xref: /openbmc/qemu/hw/i386/pc.c (revision 316f239c294101d63a2b336c86c7970d8ae15213)
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 
25 #include "qemu/osdep.h"
26 #include "qemu/units.h"
27 #include "hw/i386/pc.h"
28 #include "hw/char/serial.h"
29 #include "hw/char/parallel.h"
30 #include "hw/i386/apic.h"
31 #include "hw/i386/topology.h"
32 #include "hw/i386/fw_cfg.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 "migration/vmstate.h"
44 #include "multiboot.h"
45 #include "hw/timer/mc146818rtc.h"
46 #include "hw/dma/i8257.h"
47 #include "hw/timer/i8254.h"
48 #include "hw/input/i8042.h"
49 #include "hw/irq.h"
50 #include "hw/audio/pcspk.h"
51 #include "hw/pci/msi.h"
52 #include "hw/sysbus.h"
53 #include "sysemu/sysemu.h"
54 #include "sysemu/tcg.h"
55 #include "sysemu/numa.h"
56 #include "sysemu/kvm.h"
57 #include "sysemu/qtest.h"
58 #include "sysemu/reset.h"
59 #include "sysemu/runstate.h"
60 #include "kvm_i386.h"
61 #include "hw/xen/xen.h"
62 #include "hw/xen/start_info.h"
63 #include "ui/qemu-spice.h"
64 #include "exec/memory.h"
65 #include "exec/address-spaces.h"
66 #include "sysemu/arch_init.h"
67 #include "qemu/bitmap.h"
68 #include "qemu/config-file.h"
69 #include "qemu/error-report.h"
70 #include "qemu/option.h"
71 #include "hw/acpi/acpi.h"
72 #include "hw/acpi/cpu_hotplug.h"
73 #include "hw/boards.h"
74 #include "acpi-build.h"
75 #include "hw/mem/pc-dimm.h"
76 #include "qapi/error.h"
77 #include "qapi/qapi-visit-common.h"
78 #include "qapi/visitor.h"
79 #include "qom/cpu.h"
80 #include "hw/nmi.h"
81 #include "hw/usb.h"
82 #include "hw/i386/intel_iommu.h"
83 #include "hw/net/ne2000-isa.h"
84 #include "standard-headers/asm-x86/bootparam.h"
85 #include "hw/virtio/virtio-pmem-pci.h"
86 #include "hw/mem/memory-device.h"
87 #include "sysemu/replay.h"
88 #include "qapi/qmp/qerror.h"
89 #include "config-devices.h"
90 
91 /* debug PC/ISA interrupts */
92 //#define DEBUG_IRQ
93 
94 #ifdef DEBUG_IRQ
95 #define DPRINTF(fmt, ...)                                       \
96     do { printf("CPUIRQ: " fmt , ## __VA_ARGS__); } while (0)
97 #else
98 #define DPRINTF(fmt, ...)
99 #endif
100 
101 #define E820_NR_ENTRIES		16
102 
103 struct e820_entry {
104     uint64_t address;
105     uint64_t length;
106     uint32_t type;
107 } QEMU_PACKED __attribute((__aligned__(4)));
108 
109 struct e820_table {
110     uint32_t count;
111     struct e820_entry entry[E820_NR_ENTRIES];
112 } QEMU_PACKED __attribute((__aligned__(4)));
113 
114 static struct e820_table e820_reserve;
115 static struct e820_entry *e820_table;
116 static unsigned e820_entries;
117 struct hpet_fw_config hpet_cfg = {.count = UINT8_MAX};
118 
119 /* Physical Address of PVH entry point read from kernel ELF NOTE */
120 static size_t pvh_start_addr;
121 
122 GlobalProperty pc_compat_4_1[] = {};
123 const size_t pc_compat_4_1_len = G_N_ELEMENTS(pc_compat_4_1);
124 
125 GlobalProperty pc_compat_4_0[] = {};
126 const size_t pc_compat_4_0_len = G_N_ELEMENTS(pc_compat_4_0);
127 
128 GlobalProperty pc_compat_3_1[] = {
129     { "intel-iommu", "dma-drain", "off" },
130     { "Opteron_G3" "-" TYPE_X86_CPU, "rdtscp", "off" },
131     { "Opteron_G4" "-" TYPE_X86_CPU, "rdtscp", "off" },
132     { "Opteron_G4" "-" TYPE_X86_CPU, "npt", "off" },
133     { "Opteron_G4" "-" TYPE_X86_CPU, "nrip-save", "off" },
134     { "Opteron_G5" "-" TYPE_X86_CPU, "rdtscp", "off" },
135     { "Opteron_G5" "-" TYPE_X86_CPU, "npt", "off" },
136     { "Opteron_G5" "-" TYPE_X86_CPU, "nrip-save", "off" },
137     { "EPYC" "-" TYPE_X86_CPU, "npt", "off" },
138     { "EPYC" "-" TYPE_X86_CPU, "nrip-save", "off" },
139     { "EPYC-IBPB" "-" TYPE_X86_CPU, "npt", "off" },
140     { "EPYC-IBPB" "-" TYPE_X86_CPU, "nrip-save", "off" },
141     { "Skylake-Client" "-" TYPE_X86_CPU,      "mpx", "on" },
142     { "Skylake-Client-IBRS" "-" TYPE_X86_CPU, "mpx", "on" },
143     { "Skylake-Server" "-" TYPE_X86_CPU,      "mpx", "on" },
144     { "Skylake-Server-IBRS" "-" TYPE_X86_CPU, "mpx", "on" },
145     { "Cascadelake-Server" "-" TYPE_X86_CPU,  "mpx", "on" },
146     { "Icelake-Client" "-" TYPE_X86_CPU,      "mpx", "on" },
147     { "Icelake-Server" "-" TYPE_X86_CPU,      "mpx", "on" },
148     { "Cascadelake-Server" "-" TYPE_X86_CPU, "stepping", "5" },
149     { TYPE_X86_CPU, "x-intel-pt-auto-level", "off" },
150 };
151 const size_t pc_compat_3_1_len = G_N_ELEMENTS(pc_compat_3_1);
152 
153 GlobalProperty pc_compat_3_0[] = {
154     { TYPE_X86_CPU, "x-hv-synic-kvm-only", "on" },
155     { "Skylake-Server" "-" TYPE_X86_CPU, "pku", "off" },
156     { "Skylake-Server-IBRS" "-" TYPE_X86_CPU, "pku", "off" },
157 };
158 const size_t pc_compat_3_0_len = G_N_ELEMENTS(pc_compat_3_0);
159 
160 GlobalProperty pc_compat_2_12[] = {
161     { TYPE_X86_CPU, "legacy-cache", "on" },
162     { TYPE_X86_CPU, "topoext", "off" },
163     { "EPYC-" TYPE_X86_CPU, "xlevel", "0x8000000a" },
164     { "EPYC-IBPB-" TYPE_X86_CPU, "xlevel", "0x8000000a" },
165 };
166 const size_t pc_compat_2_12_len = G_N_ELEMENTS(pc_compat_2_12);
167 
168 GlobalProperty pc_compat_2_11[] = {
169     { TYPE_X86_CPU, "x-migrate-smi-count", "off" },
170     { "Skylake-Server" "-" TYPE_X86_CPU, "clflushopt", "off" },
171 };
172 const size_t pc_compat_2_11_len = G_N_ELEMENTS(pc_compat_2_11);
173 
174 GlobalProperty pc_compat_2_10[] = {
175     { TYPE_X86_CPU, "x-hv-max-vps", "0x40" },
176     { "i440FX-pcihost", "x-pci-hole64-fix", "off" },
177     { "q35-pcihost", "x-pci-hole64-fix", "off" },
178 };
179 const size_t pc_compat_2_10_len = G_N_ELEMENTS(pc_compat_2_10);
180 
181 GlobalProperty pc_compat_2_9[] = {
182     { "mch", "extended-tseg-mbytes", "0" },
183 };
184 const size_t pc_compat_2_9_len = G_N_ELEMENTS(pc_compat_2_9);
185 
186 GlobalProperty pc_compat_2_8[] = {
187     { TYPE_X86_CPU, "tcg-cpuid", "off" },
188     { "kvmclock", "x-mach-use-reliable-get-clock", "off" },
189     { "ICH9-LPC", "x-smi-broadcast", "off" },
190     { TYPE_X86_CPU, "vmware-cpuid-freq", "off" },
191     { "Haswell-" TYPE_X86_CPU, "stepping", "1" },
192 };
193 const size_t pc_compat_2_8_len = G_N_ELEMENTS(pc_compat_2_8);
194 
195 GlobalProperty pc_compat_2_7[] = {
196     { TYPE_X86_CPU, "l3-cache", "off" },
197     { TYPE_X86_CPU, "full-cpuid-auto-level", "off" },
198     { "Opteron_G3" "-" TYPE_X86_CPU, "family", "15" },
199     { "Opteron_G3" "-" TYPE_X86_CPU, "model", "6" },
200     { "Opteron_G3" "-" TYPE_X86_CPU, "stepping", "1" },
201     { "isa-pcspk", "migrate", "off" },
202 };
203 const size_t pc_compat_2_7_len = G_N_ELEMENTS(pc_compat_2_7);
204 
205 GlobalProperty pc_compat_2_6[] = {
206     { TYPE_X86_CPU, "cpuid-0xb", "off" },
207     { "vmxnet3", "romfile", "" },
208     { TYPE_X86_CPU, "fill-mtrr-mask", "off" },
209     { "apic-common", "legacy-instance-id", "on", }
210 };
211 const size_t pc_compat_2_6_len = G_N_ELEMENTS(pc_compat_2_6);
212 
213 GlobalProperty pc_compat_2_5[] = {};
214 const size_t pc_compat_2_5_len = G_N_ELEMENTS(pc_compat_2_5);
215 
216 GlobalProperty pc_compat_2_4[] = {
217     PC_CPU_MODEL_IDS("2.4.0")
218     { "Haswell-" TYPE_X86_CPU, "abm", "off" },
219     { "Haswell-noTSX-" TYPE_X86_CPU, "abm", "off" },
220     { "Broadwell-" TYPE_X86_CPU, "abm", "off" },
221     { "Broadwell-noTSX-" TYPE_X86_CPU, "abm", "off" },
222     { "host" "-" TYPE_X86_CPU, "host-cache-info", "on" },
223     { TYPE_X86_CPU, "check", "off" },
224     { "qemu64" "-" TYPE_X86_CPU, "sse4a", "on" },
225     { "qemu64" "-" TYPE_X86_CPU, "abm", "on" },
226     { "qemu64" "-" TYPE_X86_CPU, "popcnt", "on" },
227     { "qemu32" "-" TYPE_X86_CPU, "popcnt", "on" },
228     { "Opteron_G2" "-" TYPE_X86_CPU, "rdtscp", "on" },
229     { "Opteron_G3" "-" TYPE_X86_CPU, "rdtscp", "on" },
230     { "Opteron_G4" "-" TYPE_X86_CPU, "rdtscp", "on" },
231     { "Opteron_G5" "-" TYPE_X86_CPU, "rdtscp", "on", }
232 };
233 const size_t pc_compat_2_4_len = G_N_ELEMENTS(pc_compat_2_4);
234 
235 GlobalProperty pc_compat_2_3[] = {
236     PC_CPU_MODEL_IDS("2.3.0")
237     { TYPE_X86_CPU, "arat", "off" },
238     { "qemu64" "-" TYPE_X86_CPU, "min-level", "4" },
239     { "kvm64" "-" TYPE_X86_CPU, "min-level", "5" },
240     { "pentium3" "-" TYPE_X86_CPU, "min-level", "2" },
241     { "n270" "-" TYPE_X86_CPU, "min-level", "5" },
242     { "Conroe" "-" TYPE_X86_CPU, "min-level", "4" },
243     { "Penryn" "-" TYPE_X86_CPU, "min-level", "4" },
244     { "Nehalem" "-" TYPE_X86_CPU, "min-level", "4" },
245     { "n270" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
246     { "Penryn" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
247     { "Conroe" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
248     { "Nehalem" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
249     { "Westmere" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
250     { "SandyBridge" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
251     { "IvyBridge" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
252     { "Haswell" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
253     { "Haswell-noTSX" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
254     { "Broadwell" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
255     { "Broadwell-noTSX" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
256     { TYPE_X86_CPU, "kvm-no-smi-migration", "on" },
257 };
258 const size_t pc_compat_2_3_len = G_N_ELEMENTS(pc_compat_2_3);
259 
260 GlobalProperty pc_compat_2_2[] = {
261     PC_CPU_MODEL_IDS("2.2.0")
262     { "kvm64" "-" TYPE_X86_CPU, "vme", "off" },
263     { "kvm32" "-" TYPE_X86_CPU, "vme", "off" },
264     { "Conroe" "-" TYPE_X86_CPU, "vme", "off" },
265     { "Penryn" "-" TYPE_X86_CPU, "vme", "off" },
266     { "Nehalem" "-" TYPE_X86_CPU, "vme", "off" },
267     { "Westmere" "-" TYPE_X86_CPU, "vme", "off" },
268     { "SandyBridge" "-" TYPE_X86_CPU, "vme", "off" },
269     { "Haswell" "-" TYPE_X86_CPU, "vme", "off" },
270     { "Broadwell" "-" TYPE_X86_CPU, "vme", "off" },
271     { "Opteron_G1" "-" TYPE_X86_CPU, "vme", "off" },
272     { "Opteron_G2" "-" TYPE_X86_CPU, "vme", "off" },
273     { "Opteron_G3" "-" TYPE_X86_CPU, "vme", "off" },
274     { "Opteron_G4" "-" TYPE_X86_CPU, "vme", "off" },
275     { "Opteron_G5" "-" TYPE_X86_CPU, "vme", "off" },
276     { "Haswell" "-" TYPE_X86_CPU, "f16c", "off" },
277     { "Haswell" "-" TYPE_X86_CPU, "rdrand", "off" },
278     { "Broadwell" "-" TYPE_X86_CPU, "f16c", "off" },
279     { "Broadwell" "-" TYPE_X86_CPU, "rdrand", "off" },
280 };
281 const size_t pc_compat_2_2_len = G_N_ELEMENTS(pc_compat_2_2);
282 
283 GlobalProperty pc_compat_2_1[] = {
284     PC_CPU_MODEL_IDS("2.1.0")
285     { "coreduo" "-" TYPE_X86_CPU, "vmx", "on" },
286     { "core2duo" "-" TYPE_X86_CPU, "vmx", "on" },
287 };
288 const size_t pc_compat_2_1_len = G_N_ELEMENTS(pc_compat_2_1);
289 
290 GlobalProperty pc_compat_2_0[] = {
291     PC_CPU_MODEL_IDS("2.0.0")
292     { "virtio-scsi-pci", "any_layout", "off" },
293     { "PIIX4_PM", "memory-hotplug-support", "off" },
294     { "apic", "version", "0x11" },
295     { "nec-usb-xhci", "superspeed-ports-first", "off" },
296     { "nec-usb-xhci", "force-pcie-endcap", "on" },
297     { "pci-serial", "prog_if", "0" },
298     { "pci-serial-2x", "prog_if", "0" },
299     { "pci-serial-4x", "prog_if", "0" },
300     { "virtio-net-pci", "guest_announce", "off" },
301     { "ICH9-LPC", "memory-hotplug-support", "off" },
302     { "xio3130-downstream", COMPAT_PROP_PCP, "off" },
303     { "ioh3420", COMPAT_PROP_PCP, "off" },
304 };
305 const size_t pc_compat_2_0_len = G_N_ELEMENTS(pc_compat_2_0);
306 
307 GlobalProperty pc_compat_1_7[] = {
308     PC_CPU_MODEL_IDS("1.7.0")
309     { TYPE_USB_DEVICE, "msos-desc", "no" },
310     { "PIIX4_PM", "acpi-pci-hotplug-with-bridge-support", "off" },
311     { "hpet", HPET_INTCAP, "4" },
312 };
313 const size_t pc_compat_1_7_len = G_N_ELEMENTS(pc_compat_1_7);
314 
315 GlobalProperty pc_compat_1_6[] = {
316     PC_CPU_MODEL_IDS("1.6.0")
317     { "e1000", "mitigation", "off" },
318     { "qemu64-" TYPE_X86_CPU, "model", "2" },
319     { "qemu32-" TYPE_X86_CPU, "model", "3" },
320     { "i440FX-pcihost", "short_root_bus", "1" },
321     { "q35-pcihost", "short_root_bus", "1" },
322 };
323 const size_t pc_compat_1_6_len = G_N_ELEMENTS(pc_compat_1_6);
324 
325 GlobalProperty pc_compat_1_5[] = {
326     PC_CPU_MODEL_IDS("1.5.0")
327     { "Conroe-" TYPE_X86_CPU, "model", "2" },
328     { "Conroe-" TYPE_X86_CPU, "min-level", "2" },
329     { "Penryn-" TYPE_X86_CPU, "model", "2" },
330     { "Penryn-" TYPE_X86_CPU, "min-level", "2" },
331     { "Nehalem-" TYPE_X86_CPU, "model", "2" },
332     { "Nehalem-" TYPE_X86_CPU, "min-level", "2" },
333     { "virtio-net-pci", "any_layout", "off" },
334     { TYPE_X86_CPU, "pmu", "on" },
335     { "i440FX-pcihost", "short_root_bus", "0" },
336     { "q35-pcihost", "short_root_bus", "0" },
337 };
338 const size_t pc_compat_1_5_len = G_N_ELEMENTS(pc_compat_1_5);
339 
340 GlobalProperty pc_compat_1_4[] = {
341     PC_CPU_MODEL_IDS("1.4.0")
342     { "scsi-hd", "discard_granularity", "0" },
343     { "scsi-cd", "discard_granularity", "0" },
344     { "scsi-disk", "discard_granularity", "0" },
345     { "ide-hd", "discard_granularity", "0" },
346     { "ide-cd", "discard_granularity", "0" },
347     { "ide-drive", "discard_granularity", "0" },
348     { "virtio-blk-pci", "discard_granularity", "0" },
349     /* DEV_NVECTORS_UNSPECIFIED as a uint32_t string: */
350     { "virtio-serial-pci", "vectors", "0xFFFFFFFF" },
351     { "virtio-net-pci", "ctrl_guest_offloads", "off" },
352     { "e1000", "romfile", "pxe-e1000.rom" },
353     { "ne2k_pci", "romfile", "pxe-ne2k_pci.rom" },
354     { "pcnet", "romfile", "pxe-pcnet.rom" },
355     { "rtl8139", "romfile", "pxe-rtl8139.rom" },
356     { "virtio-net-pci", "romfile", "pxe-virtio.rom" },
357     { "486-" TYPE_X86_CPU, "model", "0" },
358     { "n270" "-" TYPE_X86_CPU, "movbe", "off" },
359     { "Westmere" "-" TYPE_X86_CPU, "pclmulqdq", "off" },
360 };
361 const size_t pc_compat_1_4_len = G_N_ELEMENTS(pc_compat_1_4);
362 
363 void gsi_handler(void *opaque, int n, int level)
364 {
365     GSIState *s = opaque;
366 
367     DPRINTF("pc: %s GSI %d\n", level ? "raising" : "lowering", n);
368     if (n < ISA_NUM_IRQS) {
369         qemu_set_irq(s->i8259_irq[n], level);
370     }
371     qemu_set_irq(s->ioapic_irq[n], level);
372 }
373 
374 static void ioport80_write(void *opaque, hwaddr addr, uint64_t data,
375                            unsigned size)
376 {
377 }
378 
379 static uint64_t ioport80_read(void *opaque, hwaddr addr, unsigned size)
380 {
381     return 0xffffffffffffffffULL;
382 }
383 
384 /* MSDOS compatibility mode FPU exception support */
385 static qemu_irq ferr_irq;
386 
387 void pc_register_ferr_irq(qemu_irq irq)
388 {
389     ferr_irq = irq;
390 }
391 
392 /* XXX: add IGNNE support */
393 void cpu_set_ferr(CPUX86State *s)
394 {
395     qemu_irq_raise(ferr_irq);
396 }
397 
398 static void ioportF0_write(void *opaque, hwaddr addr, uint64_t data,
399                            unsigned size)
400 {
401     qemu_irq_lower(ferr_irq);
402 }
403 
404 static uint64_t ioportF0_read(void *opaque, hwaddr addr, unsigned size)
405 {
406     return 0xffffffffffffffffULL;
407 }
408 
409 /* TSC handling */
410 uint64_t cpu_get_tsc(CPUX86State *env)
411 {
412     return cpu_get_ticks();
413 }
414 
415 /* IRQ handling */
416 int cpu_get_pic_interrupt(CPUX86State *env)
417 {
418     X86CPU *cpu = env_archcpu(env);
419     int intno;
420 
421     if (!kvm_irqchip_in_kernel()) {
422         intno = apic_get_interrupt(cpu->apic_state);
423         if (intno >= 0) {
424             return intno;
425         }
426         /* read the irq from the PIC */
427         if (!apic_accept_pic_intr(cpu->apic_state)) {
428             return -1;
429         }
430     }
431 
432     intno = pic_read_irq(isa_pic);
433     return intno;
434 }
435 
436 static void pic_irq_request(void *opaque, int irq, int level)
437 {
438     CPUState *cs = first_cpu;
439     X86CPU *cpu = X86_CPU(cs);
440 
441     DPRINTF("pic_irqs: %s irq %d\n", level? "raise" : "lower", irq);
442     if (cpu->apic_state && !kvm_irqchip_in_kernel()) {
443         CPU_FOREACH(cs) {
444             cpu = X86_CPU(cs);
445             if (apic_accept_pic_intr(cpu->apic_state)) {
446                 apic_deliver_pic_intr(cpu->apic_state, level);
447             }
448         }
449     } else {
450         if (level) {
451             cpu_interrupt(cs, CPU_INTERRUPT_HARD);
452         } else {
453             cpu_reset_interrupt(cs, CPU_INTERRUPT_HARD);
454         }
455     }
456 }
457 
458 /* PC cmos mappings */
459 
460 #define REG_EQUIPMENT_BYTE          0x14
461 
462 int cmos_get_fd_drive_type(FloppyDriveType fd0)
463 {
464     int val;
465 
466     switch (fd0) {
467     case FLOPPY_DRIVE_TYPE_144:
468         /* 1.44 Mb 3"5 drive */
469         val = 4;
470         break;
471     case FLOPPY_DRIVE_TYPE_288:
472         /* 2.88 Mb 3"5 drive */
473         val = 5;
474         break;
475     case FLOPPY_DRIVE_TYPE_120:
476         /* 1.2 Mb 5"5 drive */
477         val = 2;
478         break;
479     case FLOPPY_DRIVE_TYPE_NONE:
480     default:
481         val = 0;
482         break;
483     }
484     return val;
485 }
486 
487 static void cmos_init_hd(ISADevice *s, int type_ofs, int info_ofs,
488                          int16_t cylinders, int8_t heads, int8_t sectors)
489 {
490     rtc_set_memory(s, type_ofs, 47);
491     rtc_set_memory(s, info_ofs, cylinders);
492     rtc_set_memory(s, info_ofs + 1, cylinders >> 8);
493     rtc_set_memory(s, info_ofs + 2, heads);
494     rtc_set_memory(s, info_ofs + 3, 0xff);
495     rtc_set_memory(s, info_ofs + 4, 0xff);
496     rtc_set_memory(s, info_ofs + 5, 0xc0 | ((heads > 8) << 3));
497     rtc_set_memory(s, info_ofs + 6, cylinders);
498     rtc_set_memory(s, info_ofs + 7, cylinders >> 8);
499     rtc_set_memory(s, info_ofs + 8, sectors);
500 }
501 
502 /* convert boot_device letter to something recognizable by the bios */
503 static int boot_device2nibble(char boot_device)
504 {
505     switch(boot_device) {
506     case 'a':
507     case 'b':
508         return 0x01; /* floppy boot */
509     case 'c':
510         return 0x02; /* hard drive boot */
511     case 'd':
512         return 0x03; /* CD-ROM boot */
513     case 'n':
514         return 0x04; /* Network boot */
515     }
516     return 0;
517 }
518 
519 static void set_boot_dev(ISADevice *s, const char *boot_device, Error **errp)
520 {
521 #define PC_MAX_BOOT_DEVICES 3
522     int nbds, bds[3] = { 0, };
523     int i;
524 
525     nbds = strlen(boot_device);
526     if (nbds > PC_MAX_BOOT_DEVICES) {
527         error_setg(errp, "Too many boot devices for PC");
528         return;
529     }
530     for (i = 0; i < nbds; i++) {
531         bds[i] = boot_device2nibble(boot_device[i]);
532         if (bds[i] == 0) {
533             error_setg(errp, "Invalid boot device for PC: '%c'",
534                        boot_device[i]);
535             return;
536         }
537     }
538     rtc_set_memory(s, 0x3d, (bds[1] << 4) | bds[0]);
539     rtc_set_memory(s, 0x38, (bds[2] << 4) | (fd_bootchk ? 0x0 : 0x1));
540 }
541 
542 static void pc_boot_set(void *opaque, const char *boot_device, Error **errp)
543 {
544     set_boot_dev(opaque, boot_device, errp);
545 }
546 
547 static void pc_cmos_init_floppy(ISADevice *rtc_state, ISADevice *floppy)
548 {
549     int val, nb, i;
550     FloppyDriveType fd_type[2] = { FLOPPY_DRIVE_TYPE_NONE,
551                                    FLOPPY_DRIVE_TYPE_NONE };
552 
553     /* floppy type */
554     if (floppy) {
555         for (i = 0; i < 2; i++) {
556             fd_type[i] = isa_fdc_get_drive_type(floppy, i);
557         }
558     }
559     val = (cmos_get_fd_drive_type(fd_type[0]) << 4) |
560         cmos_get_fd_drive_type(fd_type[1]);
561     rtc_set_memory(rtc_state, 0x10, val);
562 
563     val = rtc_get_memory(rtc_state, REG_EQUIPMENT_BYTE);
564     nb = 0;
565     if (fd_type[0] != FLOPPY_DRIVE_TYPE_NONE) {
566         nb++;
567     }
568     if (fd_type[1] != FLOPPY_DRIVE_TYPE_NONE) {
569         nb++;
570     }
571     switch (nb) {
572     case 0:
573         break;
574     case 1:
575         val |= 0x01; /* 1 drive, ready for boot */
576         break;
577     case 2:
578         val |= 0x41; /* 2 drives, ready for boot */
579         break;
580     }
581     rtc_set_memory(rtc_state, REG_EQUIPMENT_BYTE, val);
582 }
583 
584 typedef struct pc_cmos_init_late_arg {
585     ISADevice *rtc_state;
586     BusState *idebus[2];
587 } pc_cmos_init_late_arg;
588 
589 typedef struct check_fdc_state {
590     ISADevice *floppy;
591     bool multiple;
592 } CheckFdcState;
593 
594 static int check_fdc(Object *obj, void *opaque)
595 {
596     CheckFdcState *state = opaque;
597     Object *fdc;
598     uint32_t iobase;
599     Error *local_err = NULL;
600 
601     fdc = object_dynamic_cast(obj, TYPE_ISA_FDC);
602     if (!fdc) {
603         return 0;
604     }
605 
606     iobase = object_property_get_uint(obj, "iobase", &local_err);
607     if (local_err || iobase != 0x3f0) {
608         error_free(local_err);
609         return 0;
610     }
611 
612     if (state->floppy) {
613         state->multiple = true;
614     } else {
615         state->floppy = ISA_DEVICE(obj);
616     }
617     return 0;
618 }
619 
620 static const char * const fdc_container_path[] = {
621     "/unattached", "/peripheral", "/peripheral-anon"
622 };
623 
624 /*
625  * Locate the FDC at IO address 0x3f0, in order to configure the CMOS registers
626  * and ACPI objects.
627  */
628 ISADevice *pc_find_fdc0(void)
629 {
630     int i;
631     Object *container;
632     CheckFdcState state = { 0 };
633 
634     for (i = 0; i < ARRAY_SIZE(fdc_container_path); i++) {
635         container = container_get(qdev_get_machine(), fdc_container_path[i]);
636         object_child_foreach(container, check_fdc, &state);
637     }
638 
639     if (state.multiple) {
640         warn_report("multiple floppy disk controllers with "
641                     "iobase=0x3f0 have been found");
642         error_printf("the one being picked for CMOS setup might not reflect "
643                      "your intent");
644     }
645 
646     return state.floppy;
647 }
648 
649 static void pc_cmos_init_late(void *opaque)
650 {
651     pc_cmos_init_late_arg *arg = opaque;
652     ISADevice *s = arg->rtc_state;
653     int16_t cylinders;
654     int8_t heads, sectors;
655     int val;
656     int i, trans;
657 
658     val = 0;
659     if (arg->idebus[0] && ide_get_geometry(arg->idebus[0], 0,
660                                            &cylinders, &heads, &sectors) >= 0) {
661         cmos_init_hd(s, 0x19, 0x1b, cylinders, heads, sectors);
662         val |= 0xf0;
663     }
664     if (arg->idebus[0] && ide_get_geometry(arg->idebus[0], 1,
665                                            &cylinders, &heads, &sectors) >= 0) {
666         cmos_init_hd(s, 0x1a, 0x24, cylinders, heads, sectors);
667         val |= 0x0f;
668     }
669     rtc_set_memory(s, 0x12, val);
670 
671     val = 0;
672     for (i = 0; i < 4; i++) {
673         /* NOTE: ide_get_geometry() returns the physical
674            geometry.  It is always such that: 1 <= sects <= 63, 1
675            <= heads <= 16, 1 <= cylinders <= 16383. The BIOS
676            geometry can be different if a translation is done. */
677         if (arg->idebus[i / 2] &&
678             ide_get_geometry(arg->idebus[i / 2], i % 2,
679                              &cylinders, &heads, &sectors) >= 0) {
680             trans = ide_get_bios_chs_trans(arg->idebus[i / 2], i % 2) - 1;
681             assert((trans & ~3) == 0);
682             val |= trans << (i * 2);
683         }
684     }
685     rtc_set_memory(s, 0x39, val);
686 
687     pc_cmos_init_floppy(s, pc_find_fdc0());
688 
689     qemu_unregister_reset(pc_cmos_init_late, opaque);
690 }
691 
692 void pc_cmos_init(PCMachineState *pcms,
693                   BusState *idebus0, BusState *idebus1,
694                   ISADevice *s)
695 {
696     int val;
697     static pc_cmos_init_late_arg arg;
698 
699     /* various important CMOS locations needed by PC/Bochs bios */
700 
701     /* memory size */
702     /* base memory (first MiB) */
703     val = MIN(pcms->below_4g_mem_size / KiB, 640);
704     rtc_set_memory(s, 0x15, val);
705     rtc_set_memory(s, 0x16, val >> 8);
706     /* extended memory (next 64MiB) */
707     if (pcms->below_4g_mem_size > 1 * MiB) {
708         val = (pcms->below_4g_mem_size - 1 * MiB) / KiB;
709     } else {
710         val = 0;
711     }
712     if (val > 65535)
713         val = 65535;
714     rtc_set_memory(s, 0x17, val);
715     rtc_set_memory(s, 0x18, val >> 8);
716     rtc_set_memory(s, 0x30, val);
717     rtc_set_memory(s, 0x31, val >> 8);
718     /* memory between 16MiB and 4GiB */
719     if (pcms->below_4g_mem_size > 16 * MiB) {
720         val = (pcms->below_4g_mem_size - 16 * MiB) / (64 * KiB);
721     } else {
722         val = 0;
723     }
724     if (val > 65535)
725         val = 65535;
726     rtc_set_memory(s, 0x34, val);
727     rtc_set_memory(s, 0x35, val >> 8);
728     /* memory above 4GiB */
729     val = pcms->above_4g_mem_size / 65536;
730     rtc_set_memory(s, 0x5b, val);
731     rtc_set_memory(s, 0x5c, val >> 8);
732     rtc_set_memory(s, 0x5d, val >> 16);
733 
734     object_property_add_link(OBJECT(pcms), "rtc_state",
735                              TYPE_ISA_DEVICE,
736                              (Object **)&pcms->rtc,
737                              object_property_allow_set_link,
738                              OBJ_PROP_LINK_STRONG, &error_abort);
739     object_property_set_link(OBJECT(pcms), OBJECT(s),
740                              "rtc_state", &error_abort);
741 
742     set_boot_dev(s, MACHINE(pcms)->boot_order, &error_fatal);
743 
744     val = 0;
745     val |= 0x02; /* FPU is there */
746     val |= 0x04; /* PS/2 mouse installed */
747     rtc_set_memory(s, REG_EQUIPMENT_BYTE, val);
748 
749     /* hard drives and FDC */
750     arg.rtc_state = s;
751     arg.idebus[0] = idebus0;
752     arg.idebus[1] = idebus1;
753     qemu_register_reset(pc_cmos_init_late, &arg);
754 }
755 
756 #define TYPE_PORT92 "port92"
757 #define PORT92(obj) OBJECT_CHECK(Port92State, (obj), TYPE_PORT92)
758 
759 /* port 92 stuff: could be split off */
760 typedef struct Port92State {
761     ISADevice parent_obj;
762 
763     MemoryRegion io;
764     uint8_t outport;
765     qemu_irq a20_out;
766 } Port92State;
767 
768 static void port92_write(void *opaque, hwaddr addr, uint64_t val,
769                          unsigned size)
770 {
771     Port92State *s = opaque;
772     int oldval = s->outport;
773 
774     DPRINTF("port92: write 0x%02" PRIx64 "\n", val);
775     s->outport = val;
776     qemu_set_irq(s->a20_out, (val >> 1) & 1);
777     if ((val & 1) && !(oldval & 1)) {
778         qemu_system_reset_request(SHUTDOWN_CAUSE_GUEST_RESET);
779     }
780 }
781 
782 static uint64_t port92_read(void *opaque, hwaddr addr,
783                             unsigned size)
784 {
785     Port92State *s = opaque;
786     uint32_t ret;
787 
788     ret = s->outport;
789     DPRINTF("port92: read 0x%02x\n", ret);
790     return ret;
791 }
792 
793 static void port92_init(ISADevice *dev, qemu_irq a20_out)
794 {
795     qdev_connect_gpio_out_named(DEVICE(dev), PORT92_A20_LINE, 0, a20_out);
796 }
797 
798 static const VMStateDescription vmstate_port92_isa = {
799     .name = "port92",
800     .version_id = 1,
801     .minimum_version_id = 1,
802     .fields = (VMStateField[]) {
803         VMSTATE_UINT8(outport, Port92State),
804         VMSTATE_END_OF_LIST()
805     }
806 };
807 
808 static void port92_reset(DeviceState *d)
809 {
810     Port92State *s = PORT92(d);
811 
812     s->outport &= ~1;
813 }
814 
815 static const MemoryRegionOps port92_ops = {
816     .read = port92_read,
817     .write = port92_write,
818     .impl = {
819         .min_access_size = 1,
820         .max_access_size = 1,
821     },
822     .endianness = DEVICE_LITTLE_ENDIAN,
823 };
824 
825 static void port92_initfn(Object *obj)
826 {
827     Port92State *s = PORT92(obj);
828 
829     memory_region_init_io(&s->io, OBJECT(s), &port92_ops, s, "port92", 1);
830 
831     s->outport = 0;
832 
833     qdev_init_gpio_out_named(DEVICE(obj), &s->a20_out, PORT92_A20_LINE, 1);
834 }
835 
836 static void port92_realizefn(DeviceState *dev, Error **errp)
837 {
838     ISADevice *isadev = ISA_DEVICE(dev);
839     Port92State *s = PORT92(dev);
840 
841     isa_register_ioport(isadev, &s->io, 0x92);
842 }
843 
844 static void port92_class_initfn(ObjectClass *klass, void *data)
845 {
846     DeviceClass *dc = DEVICE_CLASS(klass);
847 
848     dc->realize = port92_realizefn;
849     dc->reset = port92_reset;
850     dc->vmsd = &vmstate_port92_isa;
851     /*
852      * Reason: unlike ordinary ISA devices, this one needs additional
853      * wiring: its A20 output line needs to be wired up by
854      * port92_init().
855      */
856     dc->user_creatable = false;
857 }
858 
859 static const TypeInfo port92_info = {
860     .name          = TYPE_PORT92,
861     .parent        = TYPE_ISA_DEVICE,
862     .instance_size = sizeof(Port92State),
863     .instance_init = port92_initfn,
864     .class_init    = port92_class_initfn,
865 };
866 
867 static void port92_register_types(void)
868 {
869     type_register_static(&port92_info);
870 }
871 
872 type_init(port92_register_types)
873 
874 static void handle_a20_line_change(void *opaque, int irq, int level)
875 {
876     X86CPU *cpu = opaque;
877 
878     /* XXX: send to all CPUs ? */
879     /* XXX: add logic to handle multiple A20 line sources */
880     x86_cpu_set_a20(cpu, level);
881 }
882 
883 int e820_add_entry(uint64_t address, uint64_t length, uint32_t type)
884 {
885     int index = le32_to_cpu(e820_reserve.count);
886     struct e820_entry *entry;
887 
888     if (type != E820_RAM) {
889         /* old FW_CFG_E820_TABLE entry -- reservations only */
890         if (index >= E820_NR_ENTRIES) {
891             return -EBUSY;
892         }
893         entry = &e820_reserve.entry[index++];
894 
895         entry->address = cpu_to_le64(address);
896         entry->length = cpu_to_le64(length);
897         entry->type = cpu_to_le32(type);
898 
899         e820_reserve.count = cpu_to_le32(index);
900     }
901 
902     /* new "etc/e820" file -- include ram too */
903     e820_table = g_renew(struct e820_entry, e820_table, e820_entries + 1);
904     e820_table[e820_entries].address = cpu_to_le64(address);
905     e820_table[e820_entries].length = cpu_to_le64(length);
906     e820_table[e820_entries].type = cpu_to_le32(type);
907     e820_entries++;
908 
909     return e820_entries;
910 }
911 
912 int e820_get_num_entries(void)
913 {
914     return e820_entries;
915 }
916 
917 bool e820_get_entry(int idx, uint32_t type, uint64_t *address, uint64_t *length)
918 {
919     if (idx < e820_entries && e820_table[idx].type == cpu_to_le32(type)) {
920         *address = le64_to_cpu(e820_table[idx].address);
921         *length = le64_to_cpu(e820_table[idx].length);
922         return true;
923     }
924     return false;
925 }
926 
927 /* Calculates initial APIC ID for a specific CPU index
928  *
929  * Currently we need to be able to calculate the APIC ID from the CPU index
930  * alone (without requiring a CPU object), as the QEMU<->Seabios interfaces have
931  * no concept of "CPU index", and the NUMA tables on fw_cfg need the APIC ID of
932  * all CPUs up to max_cpus.
933  */
934 static uint32_t x86_cpu_apic_id_from_index(PCMachineState *pcms,
935                                            unsigned int cpu_index)
936 {
937     MachineState *ms = MACHINE(pcms);
938     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
939     uint32_t correct_id;
940     static bool warned;
941 
942     correct_id = x86_apicid_from_cpu_idx(pcms->smp_dies, ms->smp.cores,
943                                          ms->smp.threads, cpu_index);
944     if (pcmc->compat_apic_id_mode) {
945         if (cpu_index != correct_id && !warned && !qtest_enabled()) {
946             error_report("APIC IDs set in compatibility mode, "
947                          "CPU topology won't match the configuration");
948             warned = true;
949         }
950         return cpu_index;
951     } else {
952         return correct_id;
953     }
954 }
955 
956 static void pc_build_smbios(PCMachineState *pcms)
957 {
958     uint8_t *smbios_tables, *smbios_anchor;
959     size_t smbios_tables_len, smbios_anchor_len;
960     struct smbios_phys_mem_area *mem_array;
961     unsigned i, array_count;
962     MachineState *ms = MACHINE(pcms);
963     X86CPU *cpu = X86_CPU(ms->possible_cpus->cpus[0].cpu);
964 
965     /* tell smbios about cpuid version and features */
966     smbios_set_cpuid(cpu->env.cpuid_version, cpu->env.features[FEAT_1_EDX]);
967 
968     smbios_tables = smbios_get_table_legacy(ms, &smbios_tables_len);
969     if (smbios_tables) {
970         fw_cfg_add_bytes(pcms->fw_cfg, FW_CFG_SMBIOS_ENTRIES,
971                          smbios_tables, smbios_tables_len);
972     }
973 
974     /* build the array of physical mem area from e820 table */
975     mem_array = g_malloc0(sizeof(*mem_array) * e820_get_num_entries());
976     for (i = 0, array_count = 0; i < e820_get_num_entries(); i++) {
977         uint64_t addr, len;
978 
979         if (e820_get_entry(i, E820_RAM, &addr, &len)) {
980             mem_array[array_count].address = addr;
981             mem_array[array_count].length = len;
982             array_count++;
983         }
984     }
985     smbios_get_tables(ms, mem_array, array_count,
986                       &smbios_tables, &smbios_tables_len,
987                       &smbios_anchor, &smbios_anchor_len);
988     g_free(mem_array);
989 
990     if (smbios_anchor) {
991         fw_cfg_add_file(pcms->fw_cfg, "etc/smbios/smbios-tables",
992                         smbios_tables, smbios_tables_len);
993         fw_cfg_add_file(pcms->fw_cfg, "etc/smbios/smbios-anchor",
994                         smbios_anchor, smbios_anchor_len);
995     }
996 }
997 
998 static FWCfgState *bochs_bios_init(AddressSpace *as, PCMachineState *pcms)
999 {
1000     FWCfgState *fw_cfg;
1001     uint64_t *numa_fw_cfg;
1002     int i;
1003     const CPUArchIdList *cpus;
1004     MachineClass *mc = MACHINE_GET_CLASS(pcms);
1005 
1006     fw_cfg = fw_cfg_init_io_dma(FW_CFG_IO_BASE, FW_CFG_IO_BASE + 4, as);
1007     fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
1008 
1009     /* FW_CFG_MAX_CPUS is a bit confusing/problematic on x86:
1010      *
1011      * For machine types prior to 1.8, SeaBIOS needs FW_CFG_MAX_CPUS for
1012      * building MPTable, ACPI MADT, ACPI CPU hotplug and ACPI SRAT table,
1013      * that tables are based on xAPIC ID and QEMU<->SeaBIOS interface
1014      * for CPU hotplug also uses APIC ID and not "CPU index".
1015      * This means that FW_CFG_MAX_CPUS is not the "maximum number of CPUs",
1016      * but the "limit to the APIC ID values SeaBIOS may see".
1017      *
1018      * So for compatibility reasons with old BIOSes we are stuck with
1019      * "etc/max-cpus" actually being apic_id_limit
1020      */
1021     fw_cfg_add_i16(fw_cfg, FW_CFG_MAX_CPUS, (uint16_t)pcms->apic_id_limit);
1022     fw_cfg_add_i64(fw_cfg, FW_CFG_RAM_SIZE, (uint64_t)ram_size);
1023     fw_cfg_add_bytes(fw_cfg, FW_CFG_ACPI_TABLES,
1024                      acpi_tables, acpi_tables_len);
1025     fw_cfg_add_i32(fw_cfg, FW_CFG_IRQ0_OVERRIDE, kvm_allows_irq0_override());
1026 
1027     fw_cfg_add_bytes(fw_cfg, FW_CFG_E820_TABLE,
1028                      &e820_reserve, sizeof(e820_reserve));
1029     fw_cfg_add_file(fw_cfg, "etc/e820", e820_table,
1030                     sizeof(struct e820_entry) * e820_entries);
1031 
1032     fw_cfg_add_bytes(fw_cfg, FW_CFG_HPET, &hpet_cfg, sizeof(hpet_cfg));
1033     /* allocate memory for the NUMA channel: one (64bit) word for the number
1034      * of nodes, one word for each VCPU->node and one word for each node to
1035      * hold the amount of memory.
1036      */
1037     numa_fw_cfg = g_new0(uint64_t, 1 + pcms->apic_id_limit + nb_numa_nodes);
1038     numa_fw_cfg[0] = cpu_to_le64(nb_numa_nodes);
1039     cpus = mc->possible_cpu_arch_ids(MACHINE(pcms));
1040     for (i = 0; i < cpus->len; i++) {
1041         unsigned int apic_id = cpus->cpus[i].arch_id;
1042         assert(apic_id < pcms->apic_id_limit);
1043         numa_fw_cfg[apic_id + 1] = cpu_to_le64(cpus->cpus[i].props.node_id);
1044     }
1045     for (i = 0; i < nb_numa_nodes; i++) {
1046         numa_fw_cfg[pcms->apic_id_limit + 1 + i] =
1047             cpu_to_le64(numa_info[i].node_mem);
1048     }
1049     fw_cfg_add_bytes(fw_cfg, FW_CFG_NUMA, numa_fw_cfg,
1050                      (1 + pcms->apic_id_limit + nb_numa_nodes) *
1051                      sizeof(*numa_fw_cfg));
1052 
1053     return fw_cfg;
1054 }
1055 
1056 static long get_file_size(FILE *f)
1057 {
1058     long where, size;
1059 
1060     /* XXX: on Unix systems, using fstat() probably makes more sense */
1061 
1062     where = ftell(f);
1063     fseek(f, 0, SEEK_END);
1064     size = ftell(f);
1065     fseek(f, where, SEEK_SET);
1066 
1067     return size;
1068 }
1069 
1070 struct setup_data {
1071     uint64_t next;
1072     uint32_t type;
1073     uint32_t len;
1074     uint8_t data[0];
1075 } __attribute__((packed));
1076 
1077 
1078 /*
1079  * The entry point into the kernel for PVH boot is different from
1080  * the native entry point.  The PVH entry is defined by the x86/HVM
1081  * direct boot ABI and is available in an ELFNOTE in the kernel binary.
1082  *
1083  * This function is passed to load_elf() when it is called from
1084  * load_elfboot() which then additionally checks for an ELF Note of
1085  * type XEN_ELFNOTE_PHYS32_ENTRY and passes it to this function to
1086  * parse the PVH entry address from the ELF Note.
1087  *
1088  * Due to trickery in elf_opts.h, load_elf() is actually available as
1089  * load_elf32() or load_elf64() and this routine needs to be able
1090  * to deal with being called as 32 or 64 bit.
1091  *
1092  * The address of the PVH entry point is saved to the 'pvh_start_addr'
1093  * global variable.  (although the entry point is 32-bit, the kernel
1094  * binary can be either 32-bit or 64-bit).
1095  */
1096 static uint64_t read_pvh_start_addr(void *arg1, void *arg2, bool is64)
1097 {
1098     size_t *elf_note_data_addr;
1099 
1100     /* Check if ELF Note header passed in is valid */
1101     if (arg1 == NULL) {
1102         return 0;
1103     }
1104 
1105     if (is64) {
1106         struct elf64_note *nhdr64 = (struct elf64_note *)arg1;
1107         uint64_t nhdr_size64 = sizeof(struct elf64_note);
1108         uint64_t phdr_align = *(uint64_t *)arg2;
1109         uint64_t nhdr_namesz = nhdr64->n_namesz;
1110 
1111         elf_note_data_addr =
1112             ((void *)nhdr64) + nhdr_size64 +
1113             QEMU_ALIGN_UP(nhdr_namesz, phdr_align);
1114     } else {
1115         struct elf32_note *nhdr32 = (struct elf32_note *)arg1;
1116         uint32_t nhdr_size32 = sizeof(struct elf32_note);
1117         uint32_t phdr_align = *(uint32_t *)arg2;
1118         uint32_t nhdr_namesz = nhdr32->n_namesz;
1119 
1120         elf_note_data_addr =
1121             ((void *)nhdr32) + nhdr_size32 +
1122             QEMU_ALIGN_UP(nhdr_namesz, phdr_align);
1123     }
1124 
1125     pvh_start_addr = *elf_note_data_addr;
1126 
1127     return pvh_start_addr;
1128 }
1129 
1130 static bool load_elfboot(const char *kernel_filename,
1131                    int kernel_file_size,
1132                    uint8_t *header,
1133                    size_t pvh_xen_start_addr,
1134                    FWCfgState *fw_cfg)
1135 {
1136     uint32_t flags = 0;
1137     uint32_t mh_load_addr = 0;
1138     uint32_t elf_kernel_size = 0;
1139     uint64_t elf_entry;
1140     uint64_t elf_low, elf_high;
1141     int kernel_size;
1142 
1143     if (ldl_p(header) != 0x464c457f) {
1144         return false; /* no elfboot */
1145     }
1146 
1147     bool elf_is64 = header[EI_CLASS] == ELFCLASS64;
1148     flags = elf_is64 ?
1149         ((Elf64_Ehdr *)header)->e_flags : ((Elf32_Ehdr *)header)->e_flags;
1150 
1151     if (flags & 0x00010004) { /* LOAD_ELF_HEADER_HAS_ADDR */
1152         error_report("elfboot unsupported flags = %x", flags);
1153         exit(1);
1154     }
1155 
1156     uint64_t elf_note_type = XEN_ELFNOTE_PHYS32_ENTRY;
1157     kernel_size = load_elf(kernel_filename, read_pvh_start_addr,
1158                            NULL, &elf_note_type, &elf_entry,
1159                            &elf_low, &elf_high, 0, I386_ELF_MACHINE,
1160                            0, 0);
1161 
1162     if (kernel_size < 0) {
1163         error_report("Error while loading elf kernel");
1164         exit(1);
1165     }
1166     mh_load_addr = elf_low;
1167     elf_kernel_size = elf_high - elf_low;
1168 
1169     if (pvh_start_addr == 0) {
1170         error_report("Error loading uncompressed kernel without PVH ELF Note");
1171         exit(1);
1172     }
1173     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ENTRY, pvh_start_addr);
1174     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ADDR, mh_load_addr);
1175     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_SIZE, elf_kernel_size);
1176 
1177     return true;
1178 }
1179 
1180 static void load_linux(PCMachineState *pcms,
1181                        FWCfgState *fw_cfg)
1182 {
1183     uint16_t protocol;
1184     int setup_size, kernel_size, cmdline_size;
1185     int dtb_size, setup_data_offset;
1186     uint32_t initrd_max;
1187     uint8_t header[8192], *setup, *kernel;
1188     hwaddr real_addr, prot_addr, cmdline_addr, initrd_addr = 0;
1189     FILE *f;
1190     char *vmode;
1191     MachineState *machine = MACHINE(pcms);
1192     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1193     struct setup_data *setup_data;
1194     const char *kernel_filename = machine->kernel_filename;
1195     const char *initrd_filename = machine->initrd_filename;
1196     const char *dtb_filename = machine->dtb;
1197     const char *kernel_cmdline = machine->kernel_cmdline;
1198 
1199     /* Align to 16 bytes as a paranoia measure */
1200     cmdline_size = (strlen(kernel_cmdline)+16) & ~15;
1201 
1202     /* load the kernel header */
1203     f = fopen(kernel_filename, "rb");
1204     if (!f || !(kernel_size = get_file_size(f)) ||
1205         fread(header, 1, MIN(ARRAY_SIZE(header), kernel_size), f) !=
1206         MIN(ARRAY_SIZE(header), kernel_size)) {
1207         fprintf(stderr, "qemu: could not load kernel '%s': %s\n",
1208                 kernel_filename, strerror(errno));
1209         exit(1);
1210     }
1211 
1212     /* kernel protocol version */
1213 #if 0
1214     fprintf(stderr, "header magic: %#x\n", ldl_p(header+0x202));
1215 #endif
1216     if (ldl_p(header+0x202) == 0x53726448) {
1217         protocol = lduw_p(header+0x206);
1218     } else {
1219         /*
1220          * This could be a multiboot kernel. If it is, let's stop treating it
1221          * like a Linux kernel.
1222          * Note: some multiboot images could be in the ELF format (the same of
1223          * PVH), so we try multiboot first since we check the multiboot magic
1224          * header before to load it.
1225          */
1226         if (load_multiboot(fw_cfg, f, kernel_filename, initrd_filename,
1227                            kernel_cmdline, kernel_size, header)) {
1228             return;
1229         }
1230         /*
1231          * Check if the file is an uncompressed kernel file (ELF) and load it,
1232          * saving the PVH entry point used by the x86/HVM direct boot ABI.
1233          * If load_elfboot() is successful, populate the fw_cfg info.
1234          */
1235         if (pcmc->pvh_enabled &&
1236             load_elfboot(kernel_filename, kernel_size,
1237                          header, pvh_start_addr, fw_cfg)) {
1238             fclose(f);
1239 
1240             fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_SIZE,
1241                 strlen(kernel_cmdline) + 1);
1242             fw_cfg_add_string(fw_cfg, FW_CFG_CMDLINE_DATA, kernel_cmdline);
1243 
1244             fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_SIZE, sizeof(header));
1245             fw_cfg_add_bytes(fw_cfg, FW_CFG_SETUP_DATA,
1246                              header, sizeof(header));
1247 
1248             /* load initrd */
1249             if (initrd_filename) {
1250                 gsize initrd_size;
1251                 gchar *initrd_data;
1252                 GError *gerr = NULL;
1253 
1254                 if (!g_file_get_contents(initrd_filename, &initrd_data,
1255                             &initrd_size, &gerr)) {
1256                     fprintf(stderr, "qemu: error reading initrd %s: %s\n",
1257                             initrd_filename, gerr->message);
1258                     exit(1);
1259                 }
1260 
1261                 initrd_max = pcms->below_4g_mem_size - pcmc->acpi_data_size - 1;
1262                 if (initrd_size >= initrd_max) {
1263                     fprintf(stderr, "qemu: initrd is too large, cannot support."
1264                             "(max: %"PRIu32", need %"PRId64")\n",
1265                             initrd_max, (uint64_t)initrd_size);
1266                     exit(1);
1267                 }
1268 
1269                 initrd_addr = (initrd_max - initrd_size) & ~4095;
1270 
1271                 fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_ADDR, initrd_addr);
1272                 fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_SIZE, initrd_size);
1273                 fw_cfg_add_bytes(fw_cfg, FW_CFG_INITRD_DATA, initrd_data,
1274                                  initrd_size);
1275             }
1276 
1277             option_rom[nb_option_roms].bootindex = 0;
1278             option_rom[nb_option_roms].name = "pvh.bin";
1279             nb_option_roms++;
1280 
1281             return;
1282         }
1283         protocol = 0;
1284     }
1285 
1286     if (protocol < 0x200 || !(header[0x211] & 0x01)) {
1287         /* Low kernel */
1288         real_addr    = 0x90000;
1289         cmdline_addr = 0x9a000 - cmdline_size;
1290         prot_addr    = 0x10000;
1291     } else if (protocol < 0x202) {
1292         /* High but ancient kernel */
1293         real_addr    = 0x90000;
1294         cmdline_addr = 0x9a000 - cmdline_size;
1295         prot_addr    = 0x100000;
1296     } else {
1297         /* High and recent kernel */
1298         real_addr    = 0x10000;
1299         cmdline_addr = 0x20000;
1300         prot_addr    = 0x100000;
1301     }
1302 
1303 #if 0
1304     fprintf(stderr,
1305             "qemu: real_addr     = 0x" TARGET_FMT_plx "\n"
1306             "qemu: cmdline_addr  = 0x" TARGET_FMT_plx "\n"
1307             "qemu: prot_addr     = 0x" TARGET_FMT_plx "\n",
1308             real_addr,
1309             cmdline_addr,
1310             prot_addr);
1311 #endif
1312 
1313     /* highest address for loading the initrd */
1314     if (protocol >= 0x20c &&
1315         lduw_p(header+0x236) & XLF_CAN_BE_LOADED_ABOVE_4G) {
1316         /*
1317          * Linux has supported initrd up to 4 GB for a very long time (2007,
1318          * long before XLF_CAN_BE_LOADED_ABOVE_4G which was added in 2013),
1319          * though it only sets initrd_max to 2 GB to "work around bootloader
1320          * bugs". Luckily, QEMU firmware(which does something like bootloader)
1321          * has supported this.
1322          *
1323          * It's believed that if XLF_CAN_BE_LOADED_ABOVE_4G is set, initrd can
1324          * be loaded into any address.
1325          *
1326          * In addition, initrd_max is uint32_t simply because QEMU doesn't
1327          * support the 64-bit boot protocol (specifically the ext_ramdisk_image
1328          * field).
1329          *
1330          * Therefore here just limit initrd_max to UINT32_MAX simply as well.
1331          */
1332         initrd_max = UINT32_MAX;
1333     } else if (protocol >= 0x203) {
1334         initrd_max = ldl_p(header+0x22c);
1335     } else {
1336         initrd_max = 0x37ffffff;
1337     }
1338 
1339     if (initrd_max >= pcms->below_4g_mem_size - pcmc->acpi_data_size) {
1340         initrd_max = pcms->below_4g_mem_size - pcmc->acpi_data_size - 1;
1341     }
1342 
1343     fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_ADDR, cmdline_addr);
1344     fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_SIZE, strlen(kernel_cmdline)+1);
1345     fw_cfg_add_string(fw_cfg, FW_CFG_CMDLINE_DATA, kernel_cmdline);
1346 
1347     if (protocol >= 0x202) {
1348         stl_p(header+0x228, cmdline_addr);
1349     } else {
1350         stw_p(header+0x20, 0xA33F);
1351         stw_p(header+0x22, cmdline_addr-real_addr);
1352     }
1353 
1354     /* handle vga= parameter */
1355     vmode = strstr(kernel_cmdline, "vga=");
1356     if (vmode) {
1357         unsigned int video_mode;
1358         /* skip "vga=" */
1359         vmode += 4;
1360         if (!strncmp(vmode, "normal", 6)) {
1361             video_mode = 0xffff;
1362         } else if (!strncmp(vmode, "ext", 3)) {
1363             video_mode = 0xfffe;
1364         } else if (!strncmp(vmode, "ask", 3)) {
1365             video_mode = 0xfffd;
1366         } else {
1367             video_mode = strtol(vmode, NULL, 0);
1368         }
1369         stw_p(header+0x1fa, video_mode);
1370     }
1371 
1372     /* loader type */
1373     /* High nybble = B reserved for QEMU; low nybble is revision number.
1374        If this code is substantially changed, you may want to consider
1375        incrementing the revision. */
1376     if (protocol >= 0x200) {
1377         header[0x210] = 0xB0;
1378     }
1379     /* heap */
1380     if (protocol >= 0x201) {
1381         header[0x211] |= 0x80;	/* CAN_USE_HEAP */
1382         stw_p(header+0x224, cmdline_addr-real_addr-0x200);
1383     }
1384 
1385     /* load initrd */
1386     if (initrd_filename) {
1387         gsize initrd_size;
1388         gchar *initrd_data;
1389         GError *gerr = NULL;
1390 
1391         if (protocol < 0x200) {
1392             fprintf(stderr, "qemu: linux kernel too old to load a ram disk\n");
1393             exit(1);
1394         }
1395 
1396         if (!g_file_get_contents(initrd_filename, &initrd_data,
1397                                  &initrd_size, &gerr)) {
1398             fprintf(stderr, "qemu: error reading initrd %s: %s\n",
1399                     initrd_filename, gerr->message);
1400             exit(1);
1401         }
1402         if (initrd_size >= initrd_max) {
1403             fprintf(stderr, "qemu: initrd is too large, cannot support."
1404                     "(max: %"PRIu32", need %"PRId64")\n",
1405                     initrd_max, (uint64_t)initrd_size);
1406             exit(1);
1407         }
1408 
1409         initrd_addr = (initrd_max-initrd_size) & ~4095;
1410 
1411         fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_ADDR, initrd_addr);
1412         fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_SIZE, initrd_size);
1413         fw_cfg_add_bytes(fw_cfg, FW_CFG_INITRD_DATA, initrd_data, initrd_size);
1414 
1415         stl_p(header+0x218, initrd_addr);
1416         stl_p(header+0x21c, initrd_size);
1417     }
1418 
1419     /* load kernel and setup */
1420     setup_size = header[0x1f1];
1421     if (setup_size == 0) {
1422         setup_size = 4;
1423     }
1424     setup_size = (setup_size+1)*512;
1425     if (setup_size > kernel_size) {
1426         fprintf(stderr, "qemu: invalid kernel header\n");
1427         exit(1);
1428     }
1429     kernel_size -= setup_size;
1430 
1431     setup  = g_malloc(setup_size);
1432     kernel = g_malloc(kernel_size);
1433     fseek(f, 0, SEEK_SET);
1434     if (fread(setup, 1, setup_size, f) != setup_size) {
1435         fprintf(stderr, "fread() failed\n");
1436         exit(1);
1437     }
1438     if (fread(kernel, 1, kernel_size, f) != kernel_size) {
1439         fprintf(stderr, "fread() failed\n");
1440         exit(1);
1441     }
1442     fclose(f);
1443 
1444     /* append dtb to kernel */
1445     if (dtb_filename) {
1446         if (protocol < 0x209) {
1447             fprintf(stderr, "qemu: Linux kernel too old to load a dtb\n");
1448             exit(1);
1449         }
1450 
1451         dtb_size = get_image_size(dtb_filename);
1452         if (dtb_size <= 0) {
1453             fprintf(stderr, "qemu: error reading dtb %s: %s\n",
1454                     dtb_filename, strerror(errno));
1455             exit(1);
1456         }
1457 
1458         setup_data_offset = QEMU_ALIGN_UP(kernel_size, 16);
1459         kernel_size = setup_data_offset + sizeof(struct setup_data) + dtb_size;
1460         kernel = g_realloc(kernel, kernel_size);
1461 
1462         stq_p(header+0x250, prot_addr + setup_data_offset);
1463 
1464         setup_data = (struct setup_data *)(kernel + setup_data_offset);
1465         setup_data->next = 0;
1466         setup_data->type = cpu_to_le32(SETUP_DTB);
1467         setup_data->len = cpu_to_le32(dtb_size);
1468 
1469         load_image_size(dtb_filename, setup_data->data, dtb_size);
1470     }
1471 
1472     memcpy(setup, header, MIN(sizeof(header), setup_size));
1473 
1474     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ADDR, prot_addr);
1475     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_SIZE, kernel_size);
1476     fw_cfg_add_bytes(fw_cfg, FW_CFG_KERNEL_DATA, kernel, kernel_size);
1477 
1478     fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_ADDR, real_addr);
1479     fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_SIZE, setup_size);
1480     fw_cfg_add_bytes(fw_cfg, FW_CFG_SETUP_DATA, setup, setup_size);
1481 
1482     option_rom[nb_option_roms].bootindex = 0;
1483     option_rom[nb_option_roms].name = "linuxboot.bin";
1484     if (pcmc->linuxboot_dma_enabled && fw_cfg_dma_enabled(fw_cfg)) {
1485         option_rom[nb_option_roms].name = "linuxboot_dma.bin";
1486     }
1487     nb_option_roms++;
1488 }
1489 
1490 #define NE2000_NB_MAX 6
1491 
1492 static const int ne2000_io[NE2000_NB_MAX] = { 0x300, 0x320, 0x340, 0x360,
1493                                               0x280, 0x380 };
1494 static const int ne2000_irq[NE2000_NB_MAX] = { 9, 10, 11, 3, 4, 5 };
1495 
1496 void pc_init_ne2k_isa(ISABus *bus, NICInfo *nd)
1497 {
1498     static int nb_ne2k = 0;
1499 
1500     if (nb_ne2k == NE2000_NB_MAX)
1501         return;
1502     isa_ne2000_init(bus, ne2000_io[nb_ne2k],
1503                     ne2000_irq[nb_ne2k], nd);
1504     nb_ne2k++;
1505 }
1506 
1507 DeviceState *cpu_get_current_apic(void)
1508 {
1509     if (current_cpu) {
1510         X86CPU *cpu = X86_CPU(current_cpu);
1511         return cpu->apic_state;
1512     } else {
1513         return NULL;
1514     }
1515 }
1516 
1517 void pc_acpi_smi_interrupt(void *opaque, int irq, int level)
1518 {
1519     X86CPU *cpu = opaque;
1520 
1521     if (level) {
1522         cpu_interrupt(CPU(cpu), CPU_INTERRUPT_SMI);
1523     }
1524 }
1525 
1526 static void pc_new_cpu(PCMachineState *pcms, int64_t apic_id, Error **errp)
1527 {
1528     Object *cpu = NULL;
1529     Error *local_err = NULL;
1530     CPUX86State *env = NULL;
1531 
1532     cpu = object_new(MACHINE(pcms)->cpu_type);
1533 
1534     env = &X86_CPU(cpu)->env;
1535     env->nr_dies = pcms->smp_dies;
1536 
1537     object_property_set_uint(cpu, apic_id, "apic-id", &local_err);
1538     object_property_set_bool(cpu, true, "realized", &local_err);
1539 
1540     object_unref(cpu);
1541     error_propagate(errp, local_err);
1542 }
1543 
1544 /*
1545  * This function is very similar to smp_parse()
1546  * in hw/core/machine.c but includes CPU die support.
1547  */
1548 void pc_smp_parse(MachineState *ms, QemuOpts *opts)
1549 {
1550     PCMachineState *pcms = PC_MACHINE(ms);
1551 
1552     if (opts) {
1553         unsigned cpus    = qemu_opt_get_number(opts, "cpus", 0);
1554         unsigned sockets = qemu_opt_get_number(opts, "sockets", 0);
1555         unsigned dies = qemu_opt_get_number(opts, "dies", 1);
1556         unsigned cores   = qemu_opt_get_number(opts, "cores", 0);
1557         unsigned threads = qemu_opt_get_number(opts, "threads", 0);
1558 
1559         /* compute missing values, prefer sockets over cores over threads */
1560         if (cpus == 0 || sockets == 0) {
1561             cores = cores > 0 ? cores : 1;
1562             threads = threads > 0 ? threads : 1;
1563             if (cpus == 0) {
1564                 sockets = sockets > 0 ? sockets : 1;
1565                 cpus = cores * threads * dies * sockets;
1566             } else {
1567                 ms->smp.max_cpus =
1568                         qemu_opt_get_number(opts, "maxcpus", cpus);
1569                 sockets = ms->smp.max_cpus / (cores * threads * dies);
1570             }
1571         } else if (cores == 0) {
1572             threads = threads > 0 ? threads : 1;
1573             cores = cpus / (sockets * dies * threads);
1574             cores = cores > 0 ? cores : 1;
1575         } else if (threads == 0) {
1576             threads = cpus / (cores * dies * sockets);
1577             threads = threads > 0 ? threads : 1;
1578         } else if (sockets * dies * cores * threads < cpus) {
1579             error_report("cpu topology: "
1580                          "sockets (%u) * dies (%u) * cores (%u) * threads (%u) < "
1581                          "smp_cpus (%u)",
1582                          sockets, dies, cores, threads, cpus);
1583             exit(1);
1584         }
1585 
1586         ms->smp.max_cpus =
1587                 qemu_opt_get_number(opts, "maxcpus", cpus);
1588 
1589         if (ms->smp.max_cpus < cpus) {
1590             error_report("maxcpus must be equal to or greater than smp");
1591             exit(1);
1592         }
1593 
1594         if (sockets * dies * cores * threads > ms->smp.max_cpus) {
1595             error_report("cpu topology: "
1596                          "sockets (%u) * dies (%u) * cores (%u) * threads (%u) > "
1597                          "maxcpus (%u)",
1598                          sockets, dies, cores, threads,
1599                          ms->smp.max_cpus);
1600             exit(1);
1601         }
1602 
1603         if (sockets * dies * cores * threads != ms->smp.max_cpus) {
1604             warn_report("Invalid CPU topology deprecated: "
1605                         "sockets (%u) * dies (%u) * cores (%u) * threads (%u) "
1606                         "!= maxcpus (%u)",
1607                         sockets, dies, cores, threads,
1608                         ms->smp.max_cpus);
1609         }
1610 
1611         ms->smp.cpus = cpus;
1612         ms->smp.cores = cores;
1613         ms->smp.threads = threads;
1614         pcms->smp_dies = dies;
1615     }
1616 
1617     if (ms->smp.cpus > 1) {
1618         Error *blocker = NULL;
1619         error_setg(&blocker, QERR_REPLAY_NOT_SUPPORTED, "smp");
1620         replay_add_blocker(blocker);
1621     }
1622 }
1623 
1624 void pc_hot_add_cpu(MachineState *ms, const int64_t id, Error **errp)
1625 {
1626     PCMachineState *pcms = PC_MACHINE(ms);
1627     int64_t apic_id = x86_cpu_apic_id_from_index(pcms, id);
1628     Error *local_err = NULL;
1629 
1630     if (id < 0) {
1631         error_setg(errp, "Invalid CPU id: %" PRIi64, id);
1632         return;
1633     }
1634 
1635     if (apic_id >= ACPI_CPU_HOTPLUG_ID_LIMIT) {
1636         error_setg(errp, "Unable to add CPU: %" PRIi64
1637                    ", resulting APIC ID (%" PRIi64 ") is too large",
1638                    id, apic_id);
1639         return;
1640     }
1641 
1642     pc_new_cpu(PC_MACHINE(ms), apic_id, &local_err);
1643     if (local_err) {
1644         error_propagate(errp, local_err);
1645         return;
1646     }
1647 }
1648 
1649 void pc_cpus_init(PCMachineState *pcms)
1650 {
1651     int i;
1652     const CPUArchIdList *possible_cpus;
1653     MachineState *ms = MACHINE(pcms);
1654     MachineClass *mc = MACHINE_GET_CLASS(pcms);
1655     PCMachineClass *pcmc = PC_MACHINE_CLASS(mc);
1656 
1657     x86_cpu_set_default_version(pcmc->default_cpu_version);
1658 
1659     /* Calculates the limit to CPU APIC ID values
1660      *
1661      * Limit for the APIC ID value, so that all
1662      * CPU APIC IDs are < pcms->apic_id_limit.
1663      *
1664      * This is used for FW_CFG_MAX_CPUS. See comments on bochs_bios_init().
1665      */
1666     pcms->apic_id_limit = x86_cpu_apic_id_from_index(pcms,
1667                                                      ms->smp.max_cpus - 1) + 1;
1668     possible_cpus = mc->possible_cpu_arch_ids(ms);
1669     for (i = 0; i < ms->smp.cpus; i++) {
1670         pc_new_cpu(pcms, possible_cpus->cpus[i].arch_id, &error_fatal);
1671     }
1672 }
1673 
1674 static void pc_build_feature_control_file(PCMachineState *pcms)
1675 {
1676     MachineState *ms = MACHINE(pcms);
1677     X86CPU *cpu = X86_CPU(ms->possible_cpus->cpus[0].cpu);
1678     CPUX86State *env = &cpu->env;
1679     uint32_t unused, ecx, edx;
1680     uint64_t feature_control_bits = 0;
1681     uint64_t *val;
1682 
1683     cpu_x86_cpuid(env, 1, 0, &unused, &unused, &ecx, &edx);
1684     if (ecx & CPUID_EXT_VMX) {
1685         feature_control_bits |= FEATURE_CONTROL_VMXON_ENABLED_OUTSIDE_SMX;
1686     }
1687 
1688     if ((edx & (CPUID_EXT2_MCE | CPUID_EXT2_MCA)) ==
1689         (CPUID_EXT2_MCE | CPUID_EXT2_MCA) &&
1690         (env->mcg_cap & MCG_LMCE_P)) {
1691         feature_control_bits |= FEATURE_CONTROL_LMCE;
1692     }
1693 
1694     if (!feature_control_bits) {
1695         return;
1696     }
1697 
1698     val = g_malloc(sizeof(*val));
1699     *val = cpu_to_le64(feature_control_bits | FEATURE_CONTROL_LOCKED);
1700     fw_cfg_add_file(pcms->fw_cfg, "etc/msr_feature_control", val, sizeof(*val));
1701 }
1702 
1703 static void rtc_set_cpus_count(ISADevice *rtc, uint16_t cpus_count)
1704 {
1705     if (cpus_count > 0xff) {
1706         /* If the number of CPUs can't be represented in 8 bits, the
1707          * BIOS must use "FW_CFG_NB_CPUS". Set RTC field to 0 just
1708          * to make old BIOSes fail more predictably.
1709          */
1710         rtc_set_memory(rtc, 0x5f, 0);
1711     } else {
1712         rtc_set_memory(rtc, 0x5f, cpus_count - 1);
1713     }
1714 }
1715 
1716 static
1717 void pc_machine_done(Notifier *notifier, void *data)
1718 {
1719     PCMachineState *pcms = container_of(notifier,
1720                                         PCMachineState, machine_done);
1721     PCIBus *bus = pcms->bus;
1722 
1723     /* set the number of CPUs */
1724     rtc_set_cpus_count(pcms->rtc, pcms->boot_cpus);
1725 
1726     if (bus) {
1727         int extra_hosts = 0;
1728 
1729         QLIST_FOREACH(bus, &bus->child, sibling) {
1730             /* look for expander root buses */
1731             if (pci_bus_is_root(bus)) {
1732                 extra_hosts++;
1733             }
1734         }
1735         if (extra_hosts && pcms->fw_cfg) {
1736             uint64_t *val = g_malloc(sizeof(*val));
1737             *val = cpu_to_le64(extra_hosts);
1738             fw_cfg_add_file(pcms->fw_cfg,
1739                     "etc/extra-pci-roots", val, sizeof(*val));
1740         }
1741     }
1742 
1743     acpi_setup();
1744     if (pcms->fw_cfg) {
1745         pc_build_smbios(pcms);
1746         pc_build_feature_control_file(pcms);
1747         /* update FW_CFG_NB_CPUS to account for -device added CPUs */
1748         fw_cfg_modify_i16(pcms->fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
1749     }
1750 
1751     if (pcms->apic_id_limit > 255 && !xen_enabled()) {
1752         IntelIOMMUState *iommu = INTEL_IOMMU_DEVICE(x86_iommu_get_default());
1753 
1754         if (!iommu || !x86_iommu_ir_supported(X86_IOMMU_DEVICE(iommu)) ||
1755             iommu->intr_eim != ON_OFF_AUTO_ON) {
1756             error_report("current -smp configuration requires "
1757                          "Extended Interrupt Mode enabled. "
1758                          "You can add an IOMMU using: "
1759                          "-device intel-iommu,intremap=on,eim=on");
1760             exit(EXIT_FAILURE);
1761         }
1762     }
1763 }
1764 
1765 void pc_guest_info_init(PCMachineState *pcms)
1766 {
1767     int i;
1768 
1769     pcms->apic_xrupt_override = kvm_allows_irq0_override();
1770     pcms->numa_nodes = nb_numa_nodes;
1771     pcms->node_mem = g_malloc0(pcms->numa_nodes *
1772                                     sizeof *pcms->node_mem);
1773     for (i = 0; i < nb_numa_nodes; i++) {
1774         pcms->node_mem[i] = numa_info[i].node_mem;
1775     }
1776 
1777     pcms->machine_done.notify = pc_machine_done;
1778     qemu_add_machine_init_done_notifier(&pcms->machine_done);
1779 }
1780 
1781 /* setup pci memory address space mapping into system address space */
1782 void pc_pci_as_mapping_init(Object *owner, MemoryRegion *system_memory,
1783                             MemoryRegion *pci_address_space)
1784 {
1785     /* Set to lower priority than RAM */
1786     memory_region_add_subregion_overlap(system_memory, 0x0,
1787                                         pci_address_space, -1);
1788 }
1789 
1790 void xen_load_linux(PCMachineState *pcms)
1791 {
1792     int i;
1793     FWCfgState *fw_cfg;
1794 
1795     assert(MACHINE(pcms)->kernel_filename != NULL);
1796 
1797     fw_cfg = fw_cfg_init_io(FW_CFG_IO_BASE);
1798     fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
1799     rom_set_fw(fw_cfg);
1800 
1801     load_linux(pcms, fw_cfg);
1802     for (i = 0; i < nb_option_roms; i++) {
1803         assert(!strcmp(option_rom[i].name, "linuxboot.bin") ||
1804                !strcmp(option_rom[i].name, "linuxboot_dma.bin") ||
1805                !strcmp(option_rom[i].name, "pvh.bin") ||
1806                !strcmp(option_rom[i].name, "multiboot.bin"));
1807         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1808     }
1809     pcms->fw_cfg = fw_cfg;
1810 }
1811 
1812 void pc_memory_init(PCMachineState *pcms,
1813                     MemoryRegion *system_memory,
1814                     MemoryRegion *rom_memory,
1815                     MemoryRegion **ram_memory)
1816 {
1817     int linux_boot, i;
1818     MemoryRegion *ram, *option_rom_mr;
1819     MemoryRegion *ram_below_4g, *ram_above_4g;
1820     FWCfgState *fw_cfg;
1821     MachineState *machine = MACHINE(pcms);
1822     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1823 
1824     assert(machine->ram_size == pcms->below_4g_mem_size +
1825                                 pcms->above_4g_mem_size);
1826 
1827     linux_boot = (machine->kernel_filename != NULL);
1828 
1829     /* Allocate RAM.  We allocate it as a single memory region and use
1830      * aliases to address portions of it, mostly for backwards compatibility
1831      * with older qemus that used qemu_ram_alloc().
1832      */
1833     ram = g_malloc(sizeof(*ram));
1834     memory_region_allocate_system_memory(ram, NULL, "pc.ram",
1835                                          machine->ram_size);
1836     *ram_memory = ram;
1837     ram_below_4g = g_malloc(sizeof(*ram_below_4g));
1838     memory_region_init_alias(ram_below_4g, NULL, "ram-below-4g", ram,
1839                              0, pcms->below_4g_mem_size);
1840     memory_region_add_subregion(system_memory, 0, ram_below_4g);
1841     e820_add_entry(0, pcms->below_4g_mem_size, E820_RAM);
1842     if (pcms->above_4g_mem_size > 0) {
1843         ram_above_4g = g_malloc(sizeof(*ram_above_4g));
1844         memory_region_init_alias(ram_above_4g, NULL, "ram-above-4g", ram,
1845                                  pcms->below_4g_mem_size,
1846                                  pcms->above_4g_mem_size);
1847         memory_region_add_subregion(system_memory, 0x100000000ULL,
1848                                     ram_above_4g);
1849         e820_add_entry(0x100000000ULL, pcms->above_4g_mem_size, E820_RAM);
1850     }
1851 
1852     if (!pcmc->has_reserved_memory &&
1853         (machine->ram_slots ||
1854          (machine->maxram_size > machine->ram_size))) {
1855         MachineClass *mc = MACHINE_GET_CLASS(machine);
1856 
1857         error_report("\"-memory 'slots|maxmem'\" is not supported by: %s",
1858                      mc->name);
1859         exit(EXIT_FAILURE);
1860     }
1861 
1862     /* always allocate the device memory information */
1863     machine->device_memory = g_malloc0(sizeof(*machine->device_memory));
1864 
1865     /* initialize device memory address space */
1866     if (pcmc->has_reserved_memory &&
1867         (machine->ram_size < machine->maxram_size)) {
1868         ram_addr_t device_mem_size = machine->maxram_size - machine->ram_size;
1869 
1870         if (machine->ram_slots > ACPI_MAX_RAM_SLOTS) {
1871             error_report("unsupported amount of memory slots: %"PRIu64,
1872                          machine->ram_slots);
1873             exit(EXIT_FAILURE);
1874         }
1875 
1876         if (QEMU_ALIGN_UP(machine->maxram_size,
1877                           TARGET_PAGE_SIZE) != machine->maxram_size) {
1878             error_report("maximum memory size must by aligned to multiple of "
1879                          "%d bytes", TARGET_PAGE_SIZE);
1880             exit(EXIT_FAILURE);
1881         }
1882 
1883         machine->device_memory->base =
1884             ROUND_UP(0x100000000ULL + pcms->above_4g_mem_size, 1 * GiB);
1885 
1886         if (pcmc->enforce_aligned_dimm) {
1887             /* size device region assuming 1G page max alignment per slot */
1888             device_mem_size += (1 * GiB) * machine->ram_slots;
1889         }
1890 
1891         if ((machine->device_memory->base + device_mem_size) <
1892             device_mem_size) {
1893             error_report("unsupported amount of maximum memory: " RAM_ADDR_FMT,
1894                          machine->maxram_size);
1895             exit(EXIT_FAILURE);
1896         }
1897 
1898         memory_region_init(&machine->device_memory->mr, OBJECT(pcms),
1899                            "device-memory", device_mem_size);
1900         memory_region_add_subregion(system_memory, machine->device_memory->base,
1901                                     &machine->device_memory->mr);
1902     }
1903 
1904     /* Initialize PC system firmware */
1905     pc_system_firmware_init(pcms, rom_memory);
1906 
1907     option_rom_mr = g_malloc(sizeof(*option_rom_mr));
1908     memory_region_init_ram(option_rom_mr, NULL, "pc.rom", PC_ROM_SIZE,
1909                            &error_fatal);
1910     if (pcmc->pci_enabled) {
1911         memory_region_set_readonly(option_rom_mr, true);
1912     }
1913     memory_region_add_subregion_overlap(rom_memory,
1914                                         PC_ROM_MIN_VGA,
1915                                         option_rom_mr,
1916                                         1);
1917 
1918     fw_cfg = bochs_bios_init(&address_space_memory, pcms);
1919 
1920     rom_set_fw(fw_cfg);
1921 
1922     if (pcmc->has_reserved_memory && machine->device_memory->base) {
1923         uint64_t *val = g_malloc(sizeof(*val));
1924         PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1925         uint64_t res_mem_end = machine->device_memory->base;
1926 
1927         if (!pcmc->broken_reserved_end) {
1928             res_mem_end += memory_region_size(&machine->device_memory->mr);
1929         }
1930         *val = cpu_to_le64(ROUND_UP(res_mem_end, 1 * GiB));
1931         fw_cfg_add_file(fw_cfg, "etc/reserved-memory-end", val, sizeof(*val));
1932     }
1933 
1934     if (linux_boot) {
1935         load_linux(pcms, fw_cfg);
1936     }
1937 
1938     for (i = 0; i < nb_option_roms; i++) {
1939         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1940     }
1941     pcms->fw_cfg = fw_cfg;
1942 
1943     /* Init default IOAPIC address space */
1944     pcms->ioapic_as = &address_space_memory;
1945 }
1946 
1947 /*
1948  * The 64bit pci hole starts after "above 4G RAM" and
1949  * potentially the space reserved for memory hotplug.
1950  */
1951 uint64_t pc_pci_hole64_start(void)
1952 {
1953     PCMachineState *pcms = PC_MACHINE(qdev_get_machine());
1954     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1955     MachineState *ms = MACHINE(pcms);
1956     uint64_t hole64_start = 0;
1957 
1958     if (pcmc->has_reserved_memory && ms->device_memory->base) {
1959         hole64_start = ms->device_memory->base;
1960         if (!pcmc->broken_reserved_end) {
1961             hole64_start += memory_region_size(&ms->device_memory->mr);
1962         }
1963     } else {
1964         hole64_start = 0x100000000ULL + pcms->above_4g_mem_size;
1965     }
1966 
1967     return ROUND_UP(hole64_start, 1 * GiB);
1968 }
1969 
1970 qemu_irq pc_allocate_cpu_irq(void)
1971 {
1972     return qemu_allocate_irq(pic_irq_request, NULL, 0);
1973 }
1974 
1975 DeviceState *pc_vga_init(ISABus *isa_bus, PCIBus *pci_bus)
1976 {
1977     DeviceState *dev = NULL;
1978 
1979     rom_set_order_override(FW_CFG_ORDER_OVERRIDE_VGA);
1980     if (pci_bus) {
1981         PCIDevice *pcidev = pci_vga_init(pci_bus);
1982         dev = pcidev ? &pcidev->qdev : NULL;
1983     } else if (isa_bus) {
1984         ISADevice *isadev = isa_vga_init(isa_bus);
1985         dev = isadev ? DEVICE(isadev) : NULL;
1986     }
1987     rom_reset_order_override();
1988     return dev;
1989 }
1990 
1991 static const MemoryRegionOps ioport80_io_ops = {
1992     .write = ioport80_write,
1993     .read = ioport80_read,
1994     .endianness = DEVICE_NATIVE_ENDIAN,
1995     .impl = {
1996         .min_access_size = 1,
1997         .max_access_size = 1,
1998     },
1999 };
2000 
2001 static const MemoryRegionOps ioportF0_io_ops = {
2002     .write = ioportF0_write,
2003     .read = ioportF0_read,
2004     .endianness = DEVICE_NATIVE_ENDIAN,
2005     .impl = {
2006         .min_access_size = 1,
2007         .max_access_size = 1,
2008     },
2009 };
2010 
2011 static void pc_superio_init(ISABus *isa_bus, bool create_fdctrl, bool no_vmport)
2012 {
2013     int i;
2014     DriveInfo *fd[MAX_FD];
2015     qemu_irq *a20_line;
2016     ISADevice *i8042, *port92, *vmmouse;
2017 
2018     serial_hds_isa_init(isa_bus, 0, MAX_ISA_SERIAL_PORTS);
2019     parallel_hds_isa_init(isa_bus, MAX_PARALLEL_PORTS);
2020 
2021     for (i = 0; i < MAX_FD; i++) {
2022         fd[i] = drive_get(IF_FLOPPY, 0, i);
2023         create_fdctrl |= !!fd[i];
2024     }
2025     if (create_fdctrl) {
2026         fdctrl_init_isa(isa_bus, fd);
2027     }
2028 
2029     i8042 = isa_create_simple(isa_bus, "i8042");
2030     if (!no_vmport) {
2031         vmport_init(isa_bus);
2032         vmmouse = isa_try_create(isa_bus, "vmmouse");
2033     } else {
2034         vmmouse = NULL;
2035     }
2036     if (vmmouse) {
2037         DeviceState *dev = DEVICE(vmmouse);
2038         qdev_prop_set_ptr(dev, "ps2_mouse", i8042);
2039         qdev_init_nofail(dev);
2040     }
2041     port92 = isa_create_simple(isa_bus, "port92");
2042 
2043     a20_line = qemu_allocate_irqs(handle_a20_line_change, first_cpu, 2);
2044     i8042_setup_a20_line(i8042, a20_line[0]);
2045     port92_init(port92, a20_line[1]);
2046     g_free(a20_line);
2047 }
2048 
2049 void pc_basic_device_init(ISABus *isa_bus, qemu_irq *gsi,
2050                           ISADevice **rtc_state,
2051                           bool create_fdctrl,
2052                           bool no_vmport,
2053                           bool has_pit,
2054                           uint32_t hpet_irqs)
2055 {
2056     int i;
2057     DeviceState *hpet = NULL;
2058     int pit_isa_irq = 0;
2059     qemu_irq pit_alt_irq = NULL;
2060     qemu_irq rtc_irq = NULL;
2061     ISADevice *pit = NULL;
2062     MemoryRegion *ioport80_io = g_new(MemoryRegion, 1);
2063     MemoryRegion *ioportF0_io = g_new(MemoryRegion, 1);
2064 
2065     memory_region_init_io(ioport80_io, NULL, &ioport80_io_ops, NULL, "ioport80", 1);
2066     memory_region_add_subregion(isa_bus->address_space_io, 0x80, ioport80_io);
2067 
2068     memory_region_init_io(ioportF0_io, NULL, &ioportF0_io_ops, NULL, "ioportF0", 1);
2069     memory_region_add_subregion(isa_bus->address_space_io, 0xf0, ioportF0_io);
2070 
2071     /*
2072      * Check if an HPET shall be created.
2073      *
2074      * Without KVM_CAP_PIT_STATE2, we cannot switch off the in-kernel PIT
2075      * when the HPET wants to take over. Thus we have to disable the latter.
2076      */
2077     if (!no_hpet && (!kvm_irqchip_in_kernel() || kvm_has_pit_state2())) {
2078         /* In order to set property, here not using sysbus_try_create_simple */
2079         hpet = qdev_try_create(NULL, TYPE_HPET);
2080         if (hpet) {
2081             /* For pc-piix-*, hpet's intcap is always IRQ2. For pc-q35-1.7
2082              * and earlier, use IRQ2 for compat. Otherwise, use IRQ16~23,
2083              * IRQ8 and IRQ2.
2084              */
2085             uint8_t compat = object_property_get_uint(OBJECT(hpet),
2086                     HPET_INTCAP, NULL);
2087             if (!compat) {
2088                 qdev_prop_set_uint32(hpet, HPET_INTCAP, hpet_irqs);
2089             }
2090             qdev_init_nofail(hpet);
2091             sysbus_mmio_map(SYS_BUS_DEVICE(hpet), 0, HPET_BASE);
2092 
2093             for (i = 0; i < GSI_NUM_PINS; i++) {
2094                 sysbus_connect_irq(SYS_BUS_DEVICE(hpet), i, gsi[i]);
2095             }
2096             pit_isa_irq = -1;
2097             pit_alt_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_PIT_INT);
2098             rtc_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_RTC_INT);
2099         }
2100     }
2101     *rtc_state = mc146818_rtc_init(isa_bus, 2000, rtc_irq);
2102 
2103     qemu_register_boot_set(pc_boot_set, *rtc_state);
2104 
2105     if (!xen_enabled() && has_pit) {
2106         if (kvm_pit_in_kernel()) {
2107             pit = kvm_pit_init(isa_bus, 0x40);
2108         } else {
2109             pit = i8254_pit_init(isa_bus, 0x40, pit_isa_irq, pit_alt_irq);
2110         }
2111         if (hpet) {
2112             /* connect PIT to output control line of the HPET */
2113             qdev_connect_gpio_out(hpet, 0, qdev_get_gpio_in(DEVICE(pit), 0));
2114         }
2115         pcspk_init(isa_bus, pit);
2116     }
2117 
2118     i8257_dma_init(isa_bus, 0);
2119 
2120     /* Super I/O */
2121     pc_superio_init(isa_bus, create_fdctrl, no_vmport);
2122 }
2123 
2124 void pc_nic_init(PCMachineClass *pcmc, ISABus *isa_bus, PCIBus *pci_bus)
2125 {
2126     int i;
2127 
2128     rom_set_order_override(FW_CFG_ORDER_OVERRIDE_NIC);
2129     for (i = 0; i < nb_nics; i++) {
2130         NICInfo *nd = &nd_table[i];
2131         const char *model = nd->model ? nd->model : pcmc->default_nic_model;
2132 
2133         if (g_str_equal(model, "ne2k_isa")) {
2134             pc_init_ne2k_isa(isa_bus, nd);
2135         } else {
2136             pci_nic_init_nofail(nd, pci_bus, model, NULL);
2137         }
2138     }
2139     rom_reset_order_override();
2140 }
2141 
2142 void ioapic_init_gsi(GSIState *gsi_state, const char *parent_name)
2143 {
2144     DeviceState *dev;
2145     SysBusDevice *d;
2146     unsigned int i;
2147 
2148     if (kvm_ioapic_in_kernel()) {
2149         dev = qdev_create(NULL, TYPE_KVM_IOAPIC);
2150     } else {
2151         dev = qdev_create(NULL, TYPE_IOAPIC);
2152     }
2153     if (parent_name) {
2154         object_property_add_child(object_resolve_path(parent_name, NULL),
2155                                   "ioapic", OBJECT(dev), NULL);
2156     }
2157     qdev_init_nofail(dev);
2158     d = SYS_BUS_DEVICE(dev);
2159     sysbus_mmio_map(d, 0, IO_APIC_DEFAULT_ADDRESS);
2160 
2161     for (i = 0; i < IOAPIC_NUM_PINS; i++) {
2162         gsi_state->ioapic_irq[i] = qdev_get_gpio_in(dev, i);
2163     }
2164 }
2165 
2166 static void pc_memory_pre_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
2167                                Error **errp)
2168 {
2169     const PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2170     const PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
2171     const MachineState *ms = MACHINE(hotplug_dev);
2172     const bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2173     const uint64_t legacy_align = TARGET_PAGE_SIZE;
2174     Error *local_err = NULL;
2175 
2176     /*
2177      * When -no-acpi is used with Q35 machine type, no ACPI is built,
2178      * but pcms->acpi_dev is still created. Check !acpi_enabled in
2179      * addition to cover this case.
2180      */
2181     if (!pcms->acpi_dev || !acpi_enabled) {
2182         error_setg(errp,
2183                    "memory hotplug is not enabled: missing acpi device or acpi disabled");
2184         return;
2185     }
2186 
2187     if (is_nvdimm && !ms->nvdimms_state->is_enabled) {
2188         error_setg(errp, "nvdimm is not enabled: missing 'nvdimm' in '-M'");
2189         return;
2190     }
2191 
2192     hotplug_handler_pre_plug(pcms->acpi_dev, dev, &local_err);
2193     if (local_err) {
2194         error_propagate(errp, local_err);
2195         return;
2196     }
2197 
2198     pc_dimm_pre_plug(PC_DIMM(dev), MACHINE(hotplug_dev),
2199                      pcmc->enforce_aligned_dimm ? NULL : &legacy_align, errp);
2200 }
2201 
2202 static void pc_memory_plug(HotplugHandler *hotplug_dev,
2203                            DeviceState *dev, Error **errp)
2204 {
2205     Error *local_err = NULL;
2206     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2207     MachineState *ms = MACHINE(hotplug_dev);
2208     bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2209 
2210     pc_dimm_plug(PC_DIMM(dev), MACHINE(pcms), &local_err);
2211     if (local_err) {
2212         goto out;
2213     }
2214 
2215     if (is_nvdimm) {
2216         nvdimm_plug(ms->nvdimms_state);
2217     }
2218 
2219     hotplug_handler_plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &error_abort);
2220 out:
2221     error_propagate(errp, local_err);
2222 }
2223 
2224 static void pc_memory_unplug_request(HotplugHandler *hotplug_dev,
2225                                      DeviceState *dev, Error **errp)
2226 {
2227     Error *local_err = NULL;
2228     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2229 
2230     /*
2231      * When -no-acpi is used with Q35 machine type, no ACPI is built,
2232      * but pcms->acpi_dev is still created. Check !acpi_enabled in
2233      * addition to cover this case.
2234      */
2235     if (!pcms->acpi_dev || !acpi_enabled) {
2236         error_setg(&local_err,
2237                    "memory hotplug is not enabled: missing acpi device or acpi disabled");
2238         goto out;
2239     }
2240 
2241     if (object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM)) {
2242         error_setg(&local_err,
2243                    "nvdimm device hot unplug is not supported yet.");
2244         goto out;
2245     }
2246 
2247     hotplug_handler_unplug_request(HOTPLUG_HANDLER(pcms->acpi_dev), dev,
2248                                    &local_err);
2249 out:
2250     error_propagate(errp, local_err);
2251 }
2252 
2253 static void pc_memory_unplug(HotplugHandler *hotplug_dev,
2254                              DeviceState *dev, Error **errp)
2255 {
2256     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2257     Error *local_err = NULL;
2258 
2259     hotplug_handler_unplug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
2260     if (local_err) {
2261         goto out;
2262     }
2263 
2264     pc_dimm_unplug(PC_DIMM(dev), MACHINE(pcms));
2265     object_property_set_bool(OBJECT(dev), false, "realized", NULL);
2266  out:
2267     error_propagate(errp, local_err);
2268 }
2269 
2270 static int pc_apic_cmp(const void *a, const void *b)
2271 {
2272    CPUArchId *apic_a = (CPUArchId *)a;
2273    CPUArchId *apic_b = (CPUArchId *)b;
2274 
2275    return apic_a->arch_id - apic_b->arch_id;
2276 }
2277 
2278 /* returns pointer to CPUArchId descriptor that matches CPU's apic_id
2279  * in ms->possible_cpus->cpus, if ms->possible_cpus->cpus has no
2280  * entry corresponding to CPU's apic_id returns NULL.
2281  */
2282 static CPUArchId *pc_find_cpu_slot(MachineState *ms, uint32_t id, int *idx)
2283 {
2284     CPUArchId apic_id, *found_cpu;
2285 
2286     apic_id.arch_id = id;
2287     found_cpu = bsearch(&apic_id, ms->possible_cpus->cpus,
2288         ms->possible_cpus->len, sizeof(*ms->possible_cpus->cpus),
2289         pc_apic_cmp);
2290     if (found_cpu && idx) {
2291         *idx = found_cpu - ms->possible_cpus->cpus;
2292     }
2293     return found_cpu;
2294 }
2295 
2296 static void pc_cpu_plug(HotplugHandler *hotplug_dev,
2297                         DeviceState *dev, Error **errp)
2298 {
2299     CPUArchId *found_cpu;
2300     Error *local_err = NULL;
2301     X86CPU *cpu = X86_CPU(dev);
2302     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2303 
2304     if (pcms->acpi_dev) {
2305         hotplug_handler_plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
2306         if (local_err) {
2307             goto out;
2308         }
2309     }
2310 
2311     /* increment the number of CPUs */
2312     pcms->boot_cpus++;
2313     if (pcms->rtc) {
2314         rtc_set_cpus_count(pcms->rtc, pcms->boot_cpus);
2315     }
2316     if (pcms->fw_cfg) {
2317         fw_cfg_modify_i16(pcms->fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
2318     }
2319 
2320     found_cpu = pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, NULL);
2321     found_cpu->cpu = OBJECT(dev);
2322 out:
2323     error_propagate(errp, local_err);
2324 }
2325 static void pc_cpu_unplug_request_cb(HotplugHandler *hotplug_dev,
2326                                      DeviceState *dev, Error **errp)
2327 {
2328     int idx = -1;
2329     Error *local_err = NULL;
2330     X86CPU *cpu = X86_CPU(dev);
2331     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2332 
2333     if (!pcms->acpi_dev) {
2334         error_setg(&local_err, "CPU hot unplug not supported without ACPI");
2335         goto out;
2336     }
2337 
2338     pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, &idx);
2339     assert(idx != -1);
2340     if (idx == 0) {
2341         error_setg(&local_err, "Boot CPU is unpluggable");
2342         goto out;
2343     }
2344 
2345     hotplug_handler_unplug_request(HOTPLUG_HANDLER(pcms->acpi_dev), dev,
2346                                    &local_err);
2347     if (local_err) {
2348         goto out;
2349     }
2350 
2351  out:
2352     error_propagate(errp, local_err);
2353 
2354 }
2355 
2356 static void pc_cpu_unplug_cb(HotplugHandler *hotplug_dev,
2357                              DeviceState *dev, Error **errp)
2358 {
2359     CPUArchId *found_cpu;
2360     Error *local_err = NULL;
2361     X86CPU *cpu = X86_CPU(dev);
2362     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2363 
2364     hotplug_handler_unplug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
2365     if (local_err) {
2366         goto out;
2367     }
2368 
2369     found_cpu = pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, NULL);
2370     found_cpu->cpu = NULL;
2371     object_property_set_bool(OBJECT(dev), false, "realized", NULL);
2372 
2373     /* decrement the number of CPUs */
2374     pcms->boot_cpus--;
2375     /* Update the number of CPUs in CMOS */
2376     rtc_set_cpus_count(pcms->rtc, pcms->boot_cpus);
2377     fw_cfg_modify_i16(pcms->fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
2378  out:
2379     error_propagate(errp, local_err);
2380 }
2381 
2382 static void pc_cpu_pre_plug(HotplugHandler *hotplug_dev,
2383                             DeviceState *dev, Error **errp)
2384 {
2385     int idx;
2386     CPUState *cs;
2387     CPUArchId *cpu_slot;
2388     X86CPUTopoInfo topo;
2389     X86CPU *cpu = X86_CPU(dev);
2390     CPUX86State *env = &cpu->env;
2391     MachineState *ms = MACHINE(hotplug_dev);
2392     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2393     unsigned int smp_cores = ms->smp.cores;
2394     unsigned int smp_threads = ms->smp.threads;
2395 
2396     if(!object_dynamic_cast(OBJECT(cpu), ms->cpu_type)) {
2397         error_setg(errp, "Invalid CPU type, expected cpu type: '%s'",
2398                    ms->cpu_type);
2399         return;
2400     }
2401 
2402     env->nr_dies = pcms->smp_dies;
2403 
2404     /*
2405      * If APIC ID is not set,
2406      * set it based on socket/die/core/thread properties.
2407      */
2408     if (cpu->apic_id == UNASSIGNED_APIC_ID) {
2409         int max_socket = (ms->smp.max_cpus - 1) /
2410                                 smp_threads / smp_cores / pcms->smp_dies;
2411 
2412         if (cpu->socket_id < 0) {
2413             error_setg(errp, "CPU socket-id is not set");
2414             return;
2415         } else if (cpu->socket_id > max_socket) {
2416             error_setg(errp, "Invalid CPU socket-id: %u must be in range 0:%u",
2417                        cpu->socket_id, max_socket);
2418             return;
2419         } else if (cpu->die_id > pcms->smp_dies - 1) {
2420             error_setg(errp, "Invalid CPU die-id: %u must be in range 0:%u",
2421                        cpu->die_id, max_socket);
2422             return;
2423         }
2424         if (cpu->core_id < 0) {
2425             error_setg(errp, "CPU core-id is not set");
2426             return;
2427         } else if (cpu->core_id > (smp_cores - 1)) {
2428             error_setg(errp, "Invalid CPU core-id: %u must be in range 0:%u",
2429                        cpu->core_id, smp_cores - 1);
2430             return;
2431         }
2432         if (cpu->thread_id < 0) {
2433             error_setg(errp, "CPU thread-id is not set");
2434             return;
2435         } else if (cpu->thread_id > (smp_threads - 1)) {
2436             error_setg(errp, "Invalid CPU thread-id: %u must be in range 0:%u",
2437                        cpu->thread_id, smp_threads - 1);
2438             return;
2439         }
2440 
2441         topo.pkg_id = cpu->socket_id;
2442         topo.die_id = cpu->die_id;
2443         topo.core_id = cpu->core_id;
2444         topo.smt_id = cpu->thread_id;
2445         cpu->apic_id = apicid_from_topo_ids(pcms->smp_dies, smp_cores,
2446                                             smp_threads, &topo);
2447     }
2448 
2449     cpu_slot = pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, &idx);
2450     if (!cpu_slot) {
2451         MachineState *ms = MACHINE(pcms);
2452 
2453         x86_topo_ids_from_apicid(cpu->apic_id, pcms->smp_dies,
2454                                  smp_cores, smp_threads, &topo);
2455         error_setg(errp,
2456             "Invalid CPU [socket: %u, die: %u, core: %u, thread: %u] with"
2457             " APIC ID %" PRIu32 ", valid index range 0:%d",
2458             topo.pkg_id, topo.die_id, topo.core_id, topo.smt_id,
2459             cpu->apic_id, ms->possible_cpus->len - 1);
2460         return;
2461     }
2462 
2463     if (cpu_slot->cpu) {
2464         error_setg(errp, "CPU[%d] with APIC ID %" PRIu32 " exists",
2465                    idx, cpu->apic_id);
2466         return;
2467     }
2468 
2469     /* if 'address' properties socket-id/core-id/thread-id are not set, set them
2470      * so that machine_query_hotpluggable_cpus would show correct values
2471      */
2472     /* TODO: move socket_id/core_id/thread_id checks into x86_cpu_realizefn()
2473      * once -smp refactoring is complete and there will be CPU private
2474      * CPUState::nr_cores and CPUState::nr_threads fields instead of globals */
2475     x86_topo_ids_from_apicid(cpu->apic_id, pcms->smp_dies,
2476                              smp_cores, smp_threads, &topo);
2477     if (cpu->socket_id != -1 && cpu->socket_id != topo.pkg_id) {
2478         error_setg(errp, "property socket-id: %u doesn't match set apic-id:"
2479             " 0x%x (socket-id: %u)", cpu->socket_id, cpu->apic_id, topo.pkg_id);
2480         return;
2481     }
2482     cpu->socket_id = topo.pkg_id;
2483 
2484     if (cpu->die_id != -1 && cpu->die_id != topo.die_id) {
2485         error_setg(errp, "property die-id: %u doesn't match set apic-id:"
2486             " 0x%x (die-id: %u)", cpu->die_id, cpu->apic_id, topo.die_id);
2487         return;
2488     }
2489     cpu->die_id = topo.die_id;
2490 
2491     if (cpu->core_id != -1 && cpu->core_id != topo.core_id) {
2492         error_setg(errp, "property core-id: %u doesn't match set apic-id:"
2493             " 0x%x (core-id: %u)", cpu->core_id, cpu->apic_id, topo.core_id);
2494         return;
2495     }
2496     cpu->core_id = topo.core_id;
2497 
2498     if (cpu->thread_id != -1 && cpu->thread_id != topo.smt_id) {
2499         error_setg(errp, "property thread-id: %u doesn't match set apic-id:"
2500             " 0x%x (thread-id: %u)", cpu->thread_id, cpu->apic_id, topo.smt_id);
2501         return;
2502     }
2503     cpu->thread_id = topo.smt_id;
2504 
2505     if (hyperv_feat_enabled(cpu, HYPERV_FEAT_VPINDEX) &&
2506         !kvm_hv_vpindex_settable()) {
2507         error_setg(errp, "kernel doesn't allow setting HyperV VP_INDEX");
2508         return;
2509     }
2510 
2511     cs = CPU(cpu);
2512     cs->cpu_index = idx;
2513 
2514     numa_cpu_pre_plug(cpu_slot, dev, errp);
2515 }
2516 
2517 static void pc_virtio_pmem_pci_pre_plug(HotplugHandler *hotplug_dev,
2518                                         DeviceState *dev, Error **errp)
2519 {
2520     HotplugHandler *hotplug_dev2 = qdev_get_bus_hotplug_handler(dev);
2521     Error *local_err = NULL;
2522 
2523     if (!hotplug_dev2) {
2524         /*
2525          * Without a bus hotplug handler, we cannot control the plug/unplug
2526          * order. This should never be the case on x86, however better add
2527          * a safety net.
2528          */
2529         error_setg(errp, "virtio-pmem-pci not supported on this bus.");
2530         return;
2531     }
2532     /*
2533      * First, see if we can plug this memory device at all. If that
2534      * succeeds, branch of to the actual hotplug handler.
2535      */
2536     memory_device_pre_plug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev), NULL,
2537                            &local_err);
2538     if (!local_err) {
2539         hotplug_handler_pre_plug(hotplug_dev2, dev, &local_err);
2540     }
2541     error_propagate(errp, local_err);
2542 }
2543 
2544 static void pc_virtio_pmem_pci_plug(HotplugHandler *hotplug_dev,
2545                                     DeviceState *dev, Error **errp)
2546 {
2547     HotplugHandler *hotplug_dev2 = qdev_get_bus_hotplug_handler(dev);
2548     Error *local_err = NULL;
2549 
2550     /*
2551      * Plug the memory device first and then branch off to the actual
2552      * hotplug handler. If that one fails, we can easily undo the memory
2553      * device bits.
2554      */
2555     memory_device_plug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev));
2556     hotplug_handler_plug(hotplug_dev2, dev, &local_err);
2557     if (local_err) {
2558         memory_device_unplug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev));
2559     }
2560     error_propagate(errp, local_err);
2561 }
2562 
2563 static void pc_virtio_pmem_pci_unplug_request(HotplugHandler *hotplug_dev,
2564                                               DeviceState *dev, Error **errp)
2565 {
2566     /* We don't support virtio pmem hot unplug */
2567     error_setg(errp, "virtio pmem device unplug not supported.");
2568 }
2569 
2570 static void pc_virtio_pmem_pci_unplug(HotplugHandler *hotplug_dev,
2571                                       DeviceState *dev, Error **errp)
2572 {
2573     /* We don't support virtio pmem hot unplug */
2574 }
2575 
2576 static void pc_machine_device_pre_plug_cb(HotplugHandler *hotplug_dev,
2577                                           DeviceState *dev, Error **errp)
2578 {
2579     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2580         pc_memory_pre_plug(hotplug_dev, dev, errp);
2581     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2582         pc_cpu_pre_plug(hotplug_dev, dev, errp);
2583     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI)) {
2584         pc_virtio_pmem_pci_pre_plug(hotplug_dev, dev, errp);
2585     }
2586 }
2587 
2588 static void pc_machine_device_plug_cb(HotplugHandler *hotplug_dev,
2589                                       DeviceState *dev, Error **errp)
2590 {
2591     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2592         pc_memory_plug(hotplug_dev, dev, errp);
2593     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2594         pc_cpu_plug(hotplug_dev, dev, errp);
2595     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI)) {
2596         pc_virtio_pmem_pci_plug(hotplug_dev, dev, errp);
2597     }
2598 }
2599 
2600 static void pc_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
2601                                                 DeviceState *dev, Error **errp)
2602 {
2603     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2604         pc_memory_unplug_request(hotplug_dev, dev, errp);
2605     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2606         pc_cpu_unplug_request_cb(hotplug_dev, dev, errp);
2607     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI)) {
2608         pc_virtio_pmem_pci_unplug_request(hotplug_dev, dev, errp);
2609     } else {
2610         error_setg(errp, "acpi: device unplug request for not supported device"
2611                    " type: %s", object_get_typename(OBJECT(dev)));
2612     }
2613 }
2614 
2615 static void pc_machine_device_unplug_cb(HotplugHandler *hotplug_dev,
2616                                         DeviceState *dev, Error **errp)
2617 {
2618     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2619         pc_memory_unplug(hotplug_dev, dev, errp);
2620     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2621         pc_cpu_unplug_cb(hotplug_dev, dev, errp);
2622     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI)) {
2623         pc_virtio_pmem_pci_unplug(hotplug_dev, dev, errp);
2624     } else {
2625         error_setg(errp, "acpi: device unplug for not supported device"
2626                    " type: %s", object_get_typename(OBJECT(dev)));
2627     }
2628 }
2629 
2630 static HotplugHandler *pc_get_hotplug_handler(MachineState *machine,
2631                                              DeviceState *dev)
2632 {
2633     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) ||
2634         object_dynamic_cast(OBJECT(dev), TYPE_CPU) ||
2635         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI)) {
2636         return HOTPLUG_HANDLER(machine);
2637     }
2638 
2639     return NULL;
2640 }
2641 
2642 static void
2643 pc_machine_get_device_memory_region_size(Object *obj, Visitor *v,
2644                                          const char *name, void *opaque,
2645                                          Error **errp)
2646 {
2647     MachineState *ms = MACHINE(obj);
2648     int64_t value = 0;
2649 
2650     if (ms->device_memory) {
2651         value = memory_region_size(&ms->device_memory->mr);
2652     }
2653 
2654     visit_type_int(v, name, &value, errp);
2655 }
2656 
2657 static void pc_machine_get_max_ram_below_4g(Object *obj, Visitor *v,
2658                                             const char *name, void *opaque,
2659                                             Error **errp)
2660 {
2661     PCMachineState *pcms = PC_MACHINE(obj);
2662     uint64_t value = pcms->max_ram_below_4g;
2663 
2664     visit_type_size(v, name, &value, errp);
2665 }
2666 
2667 static void pc_machine_set_max_ram_below_4g(Object *obj, Visitor *v,
2668                                             const char *name, void *opaque,
2669                                             Error **errp)
2670 {
2671     PCMachineState *pcms = PC_MACHINE(obj);
2672     Error *error = NULL;
2673     uint64_t value;
2674 
2675     visit_type_size(v, name, &value, &error);
2676     if (error) {
2677         error_propagate(errp, error);
2678         return;
2679     }
2680     if (value > 4 * GiB) {
2681         error_setg(&error,
2682                    "Machine option 'max-ram-below-4g=%"PRIu64
2683                    "' expects size less than or equal to 4G", value);
2684         error_propagate(errp, error);
2685         return;
2686     }
2687 
2688     if (value < 1 * MiB) {
2689         warn_report("Only %" PRIu64 " bytes of RAM below the 4GiB boundary,"
2690                     "BIOS may not work with less than 1MiB", value);
2691     }
2692 
2693     pcms->max_ram_below_4g = value;
2694 }
2695 
2696 static void pc_machine_get_vmport(Object *obj, Visitor *v, const char *name,
2697                                   void *opaque, Error **errp)
2698 {
2699     PCMachineState *pcms = PC_MACHINE(obj);
2700     OnOffAuto vmport = pcms->vmport;
2701 
2702     visit_type_OnOffAuto(v, name, &vmport, errp);
2703 }
2704 
2705 static void pc_machine_set_vmport(Object *obj, Visitor *v, const char *name,
2706                                   void *opaque, Error **errp)
2707 {
2708     PCMachineState *pcms = PC_MACHINE(obj);
2709 
2710     visit_type_OnOffAuto(v, name, &pcms->vmport, errp);
2711 }
2712 
2713 bool pc_machine_is_smm_enabled(PCMachineState *pcms)
2714 {
2715     bool smm_available = false;
2716 
2717     if (pcms->smm == ON_OFF_AUTO_OFF) {
2718         return false;
2719     }
2720 
2721     if (tcg_enabled() || qtest_enabled()) {
2722         smm_available = true;
2723     } else if (kvm_enabled()) {
2724         smm_available = kvm_has_smm();
2725     }
2726 
2727     if (smm_available) {
2728         return true;
2729     }
2730 
2731     if (pcms->smm == ON_OFF_AUTO_ON) {
2732         error_report("System Management Mode not supported by this hypervisor.");
2733         exit(1);
2734     }
2735     return false;
2736 }
2737 
2738 static void pc_machine_get_smm(Object *obj, Visitor *v, const char *name,
2739                                void *opaque, Error **errp)
2740 {
2741     PCMachineState *pcms = PC_MACHINE(obj);
2742     OnOffAuto smm = pcms->smm;
2743 
2744     visit_type_OnOffAuto(v, name, &smm, errp);
2745 }
2746 
2747 static void pc_machine_set_smm(Object *obj, Visitor *v, const char *name,
2748                                void *opaque, Error **errp)
2749 {
2750     PCMachineState *pcms = PC_MACHINE(obj);
2751 
2752     visit_type_OnOffAuto(v, name, &pcms->smm, errp);
2753 }
2754 
2755 static bool pc_machine_get_smbus(Object *obj, Error **errp)
2756 {
2757     PCMachineState *pcms = PC_MACHINE(obj);
2758 
2759     return pcms->smbus_enabled;
2760 }
2761 
2762 static void pc_machine_set_smbus(Object *obj, bool value, Error **errp)
2763 {
2764     PCMachineState *pcms = PC_MACHINE(obj);
2765 
2766     pcms->smbus_enabled = value;
2767 }
2768 
2769 static bool pc_machine_get_sata(Object *obj, Error **errp)
2770 {
2771     PCMachineState *pcms = PC_MACHINE(obj);
2772 
2773     return pcms->sata_enabled;
2774 }
2775 
2776 static void pc_machine_set_sata(Object *obj, bool value, Error **errp)
2777 {
2778     PCMachineState *pcms = PC_MACHINE(obj);
2779 
2780     pcms->sata_enabled = value;
2781 }
2782 
2783 static bool pc_machine_get_pit(Object *obj, Error **errp)
2784 {
2785     PCMachineState *pcms = PC_MACHINE(obj);
2786 
2787     return pcms->pit_enabled;
2788 }
2789 
2790 static void pc_machine_set_pit(Object *obj, bool value, Error **errp)
2791 {
2792     PCMachineState *pcms = PC_MACHINE(obj);
2793 
2794     pcms->pit_enabled = value;
2795 }
2796 
2797 static void pc_machine_initfn(Object *obj)
2798 {
2799     PCMachineState *pcms = PC_MACHINE(obj);
2800 
2801     pcms->max_ram_below_4g = 0; /* use default */
2802     pcms->smm = ON_OFF_AUTO_AUTO;
2803 #ifdef CONFIG_VMPORT
2804     pcms->vmport = ON_OFF_AUTO_AUTO;
2805 #else
2806     pcms->vmport = ON_OFF_AUTO_OFF;
2807 #endif /* CONFIG_VMPORT */
2808     /* acpi build is enabled by default if machine supports it */
2809     pcms->acpi_build_enabled = PC_MACHINE_GET_CLASS(pcms)->has_acpi_build;
2810     pcms->smbus_enabled = true;
2811     pcms->sata_enabled = true;
2812     pcms->pit_enabled = true;
2813     pcms->smp_dies = 1;
2814 
2815     pc_system_flash_create(pcms);
2816 }
2817 
2818 static void pc_machine_reset(MachineState *machine)
2819 {
2820     CPUState *cs;
2821     X86CPU *cpu;
2822 
2823     qemu_devices_reset();
2824 
2825     /* Reset APIC after devices have been reset to cancel
2826      * any changes that qemu_devices_reset() might have done.
2827      */
2828     CPU_FOREACH(cs) {
2829         cpu = X86_CPU(cs);
2830 
2831         if (cpu->apic_state) {
2832             device_reset(cpu->apic_state);
2833         }
2834     }
2835 }
2836 
2837 static CpuInstanceProperties
2838 pc_cpu_index_to_props(MachineState *ms, unsigned cpu_index)
2839 {
2840     MachineClass *mc = MACHINE_GET_CLASS(ms);
2841     const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(ms);
2842 
2843     assert(cpu_index < possible_cpus->len);
2844     return possible_cpus->cpus[cpu_index].props;
2845 }
2846 
2847 static int64_t pc_get_default_cpu_node_id(const MachineState *ms, int idx)
2848 {
2849    X86CPUTopoInfo topo;
2850    PCMachineState *pcms = PC_MACHINE(ms);
2851 
2852    assert(idx < ms->possible_cpus->len);
2853    x86_topo_ids_from_apicid(ms->possible_cpus->cpus[idx].arch_id,
2854                             pcms->smp_dies, ms->smp.cores,
2855                             ms->smp.threads, &topo);
2856    return topo.pkg_id % nb_numa_nodes;
2857 }
2858 
2859 static const CPUArchIdList *pc_possible_cpu_arch_ids(MachineState *ms)
2860 {
2861     PCMachineState *pcms = PC_MACHINE(ms);
2862     int i;
2863     unsigned int max_cpus = ms->smp.max_cpus;
2864 
2865     if (ms->possible_cpus) {
2866         /*
2867          * make sure that max_cpus hasn't changed since the first use, i.e.
2868          * -smp hasn't been parsed after it
2869         */
2870         assert(ms->possible_cpus->len == max_cpus);
2871         return ms->possible_cpus;
2872     }
2873 
2874     ms->possible_cpus = g_malloc0(sizeof(CPUArchIdList) +
2875                                   sizeof(CPUArchId) * max_cpus);
2876     ms->possible_cpus->len = max_cpus;
2877     for (i = 0; i < ms->possible_cpus->len; i++) {
2878         X86CPUTopoInfo topo;
2879 
2880         ms->possible_cpus->cpus[i].type = ms->cpu_type;
2881         ms->possible_cpus->cpus[i].vcpus_count = 1;
2882         ms->possible_cpus->cpus[i].arch_id = x86_cpu_apic_id_from_index(pcms, i);
2883         x86_topo_ids_from_apicid(ms->possible_cpus->cpus[i].arch_id,
2884                                  pcms->smp_dies, ms->smp.cores,
2885                                  ms->smp.threads, &topo);
2886         ms->possible_cpus->cpus[i].props.has_socket_id = true;
2887         ms->possible_cpus->cpus[i].props.socket_id = topo.pkg_id;
2888         ms->possible_cpus->cpus[i].props.has_die_id = true;
2889         ms->possible_cpus->cpus[i].props.die_id = topo.die_id;
2890         ms->possible_cpus->cpus[i].props.has_core_id = true;
2891         ms->possible_cpus->cpus[i].props.core_id = topo.core_id;
2892         ms->possible_cpus->cpus[i].props.has_thread_id = true;
2893         ms->possible_cpus->cpus[i].props.thread_id = topo.smt_id;
2894     }
2895     return ms->possible_cpus;
2896 }
2897 
2898 static void x86_nmi(NMIState *n, int cpu_index, Error **errp)
2899 {
2900     /* cpu index isn't used */
2901     CPUState *cs;
2902 
2903     CPU_FOREACH(cs) {
2904         X86CPU *cpu = X86_CPU(cs);
2905 
2906         if (!cpu->apic_state) {
2907             cpu_interrupt(cs, CPU_INTERRUPT_NMI);
2908         } else {
2909             apic_deliver_nmi(cpu->apic_state);
2910         }
2911     }
2912 }
2913 
2914 static void pc_machine_class_init(ObjectClass *oc, void *data)
2915 {
2916     MachineClass *mc = MACHINE_CLASS(oc);
2917     PCMachineClass *pcmc = PC_MACHINE_CLASS(oc);
2918     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
2919     NMIClass *nc = NMI_CLASS(oc);
2920 
2921     pcmc->pci_enabled = true;
2922     pcmc->has_acpi_build = true;
2923     pcmc->rsdp_in_ram = true;
2924     pcmc->smbios_defaults = true;
2925     pcmc->smbios_uuid_encoded = true;
2926     pcmc->gigabyte_align = true;
2927     pcmc->has_reserved_memory = true;
2928     pcmc->kvmclock_enabled = true;
2929     pcmc->enforce_aligned_dimm = true;
2930     /* BIOS ACPI tables: 128K. Other BIOS datastructures: less than 4K reported
2931      * to be used at the moment, 32K should be enough for a while.  */
2932     pcmc->acpi_data_size = 0x20000 + 0x8000;
2933     pcmc->save_tsc_khz = true;
2934     pcmc->linuxboot_dma_enabled = true;
2935     pcmc->pvh_enabled = true;
2936     assert(!mc->get_hotplug_handler);
2937     mc->get_hotplug_handler = pc_get_hotplug_handler;
2938     mc->cpu_index_to_instance_props = pc_cpu_index_to_props;
2939     mc->get_default_cpu_node_id = pc_get_default_cpu_node_id;
2940     mc->possible_cpu_arch_ids = pc_possible_cpu_arch_ids;
2941     mc->auto_enable_numa_with_memhp = true;
2942     mc->has_hotpluggable_cpus = true;
2943     mc->default_boot_order = "cad";
2944     mc->hot_add_cpu = pc_hot_add_cpu;
2945     mc->smp_parse = pc_smp_parse;
2946     mc->block_default_type = IF_IDE;
2947     mc->max_cpus = 255;
2948     mc->reset = pc_machine_reset;
2949     hc->pre_plug = pc_machine_device_pre_plug_cb;
2950     hc->plug = pc_machine_device_plug_cb;
2951     hc->unplug_request = pc_machine_device_unplug_request_cb;
2952     hc->unplug = pc_machine_device_unplug_cb;
2953     nc->nmi_monitor_handler = x86_nmi;
2954     mc->default_cpu_type = TARGET_DEFAULT_CPU_TYPE;
2955     mc->nvdimm_supported = true;
2956     mc->numa_mem_supported = true;
2957 
2958     object_class_property_add(oc, PC_MACHINE_DEVMEM_REGION_SIZE, "int",
2959         pc_machine_get_device_memory_region_size, NULL,
2960         NULL, NULL, &error_abort);
2961 
2962     object_class_property_add(oc, PC_MACHINE_MAX_RAM_BELOW_4G, "size",
2963         pc_machine_get_max_ram_below_4g, pc_machine_set_max_ram_below_4g,
2964         NULL, NULL, &error_abort);
2965 
2966     object_class_property_set_description(oc, PC_MACHINE_MAX_RAM_BELOW_4G,
2967         "Maximum ram below the 4G boundary (32bit boundary)", &error_abort);
2968 
2969     object_class_property_add(oc, PC_MACHINE_SMM, "OnOffAuto",
2970         pc_machine_get_smm, pc_machine_set_smm,
2971         NULL, NULL, &error_abort);
2972     object_class_property_set_description(oc, PC_MACHINE_SMM,
2973         "Enable SMM (pc & q35)", &error_abort);
2974 
2975     object_class_property_add(oc, PC_MACHINE_VMPORT, "OnOffAuto",
2976         pc_machine_get_vmport, pc_machine_set_vmport,
2977         NULL, NULL, &error_abort);
2978     object_class_property_set_description(oc, PC_MACHINE_VMPORT,
2979         "Enable vmport (pc & q35)", &error_abort);
2980 
2981     object_class_property_add_bool(oc, PC_MACHINE_SMBUS,
2982         pc_machine_get_smbus, pc_machine_set_smbus, &error_abort);
2983 
2984     object_class_property_add_bool(oc, PC_MACHINE_SATA,
2985         pc_machine_get_sata, pc_machine_set_sata, &error_abort);
2986 
2987     object_class_property_add_bool(oc, PC_MACHINE_PIT,
2988         pc_machine_get_pit, pc_machine_set_pit, &error_abort);
2989 }
2990 
2991 static const TypeInfo pc_machine_info = {
2992     .name = TYPE_PC_MACHINE,
2993     .parent = TYPE_MACHINE,
2994     .abstract = true,
2995     .instance_size = sizeof(PCMachineState),
2996     .instance_init = pc_machine_initfn,
2997     .class_size = sizeof(PCMachineClass),
2998     .class_init = pc_machine_class_init,
2999     .interfaces = (InterfaceInfo[]) {
3000          { TYPE_HOTPLUG_HANDLER },
3001          { TYPE_NMI },
3002          { }
3003     },
3004 };
3005 
3006 static void pc_machine_register_types(void)
3007 {
3008     type_register_static(&pc_machine_info);
3009 }
3010 
3011 type_init(pc_machine_register_types)
3012