pvh: load initrd and expose it through fw_cfg
[qemu/ar7.git] / hw / i386 / pc.c
blob5d61557a2438e15988fdc31c70a8d3346e273252
1 /*
2 * QEMU PC System Emulator
4 * Copyright (c) 2003-2004 Fabrice Bellard
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:
13 * The above copyright notice and this permission notice shall be included in
14 * all copies or substantial portions of the Software.
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.
25 #include "qemu/osdep.h"
26 #include "qemu/units.h"
27 #include "hw/hw.h"
28 #include "hw/i386/pc.h"
29 #include "hw/char/serial.h"
30 #include "hw/char/parallel.h"
31 #include "hw/i386/apic.h"
32 #include "hw/i386/topology.h"
33 #include "sysemu/cpus.h"
34 #include "hw/block/fdc.h"
35 #include "hw/ide.h"
36 #include "hw/pci/pci.h"
37 #include "hw/pci/pci_bus.h"
38 #include "hw/nvram/fw_cfg.h"
39 #include "hw/timer/hpet.h"
40 #include "hw/firmware/smbios.h"
41 #include "hw/loader.h"
42 #include "elf.h"
43 #include "multiboot.h"
44 #include "hw/timer/mc146818rtc.h"
45 #include "hw/dma/i8257.h"
46 #include "hw/timer/i8254.h"
47 #include "hw/input/i8042.h"
48 #include "hw/audio/pcspk.h"
49 #include "hw/pci/msi.h"
50 #include "hw/sysbus.h"
51 #include "sysemu/sysemu.h"
52 #include "sysemu/numa.h"
53 #include "sysemu/kvm.h"
54 #include "sysemu/qtest.h"
55 #include "kvm_i386.h"
56 #include "hw/xen/xen.h"
57 #include "hw/xen/start_info.h"
58 #include "ui/qemu-spice.h"
59 #include "exec/memory.h"
60 #include "exec/address-spaces.h"
61 #include "sysemu/arch_init.h"
62 #include "qemu/bitmap.h"
63 #include "qemu/config-file.h"
64 #include "qemu/error-report.h"
65 #include "qemu/option.h"
66 #include "hw/acpi/acpi.h"
67 #include "hw/acpi/cpu_hotplug.h"
68 #include "hw/boards.h"
69 #include "acpi-build.h"
70 #include "hw/mem/pc-dimm.h"
71 #include "qapi/error.h"
72 #include "qapi/qapi-visit-common.h"
73 #include "qapi/visitor.h"
74 #include "qom/cpu.h"
75 #include "hw/nmi.h"
76 #include "hw/usb.h"
77 #include "hw/i386/intel_iommu.h"
78 #include "hw/net/ne2000-isa.h"
80 /* debug PC/ISA interrupts */
81 //#define DEBUG_IRQ
83 #ifdef DEBUG_IRQ
84 #define DPRINTF(fmt, ...) \
85 do { printf("CPUIRQ: " fmt , ## __VA_ARGS__); } while (0)
86 #else
87 #define DPRINTF(fmt, ...)
88 #endif
90 #define FW_CFG_ACPI_TABLES (FW_CFG_ARCH_LOCAL + 0)
91 #define FW_CFG_SMBIOS_ENTRIES (FW_CFG_ARCH_LOCAL + 1)
92 #define FW_CFG_IRQ0_OVERRIDE (FW_CFG_ARCH_LOCAL + 2)
93 #define FW_CFG_E820_TABLE (FW_CFG_ARCH_LOCAL + 3)
94 #define FW_CFG_HPET (FW_CFG_ARCH_LOCAL + 4)
96 #define E820_NR_ENTRIES 16
98 struct e820_entry {
99 uint64_t address;
100 uint64_t length;
101 uint32_t type;
102 } QEMU_PACKED __attribute((__aligned__(4)));
104 struct e820_table {
105 uint32_t count;
106 struct e820_entry entry[E820_NR_ENTRIES];
107 } QEMU_PACKED __attribute((__aligned__(4)));
109 static struct e820_table e820_reserve;
110 static struct e820_entry *e820_table;
111 static unsigned e820_entries;
112 struct hpet_fw_config hpet_cfg = {.count = UINT8_MAX};
114 /* Physical Address of PVH entry point read from kernel ELF NOTE */
115 static size_t pvh_start_addr;
117 GlobalProperty pc_compat_3_1[] = {
118 { "intel-iommu", "dma-drain", "off" },
119 { "Opteron_G3" "-" TYPE_X86_CPU, "rdtscp", "off" },
120 { "Opteron_G4" "-" TYPE_X86_CPU, "rdtscp", "off" },
121 { "Opteron_G4" "-" TYPE_X86_CPU, "npt", "off" },
122 { "Opteron_G4" "-" TYPE_X86_CPU, "nrip-save", "off" },
123 { "Opteron_G5" "-" TYPE_X86_CPU, "rdtscp", "off" },
124 { "Opteron_G5" "-" TYPE_X86_CPU, "npt", "off" },
125 { "Opteron_G5" "-" TYPE_X86_CPU, "nrip-save", "off" },
126 { "EPYC" "-" TYPE_X86_CPU, "npt", "off" },
127 { "EPYC" "-" TYPE_X86_CPU, "nrip-save", "off" },
128 { "EPYC-IBPB" "-" TYPE_X86_CPU, "npt", "off" },
129 { "EPYC-IBPB" "-" TYPE_X86_CPU, "nrip-save", "off" },
130 { "Skylake-Client" "-" TYPE_X86_CPU, "mpx", "on" },
131 { "Skylake-Client-IBRS" "-" TYPE_X86_CPU, "mpx", "on" },
132 { "Skylake-Server" "-" TYPE_X86_CPU, "mpx", "on" },
133 { "Skylake-Server-IBRS" "-" TYPE_X86_CPU, "mpx", "on" },
134 { "Cascadelake-Server" "-" TYPE_X86_CPU, "mpx", "on" },
135 { "Icelake-Client" "-" TYPE_X86_CPU, "mpx", "on" },
136 { "Icelake-Server" "-" TYPE_X86_CPU, "mpx", "on" },
137 { "Cascadelake-Server" "-" TYPE_X86_CPU, "stepping", "5" },
139 const size_t pc_compat_3_1_len = G_N_ELEMENTS(pc_compat_3_1);
141 GlobalProperty pc_compat_3_0[] = {
142 { TYPE_X86_CPU, "x-hv-synic-kvm-only", "on" },
143 { "Skylake-Server" "-" TYPE_X86_CPU, "pku", "off" },
144 { "Skylake-Server-IBRS" "-" TYPE_X86_CPU, "pku", "off" },
146 const size_t pc_compat_3_0_len = G_N_ELEMENTS(pc_compat_3_0);
148 GlobalProperty pc_compat_2_12[] = {
149 { TYPE_X86_CPU, "legacy-cache", "on" },
150 { TYPE_X86_CPU, "topoext", "off" },
151 { "EPYC-" TYPE_X86_CPU, "xlevel", "0x8000000a" },
152 { "EPYC-IBPB-" TYPE_X86_CPU, "xlevel", "0x8000000a" },
154 const size_t pc_compat_2_12_len = G_N_ELEMENTS(pc_compat_2_12);
156 GlobalProperty pc_compat_2_11[] = {
157 { TYPE_X86_CPU, "x-migrate-smi-count", "off" },
158 { "Skylake-Server" "-" TYPE_X86_CPU, "clflushopt", "off" },
160 const size_t pc_compat_2_11_len = G_N_ELEMENTS(pc_compat_2_11);
162 GlobalProperty pc_compat_2_10[] = {
163 { TYPE_X86_CPU, "x-hv-max-vps", "0x40" },
164 { "i440FX-pcihost", "x-pci-hole64-fix", "off" },
165 { "q35-pcihost", "x-pci-hole64-fix", "off" },
167 const size_t pc_compat_2_10_len = G_N_ELEMENTS(pc_compat_2_10);
169 GlobalProperty pc_compat_2_9[] = {
170 { "mch", "extended-tseg-mbytes", "0" },
172 const size_t pc_compat_2_9_len = G_N_ELEMENTS(pc_compat_2_9);
174 GlobalProperty pc_compat_2_8[] = {
175 { TYPE_X86_CPU, "tcg-cpuid", "off" },
176 { "kvmclock", "x-mach-use-reliable-get-clock", "off" },
177 { "ICH9-LPC", "x-smi-broadcast", "off" },
178 { TYPE_X86_CPU, "vmware-cpuid-freq", "off" },
179 { "Haswell-" TYPE_X86_CPU, "stepping", "1" },
181 const size_t pc_compat_2_8_len = G_N_ELEMENTS(pc_compat_2_8);
183 GlobalProperty pc_compat_2_7[] = {
184 { TYPE_X86_CPU, "l3-cache", "off" },
185 { TYPE_X86_CPU, "full-cpuid-auto-level", "off" },
186 { "Opteron_G3" "-" TYPE_X86_CPU, "family", "15" },
187 { "Opteron_G3" "-" TYPE_X86_CPU, "model", "6" },
188 { "Opteron_G3" "-" TYPE_X86_CPU, "stepping", "1" },
189 { "isa-pcspk", "migrate", "off" },
191 const size_t pc_compat_2_7_len = G_N_ELEMENTS(pc_compat_2_7);
193 GlobalProperty pc_compat_2_6[] = {
194 { TYPE_X86_CPU, "cpuid-0xb", "off" },
195 { "vmxnet3", "romfile", "" },
196 { TYPE_X86_CPU, "fill-mtrr-mask", "off" },
197 { "apic-common", "legacy-instance-id", "on", }
199 const size_t pc_compat_2_6_len = G_N_ELEMENTS(pc_compat_2_6);
201 GlobalProperty pc_compat_2_5[] = {};
202 const size_t pc_compat_2_5_len = G_N_ELEMENTS(pc_compat_2_5);
204 GlobalProperty pc_compat_2_4[] = {
205 PC_CPU_MODEL_IDS("2.4.0")
206 { "Haswell-" TYPE_X86_CPU, "abm", "off" },
207 { "Haswell-noTSX-" TYPE_X86_CPU, "abm", "off" },
208 { "Broadwell-" TYPE_X86_CPU, "abm", "off" },
209 { "Broadwell-noTSX-" TYPE_X86_CPU, "abm", "off" },
210 { "host" "-" TYPE_X86_CPU, "host-cache-info", "on" },
211 { TYPE_X86_CPU, "check", "off" },
212 { "qemu64" "-" TYPE_X86_CPU, "sse4a", "on" },
213 { "qemu64" "-" TYPE_X86_CPU, "abm", "on" },
214 { "qemu64" "-" TYPE_X86_CPU, "popcnt", "on" },
215 { "qemu32" "-" TYPE_X86_CPU, "popcnt", "on" },
216 { "Opteron_G2" "-" TYPE_X86_CPU, "rdtscp", "on" },
217 { "Opteron_G3" "-" TYPE_X86_CPU, "rdtscp", "on" },
218 { "Opteron_G4" "-" TYPE_X86_CPU, "rdtscp", "on" },
219 { "Opteron_G5" "-" TYPE_X86_CPU, "rdtscp", "on", }
221 const size_t pc_compat_2_4_len = G_N_ELEMENTS(pc_compat_2_4);
223 GlobalProperty pc_compat_2_3[] = {
224 PC_CPU_MODEL_IDS("2.3.0")
225 { TYPE_X86_CPU, "arat", "off" },
226 { "qemu64" "-" TYPE_X86_CPU, "min-level", "4" },
227 { "kvm64" "-" TYPE_X86_CPU, "min-level", "5" },
228 { "pentium3" "-" TYPE_X86_CPU, "min-level", "2" },
229 { "n270" "-" TYPE_X86_CPU, "min-level", "5" },
230 { "Conroe" "-" TYPE_X86_CPU, "min-level", "4" },
231 { "Penryn" "-" TYPE_X86_CPU, "min-level", "4" },
232 { "Nehalem" "-" TYPE_X86_CPU, "min-level", "4" },
233 { "n270" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
234 { "Penryn" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
235 { "Conroe" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
236 { "Nehalem" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
237 { "Westmere" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
238 { "SandyBridge" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
239 { "IvyBridge" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
240 { "Haswell" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
241 { "Haswell-noTSX" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
242 { "Broadwell" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
243 { "Broadwell-noTSX" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
244 { TYPE_X86_CPU, "kvm-no-smi-migration", "on" },
246 const size_t pc_compat_2_3_len = G_N_ELEMENTS(pc_compat_2_3);
248 GlobalProperty pc_compat_2_2[] = {
249 PC_CPU_MODEL_IDS("2.2.0")
250 { "kvm64" "-" TYPE_X86_CPU, "vme", "off" },
251 { "kvm32" "-" TYPE_X86_CPU, "vme", "off" },
252 { "Conroe" "-" TYPE_X86_CPU, "vme", "off" },
253 { "Penryn" "-" TYPE_X86_CPU, "vme", "off" },
254 { "Nehalem" "-" TYPE_X86_CPU, "vme", "off" },
255 { "Westmere" "-" TYPE_X86_CPU, "vme", "off" },
256 { "SandyBridge" "-" TYPE_X86_CPU, "vme", "off" },
257 { "Haswell" "-" TYPE_X86_CPU, "vme", "off" },
258 { "Broadwell" "-" TYPE_X86_CPU, "vme", "off" },
259 { "Opteron_G1" "-" TYPE_X86_CPU, "vme", "off" },
260 { "Opteron_G2" "-" TYPE_X86_CPU, "vme", "off" },
261 { "Opteron_G3" "-" TYPE_X86_CPU, "vme", "off" },
262 { "Opteron_G4" "-" TYPE_X86_CPU, "vme", "off" },
263 { "Opteron_G5" "-" TYPE_X86_CPU, "vme", "off" },
264 { "Haswell" "-" TYPE_X86_CPU, "f16c", "off" },
265 { "Haswell" "-" TYPE_X86_CPU, "rdrand", "off" },
266 { "Broadwell" "-" TYPE_X86_CPU, "f16c", "off" },
267 { "Broadwell" "-" TYPE_X86_CPU, "rdrand", "off" },
269 const size_t pc_compat_2_2_len = G_N_ELEMENTS(pc_compat_2_2);
271 GlobalProperty pc_compat_2_1[] = {
272 PC_CPU_MODEL_IDS("2.1.0")
273 { "coreduo" "-" TYPE_X86_CPU, "vmx", "on" },
274 { "core2duo" "-" TYPE_X86_CPU, "vmx", "on" },
276 const size_t pc_compat_2_1_len = G_N_ELEMENTS(pc_compat_2_1);
278 GlobalProperty pc_compat_2_0[] = {
279 PC_CPU_MODEL_IDS("2.0.0")
280 { "virtio-scsi-pci", "any_layout", "off" },
281 { "PIIX4_PM", "memory-hotplug-support", "off" },
282 { "apic", "version", "0x11" },
283 { "nec-usb-xhci", "superspeed-ports-first", "off" },
284 { "nec-usb-xhci", "force-pcie-endcap", "on" },
285 { "pci-serial", "prog_if", "0" },
286 { "pci-serial-2x", "prog_if", "0" },
287 { "pci-serial-4x", "prog_if", "0" },
288 { "virtio-net-pci", "guest_announce", "off" },
289 { "ICH9-LPC", "memory-hotplug-support", "off" },
290 { "xio3130-downstream", COMPAT_PROP_PCP, "off" },
291 { "ioh3420", COMPAT_PROP_PCP, "off" },
293 const size_t pc_compat_2_0_len = G_N_ELEMENTS(pc_compat_2_0);
295 GlobalProperty pc_compat_1_7[] = {
296 PC_CPU_MODEL_IDS("1.7.0")
297 { TYPE_USB_DEVICE, "msos-desc", "no" },
298 { "PIIX4_PM", "acpi-pci-hotplug-with-bridge-support", "off" },
299 { "hpet", HPET_INTCAP, "4" },
301 const size_t pc_compat_1_7_len = G_N_ELEMENTS(pc_compat_1_7);
303 GlobalProperty pc_compat_1_6[] = {
304 PC_CPU_MODEL_IDS("1.6.0")
305 { "e1000", "mitigation", "off" },
306 { "qemu64-" TYPE_X86_CPU, "model", "2" },
307 { "qemu32-" TYPE_X86_CPU, "model", "3" },
308 { "i440FX-pcihost", "short_root_bus", "1" },
309 { "q35-pcihost", "short_root_bus", "1" },
311 const size_t pc_compat_1_6_len = G_N_ELEMENTS(pc_compat_1_6);
313 GlobalProperty pc_compat_1_5[] = {
314 PC_CPU_MODEL_IDS("1.5.0")
315 { "Conroe-" TYPE_X86_CPU, "model", "2" },
316 { "Conroe-" TYPE_X86_CPU, "min-level", "2" },
317 { "Penryn-" TYPE_X86_CPU, "model", "2" },
318 { "Penryn-" TYPE_X86_CPU, "min-level", "2" },
319 { "Nehalem-" TYPE_X86_CPU, "model", "2" },
320 { "Nehalem-" TYPE_X86_CPU, "min-level", "2" },
321 { "virtio-net-pci", "any_layout", "off" },
322 { TYPE_X86_CPU, "pmu", "on" },
323 { "i440FX-pcihost", "short_root_bus", "0" },
324 { "q35-pcihost", "short_root_bus", "0" },
326 const size_t pc_compat_1_5_len = G_N_ELEMENTS(pc_compat_1_5);
328 GlobalProperty pc_compat_1_4[] = {
329 PC_CPU_MODEL_IDS("1.4.0")
330 { "scsi-hd", "discard_granularity", "0" },
331 { "scsi-cd", "discard_granularity", "0" },
332 { "scsi-disk", "discard_granularity", "0" },
333 { "ide-hd", "discard_granularity", "0" },
334 { "ide-cd", "discard_granularity", "0" },
335 { "ide-drive", "discard_granularity", "0" },
336 { "virtio-blk-pci", "discard_granularity", "0" },
337 /* DEV_NVECTORS_UNSPECIFIED as a uint32_t string: */
338 { "virtio-serial-pci", "vectors", "0xFFFFFFFF" },
339 { "virtio-net-pci", "ctrl_guest_offloads", "off" },
340 { "e1000", "romfile", "pxe-e1000.rom" },
341 { "ne2k_pci", "romfile", "pxe-ne2k_pci.rom" },
342 { "pcnet", "romfile", "pxe-pcnet.rom" },
343 { "rtl8139", "romfile", "pxe-rtl8139.rom" },
344 { "virtio-net-pci", "romfile", "pxe-virtio.rom" },
345 { "486-" TYPE_X86_CPU, "model", "0" },
346 { "n270" "-" TYPE_X86_CPU, "movbe", "off" },
347 { "Westmere" "-" TYPE_X86_CPU, "pclmulqdq", "off" },
349 const size_t pc_compat_1_4_len = G_N_ELEMENTS(pc_compat_1_4);
351 void gsi_handler(void *opaque, int n, int level)
353 GSIState *s = opaque;
355 DPRINTF("pc: %s GSI %d\n", level ? "raising" : "lowering", n);
356 if (n < ISA_NUM_IRQS) {
357 qemu_set_irq(s->i8259_irq[n], level);
359 qemu_set_irq(s->ioapic_irq[n], level);
362 static void ioport80_write(void *opaque, hwaddr addr, uint64_t data,
363 unsigned size)
367 static uint64_t ioport80_read(void *opaque, hwaddr addr, unsigned size)
369 return 0xffffffffffffffffULL;
372 /* MSDOS compatibility mode FPU exception support */
373 static qemu_irq ferr_irq;
375 void pc_register_ferr_irq(qemu_irq irq)
377 ferr_irq = irq;
380 /* XXX: add IGNNE support */
381 void cpu_set_ferr(CPUX86State *s)
383 qemu_irq_raise(ferr_irq);
386 static void ioportF0_write(void *opaque, hwaddr addr, uint64_t data,
387 unsigned size)
389 qemu_irq_lower(ferr_irq);
392 static uint64_t ioportF0_read(void *opaque, hwaddr addr, unsigned size)
394 return 0xffffffffffffffffULL;
397 /* TSC handling */
398 uint64_t cpu_get_tsc(CPUX86State *env)
400 return cpu_get_ticks();
403 /* IRQ handling */
404 int cpu_get_pic_interrupt(CPUX86State *env)
406 X86CPU *cpu = x86_env_get_cpu(env);
407 int intno;
409 if (!kvm_irqchip_in_kernel()) {
410 intno = apic_get_interrupt(cpu->apic_state);
411 if (intno >= 0) {
412 return intno;
414 /* read the irq from the PIC */
415 if (!apic_accept_pic_intr(cpu->apic_state)) {
416 return -1;
420 intno = pic_read_irq(isa_pic);
421 return intno;
424 static void pic_irq_request(void *opaque, int irq, int level)
426 CPUState *cs = first_cpu;
427 X86CPU *cpu = X86_CPU(cs);
429 DPRINTF("pic_irqs: %s irq %d\n", level? "raise" : "lower", irq);
430 if (cpu->apic_state && !kvm_irqchip_in_kernel()) {
431 CPU_FOREACH(cs) {
432 cpu = X86_CPU(cs);
433 if (apic_accept_pic_intr(cpu->apic_state)) {
434 apic_deliver_pic_intr(cpu->apic_state, level);
437 } else {
438 if (level) {
439 cpu_interrupt(cs, CPU_INTERRUPT_HARD);
440 } else {
441 cpu_reset_interrupt(cs, CPU_INTERRUPT_HARD);
446 /* PC cmos mappings */
448 #define REG_EQUIPMENT_BYTE 0x14
450 int cmos_get_fd_drive_type(FloppyDriveType fd0)
452 int val;
454 switch (fd0) {
455 case FLOPPY_DRIVE_TYPE_144:
456 /* 1.44 Mb 3"5 drive */
457 val = 4;
458 break;
459 case FLOPPY_DRIVE_TYPE_288:
460 /* 2.88 Mb 3"5 drive */
461 val = 5;
462 break;
463 case FLOPPY_DRIVE_TYPE_120:
464 /* 1.2 Mb 5"5 drive */
465 val = 2;
466 break;
467 case FLOPPY_DRIVE_TYPE_NONE:
468 default:
469 val = 0;
470 break;
472 return val;
475 static void cmos_init_hd(ISADevice *s, int type_ofs, int info_ofs,
476 int16_t cylinders, int8_t heads, int8_t sectors)
478 rtc_set_memory(s, type_ofs, 47);
479 rtc_set_memory(s, info_ofs, cylinders);
480 rtc_set_memory(s, info_ofs + 1, cylinders >> 8);
481 rtc_set_memory(s, info_ofs + 2, heads);
482 rtc_set_memory(s, info_ofs + 3, 0xff);
483 rtc_set_memory(s, info_ofs + 4, 0xff);
484 rtc_set_memory(s, info_ofs + 5, 0xc0 | ((heads > 8) << 3));
485 rtc_set_memory(s, info_ofs + 6, cylinders);
486 rtc_set_memory(s, info_ofs + 7, cylinders >> 8);
487 rtc_set_memory(s, info_ofs + 8, sectors);
490 /* convert boot_device letter to something recognizable by the bios */
491 static int boot_device2nibble(char boot_device)
493 switch(boot_device) {
494 case 'a':
495 case 'b':
496 return 0x01; /* floppy boot */
497 case 'c':
498 return 0x02; /* hard drive boot */
499 case 'd':
500 return 0x03; /* CD-ROM boot */
501 case 'n':
502 return 0x04; /* Network boot */
504 return 0;
507 static void set_boot_dev(ISADevice *s, const char *boot_device, Error **errp)
509 #define PC_MAX_BOOT_DEVICES 3
510 int nbds, bds[3] = { 0, };
511 int i;
513 nbds = strlen(boot_device);
514 if (nbds > PC_MAX_BOOT_DEVICES) {
515 error_setg(errp, "Too many boot devices for PC");
516 return;
518 for (i = 0; i < nbds; i++) {
519 bds[i] = boot_device2nibble(boot_device[i]);
520 if (bds[i] == 0) {
521 error_setg(errp, "Invalid boot device for PC: '%c'",
522 boot_device[i]);
523 return;
526 rtc_set_memory(s, 0x3d, (bds[1] << 4) | bds[0]);
527 rtc_set_memory(s, 0x38, (bds[2] << 4) | (fd_bootchk ? 0x0 : 0x1));
530 static void pc_boot_set(void *opaque, const char *boot_device, Error **errp)
532 set_boot_dev(opaque, boot_device, errp);
535 static void pc_cmos_init_floppy(ISADevice *rtc_state, ISADevice *floppy)
537 int val, nb, i;
538 FloppyDriveType fd_type[2] = { FLOPPY_DRIVE_TYPE_NONE,
539 FLOPPY_DRIVE_TYPE_NONE };
541 /* floppy type */
542 if (floppy) {
543 for (i = 0; i < 2; i++) {
544 fd_type[i] = isa_fdc_get_drive_type(floppy, i);
547 val = (cmos_get_fd_drive_type(fd_type[0]) << 4) |
548 cmos_get_fd_drive_type(fd_type[1]);
549 rtc_set_memory(rtc_state, 0x10, val);
551 val = rtc_get_memory(rtc_state, REG_EQUIPMENT_BYTE);
552 nb = 0;
553 if (fd_type[0] != FLOPPY_DRIVE_TYPE_NONE) {
554 nb++;
556 if (fd_type[1] != FLOPPY_DRIVE_TYPE_NONE) {
557 nb++;
559 switch (nb) {
560 case 0:
561 break;
562 case 1:
563 val |= 0x01; /* 1 drive, ready for boot */
564 break;
565 case 2:
566 val |= 0x41; /* 2 drives, ready for boot */
567 break;
569 rtc_set_memory(rtc_state, REG_EQUIPMENT_BYTE, val);
572 typedef struct pc_cmos_init_late_arg {
573 ISADevice *rtc_state;
574 BusState *idebus[2];
575 } pc_cmos_init_late_arg;
577 typedef struct check_fdc_state {
578 ISADevice *floppy;
579 bool multiple;
580 } CheckFdcState;
582 static int check_fdc(Object *obj, void *opaque)
584 CheckFdcState *state = opaque;
585 Object *fdc;
586 uint32_t iobase;
587 Error *local_err = NULL;
589 fdc = object_dynamic_cast(obj, TYPE_ISA_FDC);
590 if (!fdc) {
591 return 0;
594 iobase = object_property_get_uint(obj, "iobase", &local_err);
595 if (local_err || iobase != 0x3f0) {
596 error_free(local_err);
597 return 0;
600 if (state->floppy) {
601 state->multiple = true;
602 } else {
603 state->floppy = ISA_DEVICE(obj);
605 return 0;
608 static const char * const fdc_container_path[] = {
609 "/unattached", "/peripheral", "/peripheral-anon"
613 * Locate the FDC at IO address 0x3f0, in order to configure the CMOS registers
614 * and ACPI objects.
616 ISADevice *pc_find_fdc0(void)
618 int i;
619 Object *container;
620 CheckFdcState state = { 0 };
622 for (i = 0; i < ARRAY_SIZE(fdc_container_path); i++) {
623 container = container_get(qdev_get_machine(), fdc_container_path[i]);
624 object_child_foreach(container, check_fdc, &state);
627 if (state.multiple) {
628 warn_report("multiple floppy disk controllers with "
629 "iobase=0x3f0 have been found");
630 error_printf("the one being picked for CMOS setup might not reflect "
631 "your intent");
634 return state.floppy;
637 static void pc_cmos_init_late(void *opaque)
639 pc_cmos_init_late_arg *arg = opaque;
640 ISADevice *s = arg->rtc_state;
641 int16_t cylinders;
642 int8_t heads, sectors;
643 int val;
644 int i, trans;
646 val = 0;
647 if (arg->idebus[0] && ide_get_geometry(arg->idebus[0], 0,
648 &cylinders, &heads, &sectors) >= 0) {
649 cmos_init_hd(s, 0x19, 0x1b, cylinders, heads, sectors);
650 val |= 0xf0;
652 if (arg->idebus[0] && ide_get_geometry(arg->idebus[0], 1,
653 &cylinders, &heads, &sectors) >= 0) {
654 cmos_init_hd(s, 0x1a, 0x24, cylinders, heads, sectors);
655 val |= 0x0f;
657 rtc_set_memory(s, 0x12, val);
659 val = 0;
660 for (i = 0; i < 4; i++) {
661 /* NOTE: ide_get_geometry() returns the physical
662 geometry. It is always such that: 1 <= sects <= 63, 1
663 <= heads <= 16, 1 <= cylinders <= 16383. The BIOS
664 geometry can be different if a translation is done. */
665 if (arg->idebus[i / 2] &&
666 ide_get_geometry(arg->idebus[i / 2], i % 2,
667 &cylinders, &heads, &sectors) >= 0) {
668 trans = ide_get_bios_chs_trans(arg->idebus[i / 2], i % 2) - 1;
669 assert((trans & ~3) == 0);
670 val |= trans << (i * 2);
673 rtc_set_memory(s, 0x39, val);
675 pc_cmos_init_floppy(s, pc_find_fdc0());
677 qemu_unregister_reset(pc_cmos_init_late, opaque);
680 void pc_cmos_init(PCMachineState *pcms,
681 BusState *idebus0, BusState *idebus1,
682 ISADevice *s)
684 int val;
685 static pc_cmos_init_late_arg arg;
687 /* various important CMOS locations needed by PC/Bochs bios */
689 /* memory size */
690 /* base memory (first MiB) */
691 val = MIN(pcms->below_4g_mem_size / KiB, 640);
692 rtc_set_memory(s, 0x15, val);
693 rtc_set_memory(s, 0x16, val >> 8);
694 /* extended memory (next 64MiB) */
695 if (pcms->below_4g_mem_size > 1 * MiB) {
696 val = (pcms->below_4g_mem_size - 1 * MiB) / KiB;
697 } else {
698 val = 0;
700 if (val > 65535)
701 val = 65535;
702 rtc_set_memory(s, 0x17, val);
703 rtc_set_memory(s, 0x18, val >> 8);
704 rtc_set_memory(s, 0x30, val);
705 rtc_set_memory(s, 0x31, val >> 8);
706 /* memory between 16MiB and 4GiB */
707 if (pcms->below_4g_mem_size > 16 * MiB) {
708 val = (pcms->below_4g_mem_size - 16 * MiB) / (64 * KiB);
709 } else {
710 val = 0;
712 if (val > 65535)
713 val = 65535;
714 rtc_set_memory(s, 0x34, val);
715 rtc_set_memory(s, 0x35, val >> 8);
716 /* memory above 4GiB */
717 val = pcms->above_4g_mem_size / 65536;
718 rtc_set_memory(s, 0x5b, val);
719 rtc_set_memory(s, 0x5c, val >> 8);
720 rtc_set_memory(s, 0x5d, val >> 16);
722 object_property_add_link(OBJECT(pcms), "rtc_state",
723 TYPE_ISA_DEVICE,
724 (Object **)&pcms->rtc,
725 object_property_allow_set_link,
726 OBJ_PROP_LINK_STRONG, &error_abort);
727 object_property_set_link(OBJECT(pcms), OBJECT(s),
728 "rtc_state", &error_abort);
730 set_boot_dev(s, MACHINE(pcms)->boot_order, &error_fatal);
732 val = 0;
733 val |= 0x02; /* FPU is there */
734 val |= 0x04; /* PS/2 mouse installed */
735 rtc_set_memory(s, REG_EQUIPMENT_BYTE, val);
737 /* hard drives and FDC */
738 arg.rtc_state = s;
739 arg.idebus[0] = idebus0;
740 arg.idebus[1] = idebus1;
741 qemu_register_reset(pc_cmos_init_late, &arg);
744 #define TYPE_PORT92 "port92"
745 #define PORT92(obj) OBJECT_CHECK(Port92State, (obj), TYPE_PORT92)
747 /* port 92 stuff: could be split off */
748 typedef struct Port92State {
749 ISADevice parent_obj;
751 MemoryRegion io;
752 uint8_t outport;
753 qemu_irq a20_out;
754 } Port92State;
756 static void port92_write(void *opaque, hwaddr addr, uint64_t val,
757 unsigned size)
759 Port92State *s = opaque;
760 int oldval = s->outport;
762 DPRINTF("port92: write 0x%02" PRIx64 "\n", val);
763 s->outport = val;
764 qemu_set_irq(s->a20_out, (val >> 1) & 1);
765 if ((val & 1) && !(oldval & 1)) {
766 qemu_system_reset_request(SHUTDOWN_CAUSE_GUEST_RESET);
770 static uint64_t port92_read(void *opaque, hwaddr addr,
771 unsigned size)
773 Port92State *s = opaque;
774 uint32_t ret;
776 ret = s->outport;
777 DPRINTF("port92: read 0x%02x\n", ret);
778 return ret;
781 static void port92_init(ISADevice *dev, qemu_irq a20_out)
783 qdev_connect_gpio_out_named(DEVICE(dev), PORT92_A20_LINE, 0, a20_out);
786 static const VMStateDescription vmstate_port92_isa = {
787 .name = "port92",
788 .version_id = 1,
789 .minimum_version_id = 1,
790 .fields = (VMStateField[]) {
791 VMSTATE_UINT8(outport, Port92State),
792 VMSTATE_END_OF_LIST()
796 static void port92_reset(DeviceState *d)
798 Port92State *s = PORT92(d);
800 s->outport &= ~1;
803 static const MemoryRegionOps port92_ops = {
804 .read = port92_read,
805 .write = port92_write,
806 .impl = {
807 .min_access_size = 1,
808 .max_access_size = 1,
810 .endianness = DEVICE_LITTLE_ENDIAN,
813 static void port92_initfn(Object *obj)
815 Port92State *s = PORT92(obj);
817 memory_region_init_io(&s->io, OBJECT(s), &port92_ops, s, "port92", 1);
819 s->outport = 0;
821 qdev_init_gpio_out_named(DEVICE(obj), &s->a20_out, PORT92_A20_LINE, 1);
824 static void port92_realizefn(DeviceState *dev, Error **errp)
826 ISADevice *isadev = ISA_DEVICE(dev);
827 Port92State *s = PORT92(dev);
829 isa_register_ioport(isadev, &s->io, 0x92);
832 static void port92_class_initfn(ObjectClass *klass, void *data)
834 DeviceClass *dc = DEVICE_CLASS(klass);
836 dc->realize = port92_realizefn;
837 dc->reset = port92_reset;
838 dc->vmsd = &vmstate_port92_isa;
840 * Reason: unlike ordinary ISA devices, this one needs additional
841 * wiring: its A20 output line needs to be wired up by
842 * port92_init().
844 dc->user_creatable = false;
847 static const TypeInfo port92_info = {
848 .name = TYPE_PORT92,
849 .parent = TYPE_ISA_DEVICE,
850 .instance_size = sizeof(Port92State),
851 .instance_init = port92_initfn,
852 .class_init = port92_class_initfn,
855 static void port92_register_types(void)
857 type_register_static(&port92_info);
860 type_init(port92_register_types)
862 static void handle_a20_line_change(void *opaque, int irq, int level)
864 X86CPU *cpu = opaque;
866 /* XXX: send to all CPUs ? */
867 /* XXX: add logic to handle multiple A20 line sources */
868 x86_cpu_set_a20(cpu, level);
871 int e820_add_entry(uint64_t address, uint64_t length, uint32_t type)
873 int index = le32_to_cpu(e820_reserve.count);
874 struct e820_entry *entry;
876 if (type != E820_RAM) {
877 /* old FW_CFG_E820_TABLE entry -- reservations only */
878 if (index >= E820_NR_ENTRIES) {
879 return -EBUSY;
881 entry = &e820_reserve.entry[index++];
883 entry->address = cpu_to_le64(address);
884 entry->length = cpu_to_le64(length);
885 entry->type = cpu_to_le32(type);
887 e820_reserve.count = cpu_to_le32(index);
890 /* new "etc/e820" file -- include ram too */
891 e820_table = g_renew(struct e820_entry, e820_table, e820_entries + 1);
892 e820_table[e820_entries].address = cpu_to_le64(address);
893 e820_table[e820_entries].length = cpu_to_le64(length);
894 e820_table[e820_entries].type = cpu_to_le32(type);
895 e820_entries++;
897 return e820_entries;
900 int e820_get_num_entries(void)
902 return e820_entries;
905 bool e820_get_entry(int idx, uint32_t type, uint64_t *address, uint64_t *length)
907 if (idx < e820_entries && e820_table[idx].type == cpu_to_le32(type)) {
908 *address = le64_to_cpu(e820_table[idx].address);
909 *length = le64_to_cpu(e820_table[idx].length);
910 return true;
912 return false;
915 /* Enables contiguous-apic-ID mode, for compatibility */
916 static bool compat_apic_id_mode;
918 void enable_compat_apic_id_mode(void)
920 compat_apic_id_mode = true;
923 /* Calculates initial APIC ID for a specific CPU index
925 * Currently we need to be able to calculate the APIC ID from the CPU index
926 * alone (without requiring a CPU object), as the QEMU<->Seabios interfaces have
927 * no concept of "CPU index", and the NUMA tables on fw_cfg need the APIC ID of
928 * all CPUs up to max_cpus.
930 static uint32_t x86_cpu_apic_id_from_index(unsigned int cpu_index)
932 uint32_t correct_id;
933 static bool warned;
935 correct_id = x86_apicid_from_cpu_idx(smp_cores, smp_threads, cpu_index);
936 if (compat_apic_id_mode) {
937 if (cpu_index != correct_id && !warned && !qtest_enabled()) {
938 error_report("APIC IDs set in compatibility mode, "
939 "CPU topology won't match the configuration");
940 warned = true;
942 return cpu_index;
943 } else {
944 return correct_id;
948 static void pc_build_smbios(PCMachineState *pcms)
950 uint8_t *smbios_tables, *smbios_anchor;
951 size_t smbios_tables_len, smbios_anchor_len;
952 struct smbios_phys_mem_area *mem_array;
953 unsigned i, array_count;
954 MachineState *ms = MACHINE(pcms);
955 X86CPU *cpu = X86_CPU(ms->possible_cpus->cpus[0].cpu);
957 /* tell smbios about cpuid version and features */
958 smbios_set_cpuid(cpu->env.cpuid_version, cpu->env.features[FEAT_1_EDX]);
960 smbios_tables = smbios_get_table_legacy(&smbios_tables_len);
961 if (smbios_tables) {
962 fw_cfg_add_bytes(pcms->fw_cfg, FW_CFG_SMBIOS_ENTRIES,
963 smbios_tables, smbios_tables_len);
966 /* build the array of physical mem area from e820 table */
967 mem_array = g_malloc0(sizeof(*mem_array) * e820_get_num_entries());
968 for (i = 0, array_count = 0; i < e820_get_num_entries(); i++) {
969 uint64_t addr, len;
971 if (e820_get_entry(i, E820_RAM, &addr, &len)) {
972 mem_array[array_count].address = addr;
973 mem_array[array_count].length = len;
974 array_count++;
977 smbios_get_tables(mem_array, array_count,
978 &smbios_tables, &smbios_tables_len,
979 &smbios_anchor, &smbios_anchor_len);
980 g_free(mem_array);
982 if (smbios_anchor) {
983 fw_cfg_add_file(pcms->fw_cfg, "etc/smbios/smbios-tables",
984 smbios_tables, smbios_tables_len);
985 fw_cfg_add_file(pcms->fw_cfg, "etc/smbios/smbios-anchor",
986 smbios_anchor, smbios_anchor_len);
990 static FWCfgState *bochs_bios_init(AddressSpace *as, PCMachineState *pcms)
992 FWCfgState *fw_cfg;
993 uint64_t *numa_fw_cfg;
994 int i;
995 const CPUArchIdList *cpus;
996 MachineClass *mc = MACHINE_GET_CLASS(pcms);
998 fw_cfg = fw_cfg_init_io_dma(FW_CFG_IO_BASE, FW_CFG_IO_BASE + 4, as);
999 fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
1001 /* FW_CFG_MAX_CPUS is a bit confusing/problematic on x86:
1003 * For machine types prior to 1.8, SeaBIOS needs FW_CFG_MAX_CPUS for
1004 * building MPTable, ACPI MADT, ACPI CPU hotplug and ACPI SRAT table,
1005 * that tables are based on xAPIC ID and QEMU<->SeaBIOS interface
1006 * for CPU hotplug also uses APIC ID and not "CPU index".
1007 * This means that FW_CFG_MAX_CPUS is not the "maximum number of CPUs",
1008 * but the "limit to the APIC ID values SeaBIOS may see".
1010 * So for compatibility reasons with old BIOSes we are stuck with
1011 * "etc/max-cpus" actually being apic_id_limit
1013 fw_cfg_add_i16(fw_cfg, FW_CFG_MAX_CPUS, (uint16_t)pcms->apic_id_limit);
1014 fw_cfg_add_i64(fw_cfg, FW_CFG_RAM_SIZE, (uint64_t)ram_size);
1015 fw_cfg_add_bytes(fw_cfg, FW_CFG_ACPI_TABLES,
1016 acpi_tables, acpi_tables_len);
1017 fw_cfg_add_i32(fw_cfg, FW_CFG_IRQ0_OVERRIDE, kvm_allows_irq0_override());
1019 fw_cfg_add_bytes(fw_cfg, FW_CFG_E820_TABLE,
1020 &e820_reserve, sizeof(e820_reserve));
1021 fw_cfg_add_file(fw_cfg, "etc/e820", e820_table,
1022 sizeof(struct e820_entry) * e820_entries);
1024 fw_cfg_add_bytes(fw_cfg, FW_CFG_HPET, &hpet_cfg, sizeof(hpet_cfg));
1025 /* allocate memory for the NUMA channel: one (64bit) word for the number
1026 * of nodes, one word for each VCPU->node and one word for each node to
1027 * hold the amount of memory.
1029 numa_fw_cfg = g_new0(uint64_t, 1 + pcms->apic_id_limit + nb_numa_nodes);
1030 numa_fw_cfg[0] = cpu_to_le64(nb_numa_nodes);
1031 cpus = mc->possible_cpu_arch_ids(MACHINE(pcms));
1032 for (i = 0; i < cpus->len; i++) {
1033 unsigned int apic_id = cpus->cpus[i].arch_id;
1034 assert(apic_id < pcms->apic_id_limit);
1035 numa_fw_cfg[apic_id + 1] = cpu_to_le64(cpus->cpus[i].props.node_id);
1037 for (i = 0; i < nb_numa_nodes; i++) {
1038 numa_fw_cfg[pcms->apic_id_limit + 1 + i] =
1039 cpu_to_le64(numa_info[i].node_mem);
1041 fw_cfg_add_bytes(fw_cfg, FW_CFG_NUMA, numa_fw_cfg,
1042 (1 + pcms->apic_id_limit + nb_numa_nodes) *
1043 sizeof(*numa_fw_cfg));
1045 return fw_cfg;
1048 static long get_file_size(FILE *f)
1050 long where, size;
1052 /* XXX: on Unix systems, using fstat() probably makes more sense */
1054 where = ftell(f);
1055 fseek(f, 0, SEEK_END);
1056 size = ftell(f);
1057 fseek(f, where, SEEK_SET);
1059 return size;
1062 /* setup_data types */
1063 #define SETUP_NONE 0
1064 #define SETUP_E820_EXT 1
1065 #define SETUP_DTB 2
1066 #define SETUP_PCI 3
1067 #define SETUP_EFI 4
1069 struct setup_data {
1070 uint64_t next;
1071 uint32_t type;
1072 uint32_t len;
1073 uint8_t data[0];
1074 } __attribute__((packed));
1078 * The entry point into the kernel for PVH boot is different from
1079 * the native entry point. The PVH entry is defined by the x86/HVM
1080 * direct boot ABI and is available in an ELFNOTE in the kernel binary.
1082 * This function is passed to load_elf() when it is called from
1083 * load_elfboot() which then additionally checks for an ELF Note of
1084 * type XEN_ELFNOTE_PHYS32_ENTRY and passes it to this function to
1085 * parse the PVH entry address from the ELF Note.
1087 * Due to trickery in elf_opts.h, load_elf() is actually available as
1088 * load_elf32() or load_elf64() and this routine needs to be able
1089 * to deal with being called as 32 or 64 bit.
1091 * The address of the PVH entry point is saved to the 'pvh_start_addr'
1092 * global variable. (although the entry point is 32-bit, the kernel
1093 * binary can be either 32-bit or 64-bit).
1095 static uint64_t read_pvh_start_addr(void *arg1, void *arg2, bool is64)
1097 size_t *elf_note_data_addr;
1099 /* Check if ELF Note header passed in is valid */
1100 if (arg1 == NULL) {
1101 return 0;
1104 if (is64) {
1105 struct elf64_note *nhdr64 = (struct elf64_note *)arg1;
1106 uint64_t nhdr_size64 = sizeof(struct elf64_note);
1107 uint64_t phdr_align = *(uint64_t *)arg2;
1108 uint64_t nhdr_namesz = nhdr64->n_namesz;
1110 elf_note_data_addr =
1111 ((void *)nhdr64) + nhdr_size64 +
1112 QEMU_ALIGN_UP(nhdr_namesz, phdr_align);
1113 } else {
1114 struct elf32_note *nhdr32 = (struct elf32_note *)arg1;
1115 uint32_t nhdr_size32 = sizeof(struct elf32_note);
1116 uint32_t phdr_align = *(uint32_t *)arg2;
1117 uint32_t nhdr_namesz = nhdr32->n_namesz;
1119 elf_note_data_addr =
1120 ((void *)nhdr32) + nhdr_size32 +
1121 QEMU_ALIGN_UP(nhdr_namesz, phdr_align);
1124 pvh_start_addr = *elf_note_data_addr;
1126 return pvh_start_addr;
1129 static bool load_elfboot(const char *kernel_filename,
1130 int kernel_file_size,
1131 uint8_t *header,
1132 size_t pvh_xen_start_addr,
1133 FWCfgState *fw_cfg)
1135 uint32_t flags = 0;
1136 uint32_t mh_load_addr = 0;
1137 uint32_t elf_kernel_size = 0;
1138 uint64_t elf_entry;
1139 uint64_t elf_low, elf_high;
1140 int kernel_size;
1142 if (ldl_p(header) != 0x464c457f) {
1143 return false; /* no elfboot */
1146 bool elf_is64 = header[EI_CLASS] == ELFCLASS64;
1147 flags = elf_is64 ?
1148 ((Elf64_Ehdr *)header)->e_flags : ((Elf32_Ehdr *)header)->e_flags;
1150 if (flags & 0x00010004) { /* LOAD_ELF_HEADER_HAS_ADDR */
1151 error_report("elfboot unsupported flags = %x", flags);
1152 exit(1);
1155 uint64_t elf_note_type = XEN_ELFNOTE_PHYS32_ENTRY;
1156 kernel_size = load_elf(kernel_filename, read_pvh_start_addr,
1157 NULL, &elf_note_type, &elf_entry,
1158 &elf_low, &elf_high, 0, I386_ELF_MACHINE,
1159 0, 0);
1161 if (kernel_size < 0) {
1162 error_report("Error while loading elf kernel");
1163 exit(1);
1165 mh_load_addr = elf_low;
1166 elf_kernel_size = elf_high - elf_low;
1168 if (pvh_start_addr == 0) {
1169 error_report("Error loading uncompressed kernel without PVH ELF Note");
1170 exit(1);
1172 fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ENTRY, pvh_start_addr);
1173 fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ADDR, mh_load_addr);
1174 fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_SIZE, elf_kernel_size);
1176 return true;
1179 static void load_linux(PCMachineState *pcms,
1180 FWCfgState *fw_cfg)
1182 uint16_t protocol;
1183 int setup_size, kernel_size, cmdline_size;
1184 int dtb_size, setup_data_offset;
1185 uint32_t initrd_max;
1186 uint8_t header[8192], *setup, *kernel;
1187 hwaddr real_addr, prot_addr, cmdline_addr, initrd_addr = 0;
1188 FILE *f;
1189 char *vmode;
1190 MachineState *machine = MACHINE(pcms);
1191 PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1192 struct setup_data *setup_data;
1193 const char *kernel_filename = machine->kernel_filename;
1194 const char *initrd_filename = machine->initrd_filename;
1195 const char *dtb_filename = machine->dtb;
1196 const char *kernel_cmdline = machine->kernel_cmdline;
1198 /* Align to 16 bytes as a paranoia measure */
1199 cmdline_size = (strlen(kernel_cmdline)+16) & ~15;
1201 /* load the kernel header */
1202 f = fopen(kernel_filename, "rb");
1203 if (!f || !(kernel_size = get_file_size(f)) ||
1204 fread(header, 1, MIN(ARRAY_SIZE(header), kernel_size), f) !=
1205 MIN(ARRAY_SIZE(header), kernel_size)) {
1206 fprintf(stderr, "qemu: could not load kernel '%s': %s\n",
1207 kernel_filename, strerror(errno));
1208 exit(1);
1211 /* kernel protocol version */
1212 #if 0
1213 fprintf(stderr, "header magic: %#x\n", ldl_p(header+0x202));
1214 #endif
1215 if (ldl_p(header+0x202) == 0x53726448) {
1216 protocol = lduw_p(header+0x206);
1217 } else {
1219 * Check if the file is an uncompressed kernel file (ELF) and load it,
1220 * saving the PVH entry point used by the x86/HVM direct boot ABI.
1221 * If load_elfboot() is successful, populate the fw_cfg info.
1223 if (load_elfboot(kernel_filename, kernel_size,
1224 header, pvh_start_addr, fw_cfg)) {
1225 fclose(f);
1227 fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_SIZE,
1228 strlen(kernel_cmdline) + 1);
1229 fw_cfg_add_string(fw_cfg, FW_CFG_CMDLINE_DATA, kernel_cmdline);
1231 fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_SIZE, sizeof(header));
1232 fw_cfg_add_bytes(fw_cfg, FW_CFG_SETUP_DATA,
1233 header, sizeof(header));
1235 /* load initrd */
1236 if (initrd_filename) {
1237 gsize initrd_size;
1238 gchar *initrd_data;
1239 GError *gerr = NULL;
1241 if (!g_file_get_contents(initrd_filename, &initrd_data,
1242 &initrd_size, &gerr)) {
1243 fprintf(stderr, "qemu: error reading initrd %s: %s\n",
1244 initrd_filename, gerr->message);
1245 exit(1);
1248 initrd_max = pcms->below_4g_mem_size - pcmc->acpi_data_size - 1;
1249 if (initrd_size >= initrd_max) {
1250 fprintf(stderr, "qemu: initrd is too large, cannot support."
1251 "(max: %"PRIu32", need %"PRId64")\n",
1252 initrd_max, (uint64_t)initrd_size);
1253 exit(1);
1256 initrd_addr = (initrd_max - initrd_size) & ~4095;
1258 fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_ADDR, initrd_addr);
1259 fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_SIZE, initrd_size);
1260 fw_cfg_add_bytes(fw_cfg, FW_CFG_INITRD_DATA, initrd_data,
1261 initrd_size);
1264 return;
1266 /* This looks like a multiboot kernel. If it is, let's stop
1267 treating it like a Linux kernel. */
1268 if (load_multiboot(fw_cfg, f, kernel_filename, initrd_filename,
1269 kernel_cmdline, kernel_size, header)) {
1270 return;
1272 protocol = 0;
1275 if (protocol < 0x200 || !(header[0x211] & 0x01)) {
1276 /* Low kernel */
1277 real_addr = 0x90000;
1278 cmdline_addr = 0x9a000 - cmdline_size;
1279 prot_addr = 0x10000;
1280 } else if (protocol < 0x202) {
1281 /* High but ancient kernel */
1282 real_addr = 0x90000;
1283 cmdline_addr = 0x9a000 - cmdline_size;
1284 prot_addr = 0x100000;
1285 } else {
1286 /* High and recent kernel */
1287 real_addr = 0x10000;
1288 cmdline_addr = 0x20000;
1289 prot_addr = 0x100000;
1292 #if 0
1293 fprintf(stderr,
1294 "qemu: real_addr = 0x" TARGET_FMT_plx "\n"
1295 "qemu: cmdline_addr = 0x" TARGET_FMT_plx "\n"
1296 "qemu: prot_addr = 0x" TARGET_FMT_plx "\n",
1297 real_addr,
1298 cmdline_addr,
1299 prot_addr);
1300 #endif
1302 /* highest address for loading the initrd */
1303 if (protocol >= 0x203) {
1304 initrd_max = ldl_p(header+0x22c);
1305 } else {
1306 initrd_max = 0x37ffffff;
1309 if (initrd_max >= pcms->below_4g_mem_size - pcmc->acpi_data_size) {
1310 initrd_max = pcms->below_4g_mem_size - pcmc->acpi_data_size - 1;
1313 fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_ADDR, cmdline_addr);
1314 fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_SIZE, strlen(kernel_cmdline)+1);
1315 fw_cfg_add_string(fw_cfg, FW_CFG_CMDLINE_DATA, kernel_cmdline);
1317 if (protocol >= 0x202) {
1318 stl_p(header+0x228, cmdline_addr);
1319 } else {
1320 stw_p(header+0x20, 0xA33F);
1321 stw_p(header+0x22, cmdline_addr-real_addr);
1324 /* handle vga= parameter */
1325 vmode = strstr(kernel_cmdline, "vga=");
1326 if (vmode) {
1327 unsigned int video_mode;
1328 /* skip "vga=" */
1329 vmode += 4;
1330 if (!strncmp(vmode, "normal", 6)) {
1331 video_mode = 0xffff;
1332 } else if (!strncmp(vmode, "ext", 3)) {
1333 video_mode = 0xfffe;
1334 } else if (!strncmp(vmode, "ask", 3)) {
1335 video_mode = 0xfffd;
1336 } else {
1337 video_mode = strtol(vmode, NULL, 0);
1339 stw_p(header+0x1fa, video_mode);
1342 /* loader type */
1343 /* High nybble = B reserved for QEMU; low nybble is revision number.
1344 If this code is substantially changed, you may want to consider
1345 incrementing the revision. */
1346 if (protocol >= 0x200) {
1347 header[0x210] = 0xB0;
1349 /* heap */
1350 if (protocol >= 0x201) {
1351 header[0x211] |= 0x80; /* CAN_USE_HEAP */
1352 stw_p(header+0x224, cmdline_addr-real_addr-0x200);
1355 /* load initrd */
1356 if (initrd_filename) {
1357 gsize initrd_size;
1358 gchar *initrd_data;
1359 GError *gerr = NULL;
1361 if (protocol < 0x200) {
1362 fprintf(stderr, "qemu: linux kernel too old to load a ram disk\n");
1363 exit(1);
1366 if (!g_file_get_contents(initrd_filename, &initrd_data,
1367 &initrd_size, &gerr)) {
1368 fprintf(stderr, "qemu: error reading initrd %s: %s\n",
1369 initrd_filename, gerr->message);
1370 exit(1);
1372 if (initrd_size >= initrd_max) {
1373 fprintf(stderr, "qemu: initrd is too large, cannot support."
1374 "(max: %"PRIu32", need %"PRId64")\n",
1375 initrd_max, (uint64_t)initrd_size);
1376 exit(1);
1379 initrd_addr = (initrd_max-initrd_size) & ~4095;
1381 fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_ADDR, initrd_addr);
1382 fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_SIZE, initrd_size);
1383 fw_cfg_add_bytes(fw_cfg, FW_CFG_INITRD_DATA, initrd_data, initrd_size);
1385 stl_p(header+0x218, initrd_addr);
1386 stl_p(header+0x21c, initrd_size);
1389 /* load kernel and setup */
1390 setup_size = header[0x1f1];
1391 if (setup_size == 0) {
1392 setup_size = 4;
1394 setup_size = (setup_size+1)*512;
1395 if (setup_size > kernel_size) {
1396 fprintf(stderr, "qemu: invalid kernel header\n");
1397 exit(1);
1399 kernel_size -= setup_size;
1401 setup = g_malloc(setup_size);
1402 kernel = g_malloc(kernel_size);
1403 fseek(f, 0, SEEK_SET);
1404 if (fread(setup, 1, setup_size, f) != setup_size) {
1405 fprintf(stderr, "fread() failed\n");
1406 exit(1);
1408 if (fread(kernel, 1, kernel_size, f) != kernel_size) {
1409 fprintf(stderr, "fread() failed\n");
1410 exit(1);
1412 fclose(f);
1414 /* append dtb to kernel */
1415 if (dtb_filename) {
1416 if (protocol < 0x209) {
1417 fprintf(stderr, "qemu: Linux kernel too old to load a dtb\n");
1418 exit(1);
1421 dtb_size = get_image_size(dtb_filename);
1422 if (dtb_size <= 0) {
1423 fprintf(stderr, "qemu: error reading dtb %s: %s\n",
1424 dtb_filename, strerror(errno));
1425 exit(1);
1428 setup_data_offset = QEMU_ALIGN_UP(kernel_size, 16);
1429 kernel_size = setup_data_offset + sizeof(struct setup_data) + dtb_size;
1430 kernel = g_realloc(kernel, kernel_size);
1432 stq_p(header+0x250, prot_addr + setup_data_offset);
1434 setup_data = (struct setup_data *)(kernel + setup_data_offset);
1435 setup_data->next = 0;
1436 setup_data->type = cpu_to_le32(SETUP_DTB);
1437 setup_data->len = cpu_to_le32(dtb_size);
1439 load_image_size(dtb_filename, setup_data->data, dtb_size);
1442 memcpy(setup, header, MIN(sizeof(header), setup_size));
1444 fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ADDR, prot_addr);
1445 fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_SIZE, kernel_size);
1446 fw_cfg_add_bytes(fw_cfg, FW_CFG_KERNEL_DATA, kernel, kernel_size);
1448 fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_ADDR, real_addr);
1449 fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_SIZE, setup_size);
1450 fw_cfg_add_bytes(fw_cfg, FW_CFG_SETUP_DATA, setup, setup_size);
1452 option_rom[nb_option_roms].bootindex = 0;
1453 option_rom[nb_option_roms].name = "linuxboot.bin";
1454 if (pcmc->linuxboot_dma_enabled && fw_cfg_dma_enabled(fw_cfg)) {
1455 option_rom[nb_option_roms].name = "linuxboot_dma.bin";
1457 nb_option_roms++;
1460 #define NE2000_NB_MAX 6
1462 static const int ne2000_io[NE2000_NB_MAX] = { 0x300, 0x320, 0x340, 0x360,
1463 0x280, 0x380 };
1464 static const int ne2000_irq[NE2000_NB_MAX] = { 9, 10, 11, 3, 4, 5 };
1466 void pc_init_ne2k_isa(ISABus *bus, NICInfo *nd)
1468 static int nb_ne2k = 0;
1470 if (nb_ne2k == NE2000_NB_MAX)
1471 return;
1472 isa_ne2000_init(bus, ne2000_io[nb_ne2k],
1473 ne2000_irq[nb_ne2k], nd);
1474 nb_ne2k++;
1477 DeviceState *cpu_get_current_apic(void)
1479 if (current_cpu) {
1480 X86CPU *cpu = X86_CPU(current_cpu);
1481 return cpu->apic_state;
1482 } else {
1483 return NULL;
1487 void pc_acpi_smi_interrupt(void *opaque, int irq, int level)
1489 X86CPU *cpu = opaque;
1491 if (level) {
1492 cpu_interrupt(CPU(cpu), CPU_INTERRUPT_SMI);
1496 static void pc_new_cpu(const char *typename, int64_t apic_id, Error **errp)
1498 Object *cpu = NULL;
1499 Error *local_err = NULL;
1501 cpu = object_new(typename);
1503 object_property_set_uint(cpu, apic_id, "apic-id", &local_err);
1504 object_property_set_bool(cpu, true, "realized", &local_err);
1506 object_unref(cpu);
1507 error_propagate(errp, local_err);
1510 void pc_hot_add_cpu(const int64_t id, Error **errp)
1512 MachineState *ms = MACHINE(qdev_get_machine());
1513 int64_t apic_id = x86_cpu_apic_id_from_index(id);
1514 Error *local_err = NULL;
1516 if (id < 0) {
1517 error_setg(errp, "Invalid CPU id: %" PRIi64, id);
1518 return;
1521 if (apic_id >= ACPI_CPU_HOTPLUG_ID_LIMIT) {
1522 error_setg(errp, "Unable to add CPU: %" PRIi64
1523 ", resulting APIC ID (%" PRIi64 ") is too large",
1524 id, apic_id);
1525 return;
1528 pc_new_cpu(ms->cpu_type, apic_id, &local_err);
1529 if (local_err) {
1530 error_propagate(errp, local_err);
1531 return;
1535 void pc_cpus_init(PCMachineState *pcms)
1537 int i;
1538 const CPUArchIdList *possible_cpus;
1539 MachineState *ms = MACHINE(pcms);
1540 MachineClass *mc = MACHINE_GET_CLASS(pcms);
1542 /* Calculates the limit to CPU APIC ID values
1544 * Limit for the APIC ID value, so that all
1545 * CPU APIC IDs are < pcms->apic_id_limit.
1547 * This is used for FW_CFG_MAX_CPUS. See comments on bochs_bios_init().
1549 pcms->apic_id_limit = x86_cpu_apic_id_from_index(max_cpus - 1) + 1;
1550 possible_cpus = mc->possible_cpu_arch_ids(ms);
1551 for (i = 0; i < smp_cpus; i++) {
1552 pc_new_cpu(possible_cpus->cpus[i].type, possible_cpus->cpus[i].arch_id,
1553 &error_fatal);
1557 static void pc_build_feature_control_file(PCMachineState *pcms)
1559 MachineState *ms = MACHINE(pcms);
1560 X86CPU *cpu = X86_CPU(ms->possible_cpus->cpus[0].cpu);
1561 CPUX86State *env = &cpu->env;
1562 uint32_t unused, ecx, edx;
1563 uint64_t feature_control_bits = 0;
1564 uint64_t *val;
1566 cpu_x86_cpuid(env, 1, 0, &unused, &unused, &ecx, &edx);
1567 if (ecx & CPUID_EXT_VMX) {
1568 feature_control_bits |= FEATURE_CONTROL_VMXON_ENABLED_OUTSIDE_SMX;
1571 if ((edx & (CPUID_EXT2_MCE | CPUID_EXT2_MCA)) ==
1572 (CPUID_EXT2_MCE | CPUID_EXT2_MCA) &&
1573 (env->mcg_cap & MCG_LMCE_P)) {
1574 feature_control_bits |= FEATURE_CONTROL_LMCE;
1577 if (!feature_control_bits) {
1578 return;
1581 val = g_malloc(sizeof(*val));
1582 *val = cpu_to_le64(feature_control_bits | FEATURE_CONTROL_LOCKED);
1583 fw_cfg_add_file(pcms->fw_cfg, "etc/msr_feature_control", val, sizeof(*val));
1586 static void rtc_set_cpus_count(ISADevice *rtc, uint16_t cpus_count)
1588 if (cpus_count > 0xff) {
1589 /* If the number of CPUs can't be represented in 8 bits, the
1590 * BIOS must use "FW_CFG_NB_CPUS". Set RTC field to 0 just
1591 * to make old BIOSes fail more predictably.
1593 rtc_set_memory(rtc, 0x5f, 0);
1594 } else {
1595 rtc_set_memory(rtc, 0x5f, cpus_count - 1);
1599 static
1600 void pc_machine_done(Notifier *notifier, void *data)
1602 PCMachineState *pcms = container_of(notifier,
1603 PCMachineState, machine_done);
1604 PCIBus *bus = pcms->bus;
1606 /* set the number of CPUs */
1607 rtc_set_cpus_count(pcms->rtc, pcms->boot_cpus);
1609 if (bus) {
1610 int extra_hosts = 0;
1612 QLIST_FOREACH(bus, &bus->child, sibling) {
1613 /* look for expander root buses */
1614 if (pci_bus_is_root(bus)) {
1615 extra_hosts++;
1618 if (extra_hosts && pcms->fw_cfg) {
1619 uint64_t *val = g_malloc(sizeof(*val));
1620 *val = cpu_to_le64(extra_hosts);
1621 fw_cfg_add_file(pcms->fw_cfg,
1622 "etc/extra-pci-roots", val, sizeof(*val));
1626 acpi_setup();
1627 if (pcms->fw_cfg) {
1628 pc_build_smbios(pcms);
1629 pc_build_feature_control_file(pcms);
1630 /* update FW_CFG_NB_CPUS to account for -device added CPUs */
1631 fw_cfg_modify_i16(pcms->fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
1634 if (pcms->apic_id_limit > 255 && !xen_enabled()) {
1635 IntelIOMMUState *iommu = INTEL_IOMMU_DEVICE(x86_iommu_get_default());
1637 if (!iommu || !x86_iommu_ir_supported(X86_IOMMU_DEVICE(iommu)) ||
1638 iommu->intr_eim != ON_OFF_AUTO_ON) {
1639 error_report("current -smp configuration requires "
1640 "Extended Interrupt Mode enabled. "
1641 "You can add an IOMMU using: "
1642 "-device intel-iommu,intremap=on,eim=on");
1643 exit(EXIT_FAILURE);
1648 void pc_guest_info_init(PCMachineState *pcms)
1650 int i;
1652 pcms->apic_xrupt_override = kvm_allows_irq0_override();
1653 pcms->numa_nodes = nb_numa_nodes;
1654 pcms->node_mem = g_malloc0(pcms->numa_nodes *
1655 sizeof *pcms->node_mem);
1656 for (i = 0; i < nb_numa_nodes; i++) {
1657 pcms->node_mem[i] = numa_info[i].node_mem;
1660 pcms->machine_done.notify = pc_machine_done;
1661 qemu_add_machine_init_done_notifier(&pcms->machine_done);
1664 /* setup pci memory address space mapping into system address space */
1665 void pc_pci_as_mapping_init(Object *owner, MemoryRegion *system_memory,
1666 MemoryRegion *pci_address_space)
1668 /* Set to lower priority than RAM */
1669 memory_region_add_subregion_overlap(system_memory, 0x0,
1670 pci_address_space, -1);
1673 void pc_acpi_init(const char *default_dsdt)
1675 char *filename;
1677 if (acpi_tables != NULL) {
1678 /* manually set via -acpitable, leave it alone */
1679 return;
1682 filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, default_dsdt);
1683 if (filename == NULL) {
1684 warn_report("failed to find %s", default_dsdt);
1685 } else {
1686 QemuOpts *opts = qemu_opts_create(qemu_find_opts("acpi"), NULL, 0,
1687 &error_abort);
1688 Error *err = NULL;
1690 qemu_opt_set(opts, "file", filename, &error_abort);
1692 acpi_table_add_builtin(opts, &err);
1693 if (err) {
1694 warn_reportf_err(err, "failed to load %s: ", filename);
1696 g_free(filename);
1700 void xen_load_linux(PCMachineState *pcms)
1702 int i;
1703 FWCfgState *fw_cfg;
1705 assert(MACHINE(pcms)->kernel_filename != NULL);
1707 fw_cfg = fw_cfg_init_io(FW_CFG_IO_BASE);
1708 fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
1709 rom_set_fw(fw_cfg);
1711 load_linux(pcms, fw_cfg);
1712 for (i = 0; i < nb_option_roms; i++) {
1713 assert(!strcmp(option_rom[i].name, "linuxboot.bin") ||
1714 !strcmp(option_rom[i].name, "linuxboot_dma.bin") ||
1715 !strcmp(option_rom[i].name, "multiboot.bin"));
1716 rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1718 pcms->fw_cfg = fw_cfg;
1721 void pc_memory_init(PCMachineState *pcms,
1722 MemoryRegion *system_memory,
1723 MemoryRegion *rom_memory,
1724 MemoryRegion **ram_memory)
1726 int linux_boot, i;
1727 MemoryRegion *ram, *option_rom_mr;
1728 MemoryRegion *ram_below_4g, *ram_above_4g;
1729 FWCfgState *fw_cfg;
1730 MachineState *machine = MACHINE(pcms);
1731 PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1733 assert(machine->ram_size == pcms->below_4g_mem_size +
1734 pcms->above_4g_mem_size);
1736 linux_boot = (machine->kernel_filename != NULL);
1738 /* Allocate RAM. We allocate it as a single memory region and use
1739 * aliases to address portions of it, mostly for backwards compatibility
1740 * with older qemus that used qemu_ram_alloc().
1742 ram = g_malloc(sizeof(*ram));
1743 memory_region_allocate_system_memory(ram, NULL, "pc.ram",
1744 machine->ram_size);
1745 *ram_memory = ram;
1746 ram_below_4g = g_malloc(sizeof(*ram_below_4g));
1747 memory_region_init_alias(ram_below_4g, NULL, "ram-below-4g", ram,
1748 0, pcms->below_4g_mem_size);
1749 memory_region_add_subregion(system_memory, 0, ram_below_4g);
1750 e820_add_entry(0, pcms->below_4g_mem_size, E820_RAM);
1751 if (pcms->above_4g_mem_size > 0) {
1752 ram_above_4g = g_malloc(sizeof(*ram_above_4g));
1753 memory_region_init_alias(ram_above_4g, NULL, "ram-above-4g", ram,
1754 pcms->below_4g_mem_size,
1755 pcms->above_4g_mem_size);
1756 memory_region_add_subregion(system_memory, 0x100000000ULL,
1757 ram_above_4g);
1758 e820_add_entry(0x100000000ULL, pcms->above_4g_mem_size, E820_RAM);
1761 if (!pcmc->has_reserved_memory &&
1762 (machine->ram_slots ||
1763 (machine->maxram_size > machine->ram_size))) {
1764 MachineClass *mc = MACHINE_GET_CLASS(machine);
1766 error_report("\"-memory 'slots|maxmem'\" is not supported by: %s",
1767 mc->name);
1768 exit(EXIT_FAILURE);
1771 /* always allocate the device memory information */
1772 machine->device_memory = g_malloc0(sizeof(*machine->device_memory));
1774 /* initialize device memory address space */
1775 if (pcmc->has_reserved_memory &&
1776 (machine->ram_size < machine->maxram_size)) {
1777 ram_addr_t device_mem_size = machine->maxram_size - machine->ram_size;
1779 if (machine->ram_slots > ACPI_MAX_RAM_SLOTS) {
1780 error_report("unsupported amount of memory slots: %"PRIu64,
1781 machine->ram_slots);
1782 exit(EXIT_FAILURE);
1785 if (QEMU_ALIGN_UP(machine->maxram_size,
1786 TARGET_PAGE_SIZE) != machine->maxram_size) {
1787 error_report("maximum memory size must by aligned to multiple of "
1788 "%d bytes", TARGET_PAGE_SIZE);
1789 exit(EXIT_FAILURE);
1792 machine->device_memory->base =
1793 ROUND_UP(0x100000000ULL + pcms->above_4g_mem_size, 1 * GiB);
1795 if (pcmc->enforce_aligned_dimm) {
1796 /* size device region assuming 1G page max alignment per slot */
1797 device_mem_size += (1 * GiB) * machine->ram_slots;
1800 if ((machine->device_memory->base + device_mem_size) <
1801 device_mem_size) {
1802 error_report("unsupported amount of maximum memory: " RAM_ADDR_FMT,
1803 machine->maxram_size);
1804 exit(EXIT_FAILURE);
1807 memory_region_init(&machine->device_memory->mr, OBJECT(pcms),
1808 "device-memory", device_mem_size);
1809 memory_region_add_subregion(system_memory, machine->device_memory->base,
1810 &machine->device_memory->mr);
1813 /* Initialize PC system firmware */
1814 pc_system_firmware_init(rom_memory, !pcmc->pci_enabled);
1816 option_rom_mr = g_malloc(sizeof(*option_rom_mr));
1817 memory_region_init_ram(option_rom_mr, NULL, "pc.rom", PC_ROM_SIZE,
1818 &error_fatal);
1819 if (pcmc->pci_enabled) {
1820 memory_region_set_readonly(option_rom_mr, true);
1822 memory_region_add_subregion_overlap(rom_memory,
1823 PC_ROM_MIN_VGA,
1824 option_rom_mr,
1827 fw_cfg = bochs_bios_init(&address_space_memory, pcms);
1829 rom_set_fw(fw_cfg);
1831 if (pcmc->has_reserved_memory && machine->device_memory->base) {
1832 uint64_t *val = g_malloc(sizeof(*val));
1833 PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1834 uint64_t res_mem_end = machine->device_memory->base;
1836 if (!pcmc->broken_reserved_end) {
1837 res_mem_end += memory_region_size(&machine->device_memory->mr);
1839 *val = cpu_to_le64(ROUND_UP(res_mem_end, 1 * GiB));
1840 fw_cfg_add_file(fw_cfg, "etc/reserved-memory-end", val, sizeof(*val));
1843 if (linux_boot) {
1844 load_linux(pcms, fw_cfg);
1847 for (i = 0; i < nb_option_roms; i++) {
1848 rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1850 pcms->fw_cfg = fw_cfg;
1852 /* Init default IOAPIC address space */
1853 pcms->ioapic_as = &address_space_memory;
1857 * The 64bit pci hole starts after "above 4G RAM" and
1858 * potentially the space reserved for memory hotplug.
1860 uint64_t pc_pci_hole64_start(void)
1862 PCMachineState *pcms = PC_MACHINE(qdev_get_machine());
1863 PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1864 MachineState *ms = MACHINE(pcms);
1865 uint64_t hole64_start = 0;
1867 if (pcmc->has_reserved_memory && ms->device_memory->base) {
1868 hole64_start = ms->device_memory->base;
1869 if (!pcmc->broken_reserved_end) {
1870 hole64_start += memory_region_size(&ms->device_memory->mr);
1872 } else {
1873 hole64_start = 0x100000000ULL + pcms->above_4g_mem_size;
1876 return ROUND_UP(hole64_start, 1 * GiB);
1879 qemu_irq pc_allocate_cpu_irq(void)
1881 return qemu_allocate_irq(pic_irq_request, NULL, 0);
1884 DeviceState *pc_vga_init(ISABus *isa_bus, PCIBus *pci_bus)
1886 DeviceState *dev = NULL;
1888 rom_set_order_override(FW_CFG_ORDER_OVERRIDE_VGA);
1889 if (pci_bus) {
1890 PCIDevice *pcidev = pci_vga_init(pci_bus);
1891 dev = pcidev ? &pcidev->qdev : NULL;
1892 } else if (isa_bus) {
1893 ISADevice *isadev = isa_vga_init(isa_bus);
1894 dev = isadev ? DEVICE(isadev) : NULL;
1896 rom_reset_order_override();
1897 return dev;
1900 static const MemoryRegionOps ioport80_io_ops = {
1901 .write = ioport80_write,
1902 .read = ioport80_read,
1903 .endianness = DEVICE_NATIVE_ENDIAN,
1904 .impl = {
1905 .min_access_size = 1,
1906 .max_access_size = 1,
1910 static const MemoryRegionOps ioportF0_io_ops = {
1911 .write = ioportF0_write,
1912 .read = ioportF0_read,
1913 .endianness = DEVICE_NATIVE_ENDIAN,
1914 .impl = {
1915 .min_access_size = 1,
1916 .max_access_size = 1,
1920 static void pc_superio_init(ISABus *isa_bus, bool create_fdctrl, bool no_vmport)
1922 int i;
1923 DriveInfo *fd[MAX_FD];
1924 qemu_irq *a20_line;
1925 ISADevice *i8042, *port92, *vmmouse;
1927 serial_hds_isa_init(isa_bus, 0, MAX_ISA_SERIAL_PORTS);
1928 parallel_hds_isa_init(isa_bus, MAX_PARALLEL_PORTS);
1930 for (i = 0; i < MAX_FD; i++) {
1931 fd[i] = drive_get(IF_FLOPPY, 0, i);
1932 create_fdctrl |= !!fd[i];
1934 if (create_fdctrl) {
1935 fdctrl_init_isa(isa_bus, fd);
1938 i8042 = isa_create_simple(isa_bus, "i8042");
1939 if (!no_vmport) {
1940 vmport_init(isa_bus);
1941 vmmouse = isa_try_create(isa_bus, "vmmouse");
1942 } else {
1943 vmmouse = NULL;
1945 if (vmmouse) {
1946 DeviceState *dev = DEVICE(vmmouse);
1947 qdev_prop_set_ptr(dev, "ps2_mouse", i8042);
1948 qdev_init_nofail(dev);
1950 port92 = isa_create_simple(isa_bus, "port92");
1952 a20_line = qemu_allocate_irqs(handle_a20_line_change, first_cpu, 2);
1953 i8042_setup_a20_line(i8042, a20_line[0]);
1954 port92_init(port92, a20_line[1]);
1955 g_free(a20_line);
1958 void pc_basic_device_init(ISABus *isa_bus, qemu_irq *gsi,
1959 ISADevice **rtc_state,
1960 bool create_fdctrl,
1961 bool no_vmport,
1962 bool has_pit,
1963 uint32_t hpet_irqs)
1965 int i;
1966 DeviceState *hpet = NULL;
1967 int pit_isa_irq = 0;
1968 qemu_irq pit_alt_irq = NULL;
1969 qemu_irq rtc_irq = NULL;
1970 ISADevice *pit = NULL;
1971 MemoryRegion *ioport80_io = g_new(MemoryRegion, 1);
1972 MemoryRegion *ioportF0_io = g_new(MemoryRegion, 1);
1974 memory_region_init_io(ioport80_io, NULL, &ioport80_io_ops, NULL, "ioport80", 1);
1975 memory_region_add_subregion(isa_bus->address_space_io, 0x80, ioport80_io);
1977 memory_region_init_io(ioportF0_io, NULL, &ioportF0_io_ops, NULL, "ioportF0", 1);
1978 memory_region_add_subregion(isa_bus->address_space_io, 0xf0, ioportF0_io);
1981 * Check if an HPET shall be created.
1983 * Without KVM_CAP_PIT_STATE2, we cannot switch off the in-kernel PIT
1984 * when the HPET wants to take over. Thus we have to disable the latter.
1986 if (!no_hpet && (!kvm_irqchip_in_kernel() || kvm_has_pit_state2())) {
1987 /* In order to set property, here not using sysbus_try_create_simple */
1988 hpet = qdev_try_create(NULL, TYPE_HPET);
1989 if (hpet) {
1990 /* For pc-piix-*, hpet's intcap is always IRQ2. For pc-q35-1.7
1991 * and earlier, use IRQ2 for compat. Otherwise, use IRQ16~23,
1992 * IRQ8 and IRQ2.
1994 uint8_t compat = object_property_get_uint(OBJECT(hpet),
1995 HPET_INTCAP, NULL);
1996 if (!compat) {
1997 qdev_prop_set_uint32(hpet, HPET_INTCAP, hpet_irqs);
1999 qdev_init_nofail(hpet);
2000 sysbus_mmio_map(SYS_BUS_DEVICE(hpet), 0, HPET_BASE);
2002 for (i = 0; i < GSI_NUM_PINS; i++) {
2003 sysbus_connect_irq(SYS_BUS_DEVICE(hpet), i, gsi[i]);
2005 pit_isa_irq = -1;
2006 pit_alt_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_PIT_INT);
2007 rtc_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_RTC_INT);
2010 *rtc_state = mc146818_rtc_init(isa_bus, 2000, rtc_irq);
2012 qemu_register_boot_set(pc_boot_set, *rtc_state);
2014 if (!xen_enabled() && has_pit) {
2015 if (kvm_pit_in_kernel()) {
2016 pit = kvm_pit_init(isa_bus, 0x40);
2017 } else {
2018 pit = i8254_pit_init(isa_bus, 0x40, pit_isa_irq, pit_alt_irq);
2020 if (hpet) {
2021 /* connect PIT to output control line of the HPET */
2022 qdev_connect_gpio_out(hpet, 0, qdev_get_gpio_in(DEVICE(pit), 0));
2024 pcspk_init(isa_bus, pit);
2027 i8257_dma_init(isa_bus, 0);
2029 /* Super I/O */
2030 pc_superio_init(isa_bus, create_fdctrl, no_vmport);
2033 void pc_nic_init(PCMachineClass *pcmc, ISABus *isa_bus, PCIBus *pci_bus)
2035 int i;
2037 rom_set_order_override(FW_CFG_ORDER_OVERRIDE_NIC);
2038 for (i = 0; i < nb_nics; i++) {
2039 NICInfo *nd = &nd_table[i];
2040 const char *model = nd->model ? nd->model : pcmc->default_nic_model;
2042 if (g_str_equal(model, "ne2k_isa")) {
2043 pc_init_ne2k_isa(isa_bus, nd);
2044 } else {
2045 pci_nic_init_nofail(nd, pci_bus, model, NULL);
2048 rom_reset_order_override();
2051 void ioapic_init_gsi(GSIState *gsi_state, const char *parent_name)
2053 DeviceState *dev;
2054 SysBusDevice *d;
2055 unsigned int i;
2057 if (kvm_ioapic_in_kernel()) {
2058 dev = qdev_create(NULL, TYPE_KVM_IOAPIC);
2059 } else {
2060 dev = qdev_create(NULL, TYPE_IOAPIC);
2062 if (parent_name) {
2063 object_property_add_child(object_resolve_path(parent_name, NULL),
2064 "ioapic", OBJECT(dev), NULL);
2066 qdev_init_nofail(dev);
2067 d = SYS_BUS_DEVICE(dev);
2068 sysbus_mmio_map(d, 0, IO_APIC_DEFAULT_ADDRESS);
2070 for (i = 0; i < IOAPIC_NUM_PINS; i++) {
2071 gsi_state->ioapic_irq[i] = qdev_get_gpio_in(dev, i);
2075 static void pc_memory_pre_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
2076 Error **errp)
2078 const PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2079 const PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
2080 const bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2081 const uint64_t legacy_align = TARGET_PAGE_SIZE;
2084 * When -no-acpi is used with Q35 machine type, no ACPI is built,
2085 * but pcms->acpi_dev is still created. Check !acpi_enabled in
2086 * addition to cover this case.
2088 if (!pcms->acpi_dev || !acpi_enabled) {
2089 error_setg(errp,
2090 "memory hotplug is not enabled: missing acpi device or acpi disabled");
2091 return;
2094 if (is_nvdimm && !pcms->acpi_nvdimm_state.is_enabled) {
2095 error_setg(errp, "nvdimm is not enabled: missing 'nvdimm' in '-M'");
2096 return;
2099 pc_dimm_pre_plug(PC_DIMM(dev), MACHINE(hotplug_dev),
2100 pcmc->enforce_aligned_dimm ? NULL : &legacy_align, errp);
2103 static void pc_memory_plug(HotplugHandler *hotplug_dev,
2104 DeviceState *dev, Error **errp)
2106 HotplugHandlerClass *hhc;
2107 Error *local_err = NULL;
2108 PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2109 bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2111 pc_dimm_plug(PC_DIMM(dev), MACHINE(pcms), &local_err);
2112 if (local_err) {
2113 goto out;
2116 if (is_nvdimm) {
2117 nvdimm_plug(&pcms->acpi_nvdimm_state);
2120 hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
2121 hhc->plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &error_abort);
2122 out:
2123 error_propagate(errp, local_err);
2126 static void pc_memory_unplug_request(HotplugHandler *hotplug_dev,
2127 DeviceState *dev, Error **errp)
2129 HotplugHandlerClass *hhc;
2130 Error *local_err = NULL;
2131 PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2134 * When -no-acpi is used with Q35 machine type, no ACPI is built,
2135 * but pcms->acpi_dev is still created. Check !acpi_enabled in
2136 * addition to cover this case.
2138 if (!pcms->acpi_dev || !acpi_enabled) {
2139 error_setg(&local_err,
2140 "memory hotplug is not enabled: missing acpi device or acpi disabled");
2141 goto out;
2144 if (object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM)) {
2145 error_setg(&local_err,
2146 "nvdimm device hot unplug is not supported yet.");
2147 goto out;
2150 hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
2151 hhc->unplug_request(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
2153 out:
2154 error_propagate(errp, local_err);
2157 static void pc_memory_unplug(HotplugHandler *hotplug_dev,
2158 DeviceState *dev, Error **errp)
2160 PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2161 HotplugHandlerClass *hhc;
2162 Error *local_err = NULL;
2164 hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
2165 hhc->unplug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
2167 if (local_err) {
2168 goto out;
2171 pc_dimm_unplug(PC_DIMM(dev), MACHINE(pcms));
2172 object_unparent(OBJECT(dev));
2174 out:
2175 error_propagate(errp, local_err);
2178 static int pc_apic_cmp(const void *a, const void *b)
2180 CPUArchId *apic_a = (CPUArchId *)a;
2181 CPUArchId *apic_b = (CPUArchId *)b;
2183 return apic_a->arch_id - apic_b->arch_id;
2186 /* returns pointer to CPUArchId descriptor that matches CPU's apic_id
2187 * in ms->possible_cpus->cpus, if ms->possible_cpus->cpus has no
2188 * entry corresponding to CPU's apic_id returns NULL.
2190 static CPUArchId *pc_find_cpu_slot(MachineState *ms, uint32_t id, int *idx)
2192 CPUArchId apic_id, *found_cpu;
2194 apic_id.arch_id = id;
2195 found_cpu = bsearch(&apic_id, ms->possible_cpus->cpus,
2196 ms->possible_cpus->len, sizeof(*ms->possible_cpus->cpus),
2197 pc_apic_cmp);
2198 if (found_cpu && idx) {
2199 *idx = found_cpu - ms->possible_cpus->cpus;
2201 return found_cpu;
2204 static void pc_cpu_plug(HotplugHandler *hotplug_dev,
2205 DeviceState *dev, Error **errp)
2207 CPUArchId *found_cpu;
2208 HotplugHandlerClass *hhc;
2209 Error *local_err = NULL;
2210 X86CPU *cpu = X86_CPU(dev);
2211 PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2213 if (pcms->acpi_dev) {
2214 hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
2215 hhc->plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
2216 if (local_err) {
2217 goto out;
2221 /* increment the number of CPUs */
2222 pcms->boot_cpus++;
2223 if (pcms->rtc) {
2224 rtc_set_cpus_count(pcms->rtc, pcms->boot_cpus);
2226 if (pcms->fw_cfg) {
2227 fw_cfg_modify_i16(pcms->fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
2230 found_cpu = pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, NULL);
2231 found_cpu->cpu = OBJECT(dev);
2232 out:
2233 error_propagate(errp, local_err);
2235 static void pc_cpu_unplug_request_cb(HotplugHandler *hotplug_dev,
2236 DeviceState *dev, Error **errp)
2238 int idx = -1;
2239 HotplugHandlerClass *hhc;
2240 Error *local_err = NULL;
2241 X86CPU *cpu = X86_CPU(dev);
2242 PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2244 if (!pcms->acpi_dev) {
2245 error_setg(&local_err, "CPU hot unplug not supported without ACPI");
2246 goto out;
2249 pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, &idx);
2250 assert(idx != -1);
2251 if (idx == 0) {
2252 error_setg(&local_err, "Boot CPU is unpluggable");
2253 goto out;
2256 hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
2257 hhc->unplug_request(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
2259 if (local_err) {
2260 goto out;
2263 out:
2264 error_propagate(errp, local_err);
2268 static void pc_cpu_unplug_cb(HotplugHandler *hotplug_dev,
2269 DeviceState *dev, Error **errp)
2271 CPUArchId *found_cpu;
2272 HotplugHandlerClass *hhc;
2273 Error *local_err = NULL;
2274 X86CPU *cpu = X86_CPU(dev);
2275 PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2277 hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
2278 hhc->unplug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
2280 if (local_err) {
2281 goto out;
2284 found_cpu = pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, NULL);
2285 found_cpu->cpu = NULL;
2286 object_unparent(OBJECT(dev));
2288 /* decrement the number of CPUs */
2289 pcms->boot_cpus--;
2290 /* Update the number of CPUs in CMOS */
2291 rtc_set_cpus_count(pcms->rtc, pcms->boot_cpus);
2292 fw_cfg_modify_i16(pcms->fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
2293 out:
2294 error_propagate(errp, local_err);
2297 static void pc_cpu_pre_plug(HotplugHandler *hotplug_dev,
2298 DeviceState *dev, Error **errp)
2300 int idx;
2301 CPUState *cs;
2302 CPUArchId *cpu_slot;
2303 X86CPUTopoInfo topo;
2304 X86CPU *cpu = X86_CPU(dev);
2305 MachineState *ms = MACHINE(hotplug_dev);
2306 PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2308 if(!object_dynamic_cast(OBJECT(cpu), ms->cpu_type)) {
2309 error_setg(errp, "Invalid CPU type, expected cpu type: '%s'",
2310 ms->cpu_type);
2311 return;
2314 /* if APIC ID is not set, set it based on socket/core/thread properties */
2315 if (cpu->apic_id == UNASSIGNED_APIC_ID) {
2316 int max_socket = (max_cpus - 1) / smp_threads / smp_cores;
2318 if (cpu->socket_id < 0) {
2319 error_setg(errp, "CPU socket-id is not set");
2320 return;
2321 } else if (cpu->socket_id > max_socket) {
2322 error_setg(errp, "Invalid CPU socket-id: %u must be in range 0:%u",
2323 cpu->socket_id, max_socket);
2324 return;
2326 if (cpu->core_id < 0) {
2327 error_setg(errp, "CPU core-id is not set");
2328 return;
2329 } else if (cpu->core_id > (smp_cores - 1)) {
2330 error_setg(errp, "Invalid CPU core-id: %u must be in range 0:%u",
2331 cpu->core_id, smp_cores - 1);
2332 return;
2334 if (cpu->thread_id < 0) {
2335 error_setg(errp, "CPU thread-id is not set");
2336 return;
2337 } else if (cpu->thread_id > (smp_threads - 1)) {
2338 error_setg(errp, "Invalid CPU thread-id: %u must be in range 0:%u",
2339 cpu->thread_id, smp_threads - 1);
2340 return;
2343 topo.pkg_id = cpu->socket_id;
2344 topo.core_id = cpu->core_id;
2345 topo.smt_id = cpu->thread_id;
2346 cpu->apic_id = apicid_from_topo_ids(smp_cores, smp_threads, &topo);
2349 cpu_slot = pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, &idx);
2350 if (!cpu_slot) {
2351 MachineState *ms = MACHINE(pcms);
2353 x86_topo_ids_from_apicid(cpu->apic_id, smp_cores, smp_threads, &topo);
2354 error_setg(errp, "Invalid CPU [socket: %u, core: %u, thread: %u] with"
2355 " APIC ID %" PRIu32 ", valid index range 0:%d",
2356 topo.pkg_id, topo.core_id, topo.smt_id, cpu->apic_id,
2357 ms->possible_cpus->len - 1);
2358 return;
2361 if (cpu_slot->cpu) {
2362 error_setg(errp, "CPU[%d] with APIC ID %" PRIu32 " exists",
2363 idx, cpu->apic_id);
2364 return;
2367 /* if 'address' properties socket-id/core-id/thread-id are not set, set them
2368 * so that machine_query_hotpluggable_cpus would show correct values
2370 /* TODO: move socket_id/core_id/thread_id checks into x86_cpu_realizefn()
2371 * once -smp refactoring is complete and there will be CPU private
2372 * CPUState::nr_cores and CPUState::nr_threads fields instead of globals */
2373 x86_topo_ids_from_apicid(cpu->apic_id, smp_cores, smp_threads, &topo);
2374 if (cpu->socket_id != -1 && cpu->socket_id != topo.pkg_id) {
2375 error_setg(errp, "property socket-id: %u doesn't match set apic-id:"
2376 " 0x%x (socket-id: %u)", cpu->socket_id, cpu->apic_id, topo.pkg_id);
2377 return;
2379 cpu->socket_id = topo.pkg_id;
2381 if (cpu->core_id != -1 && cpu->core_id != topo.core_id) {
2382 error_setg(errp, "property core-id: %u doesn't match set apic-id:"
2383 " 0x%x (core-id: %u)", cpu->core_id, cpu->apic_id, topo.core_id);
2384 return;
2386 cpu->core_id = topo.core_id;
2388 if (cpu->thread_id != -1 && cpu->thread_id != topo.smt_id) {
2389 error_setg(errp, "property thread-id: %u doesn't match set apic-id:"
2390 " 0x%x (thread-id: %u)", cpu->thread_id, cpu->apic_id, topo.smt_id);
2391 return;
2393 cpu->thread_id = topo.smt_id;
2395 if (cpu->hyperv_vpindex && !kvm_hv_vpindex_settable()) {
2396 error_setg(errp, "kernel doesn't allow setting HyperV VP_INDEX");
2397 return;
2400 cs = CPU(cpu);
2401 cs->cpu_index = idx;
2403 numa_cpu_pre_plug(cpu_slot, dev, errp);
2406 static void pc_machine_device_pre_plug_cb(HotplugHandler *hotplug_dev,
2407 DeviceState *dev, Error **errp)
2409 if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2410 pc_memory_pre_plug(hotplug_dev, dev, errp);
2411 } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2412 pc_cpu_pre_plug(hotplug_dev, dev, errp);
2416 static void pc_machine_device_plug_cb(HotplugHandler *hotplug_dev,
2417 DeviceState *dev, Error **errp)
2419 if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2420 pc_memory_plug(hotplug_dev, dev, errp);
2421 } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2422 pc_cpu_plug(hotplug_dev, dev, errp);
2426 static void pc_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
2427 DeviceState *dev, Error **errp)
2429 if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2430 pc_memory_unplug_request(hotplug_dev, dev, errp);
2431 } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2432 pc_cpu_unplug_request_cb(hotplug_dev, dev, errp);
2433 } else {
2434 error_setg(errp, "acpi: device unplug request for not supported device"
2435 " type: %s", object_get_typename(OBJECT(dev)));
2439 static void pc_machine_device_unplug_cb(HotplugHandler *hotplug_dev,
2440 DeviceState *dev, Error **errp)
2442 if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2443 pc_memory_unplug(hotplug_dev, dev, errp);
2444 } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2445 pc_cpu_unplug_cb(hotplug_dev, dev, errp);
2446 } else {
2447 error_setg(errp, "acpi: device unplug for not supported device"
2448 " type: %s", object_get_typename(OBJECT(dev)));
2452 static HotplugHandler *pc_get_hotplug_handler(MachineState *machine,
2453 DeviceState *dev)
2455 if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) ||
2456 object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2457 return HOTPLUG_HANDLER(machine);
2460 return NULL;
2463 static void
2464 pc_machine_get_device_memory_region_size(Object *obj, Visitor *v,
2465 const char *name, void *opaque,
2466 Error **errp)
2468 MachineState *ms = MACHINE(obj);
2469 int64_t value = memory_region_size(&ms->device_memory->mr);
2471 visit_type_int(v, name, &value, errp);
2474 static void pc_machine_get_max_ram_below_4g(Object *obj, Visitor *v,
2475 const char *name, void *opaque,
2476 Error **errp)
2478 PCMachineState *pcms = PC_MACHINE(obj);
2479 uint64_t value = pcms->max_ram_below_4g;
2481 visit_type_size(v, name, &value, errp);
2484 static void pc_machine_set_max_ram_below_4g(Object *obj, Visitor *v,
2485 const char *name, void *opaque,
2486 Error **errp)
2488 PCMachineState *pcms = PC_MACHINE(obj);
2489 Error *error = NULL;
2490 uint64_t value;
2492 visit_type_size(v, name, &value, &error);
2493 if (error) {
2494 error_propagate(errp, error);
2495 return;
2497 if (value > 4 * GiB) {
2498 error_setg(&error,
2499 "Machine option 'max-ram-below-4g=%"PRIu64
2500 "' expects size less than or equal to 4G", value);
2501 error_propagate(errp, error);
2502 return;
2505 if (value < 1 * MiB) {
2506 warn_report("Only %" PRIu64 " bytes of RAM below the 4GiB boundary,"
2507 "BIOS may not work with less than 1MiB", value);
2510 pcms->max_ram_below_4g = value;
2513 static void pc_machine_get_vmport(Object *obj, Visitor *v, const char *name,
2514 void *opaque, Error **errp)
2516 PCMachineState *pcms = PC_MACHINE(obj);
2517 OnOffAuto vmport = pcms->vmport;
2519 visit_type_OnOffAuto(v, name, &vmport, errp);
2522 static void pc_machine_set_vmport(Object *obj, Visitor *v, const char *name,
2523 void *opaque, Error **errp)
2525 PCMachineState *pcms = PC_MACHINE(obj);
2527 visit_type_OnOffAuto(v, name, &pcms->vmport, errp);
2530 bool pc_machine_is_smm_enabled(PCMachineState *pcms)
2532 bool smm_available = false;
2534 if (pcms->smm == ON_OFF_AUTO_OFF) {
2535 return false;
2538 if (tcg_enabled() || qtest_enabled()) {
2539 smm_available = true;
2540 } else if (kvm_enabled()) {
2541 smm_available = kvm_has_smm();
2544 if (smm_available) {
2545 return true;
2548 if (pcms->smm == ON_OFF_AUTO_ON) {
2549 error_report("System Management Mode not supported by this hypervisor.");
2550 exit(1);
2552 return false;
2555 static void pc_machine_get_smm(Object *obj, Visitor *v, const char *name,
2556 void *opaque, Error **errp)
2558 PCMachineState *pcms = PC_MACHINE(obj);
2559 OnOffAuto smm = pcms->smm;
2561 visit_type_OnOffAuto(v, name, &smm, errp);
2564 static void pc_machine_set_smm(Object *obj, Visitor *v, const char *name,
2565 void *opaque, Error **errp)
2567 PCMachineState *pcms = PC_MACHINE(obj);
2569 visit_type_OnOffAuto(v, name, &pcms->smm, errp);
2572 static bool pc_machine_get_nvdimm(Object *obj, Error **errp)
2574 PCMachineState *pcms = PC_MACHINE(obj);
2576 return pcms->acpi_nvdimm_state.is_enabled;
2579 static void pc_machine_set_nvdimm(Object *obj, bool value, Error **errp)
2581 PCMachineState *pcms = PC_MACHINE(obj);
2583 pcms->acpi_nvdimm_state.is_enabled = value;
2586 static char *pc_machine_get_nvdimm_persistence(Object *obj, Error **errp)
2588 PCMachineState *pcms = PC_MACHINE(obj);
2590 return g_strdup(pcms->acpi_nvdimm_state.persistence_string);
2593 static void pc_machine_set_nvdimm_persistence(Object *obj, const char *value,
2594 Error **errp)
2596 PCMachineState *pcms = PC_MACHINE(obj);
2597 AcpiNVDIMMState *nvdimm_state = &pcms->acpi_nvdimm_state;
2599 if (strcmp(value, "cpu") == 0)
2600 nvdimm_state->persistence = 3;
2601 else if (strcmp(value, "mem-ctrl") == 0)
2602 nvdimm_state->persistence = 2;
2603 else {
2604 error_setg(errp, "-machine nvdimm-persistence=%s: unsupported option",
2605 value);
2606 return;
2609 g_free(nvdimm_state->persistence_string);
2610 nvdimm_state->persistence_string = g_strdup(value);
2613 static bool pc_machine_get_smbus(Object *obj, Error **errp)
2615 PCMachineState *pcms = PC_MACHINE(obj);
2617 return pcms->smbus_enabled;
2620 static void pc_machine_set_smbus(Object *obj, bool value, Error **errp)
2622 PCMachineState *pcms = PC_MACHINE(obj);
2624 pcms->smbus_enabled = value;
2627 static bool pc_machine_get_sata(Object *obj, Error **errp)
2629 PCMachineState *pcms = PC_MACHINE(obj);
2631 return pcms->sata_enabled;
2634 static void pc_machine_set_sata(Object *obj, bool value, Error **errp)
2636 PCMachineState *pcms = PC_MACHINE(obj);
2638 pcms->sata_enabled = value;
2641 static bool pc_machine_get_pit(Object *obj, Error **errp)
2643 PCMachineState *pcms = PC_MACHINE(obj);
2645 return pcms->pit_enabled;
2648 static void pc_machine_set_pit(Object *obj, bool value, Error **errp)
2650 PCMachineState *pcms = PC_MACHINE(obj);
2652 pcms->pit_enabled = value;
2655 static void pc_machine_initfn(Object *obj)
2657 PCMachineState *pcms = PC_MACHINE(obj);
2659 pcms->max_ram_below_4g = 0; /* use default */
2660 pcms->smm = ON_OFF_AUTO_AUTO;
2661 pcms->vmport = ON_OFF_AUTO_AUTO;
2662 /* nvdimm is disabled on default. */
2663 pcms->acpi_nvdimm_state.is_enabled = false;
2664 /* acpi build is enabled by default if machine supports it */
2665 pcms->acpi_build_enabled = PC_MACHINE_GET_CLASS(pcms)->has_acpi_build;
2666 pcms->smbus_enabled = true;
2667 pcms->sata_enabled = true;
2668 pcms->pit_enabled = true;
2671 static void pc_machine_reset(void)
2673 CPUState *cs;
2674 X86CPU *cpu;
2676 qemu_devices_reset();
2678 /* Reset APIC after devices have been reset to cancel
2679 * any changes that qemu_devices_reset() might have done.
2681 CPU_FOREACH(cs) {
2682 cpu = X86_CPU(cs);
2684 if (cpu->apic_state) {
2685 device_reset(cpu->apic_state);
2690 static CpuInstanceProperties
2691 pc_cpu_index_to_props(MachineState *ms, unsigned cpu_index)
2693 MachineClass *mc = MACHINE_GET_CLASS(ms);
2694 const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(ms);
2696 assert(cpu_index < possible_cpus->len);
2697 return possible_cpus->cpus[cpu_index].props;
2700 static int64_t pc_get_default_cpu_node_id(const MachineState *ms, int idx)
2702 X86CPUTopoInfo topo;
2704 assert(idx < ms->possible_cpus->len);
2705 x86_topo_ids_from_apicid(ms->possible_cpus->cpus[idx].arch_id,
2706 smp_cores, smp_threads, &topo);
2707 return topo.pkg_id % nb_numa_nodes;
2710 static const CPUArchIdList *pc_possible_cpu_arch_ids(MachineState *ms)
2712 int i;
2714 if (ms->possible_cpus) {
2716 * make sure that max_cpus hasn't changed since the first use, i.e.
2717 * -smp hasn't been parsed after it
2719 assert(ms->possible_cpus->len == max_cpus);
2720 return ms->possible_cpus;
2723 ms->possible_cpus = g_malloc0(sizeof(CPUArchIdList) +
2724 sizeof(CPUArchId) * max_cpus);
2725 ms->possible_cpus->len = max_cpus;
2726 for (i = 0; i < ms->possible_cpus->len; i++) {
2727 X86CPUTopoInfo topo;
2729 ms->possible_cpus->cpus[i].type = ms->cpu_type;
2730 ms->possible_cpus->cpus[i].vcpus_count = 1;
2731 ms->possible_cpus->cpus[i].arch_id = x86_cpu_apic_id_from_index(i);
2732 x86_topo_ids_from_apicid(ms->possible_cpus->cpus[i].arch_id,
2733 smp_cores, smp_threads, &topo);
2734 ms->possible_cpus->cpus[i].props.has_socket_id = true;
2735 ms->possible_cpus->cpus[i].props.socket_id = topo.pkg_id;
2736 ms->possible_cpus->cpus[i].props.has_core_id = true;
2737 ms->possible_cpus->cpus[i].props.core_id = topo.core_id;
2738 ms->possible_cpus->cpus[i].props.has_thread_id = true;
2739 ms->possible_cpus->cpus[i].props.thread_id = topo.smt_id;
2741 return ms->possible_cpus;
2744 static void x86_nmi(NMIState *n, int cpu_index, Error **errp)
2746 /* cpu index isn't used */
2747 CPUState *cs;
2749 CPU_FOREACH(cs) {
2750 X86CPU *cpu = X86_CPU(cs);
2752 if (!cpu->apic_state) {
2753 cpu_interrupt(cs, CPU_INTERRUPT_NMI);
2754 } else {
2755 apic_deliver_nmi(cpu->apic_state);
2760 static void pc_machine_class_init(ObjectClass *oc, void *data)
2762 MachineClass *mc = MACHINE_CLASS(oc);
2763 PCMachineClass *pcmc = PC_MACHINE_CLASS(oc);
2764 HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
2765 NMIClass *nc = NMI_CLASS(oc);
2767 pcmc->pci_enabled = true;
2768 pcmc->has_acpi_build = true;
2769 pcmc->rsdp_in_ram = true;
2770 pcmc->smbios_defaults = true;
2771 pcmc->smbios_uuid_encoded = true;
2772 pcmc->gigabyte_align = true;
2773 pcmc->has_reserved_memory = true;
2774 pcmc->kvmclock_enabled = true;
2775 pcmc->enforce_aligned_dimm = true;
2776 /* BIOS ACPI tables: 128K. Other BIOS datastructures: less than 4K reported
2777 * to be used at the moment, 32K should be enough for a while. */
2778 pcmc->acpi_data_size = 0x20000 + 0x8000;
2779 pcmc->save_tsc_khz = true;
2780 pcmc->linuxboot_dma_enabled = true;
2781 assert(!mc->get_hotplug_handler);
2782 mc->get_hotplug_handler = pc_get_hotplug_handler;
2783 mc->cpu_index_to_instance_props = pc_cpu_index_to_props;
2784 mc->get_default_cpu_node_id = pc_get_default_cpu_node_id;
2785 mc->possible_cpu_arch_ids = pc_possible_cpu_arch_ids;
2786 mc->auto_enable_numa_with_memhp = true;
2787 mc->has_hotpluggable_cpus = true;
2788 mc->default_boot_order = "cad";
2789 mc->hot_add_cpu = pc_hot_add_cpu;
2790 mc->block_default_type = IF_IDE;
2791 mc->max_cpus = 255;
2792 mc->reset = pc_machine_reset;
2793 hc->pre_plug = pc_machine_device_pre_plug_cb;
2794 hc->plug = pc_machine_device_plug_cb;
2795 hc->unplug_request = pc_machine_device_unplug_request_cb;
2796 hc->unplug = pc_machine_device_unplug_cb;
2797 nc->nmi_monitor_handler = x86_nmi;
2798 mc->default_cpu_type = TARGET_DEFAULT_CPU_TYPE;
2800 object_class_property_add(oc, PC_MACHINE_DEVMEM_REGION_SIZE, "int",
2801 pc_machine_get_device_memory_region_size, NULL,
2802 NULL, NULL, &error_abort);
2804 object_class_property_add(oc, PC_MACHINE_MAX_RAM_BELOW_4G, "size",
2805 pc_machine_get_max_ram_below_4g, pc_machine_set_max_ram_below_4g,
2806 NULL, NULL, &error_abort);
2808 object_class_property_set_description(oc, PC_MACHINE_MAX_RAM_BELOW_4G,
2809 "Maximum ram below the 4G boundary (32bit boundary)", &error_abort);
2811 object_class_property_add(oc, PC_MACHINE_SMM, "OnOffAuto",
2812 pc_machine_get_smm, pc_machine_set_smm,
2813 NULL, NULL, &error_abort);
2814 object_class_property_set_description(oc, PC_MACHINE_SMM,
2815 "Enable SMM (pc & q35)", &error_abort);
2817 object_class_property_add(oc, PC_MACHINE_VMPORT, "OnOffAuto",
2818 pc_machine_get_vmport, pc_machine_set_vmport,
2819 NULL, NULL, &error_abort);
2820 object_class_property_set_description(oc, PC_MACHINE_VMPORT,
2821 "Enable vmport (pc & q35)", &error_abort);
2823 object_class_property_add_bool(oc, PC_MACHINE_NVDIMM,
2824 pc_machine_get_nvdimm, pc_machine_set_nvdimm, &error_abort);
2826 object_class_property_add_str(oc, PC_MACHINE_NVDIMM_PERSIST,
2827 pc_machine_get_nvdimm_persistence,
2828 pc_machine_set_nvdimm_persistence, &error_abort);
2830 object_class_property_add_bool(oc, PC_MACHINE_SMBUS,
2831 pc_machine_get_smbus, pc_machine_set_smbus, &error_abort);
2833 object_class_property_add_bool(oc, PC_MACHINE_SATA,
2834 pc_machine_get_sata, pc_machine_set_sata, &error_abort);
2836 object_class_property_add_bool(oc, PC_MACHINE_PIT,
2837 pc_machine_get_pit, pc_machine_set_pit, &error_abort);
2840 static const TypeInfo pc_machine_info = {
2841 .name = TYPE_PC_MACHINE,
2842 .parent = TYPE_MACHINE,
2843 .abstract = true,
2844 .instance_size = sizeof(PCMachineState),
2845 .instance_init = pc_machine_initfn,
2846 .class_size = sizeof(PCMachineClass),
2847 .class_init = pc_machine_class_init,
2848 .interfaces = (InterfaceInfo[]) {
2849 { TYPE_HOTPLUG_HANDLER },
2850 { TYPE_NMI },
2855 static void pc_machine_register_types(void)
2857 type_register_static(&pc_machine_info);
2860 type_init(pc_machine_register_types)