms, 433 BusState *idebus0, BusState *idebus1, 434 ISADevice *s) 435 { 436 int val; 437 static pc_cmos_init_late_arg arg; 438 439 /* various important CMOS locations needed by PC/Bochs bios */ 440 441 /* memory size */ 442 /* base memory (first MiB) */ 443 val = MIN(pcms->below_4g_mem_size / 1024, 640); 444 rtc_set_memory(s, 0x15, val); 445 rtc_set_memory(s, 0x16, val >> 8); 446 /* extended memory (next 64MiB) */ 447 if (pcms->below_4g_mem_size > 1024 * 1024) { 448 val = (pcms->below_4g_mem_size - 1024 * 1024) / 1024; 449 } else { 450 val = 0; 451 } 452 if (val > 65535) 453 val = 65535; 454 rtc_set_memory(s, 0x17, val); 455 rtc_set_memory(s, 0x18, val >> 8); 456 rtc_set_memory(s, 0x30, val); 457 rtc_set_memory(s, 0x31, val >> 8); 458 /* memory between 16MiB and 4GiB */ 459 if (pcms->below_4g_mem_size > 16 * 1024 * 1024) { 460 val = (pcms->below_4g_mem_size - 16 * 1024 * 1024) / 65536; 461 } else { 462 val = 0; 463 } 464 if (val > 65535) 465 val = 65535; 466 rtc_set_memory(s, 0x34, val); 467 rtc_set_memory(s, 0x35, val >> 8); 468 /* memory above 4GiB */ 469 val = pcms->above_4g_mem_size / 65536; 470 rtc_set_memory(s, 0x5b, val); 471 rtc_set_memory(s, 0x5c, val >> 8); 472 rtc_set_memory(s, 0x5d, val >> 16); 473 474 /* set the number of CPU */ 475 rtc_set_memory(s, 0x5f, smp_cpus - 1); 476 477 object_property_add_link(OBJECT(pcms), "rtc_state", 478 TYPE_ISA_DEVICE, 479 (Object **)&pcms->rtc, 480 object_property_allow_set_link, 481 OBJ_PROP_LINK_UNREF_ON_RELEASE, &error_abort); 482 object_property_set_link(OBJECT(pcms), OBJECT(s), 483 "rtc_state", &error_abort); 484 485 set_boot_dev(s, MACHINE(pcms)->boot_order, &error_fatal); 486 487 val = 0; 488 val |= 0x02; /* FPU is there */ 489 val |= 0x04; /* PS/2 mouse installed */ 490 rtc_set_memory(s, REG_EQUIPMENT_BYTE, val); 491 492 /* hard drives and FDC */ 493 arg.rtc_state = s; 494 arg.idebus[0] = idebus0; 495 arg.idebus[1] = idebus1; 496 qemu_register_reset(pc_cmos_init_late, &arg); 497 } 498 499 #define TYPE_PORT92 "port92" 500 #define PORT92(obj) OBJECT_CHECK(Port92State, (obj), TYPE_PORT92) 501 502 /* port 92 stuff: could be split off */ 503 typedef struct Port92State { 504 ISADevice parent_obj; 505 506 MemoryRegion io; 507 uint8_t outport; 508 qemu_irq a20_out; 509 } Port92State; 510 511 static void port92_write(void *opaque, hwaddr addr, uint64_t val, 512 unsigned size) 513 { 514 Port92State *s = opaque; 515 int oldval = s->outport; 516 517 DPRINTF("port92: write 0x%02" PRIx64 "\n", val); 518 s->outport = val; 519 qemu_set_irq(s->a20_out, (val >> 1) & 1); 520 if ((val & 1) && !(oldval & 1)) { 521 qemu_system_reset_request(); 522 } 523 } 524 525 static uint64_t port92_read(void *opaque, hwaddr addr, 526 unsigned size) 527 { 528 Port92State *s = opaque; 529 uint32_t ret; 530 531 ret = s->outport; 532 DPRINTF("port92: read 0x%02x\n", ret); 533 return ret; 534 } 535 536 static void port92_init(ISADevice *dev, qemu_irq *a20_out) 537 { 538 qdev_connect_gpio_out_named(DEVICE(dev), PORT92_A20_LINE, 0, *a20_out); 539 } 540 541 static const VMStateDescription vmstate_port92_isa = { 542 .name = "port92", 543 .version_id = 1, 544 .minimum_version_id = 1, 545 .fields = (VMStateField[]) { 546 VMSTATE_UINT8(outport, Port92State), 547 VMSTATE_END_OF_LIST() 548 } 549 }; 550 551 static void port92_reset(DeviceState *d) 552 { 553 Port92State *s = PORT92(d); 554 555 s->outport &= ~1; 556 } 557 558 static const MemoryRegionOps port92_ops = { 559 .read = port92_read, 560 .write = port92_write, 561 .impl = { 562 .min_access_size = 1, 563 .max_access_size = 1, 564 }, 565 .endianness = DEVICE_LITTLE_ENDIAN, 566 }; 567 568 static void port92_initfn(Object *obj) 569 { 570 Port92State *s = PORT92(obj); 571 572 memory_region_init_io(&s->io, OBJECT(s), &port92_ops, s, "port92", 1); 573 574 s->outport = 0; 575 576 qdev_init_gpio_out_named(DEVICE(obj), &s->a20_out, PORT92_A20_LINE, 1); 577 } 578 579 static void port92_realizefn(DeviceState *dev, Error **errp) 580 { 581 ISADevice *isadev = ISA_DEVICE(dev); 582 Port92State *s = PORT92(dev); 583 584 isa_register_ioport(isadev, &s->io, 0x92); 585 } 586 587 static void port92_class_initfn(ObjectClass *klass, void *data) 588 { 589 DeviceClass *dc = DEVICE_CLASS(klass); 590 591 dc->realize = port92_realizefn; 592 dc->reset = port92_reset; 593 dc->vmsd = &vmstate_port92_isa; 594 /* 595 * Reason: unlike ordinary ISA devices, this one needs additional 596 * wiring: its A20 output line needs to be wired up by 597 * port92_init(). 598 */ 599 dc->cannot_instantiate_with_device_add_yet = true; 600 } 601 602 static const TypeInfo port92_info = { 603 .name = TYPE_PORT92, 604 .parent = TYPE_ISA_DEVICE, 605 .instance_size = sizeof(Port92State), 606 .instance_init = port92_initfn, 607 .class_init = port92_class_initfn, 608 }; 609 610 static void port92_register_types(void) 611 { 612 type_register_static(&port92_info); 613 } 614 615 type_init(port92_register_types) 616 617 static void handle_a20_line_change(void *opaque, int irq, int level) 618 { 619 X86CPU *cpu = opaque; 620 621 /* XXX: send to all CPUs ? */ 622 /* XXX: add logic to handle multiple A20 line sources */ 623 x86_cpu_set_a20(cpu, level); 624 } 625 626 int e820_add_entry(uint64_t address, uint64_t length, uint32_t type) 627 { 628 int index = le32_to_cpu(e820_reserve.count); 629 struct e820_entry *entry; 630 631 if (type != E820_RAM) { 632 /* old FW_CFG_E820_TABLE entry -- reservations only */ 633 if (index >= E820_NR_ENTRIES) { 634 return -EBUSY; 635 } 636 entry = &e820_reserve.entry[index++]; 637 638 entry->address = cpu_to_le64(address); 639 entry->length = cpu_to_le64(length); 640 entry->type = cpu_to_le32(type); 641 642 e820_reserve.count = cpu_to_le32(index); 643 } 644 645 /* new "etc/e820" file -- include ram too */ 646 e820_table = g_renew(struct e820_entry, e820_table, e820_entries + 1); 647 e820_table[e820_entries].address = cpu_to_le64(address); 648 e820_table[e820_entries].length = cpu_to_le64(length); 649 e820_table[e820_entries].type = cpu_to_le32(type); 650 e820_entries++; 651 652 return e820_entries; 653 } 654 655 int e820_get_num_entries(void) 656 { 657 return e820_entries; 658 } 659 660 bool e820_get_entry(int idx, uint32_t type, uint64_t *address, uint64_t *length) 661 { 662 if (idx < e820_entries && e820_table[idx].type == cpu_to_le32(type)) { 663 *address = le64_to_cpu(e820_table[idx].address); 664 *length = le64_to_cpu(e820_table[idx].length); 665 return true; 666 } 667 return false; 668 } 669 670 /* Enables contiguous-apic-ID mode, for compatibility */ 671 static bool compat_apic_id_mode; 672 673 void enable_compat_apic_id_mode(void) 674 { 675 compat_apic_id_mode = true; 676 } 677 678 /* Calculates initial APIC ID for a specific CPU index 679 * 680 * Currently we need to be able to calculate the APIC ID from the CPU index 681 * alone (without requiring a CPU object), as the QEMU<->Seabios interfaces have 682 * no concept of "CPU index", and the NUMA tables on fw_cfg need the APIC ID of 683 * all CPUs up to max_cpus. 684 */ 685 static uint32_t x86_cpu_apic_id_from_index(unsigned int cpu_index) 686 { 687 uint32_t correct_id; 688 static bool warned; 689 690 correct_id = x86_apicid_from_cpu_idx(smp_cores, smp_threads, cpu_index); 691 if (compat_apic_id_mode) { 692 if (cpu_index != correct_id && !warned && !qtest_enabled()) { 693 error_report("APIC IDs set in compatibility mode, " 694 "CPU topology won't match the configuration"); 695 warned = true; 696 } 697 return cpu_index; 698 } else { 699 return correct_id; 700 } 701 } 702 703 static void pc_build_smbios(FWCfgState *fw_cfg) 704 { 705 uint8_t *smbios_tables, *smbios_anchor; 706 size_t smbios_tables_len, smbios_anchor_len; 707 struct smbios_phys_mem_area *mem_array; 708 unsigned i, array_count; 709 710 smbios_tables = smbios_get_table_legacy(&smbios_tables_len); 711 if (smbios_tables) { 712 fw_cfg_add_bytes(fw_cfg, FW_CFG_SMBIOS_ENTRIES, 713 smbios_tables, smbios_tables_len); 714 } 715 716 /* build the array of physical mem area from e820 table */ 717 mem_array = g_malloc0(sizeof(*mem_array) * e820_get_num_entries()); 718 for (i = 0, array_count = 0; i < e820_get_num_entries(); i++) { 719 uint64_t addr, len; 720 721 if (e820_get_entry(i, E820_RAM, &addr, &len)) { 722 mem_array[array_count].address = addr; 723 mem_array[array_count].length = len; 724 array_count++; 725 } 726 } 727 smbios_get_tables(mem_array, array_count, 728 &smbios_tables, &smbios_tables_len, 729 &smbios_anchor, &smbios_anchor_len); 730 g_free(mem_array); 731 732 if (smbios_anchor) { 733 fw_cfg_add_file(fw_cfg, "etc/smbios/smbios-tables", 734 smbios_tables, smbios_tables_len); 735 fw_cfg_add_file(fw_cfg, "etc/smbios/smbios-anchor", 736 smbios_anchor, smbios_anchor_len); 737 } 738 } 739 740 static FWCfgState *bochs_bios_init(AddressSpace *as, PCMachineState *pcms) 741 { 742 FWCfgState *fw_cfg; 743 uint64_t *numa_fw_cfg; 744 int i, j; 745 746 fw_cfg = fw_cfg_init_io_dma(FW_CFG_IO_BASE, FW_CFG_IO_BASE + 4, as); 747 748 /* FW_CFG_MAX_CPUS is a bit confusing/problematic on x86: 749 * 750 * SeaBIOS needs FW_CFG_MAX_CPUS for CPU hotplug, but the CPU hotplug 751 * QEMU<->SeaBIOS interface is not based on the "CPU index", but on the APIC 752 * ID of hotplugged CPUs[1]. This means that FW_CFG_MAX_CPUS is not the 753 * "maximum number of CPUs", but the "limit to the APIC ID values SeaBIOS 754 * may see". 755 * 756 * So, this means we must not use max_cpus, here, but the maximum possible 757 * APIC ID value, plus one. 758 * 759 * [1] The only kind of "CPU identifier" used between SeaBIOS and QEMU is 760 * the APIC ID, not the "CPU index" 761 */ 762 fw_cfg_add_i16(fw_cfg, FW_CFG_MAX_CPUS, (uint16_t)pcms->apic_id_limit); 763 fw_cfg_add_i64(fw_cfg, FW_CFG_RAM_SIZE, (uint64_t)ram_size); 764 fw_cfg_add_bytes(fw_cfg, FW_CFG_ACPI_TABLES, 765 acpi_tables, acpi_tables_len); 766 fw_cfg_add_i32(fw_cfg, FW_CFG_IRQ0_OVERRIDE, kvm_allows_irq0_override()); 767 768 fw_cfg_add_bytes(fw_cfg, FW_CFG_E820_TABLE, 769 &e820_reserve, sizeof(e820_reserve)); 770 fw_cfg_add_file(fw_cfg, "etc/e820", e820_table, 771 sizeof(struct e820_entry) * e820_entries); 772 773 fw_cfg_add_bytes(fw_cfg, FW_CFG_HPET, &hpet_cfg, sizeof(hpet_cfg)); 774 /* allocate memory for the NUMA channel: one (64bit) word for the number 775 * of nodes, one word for each VCPU->node and one word for each node to 776 * hold the amount of memory. 777 */ 778 numa_fw_cfg = g_new0(uint64_t, 1 + pcms->apic_id_limit + nb_numa_nodes); 779 numa_fw_cfg[0] = cpu_to_le64(nb_numa_nodes); 780 for (i = 0; i < max_cpus; i++) { 781 unsigned int apic_id = x86_cpu_apic_id_from_index(i); 782 assert(apic_id < pcms->apic_id_limit); 783 for (j = 0; j < nb_numa_nodes; j++) { 784 if (test_bit(i, numa_info[j].node_cpu)) { 785 numa_fw_cfg[apic_id + 1] = cpu_to_le64(j); 786 break; 787 } 788 } 789 } 790 for (i = 0; i < nb_numa_nodes; i++) { 791 numa_fw_cfg[pcms->apic_id_limit + 1 + i] = 792 cpu_to_le64(numa_info[i].node_mem); 793 } 794 fw_cfg_add_bytes(fw_cfg, FW_CFG_NUMA, numa_fw_cfg, 795 (1 + pcms->apic_id_limit + nb_numa_nodes) * 796 sizeof(*numa_fw_cfg)); 797 798 return fw_cfg; 799 } 800 801 static long get_file_size(FILE *f) 802 { 803 long where, size; 804 805 /* XXX: on Unix systems, using fstat() probably makes more sense */ 806 807 where = ftell(f); 808 fseek(f, 0, SEEK_END); 809 size = ftell(f); 810 fseek(f, where, SEEK_SET); 811 812 return size; 813 } 814 815 static void load_linux(PCMachineState *pcms, 816 FWCfgState *fw_cfg) 817 { 818 uint16_t protocol; 819 int setup_size, kernel_size, initrd_size = 0, cmdline_size; 820 uint32_t initrd_max; 821 uint8_t header[8192], *setup, *kernel, *initrd_data; 822 hwaddr real_addr, prot_addr, cmdline_addr, initrd_addr = 0; 823 FILE *f; 824 char *vmode; 825 MachineState *machine = MACHINE(pcms); 826 PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms); 827 const char *kernel_filename = machine->kernel_filename; 828 const char *initrd_filename = machine->initrd_filename; 829 const char *kernel_cmdline = machine->kernel_cmdline; 830 831 /* Align to 16 bytes as a paranoia measure */ 832 cmdline_size = (strlen(kernel_cmdline)+16) & ~15; 833 834 /* load the kernel header */ 835 f = fopen(kernel_filename, "rb"); 836 if (!f || !(kernel_size = get_file_size(f)) || 837 fread(header, 1, MIN(ARRAY_SIZE(header), kernel_size), f) != 838 MIN(ARRAY_SIZE(header), kernel_size)) { 839 fprintf(stderr, "qemu: could not load kernel '%s': %s\n", 840 kernel_filename, strerror(errno)); 841 exit(1); 842 } 843 844 /* kernel protocol version */ 845 #if 0 846 fprintf(stderr, "header magic: %#x\n", ldl_p(header+0x202)); 847 #endif 848 if (ldl_p(header+0x202) == 0x53726448) { 849 protocol = lduw_p(header+0x206); 850 } else { 851 /* This looks like a multiboot kernel. If it is, let's stop 852 treating it like a Linux kernel. */ 853 if (load_multiboot(fw_cfg, f, kernel_filename, initrd_filename, 854 kernel_cmdline, kernel_size, header)) { 855 return; 856 } 857 protocol = 0; 858 } 859 860 if (protocol < 0x200 || !(header[0x211] & 0x01)) { 861 /* Low kernel */ 862 real_addr = 0x90000; 863 cmdline_addr = 0x9a000 - cmdline_size; 864 prot_addr = 0x10000; 865 } else if (protocol < 0x202) { 866 /* High but ancient kernel */ 867 real_addr = 0x90000; 868 cmdline_addr = 0x9a000 - cmdline_size; 869 prot_addr = 0x100000; 870 } else { 871 /* High and recent kernel */ 872 real_addr = 0x10000; 873 cmdline_addr = 0x20000; 874 prot_addr = 0x100000; 875 } 876 877 #if 0 878 fprintf(stderr, 879 "qemu: real_addr = 0x" TARGET_FMT_plx "\n" 880 "qemu: cmdline_addr = 0x" TARGET_FMT_plx "\n" 881 "qemu: prot_addr = 0x" TARGET_FMT_plx "\n", 882 real_addr, 883 cmdline_addr, 884 prot_addr); 885 #endif 886 887 /* highest address for loading the initrd */ 888 if (protocol >= 0x203) { 889 initrd_max = ldl_p(header+0x22c); 890 } else { 891 initrd_max = 0x37ffffff; 892 } 893 894 if (initrd_max >= pcms->below_4g_mem_size - pcmc->acpi_data_size) { 895 initrd_max = pcms->below_4g_mem_size - pcmc->acpi_data_size - 1; 896 } 897 898 fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_ADDR, cmdline_addr); 899 fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_SIZE, strlen(kernel_cmdline)+1); 900 fw_cfg_add_string(fw_cfg, FW_CFG_CMDLINE_DATA, kernel_cmdline); 901 902 if (protocol >= 0x202) { 903 stl_p(header+0x228, cmdline_addr); 904 } else { 905 stw_p(header+0x20, 0xA33F); 906 stw_p(header+0x22, cmdline_addr-real_addr); 907 } 908 909 /* handle vga= parameter */ 910 vmode = strstr(kernel_cmdline, "vga="); 911 if (vmode) { 912 unsigned int video_mode; 913 /* skip "vga=" */ 914 vmode += 4; 915 if (!strncmp(vmode, "normal", 6)) { 916 video_mode = 0xffff; 917 } else if (!strncmp(vmode, "ext", 3)) { 918 video_mode = 0xfffe; 919 } else if (!strncmp(vmode, "ask", 3)) { 920 video_mode = 0xfffd; 921 } else { 922 video_mode = strtol(vmode, NULL, 0); 923 } 924 stw_p(header+0x1fa, video_mode); 925 } 926 927 /* loader type */ 928 /* High nybble = B reserved for QEMU; low nybble is revision number. 929 If this code is substantially changed, you may want to consider 930 incrementing the revision. */ 931 if (protocol >= 0x200) { 932 header[0x210] = 0xB0; 933 } 934 /* heap */ 935 if (protocol >= 0x201) { 936 header[0x211] |= 0x80; /* CAN_USE_HEAP */ 937 stw_p(header+0x224, cmdline_addr-real_addr-0x200); 938 } 939 940 /* load initrd */ 941 if (initrd_filename) { 942 if (protocol < 0x200) { 943 fprintf(stderr, "qemu: linux kernel too old to load a ram disk\n"); 944 exit(1); 945 } 946 947 initrd_size = get_image_size(initrd_filename); 948 if (initrd_size < 0) { 949 fprintf(stderr, "qemu: error reading initrd %s: %s\n", 950 initrd_filename, strerror(errno)); 951 exit(1); 952 } 953 954 initrd_addr = (initrd_max-initrd_size) & ~4095; 955 956 initrd_data = g_malloc(initrd_size); 957 load_image(initrd_filename, initrd_data); 958 959 fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_ADDR, initrd_addr); 960 fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_SIZE, initrd_size); 961 fw_cfg_add_bytes(fw_cfg, FW_CFG_INITRD_DATA, initrd_data, initrd_size); 962 963 stl_p(header+0x218, initrd_addr); 964 stl_p(header+0x21c, initrd_size); 965 } 966 967 /* load kernel and setup */ 968 setup_size = header[0x1f1]; 969 if (setup_size == 0) { 970 setup_size = 4; 971 } 972 setup_size = (setup_size+1)*512; 973 if (setup_size > kernel_size) { 974 fprintf(stderr, "qemu: invalid kernel header\n"); 975 exit(1); 976 } 977 kernel_size -= setup_size; 978 979 setup = g_malloc(setup_size); 980 kernel = g_malloc(kernel_size); 981 fseek(f, 0, SEEK_SET); 982 if (fread(setup, 1, setup_size, f) != setup_size) { 983 fprintf(stderr, "fread() failed\n"); 984 exit(1); 985 } 986 if (fread(kernel, 1, kernel_size, f) != kernel_size) { 987 fprintf(stderr, "fread() failed\n"); 988 exit(1); 989 } 990 fclose(f); 991 memcpy(setup, header, MIN(sizeof(header), setup_size)); 992 993 fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ADDR, prot_addr); 994 fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_SIZE, kernel_size); 995 fw_cfg_add_bytes(fw_cfg, FW_CFG_KERNEL_DATA, kernel, kernel_size); 996 997 fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_ADDR, real_addr); 998 fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_SIZE, setup_size); 999 fw_cfg_add_bytes(fw_cfg, FW_CFG_SETUP_DATA, setup, setup_size); 1000 1001 if (fw_cfg_dma_enabled(fw_cfg)) { 1002 option_rom[nb_option_roms].name = "linuxboot_dma.bin"; 1003 option_rom[nb_option_roms].bootindex = 0; 1004 } else { 1005 option_rom[nb_option_roms].name = "linuxboot.bin"; 1006 option_rom[nb_option_roms].bootindex = 0; 1007 } 1008 nb_option_roms++; 1009 } 1010 1011 #define NE2000_NB_MAX 6 1012 1013 static const int ne2000_io[NE2000_NB_MAX] = { 0x300, 0x320, 0x340, 0x360, 1014 0x280, 0x380 }; 1015 static const int ne2000_irq[NE2000_NB_MAX] = { 9, 10, 11, 3, 4, 5 }; 1016 1017 void pc_init_ne2k_isa(ISABus *bus, NICInfo *nd) 1018 { 1019 static int nb_ne2k = 0; 1020 1021 if (nb_ne2k == NE2000_NB_MAX) 1022 return; 1023 isa_ne2000_init(bus, ne2000_io[nb_ne2k], 1024 ne2000_irq[nb_ne2k], nd); 1025 nb_ne2k++; 1026 } 1027 1028 DeviceState *cpu_get_current_apic(void) 1029 { 1030 if (current_cpu) { 1031 X86CPU *cpu = X86_CPU(current_cpu); 1032 return cpu->apic_state; 1033 } else { 1034 return NULL; 1035 } 1036 } 1037 1038 void pc_acpi_smi_interrupt(void *opaque, int irq, int level) 1039 { 1040 X86CPU *cpu = opaque; 1041 1042 if (level) { 1043 cpu_interrupt(CPU(cpu), CPU_INTERRUPT_SMI); 1044 } 1045 } 1046 1047 static X86CPU *pc_new_cpu(const char *typename, int64_t apic_id, 1048 Error **errp) 1049 { 1050 X86CPU *cpu = NULL; 1051 Error *local_err = NULL; 1052 1053 cpu = X86_CPU(object_new(typename)); 1054 1055 object_property_set_int(OBJECT(cpu), apic_id, "apic-id", &local_err); 1056 object_property_set_bool(OBJECT(cpu), true, "realized", &local_err); 1057 1058 if (local_err) { 1059 error_propagate(errp, local_err); 1060 object_unref(OBJECT(cpu)); 1061 cpu = NULL; 1062 } 1063 return cpu; 1064 } 1065 1066 void pc_hot_add_cpu(const int64_t id, Error **errp) 1067 { 1068 X86CPU *cpu; 1069 ObjectClass *oc; 1070 PCMachineState *pcms = PC_MACHINE(qdev_get_machine()); 1071 int64_t apic_id = x86_cpu_apic_id_from_index(id); 1072 Error *local_err = NULL; 1073 1074 if (id < 0) { 1075 error_setg(errp, "Invalid CPU id: %" PRIi64, id); 1076 return; 1077 } 1078 1079 if (cpu_exists(apic_id)) { 1080 error_setg(errp, "Unable to add CPU: %" PRIi64 1081 ", it already exists", id); 1082 return; 1083 } 1084 1085 if (id >= max_cpus) { 1086 error_setg(errp, "Unable to add CPU: %" PRIi64 1087 ", max allowed: %d", id, max_cpus - 1); 1088 return; 1089 } 1090 1091 if (apic_id >= ACPI_CPU_HOTPLUG_ID_LIMIT) { 1092 error_setg(errp, "Unable to add CPU: %" PRIi64 1093 ", resulting APIC ID (%" PRIi64 ") is too large", 1094 id, apic_id); 1095 return; 1096 } 1097 1098 assert(pcms->possible_cpus->cpus[0].cpu); /* BSP is always present */ 1099 oc = OBJECT_CLASS(CPU_GET_CLASS(pcms->possible_cpus->cpus[0].cpu)); 1100 cpu = pc_new_cpu(object_class_get_name(oc), apic_id, &local_err); 1101 if (local_err) { 1102 error_propagate(errp, local_err); 1103 return; 1104 } 1105 object_unref(OBJECT(cpu)); 1106 } 1107 1108 void pc_cpus_init(PCMachineState *pcms) 1109 { 1110 int i; 1111 CPUClass *cc; 1112 ObjectClass *oc; 1113 const char *typename; 1114 gchar **model_pieces; 1115 X86CPU *cpu = NULL; 1116 MachineState *machine = MACHINE(pcms); 1117 1118 /* init CPUs */ 1119 if (machine->cpu_model == NULL) { 1120 #ifdef TARGET_X86_64 1121 machine->cpu_model = "qemu64"; 1122 #else 1123 machine->cpu_model = "qemu32"; 1124 #endif 1125 } 1126 1127 model_pieces = g_strsplit(machine->cpu_model, ",", 2); 1128 if (!model_pieces[0]) { 1129 error_report("Invalid/empty CPU model name"); 1130 exit(1); 1131 } 1132 1133 oc = cpu_class_by_name(TYPE_X86_CPU, model_pieces[0]); 1134 if (oc == NULL) { 1135 error_report("Unable to find CPU definition: %s", model_pieces[0]); 1136 exit(1); 1137 } 1138 typename = object_class_get_name(oc); 1139 cc = CPU_CLASS(oc); 1140 cc->parse_features(typename, model_pieces[1], &error_fatal); 1141 g_strfreev(model_pieces); 1142 1143 /* Calculates the limit to CPU APIC ID values 1144 * 1145 * Limit for the APIC ID value, so that all 1146 * CPU APIC IDs are < pcms->apic_id_limit. 1147 * 1148 * This is used for FW_CFG_MAX_CPUS. See comments on bochs_bios_init(). 1149 */ 1150 pcms->apic_id_limit = x86_cpu_apic_id_from_index(max_cpus - 1) + 1; 1151 if (pcms->apic_id_limit > ACPI_CPU_HOTPLUG_ID_LIMIT) { 1152 error_report("max_cpus is too large. APIC ID of last CPU is %u", 1153 pcms->apic_id_limit - 1); 1154 exit(1); 1155 } 1156 1157 pcms->possible_cpus = g_malloc0(sizeof(CPUArchIdList) + 1158 sizeof(CPUArchId) * max_cpus); 1159 for (i = 0; i < max_cpus; i++) { 1160 pcms->possible_cpus->cpus[i].arch_id = x86_cpu_apic_id_from_index(i); 1161 pcms->possible_cpus->len++; 1162 if (i < smp_cpus) { 1163 cpu = pc_new_cpu(typename, x86_cpu_apic_id_from_index(i), 1164 &error_fatal); 1165 pcms->possible_cpus->cpus[i].cpu = CPU(cpu); 1166 object_unref(OBJECT(cpu)); 1167 } 1168 } 1169 1170 /* tell smbios about cpuid version and features */ 1171 smbios_set_cpuid(cpu->env.cpuid_version, cpu->env.features[FEAT_1_EDX]); 1172 } 1173 1174 static void pc_build_feature_control_file(PCMachineState *pcms) 1175 { 1176 X86CPU *cpu = X86_CPU(pcms->possible_cpus->cpus[0].cpu); 1177 CPUX86State *env = &cpu->env; 1178 uint32_t unused, ecx, edx; 1179 uint64_t feature_control_bits = 0; 1180 uint64_t *val; 1181 1182 cpu_x86_cpuid(env, 1, 0, &unused, &unused, &ecx, &edx); 1183 if (ecx & CPUID_EXT_VMX) { 1184 feature_control_bits |= FEATURE_CONTROL_VMXON_ENABLED_OUTSIDE_SMX; 1185 } 1186 1187 if ((edx & (CPUID_EXT2_MCE | CPUID_EXT2_MCA)) == 1188 (CPUID_EXT2_MCE | CPUID_EXT2_MCA) && 1189 (env->mcg_cap & MCG_LMCE_P)) { 1190 feature_control_bits |= FEATURE_CONTROL_LMCE; 1191 } 1192 1193 if (!feature_control_bits) { 1194 return; 1195 } 1196 1197 val = g_malloc(sizeof(*val)); 1198 *val = cpu_to_le64(feature_control_bits | FEATURE_CONTROL_LOCKED); 1199 fw_cfg_add_file(pcms->fw_cfg, "etc/msr_feature_control", val, sizeof(*val)); 1200 } 1201 1202 static 1203 void pc_machine_done(Notifier *notifier, void *data) 1204 { 1205 PCMachineState *pcms = container_of(notifier, 1206 PCMachineState, machine_done); 1207 PCIBus *bus = pcms->bus; 1208 1209 if (bus) { 1210 int extra_hosts = 0; 1211 1212 QLIST_FOREACH(bus, &bus->child, sibling) { 1213 /* look for expander root buses */ 1214 if (pci_bus_is_root(bus)) { 1215 extra_hosts++; 1216 } 1217 } 1218 if (extra_hosts && pcms->fw_cfg) { 1219 uint64_t *val = g_malloc(sizeof(*val)); 1220 *val = cpu_to_le64(extra_hosts); 1221 fw_cfg_add_file(pcms->fw_cfg, 1222 "etc/extra-pci-roots", val, sizeof(*val)); 1223 } 1224 } 1225 1226 acpi_setup(); 1227 if (pcms->fw_cfg) { 1228 pc_build_smbios(pcms->fw_cfg); 1229 pc_build_feature_control_file(pcms); 1230 } 1231 } 1232 1233 void pc_guest_info_init(PCMachineState *pcms) 1234 { 1235 int i; 1236 1237 pcms->apic_xrupt_override = kvm_allows_irq0_override(); 1238 pcms->numa_nodes = nb_numa_nodes; 1239 pcms->node_mem = g_malloc0(pcms->numa_nodes * 1240 sizeof *pcms->node_mem); 1241 for (i = 0; i < nb_numa_nodes; i++) { 1242 pcms->node_mem[i] = numa_info[i].node_mem; 1243 } 1244 1245 pcms->machine_done.notify = pc_machine_done; 1246 qemu_add_machine_init_done_notifier(&pcms->machine_done); 1247 } 1248 1249 /* setup pci memory address space mapping into system address space */ 1250 void pc_pci_as_mapping_init(Object *owner, MemoryRegion *system_memory, 1251 MemoryRegion *pci_address_space) 1252 { 1253 /* Set to lower priority than RAM */ 1254 memory_region_add_subregion_overlap(system_memory, 0x0, 1255 pci_address_space, -1); 1256 } 1257 1258 void pc_acpi_init(const char *default_dsdt) 1259 { 1260 char *filename; 1261 1262 if (acpi_tables != NULL) { 1263 /* manually set via -acpitable, leave it alone */ 1264 return; 1265 } 1266 1267 filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, default_dsdt); 1268 if (filename == NULL) { 1269 fprintf(stderr, "WARNING: failed to find %s\n", default_dsdt); 1270 } else { 1271 QemuOpts *opts = qemu_opts_create(qemu_find_opts("acpi"), NULL, 0, 1272 &error_abort); 1273 Error *err = NULL; 1274 1275 qemu_opt_set(opts, "file", filename, &error_abort); 1276 1277 acpi_table_add_builtin(opts, &err); 1278 if (err) { 1279 error_reportf_err(err, "WARNING: failed to load %s: ", 1280 filename); 1281 } 1282 g_free(filename); 1283 } 1284 } 1285 1286 void xen_load_linux(PCMachineState *pcms) 1287 { 1288 int i; 1289 FWCfgState *fw_cfg; 1290 1291 assert(MACHINE(pcms)->kernel_filename != NULL); 1292 1293 fw_cfg = fw_cfg_init_io(FW_CFG_IO_BASE); 1294 rom_set_fw(fw_cfg); 1295 1296 load_linux(pcms, fw_cfg); 1297 for (i = 0; i < nb_option_roms; i++) { 1298 assert(!strcmp(option_rom[i].name, "linuxboot.bin") || 1299 !strcmp(option_rom[i].name, "linuxboot_dma.bin") || 1300 !strcmp(option_rom[i].name, "multiboot.bin")); 1301 rom_add_option(option_rom[i].name, option_rom[i].bootindex); 1302 } 1303 pcms->fw_cfg = fw_cfg; 1304 } 1305 1306 void pc_memory_init(PCMachineState *pcms, 1307 MemoryRegion *system_memory, 1308 MemoryRegion *rom_memory, 1309 MemoryRegion **ram_memory) 1310 { 1311 int linux_boot, i; 1312 MemoryRegion *ram, *option_rom_mr; 1313 MemoryRegion *ram_below_4g, *ram_above_4g; 1314 FWCfgState *fw_cfg; 1315 MachineState *machine = MACHINE(pcms); 1316 PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms); 1317 1318 assert(machine->ram_size == pcms->below_4g_mem_size + 1319 pcms->above_4g_mem_size); 1320 1321 linux_boot = (machine->kernel_filename != NULL); 1322 1323 /* Allocate RAM. We allocate it as a single memory region and use 1324 * aliases to address portions of it, mostly for backwards compatibility 1325 * with older qemus that used qemu_ram_alloc(). 1326 */ 1327 ram = g_malloc(sizeof(*ram)); 1328 memory_region_allocate_system_memory(ram, NULL, "pc.ram", 1329 machine->ram_size); 1330 *ram_memory = ram; 1331 ram_below_4g = g_malloc(sizeof(*ram_below_4g)); 1332 memory_region_init_alias(ram_below_4g, NULL, "ram-below-4g", ram, 1333 0, pcms->below_4g_mem_size); 1334 memory_region_add_subregion(system_memory, 0, ram_below_4g); 1335 e820_add_entry(0, pcms->below_4g_mem_size, E820_RAM); 1336 if (pcms->above_4g_mem_size > 0) { 1337 ram_above_4g = g_malloc(sizeof(*ram_above_4g)); 1338 memory_region_init_alias(ram_above_4g, NULL, "ram-above-4g", ram, 1339 pcms->below_4g_mem_size, 1340 pcms->above_4g_mem_size); 1341 memory_region_add_subregion(system_memory, 0x100000000ULL, 1342 ram_above_4g); 1343 e820_add_entry(0x100000000ULL, pcms->above_4g_mem_size, E820_RAM); 1344 } 1345 1346 if (!pcmc->has_reserved_memory && 1347 (machine->ram_slots || 1348 (machine->maxram_size > machine->ram_size))) { 1349 MachineClass *mc = MACHINE_GET_CLASS(machine); 1350 1351 error_report("\"-memory 'slots|maxmem'\" is not supported by: %s", 1352 mc->name); 1353 exit(EXIT_FAILURE); 1354 } 1355 1356 /* initialize hotplug memory address space */ 1357 if (pcmc->has_reserved_memory && 1358 (machine->ram_size < machine->maxram_size)) { 1359 ram_addr_t hotplug_mem_size = 1360 machine->maxram_size - machine->ram_size; 1361 1362 if (machine->ram_slots > ACPI_MAX_RAM_SLOTS) { 1363 error_report("unsupported amount of memory slots: %"PRIu64, 1364 machine->ram_slots); 1365 exit(EXIT_FAILURE); 1366 } 1367 1368 if (QEMU_ALIGN_UP(machine->maxram_size, 1369 TARGET_PAGE_SIZE) != machine->maxram_size) { 1370 error_report("maximum memory size must by aligned to multiple of " 1371 "%d bytes", TARGET_PAGE_SIZE); 1372 exit(EXIT_FAILURE); 1373 } 1374 1375 pcms->hotplug_memory.base = 1376 ROUND_UP(0x100000000ULL + pcms->above_4g_mem_size, 1ULL << 30); 1377 1378 if (pcmc->enforce_aligned_dimm) { 1379 /* size hotplug region assuming 1G page max alignment per slot */ 1380 hotplug_mem_size += (1ULL << 30) * machine->ram_slots; 1381 } 1382 1383 if ((pcms->hotplug_memory.base + hotplug_mem_size) < 1384 hotplug_mem_size) { 1385 error_report("unsupported amount of maximum memory: " RAM_ADDR_FMT, 1386 machine->maxram_size); 1387 exit(EXIT_FAILURE); 1388 } 1389 1390 memory_region_init(&pcms->hotplug_memory.mr, OBJECT(pcms), 1391 "hotplug-memory", hotplug_mem_size); 1392 memory_region_add_subregion(system_memory, pcms->hotplug_memory.base, 1393 &pcms->hotplug_memory.mr); 1394 } 1395 1396 /* Initialize PC system firmware */ 1397 pc_system_firmware_init(rom_memory, !pcmc->pci_enabled); 1398 1399 option_rom_mr = g_malloc(sizeof(*option_rom_mr)); 1400 memory_region_init_ram(option_rom_mr, NULL, "pc.rom", PC_ROM_SIZE, 1401 &error_fatal); 1402 vmstate_register_ram_global(option_rom_mr); 1403 memory_region_add_subregion_overlap(rom_memory, 1404 PC_ROM_MIN_VGA, 1405 option_rom_mr, 1406 1); 1407 1408 fw_cfg = bochs_bios_init(&address_space_memory, pcms); 1409 1410 rom_set_fw(fw_cfg); 1411 1412 if (pcmc->has_reserved_memory && pcms->hotplug_memory.base) { 1413 uint64_t *val = g_malloc(sizeof(*val)); 1414 PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms); 1415 uint64_t res_mem_end = pcms->hotplug_memory.base; 1416 1417 if (!pcmc->broken_reserved_end) { 1418 res_mem_end += memory_region_size(&pcms->hotplug_memory.mr); 1419 } 1420 *val = cpu_to_le64(ROUND_UP(res_mem_end, 0x1ULL << 30)); 1421 fw_cfg_add_file(fw_cfg, "etc/reserved-memory-end", val, sizeof(*val)); 1422 } 1423 1424 if (linux_boot) { 1425 load_linux(pcms, fw_cfg); 1426 } 1427 1428 for (i = 0; i < nb_option_roms; i++) { 1429 rom_add_option(option_rom[i].name, option_rom[i].bootindex); 1430 } 1431 pcms->fw_cfg = fw_cfg; 1432 } 1433 1434 qemu_irq pc_allocate_cpu_irq(void) 1435 { 1436 return qemu_allocate_irq(pic_irq_request, NULL, 0); 1437 } 1438 1439 DeviceState *pc_vga_init(ISABus *isa_bus, PCIBus *pci_bus) 1440 { 1441 DeviceState *dev = NULL; 1442 1443 rom_set_order_override(FW_CFG_ORDER_OVERRIDE_VGA); 1444 if (pci_bus) { 1445 PCIDevice *pcidev = pci_vga_init(pci_bus); 1446 dev = pcidev ? &pcidev->qdev : NULL; 1447 } else if (isa_bus) { 1448 ISADevice *isadev = isa_vga_init(isa_bus); 1449 dev = isadev ? DEVICE(isadev) : NULL; 1450 } 1451 rom_reset_order_override(); 1452 return dev; 1453 } 1454 1455 static const MemoryRegionOps ioport80_io_ops = { 1456 .write = ioport80_write, 1457 .read = ioport80_read, 1458 .endianness = DEVICE_NATIVE_ENDIAN, 1459 .impl = { 1460 .min_access_size = 1, 1461 .max_access_size = 1, 1462 }, 1463 }; 1464 1465 static const MemoryRegionOps ioportF0_io_ops = { 1466 .write = ioportF0_write, 1467 .read = ioportF0_read, 1468 .endianness = DEVICE_NATIVE_ENDIAN, 1469 .impl = { 1470 .min_access_size = 1, 1471 .max_access_size = 1, 1472 }, 1473 }; 1474 1475 void pc_basic_device_init(ISABus *isa_bus, qemu_irq *gsi, 1476 ISADevice **rtc_state, 1477 bool create_fdctrl, 1478 bool no_vmport, 1479 uint32_t hpet_irqs) 1480 { 1481 int i; 1482 DriveInfo *fd[MAX_FD]; 1483 DeviceState *hpet = NULL; 1484 int pit_isa_irq = 0; 1485 qemu_irq pit_alt_irq = NULL; 1486 qemu_irq rtc_irq = NULL; 1487 qemu_irq *a20_line; 1488 ISADevice *i8042, *port92, *vmmouse, *pit = NULL; 1489 MemoryRegion *ioport80_io = g_new(MemoryRegion, 1); 1490 MemoryRegion *ioportF0_io = g_new(MemoryRegion, 1); 1491 1492 memory_region_init_io(ioport80_io, NULL, &ioport80_io_ops, NULL, "ioport80", 1); 1493 memory_region_add_subregion(isa_bus->address_space_io, 0x80, ioport80_io); 1494 1495 memory_region_init_io(ioportF0_io, NULL, &ioportF0_io_ops, NULL, "ioportF0", 1); 1496 memory_region_add_subregion(isa_bus->address_space_io, 0xf0, ioportF0_io); 1497 1498 /* 1499 * Check if an HPET shall be created. 1500 * 1501 * Without KVM_CAP_PIT_STATE2, we cannot switch off the in-kernel PIT 1502 * when the HPET wants to take over. Thus we have to disable the latter. 1503 */ 1504 if (!no_hpet && (!kvm_irqchip_in_kernel() || kvm_has_pit_state2())) { 1505 /* In order to set property, here not using sysbus_try_create_simple */ 1506 hpet = qdev_try_create(NULL, TYPE_HPET); 1507 if (hpet) { 1508 /* For pc-piix-*, hpet's intcap is always IRQ2. For pc-q35-1.7 1509 * and earlier, use IRQ2 for compat. Otherwise, use IRQ16~23, 1510 * IRQ8 and IRQ2. 1511 */ 1512 uint8_t compat = object_property_get_int(OBJECT(hpet), 1513 HPET_INTCAP, NULL); 1514 if (!compat) { 1515 qdev_prop_set_uint32(hpet, HPET_INTCAP, hpet_irqs); 1516 } 1517 qdev_init_nofail(hpet); 1518 sysbus_mmio_map(SYS_BUS_DEVICE(hpet), 0, HPET_BASE); 1519 1520 for (i = 0; i < GSI_NUM_PINS; i++) { 1521 sysbus_connect_irq(SYS_BUS_DEVICE(hpet), i, gsi[i]); 1522 } 1523 pit_isa_irq = -1; 1524 pit_alt_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_PIT_INT); 1525 rtc_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_RTC_INT); 1526 } 1527 } 1528 *rtc_state = rtc_init(isa_bus, 2000, rtc_irq); 1529 1530 qemu_register_boot_set(pc_boot_set, *rtc_state); 1531 1532 if (!xen_enabled()) { 1533 if (kvm_pit_in_kernel()) { 1534 pit = kvm_pit_init(isa_bus, 0x40); 1535 } else { 1536 pit = pit_init(isa_bus, 0x40, pit_isa_irq, pit_alt_irq); 1537 } 1538 if (hpet) { 1539 /* connect PIT to output control line of the HPET */ 1540 qdev_connect_gpio_out(hpet, 0, qdev_get_gpio_in(DEVICE(pit), 0)); 1541 } 1542 pcspk_init(isa_bus, pit); 1543 } 1544 1545 serial_hds_isa_init(isa_bus, MAX_SERIAL_PORTS); 1546 parallel_hds_isa_init(isa_bus, MAX_PARALLEL_PORTS); 1547 1548 a20_line = qemu_allocate_irqs(handle_a20_line_change, first_cpu, 2); 1549 i8042 = isa_create_simple(isa_bus, "i8042"); 1550 i8042_setup_a20_line(i8042, &a20_line[0]); 1551 if (!no_vmport) { 1552 vmport_init(isa_bus); 1553 vmmouse = isa_try_create(isa_bus, "vmmouse"); 1554 } else { 1555 vmmouse = NULL; 1556 } 1557 if (vmmouse) { 1558 DeviceState *dev = DEVICE(vmmouse); 1559 qdev_prop_set_ptr(dev, "ps2_mouse", i8042); 1560 qdev_init_nofail(dev); 1561 } 1562 port92 = isa_create_simple(isa_bus, "port92"); 1563 port92_init(port92, &a20_line[1]); 1564 1565 DMA_init(isa_bus, 0); 1566 1567 for(i = 0; i < MAX_FD; i++) { 1568 fd[i] = drive_get(IF_FLOPPY, 0, i); 1569 create_fdctrl |= !!fd[i]; 1570 } 1571 if (create_fdctrl) { 1572 fdctrl_init_isa(isa_bus, fd); 1573 } 1574 } 1575 1576 void pc_nic_init(ISABus *isa_bus, PCIBus *pci_bus) 1577 { 1578 int i; 1579 1580 rom_set_order_override(FW_CFG_ORDER_OVERRIDE_NIC); 1581 for (i = 0; i < nb_nics; i++) { 1582 NICInfo *nd = &nd_table[i]; 1583 1584 if (!pci_bus || (nd->model && strcmp(nd->model, "ne2k_isa") == 0)) { 1585 pc_init_ne2k_isa(isa_bus, nd); 1586 } else { 1587 pci_nic_init_nofail(nd, pci_bus, "e1000", NULL); 1588 } 1589 } 1590 rom_reset_order_override(); 1591 } 1592 1593 void pc_pci_device_init(PCIBus *pci_bus) 1594 { 1595 int max_bus; 1596 int bus; 1597 1598 max_bus = drive_get_max_bus(IF_SCSI); 1599 for (bus = 0; bus <= max_bus; bus++) { 1600 pci_create_simple(pci_bus, -1, "lsi53c895a"); 1601 } 1602 } 1603 1604 void ioapic_init_gsi(GSIState *gsi_state, const char *parent_name) 1605 { 1606 DeviceState *dev; 1607 SysBusDevice *d; 1608 unsigned int i; 1609 1610 if (kvm_ioapic_in_kernel()) { 1611 dev = qdev_create(NULL, "kvm-ioapic"); 1612 } else { 1613 dev = qdev_create(NULL, "ioapic"); 1614 } 1615 if (parent_name) { 1616 object_property_add_child(object_resolve_path(parent_name, NULL), 1617 "ioapic", OBJECT(dev), NULL); 1618 } 1619 qdev_init_nofail(dev); 1620 d = SYS_BUS_DEVICE(dev); 1621 sysbus_mmio_map(d, 0, IO_APIC_DEFAULT_ADDRESS); 1622 1623 for (i = 0; i < IOAPIC_NUM_PINS; i++) { 1624 gsi_state->ioapic_irq[i] = qdev_get_gpio_in(dev, i); 1625 } 1626 } 1627 1628 static void pc_dimm_plug(HotplugHandler *hotplug_dev, 1629 DeviceState *dev, Error **errp) 1630 { 1631 HotplugHandlerClass *hhc; 1632 Error *local_err = NULL; 1633 PCMachineState *pcms = PC_MACHINE(hotplug_dev); 1634 PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms); 1635 PCDIMMDevice *dimm = PC_DIMM(dev); 1636 PCDIMMDeviceClass *ddc = PC_DIMM_GET_CLASS(dimm); 1637 MemoryRegion *mr = ddc->get_memory_region(dimm); 1638 uint64_t align = TARGET_PAGE_SIZE; 1639 1640 if (memory_region_get_alignment(mr) && pcmc->enforce_aligned_dimm) { 1641 align = memory_region_get_alignment(mr); 1642 } 1643 1644 if (!pcms->acpi_dev) { 1645 error_setg(&local_err, 1646 "memory hotplug is not enabled: missing acpi device"); 1647 goto out; 1648 } 1649 1650 pc_dimm_memory_plug(dev, &pcms->hotplug_memory, mr, align, &local_err); 1651 if (local_err) { 1652 goto out; 1653 } 1654 1655 hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev); 1656 hhc->plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &error_abort); 1657 out: 1658 error_propagate(errp, local_err); 1659 } 1660 1661 static void pc_dimm_unplug_request(HotplugHandler *hotplug_dev, 1662 DeviceState *dev, Error **errp) 1663 { 1664 HotplugHandlerClass *hhc; 1665 Error *local_err = NULL; 1666 PCMachineState *pcms = PC_MACHINE(hotplug_dev); 1667 1668 if (!pcms->acpi_dev) { 1669 error_setg(&local_err, 1670 "memory hotplug is not enabled: missing acpi device"); 1671 goto out; 1672 } 1673 1674 hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev); 1675 hhc->unplug_request(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err); 1676 1677 out: 1678 error_propagate(errp, local_err); 1679 } 1680 1681 static void pc_dimm_unplug(HotplugHandler *hotplug_dev, 1682 DeviceState *dev, Error **errp) 1683 { 1684 PCMachineState *pcms = PC_MACHINE(hotplug_dev); 1685 PCDIMMDevice *dimm = PC_DIMM(dev); 1686 PCDIMMDeviceClass *ddc = PC_DIMM_GET_CLASS(dimm); 1687 MemoryRegion *mr = ddc->get_memory_region(dimm); 1688 HotplugHandlerClass *hhc; 1689 Error *local_err = NULL; 1690 1691 hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev); 1692 hhc->unplug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err); 1693 1694 if (local_err) { 1695 goto out; 1696 } 1697 1698 pc_dimm_memory_unplug(dev, &pcms->hotplug_memory, mr); 1699 object_unparent(OBJECT(dev)); 1700 1701 out: 1702 error_propagate(errp, local_err); 1703 } 1704 1705 static int pc_apic_cmp(const void *a, const void *b) 1706 { 1707 CPUArchId *apic_a = (CPUArchId *)a; 1708 CPUArchId *apic_b = (CPUArchId *)b; 1709 1710 return apic_a->arch_id - apic_b->arch_id; 1711 } 1712 1713 static void pc_cpu_plug(HotplugHandler *hotplug_dev, 1714 DeviceState *dev, Error **errp) 1715 { 1716 CPUClass *cc = CPU_GET_CLASS(dev); 1717 CPUArchId apic_id, *found_cpu; 1718 HotplugHandlerClass *hhc; 1719 Error *local_err = NULL; 1720 PCMachineState *pcms = PC_MACHINE(hotplug_dev); 1721 1722 if (!dev->hotplugged) { 1723 goto out; 1724 } 1725 1726 if (!pcms->acpi_dev) { 1727 error_setg(&local_err, 1728 "cpu hotplug is not enabled: missing acpi device"); 1729 goto out; 1730 } 1731 1732 hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev); 1733 hhc->plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err); 1734 if (local_err) { 1735 goto out; 1736 } 1737 1738 /* increment the number of CPUs */ 1739 rtc_set_memory(pcms->rtc, 0x5f, rtc_get_memory(pcms->rtc, 0x5f) + 1); 1740 1741 apic_id.arch_id = cc->get_arch_id(CPU(dev)); 1742 found_cpu = bsearch(&apic_id, pcms->possible_cpus->cpus, 1743 pcms->possible_cpus->len, sizeof(*pcms->possible_cpus->cpus), 1744 pc_apic_cmp); 1745 assert(found_cpu); 1746 found_cpu->cpu = CPU(dev); 1747 out: 1748 error_propagate(errp, local_err); 1749 } 1750 static void pc_cpu_unplug_request_cb(HotplugHandler *hotplug_dev, 1751 DeviceState *dev, Error **errp) 1752 { 1753 HotplugHandlerClass *hhc; 1754 Error *local_err = NULL; 1755 PCMachineState *pcms = PC_MACHINE(hotplug_dev); 1756 1757 hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev); 1758 hhc->unplug_request(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err); 1759 1760 if (local_err) { 1761 goto out; 1762 } 1763 1764 out: 1765 error_propagate(errp, local_err); 1766 1767 } 1768 1769 static void pc_cpu_unplug_cb(HotplugHandler *hotplug_dev, 1770 DeviceState *dev, Error **errp) 1771 { 1772 HotplugHandlerClass *hhc; 1773 Error *local_err = NULL; 1774 PCMachineState *pcms = PC_MACHINE(hotplug_dev); 1775 1776 hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev); 1777 hhc->unplug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err); 1778 1779 if (local_err) { 1780 goto out; 1781 } 1782 1783 /* 1784 * TODO: enable unplug once generic CPU remove bits land 1785 * for now guest will be able to eject CPU ACPI wise but 1786 * it will come back again on machine reset. 1787 */ 1788 /* object_unparent(OBJECT(dev)); */ 1789 1790 out: 1791 error_propagate(errp, local_err); 1792 } 1793 1794 static void pc_machine_device_plug_cb(HotplugHandler *hotplug_dev, 1795 DeviceState *dev, Error **errp) 1796 { 1797 if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) { 1798 pc_dimm_plug(hotplug_dev, dev, errp); 1799 } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) { 1800 pc_cpu_plug(hotplug_dev, dev, errp); 1801 } 1802 } 1803 1804 static void pc_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev, 1805 DeviceState *dev, Error **errp) 1806 { 1807 if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) { 1808 pc_dimm_unplug_request(hotplug_dev, dev, errp); 1809 } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) { 1810 pc_cpu_unplug_request_cb(hotplug_dev, dev, errp); 1811 } else { 1812 error_setg(errp, "acpi: device unplug request for not supported device" 1813 " type: %s", object_get_typename(OBJECT(dev))); 1814 } 1815 } 1816 1817 static void pc_machine_device_unplug_cb(HotplugHandler *hotplug_dev, 1818 DeviceState *dev, Error **errp) 1819 { 1820 if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) { 1821 pc_dimm_unplug(hotplug_dev, dev, errp); 1822 } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) { 1823 pc_cpu_unplug_cb(hotplug_dev, dev, errp); 1824 } else { 1825 error_setg(errp, "acpi: device unplug for not supported device" 1826 " type: %s", object_get_typename(OBJECT(dev))); 1827 } 1828 } 1829 1830 static HotplugHandler *pc_get_hotpug_handler(MachineState *machine, 1831 DeviceState *dev) 1832 { 1833 PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(machine); 1834 1835 if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) || 1836 object_dynamic_cast(OBJECT(dev), TYPE_CPU)) { 1837 return HOTPLUG_HANDLER(machine); 1838 } 1839 1840 return pcmc->get_hotplug_handler ? 1841 pcmc->get_hotplug_handler(machine, dev) : NULL; 1842 } 1843 1844 static void 1845 pc_machine_get_hotplug_memory_region_size(Object *obj, Visitor *v, 1846 const char *name, void *opaque, 1847 Error **errp) 1848 { 1849 PCMachineState *pcms = PC_MACHINE(obj); 1850 int64_t value = memory_region_size(&pcms->hotplug_memory.mr); 1851 1852 visit_type_int(v, name, &value, errp); 1853 } 1854 1855 static void pc_machine_get_max_ram_below_4g(Object *obj, Visitor *v, 1856 const char *name, void *opaque, 1857 Error **errp) 1858 { 1859 PCMachineState *pcms = PC_MACHINE(obj); 1860 uint64_t value = pcms->max_ram_below_4g; 1861 1862 visit_type_size(v, name, &value, errp); 1863 } 1864 1865 static void pc_machine_set_max_ram_below_4g(Object *obj, Visitor *v, 1866 const char *name, void *opaque, 1867 Error **errp) 1868 { 1869 PCMachineState *pcms = PC_MACHINE(obj); 1870 Error *error = NULL; 1871 uint64_t value; 1872 1873 visit_type_size(v, name, &value, &error); 1874 if (error) { 1875 error_propagate(errp, error); 1876 return; 1877 } 1878 if (value > (1ULL << 32)) { 1879 error_setg(&error, 1880 "Machine option 'max-ram-below-4g=%"PRIu64 1881 "' expects size less than or equal to 4G", value); 1882 error_propagate(errp, error); 1883 return; 1884 } 1885 1886 if (value < (1ULL << 20)) { 1887 error_report("Warning: small max_ram_below_4g(%"PRIu64 1888 ") less than 1M. BIOS may not work..", 1889 value); 1890 } 1891 1892 pcms->max_ram_below_4g = value; 1893 } 1894 1895 static void pc_machine_get_vmport(Object *obj, Visitor *v, const char *name, 1896 void *opaque, Error **errp) 1897 { 1898 PCMachineState *pcms = PC_MACHINE(obj); 1899 OnOffAuto vmport = pcms->vmport; 1900 1901 visit_type_OnOffAuto(v, name, &vmport, errp); 1902 } 1903 1904 static void pc_machine_set_vmport(Object *obj, Visitor *v, const char *name, 1905 void *opaque, Error **errp) 1906 { 1907 PCMachineState *pcms = PC_MACHINE(obj); 1908 1909 visit_type_OnOffAuto(v, name, &pcms->vmport, errp); 1910 } 1911 1912 bool pc_machine_is_smm_enabled(PCMachineState *pcms) 1913 { 1914 bool smm_available = false; 1915 1916 if (pcms->smm == ON_OFF_AUTO_OFF) { 1917 return false; 1918 } 1919 1920 if (tcg_enabled() || qtest_enabled()) { 1921 smm_available = true; 1922 } else if (kvm_enabled()) { 1923 smm_available = kvm_has_smm(); 1924 } 1925 1926 if (smm_available) { 1927 return true; 1928 } 1929 1930 if (pcms->smm == ON_OFF_AUTO_ON) { 1931 error_report("System Management Mode not supported by this hypervisor."); 1932 exit(1); 1933 } 1934 return false; 1935 } 1936 1937 static void pc_machine_get_smm(Object *obj, Visitor *v, const char *name, 1938 void *opaque, Error **errp) 1939 { 1940 PCMachineState *pcms = PC_MACHINE(obj); 1941 OnOffAuto smm = pcms->smm; 1942 1943 visit_type_OnOffAuto(v, name, &smm, errp); 1944 } 1945 1946 static void pc_machine_set_smm(Object *obj, Visitor *v, const char *name, 1947 void *opaque, Error **errp) 1948 { 1949 PCMachineState *pcms = PC_MACHINE(obj); 1950 1951 visit_type_OnOffAuto(v, name, &pcms->smm, errp); 1952 } 1953 1954 static bool pc_machine_get_nvdimm(Object *obj, Error **errp) 1955 { 1956 PCMachineState *pcms = PC_MACHINE(obj); 1957 1958 return pcms->acpi_nvdimm_state.is_enabled; 1959 } 1960 1961 static void pc_machine_set_nvdimm(Object *obj, bool value, Error **errp) 1962 { 1963 PCMachineState *pcms = PC_MACHINE(obj); 1964 1965 pcms->acpi_nvdimm_state.is_enabled = value; 1966 } 1967 1968 static void pc_machine_initfn(Object *obj) 1969 { 1970 PCMachineState *pcms = PC_MACHINE(obj); 1971 1972 object_property_add(obj, PC_MACHINE_MEMHP_REGION_SIZE, "int", 1973 pc_machine_get_hotplug_memory_region_size, 1974 NULL, NULL, NULL, &error_abort); 1975 1976 pcms->max_ram_below_4g = 0; /* use default */ 1977 object_property_add(obj, PC_MACHINE_MAX_RAM_BELOW_4G, "size", 1978 pc_machine_get_max_ram_below_4g, 1979 pc_machine_set_max_ram_below_4g, 1980 NULL, NULL, &error_abort); 1981 object_property_set_description(obj, PC_MACHINE_MAX_RAM_BELOW_4G, 1982 "Maximum ram below the 4G boundary (32bit boundary)", 1983 &error_abort); 1984 1985 pcms->smm = ON_OFF_AUTO_AUTO; 1986 object_property_add(obj, PC_MACHINE_SMM, "OnOffAuto", 1987 pc_machine_get_smm, 1988 pc_machine_set_smm, 1989 NULL, NULL, &error_abort); 1990 object_property_set_description(obj, PC_MACHINE_SMM, 1991 "Enable SMM (pc & q35)", 1992 &error_abort); 1993 1994 pcms->vmport = ON_OFF_AUTO_AUTO; 1995 object_property_add(obj, PC_MACHINE_VMPORT, "OnOffAuto", 1996 pc_machine_get_vmport, 1997 pc_machine_set_vmport, 1998 NULL, NULL, &error_abort); 1999 object_property_set_description(obj, PC_MACHINE_VMPORT, 2000 "Enable vmport (pc & q35)", 2001 &error_abort); 2002 2003 /* nvdimm is disabled on default. */ 2004 pcms->acpi_nvdimm_state.is_enabled = false; 2005 object_property_add_bool(obj, PC_MACHINE_NVDIMM, pc_machine_get_nvdimm, 2006 pc_machine_set_nvdimm, &error_abort); 2007 } 2008 2009 static void pc_machine_reset(void) 2010 { 2011 CPUState *cs; 2012 X86CPU *cpu; 2013 2014 qemu_devices_reset(); 2015 2016 /* Reset APIC after devices have been reset to cancel 2017 * any changes that qemu_devices_reset() might have done. 2018 */ 2019 CPU_FOREACH(cs) { 2020 cpu = X86_CPU(cs); 2021 2022 if (cpu->apic_state) { 2023 device_reset(cpu->apic_state); 2024 } 2025 } 2026 } 2027 2028 static unsigned pc_cpu_index_to_socket_id(unsigned cpu_index) 2029 { 2030 X86CPUTopoInfo topo; 2031 x86_topo_ids_from_idx(smp_cores, smp_threads, cpu_index, 2032 &topo); 2033 return topo.pkg_id; 2034 } 2035 2036 static CPUArchIdList *pc_possible_cpu_arch_ids(MachineState *machine) 2037 { 2038 PCMachineState *pcms = PC_MACHINE(machine); 2039 int len = sizeof(CPUArchIdList) + 2040 sizeof(CPUArchId) * (pcms->possible_cpus->len); 2041 CPUArchIdList *list = g_malloc(len); 2042 2043 memcpy(list, pcms->possible_cpus, len); 2044 return list; 2045 } 2046 2047 static void x86_nmi(NMIState *n, int cpu_index, Error **errp) 2048 { 2049 /* cpu index isn't used */ 2050 CPUState *cs; 2051 2052 CPU_FOREACH(cs) { 2053 X86CPU *cpu = X86_CPU(cs); 2054 2055 if (!cpu->apic_state) { 2056 cpu_interrupt(cs, CPU_INTERRUPT_NMI); 2057 } else { 2058 apic_deliver_nmi(cpu->apic_state); 2059 } 2060 } 2061 } 2062 2063 static void pc_machine_class_init(ObjectClass *oc, void *data) 2064 { 2065 MachineClass *mc = MACHINE_CLASS(oc); 2066 PCMachineClass *pcmc = PC_MACHINE_CLASS(oc); 2067 HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc); 2068 NMIClass *nc = NMI_CLASS(oc); 2069 2070 pcmc->get_hotplug_handler = mc->get_hotplug_handler; 2071 pcmc->pci_enabled = true; 2072 pcmc->has_acpi_build = true; 2073 pcmc->rsdp_in_ram = true; 2074 pcmc->smbios_defaults = true; 2075 pcmc->smbios_uuid_encoded = true; 2076 pcmc->gigabyte_align = true; 2077 pcmc->has_reserved_memory = true; 2078 pcmc->kvmclock_enabled = true; 2079 pcmc->enforce_aligned_dimm = true; 2080 /* BIOS ACPI tables: 128K. Other BIOS datastructures: less than 4K reported 2081 * to be used at the moment, 32K should be enough for a while. */ 2082 pcmc->acpi_data_size = 0x20000 + 0x8000; 2083 pcmc->save_tsc_khz = true; 2084 mc->get_hotplug_handler = pc_get_hotpug_handler; 2085 mc->cpu_index_to_socket_id = pc_cpu_index_to_socket_id; 2086 mc->possible_cpu_arch_ids = pc_possible_cpu_arch_ids; 2087 mc->default_boot_order = "cad"; 2088 mc->hot_add_cpu = pc_hot_add_cpu; 2089 mc->max_cpus = 255; 2090 mc->reset = pc_machine_reset; 2091 hc->plug = pc_machine_device_plug_cb; 2092 hc->unplug_request = pc_machine_device_unplug_request_cb; 2093 hc->unplug = pc_machine_device_unplug_cb; 2094 nc->nmi_monitor_handler = x86_nmi; 2095 } 2096 2097 static const TypeInfo pc_machine_info = { 2098 .name = TYPE_PC_MACHINE, 2099 .parent = TYPE_MACHINE, 2100 .abstract = true, 2101 .instance_size = sizeof(PCMachineState), 2102 .instance_init = pc_machine_initfn, 2103 .class_size = sizeof(PCMachineClass), 2104 .class_init = pc_machine_class_init, 2105 .interfaces = (InterfaceInfo[]) { 2106 { TYPE_HOTPLUG_HANDLER }, 2107 { TYPE_NMI }, 2108 { } 2109 }, 2110 }; 2111 2112 static void pc_machine_register_types(void) 2113 { 2114 type_register_static(&pc_machine_info); 2115 } 2116 2117 type_init(pc_machine_register_types) 2118