xref: /openbmc/qemu/hw/i386/pc.c (revision 99d423e5)
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/hw.h"
28 #include "hw/i386/pc.h"
29 #include "hw/char/serial.h"
30 #include "hw/char/parallel.h"
31 #include "hw/i386/apic.h"
32 #include "hw/i386/topology.h"
33 #include "hw/i386/fw_cfg.h"
34 #include "sysemu/cpus.h"
35 #include "hw/block/fdc.h"
36 #include "hw/ide.h"
37 #include "hw/pci/pci.h"
38 #include "hw/pci/pci_bus.h"
39 #include "hw/nvram/fw_cfg.h"
40 #include "hw/timer/hpet.h"
41 #include "hw/firmware/smbios.h"
42 #include "hw/loader.h"
43 #include "elf.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/audio/pcspk.h"
50 #include "hw/pci/msi.h"
51 #include "hw/sysbus.h"
52 #include "sysemu/sysemu.h"
53 #include "sysemu/numa.h"
54 #include "sysemu/kvm.h"
55 #include "sysemu/qtest.h"
56 #include "kvm_i386.h"
57 #include "hw/xen/xen.h"
58 #include "hw/xen/start_info.h"
59 #include "ui/qemu-spice.h"
60 #include "exec/memory.h"
61 #include "exec/address-spaces.h"
62 #include "sysemu/arch_init.h"
63 #include "qemu/bitmap.h"
64 #include "qemu/config-file.h"
65 #include "qemu/error-report.h"
66 #include "qemu/option.h"
67 #include "hw/acpi/acpi.h"
68 #include "hw/acpi/cpu_hotplug.h"
69 #include "hw/boards.h"
70 #include "acpi-build.h"
71 #include "hw/mem/pc-dimm.h"
72 #include "qapi/error.h"
73 #include "qapi/qapi-visit-common.h"
74 #include "qapi/visitor.h"
75 #include "qom/cpu.h"
76 #include "hw/nmi.h"
77 #include "hw/usb.h"
78 #include "hw/i386/intel_iommu.h"
79 #include "hw/net/ne2000-isa.h"
80 #include "standard-headers/asm-x86/bootparam.h"
81 
82 /* debug PC/ISA interrupts */
83 //#define DEBUG_IRQ
84 
85 #ifdef DEBUG_IRQ
86 #define DPRINTF(fmt, ...)                                       \
87     do { printf("CPUIRQ: " fmt , ## __VA_ARGS__); } while (0)
88 #else
89 #define DPRINTF(fmt, ...)
90 #endif
91 
92 #define E820_NR_ENTRIES		16
93 
94 struct e820_entry {
95     uint64_t address;
96     uint64_t length;
97     uint32_t type;
98 } QEMU_PACKED __attribute((__aligned__(4)));
99 
100 struct e820_table {
101     uint32_t count;
102     struct e820_entry entry[E820_NR_ENTRIES];
103 } QEMU_PACKED __attribute((__aligned__(4)));
104 
105 static struct e820_table e820_reserve;
106 static struct e820_entry *e820_table;
107 static unsigned e820_entries;
108 struct hpet_fw_config hpet_cfg = {.count = UINT8_MAX};
109 
110 /* Physical Address of PVH entry point read from kernel ELF NOTE */
111 static size_t pvh_start_addr;
112 
113 GlobalProperty pc_compat_4_0_1[] = {};
114 const size_t pc_compat_4_0_1_len = G_N_ELEMENTS(pc_compat_4_0_1);
115 
116 GlobalProperty pc_compat_4_0[] = {};
117 const size_t pc_compat_4_0_len = G_N_ELEMENTS(pc_compat_4_0);
118 
119 GlobalProperty pc_compat_3_1[] = {
120     { "intel-iommu", "dma-drain", "off" },
121     { "Opteron_G3" "-" TYPE_X86_CPU, "rdtscp", "off" },
122     { "Opteron_G4" "-" TYPE_X86_CPU, "rdtscp", "off" },
123     { "Opteron_G4" "-" TYPE_X86_CPU, "npt", "off" },
124     { "Opteron_G4" "-" TYPE_X86_CPU, "nrip-save", "off" },
125     { "Opteron_G5" "-" TYPE_X86_CPU, "rdtscp", "off" },
126     { "Opteron_G5" "-" TYPE_X86_CPU, "npt", "off" },
127     { "Opteron_G5" "-" TYPE_X86_CPU, "nrip-save", "off" },
128     { "EPYC" "-" TYPE_X86_CPU, "npt", "off" },
129     { "EPYC" "-" TYPE_X86_CPU, "nrip-save", "off" },
130     { "EPYC-IBPB" "-" TYPE_X86_CPU, "npt", "off" },
131     { "EPYC-IBPB" "-" TYPE_X86_CPU, "nrip-save", "off" },
132     { "Skylake-Client" "-" TYPE_X86_CPU,      "mpx", "on" },
133     { "Skylake-Client-IBRS" "-" TYPE_X86_CPU, "mpx", "on" },
134     { "Skylake-Server" "-" TYPE_X86_CPU,      "mpx", "on" },
135     { "Skylake-Server-IBRS" "-" TYPE_X86_CPU, "mpx", "on" },
136     { "Cascadelake-Server" "-" TYPE_X86_CPU,  "mpx", "on" },
137     { "Icelake-Client" "-" TYPE_X86_CPU,      "mpx", "on" },
138     { "Icelake-Server" "-" TYPE_X86_CPU,      "mpx", "on" },
139     { "Cascadelake-Server" "-" TYPE_X86_CPU, "stepping", "5" },
140     { TYPE_X86_CPU, "x-intel-pt-auto-level", "off" },
141 };
142 const size_t pc_compat_3_1_len = G_N_ELEMENTS(pc_compat_3_1);
143 
144 GlobalProperty pc_compat_3_0[] = {
145     { TYPE_X86_CPU, "x-hv-synic-kvm-only", "on" },
146     { "Skylake-Server" "-" TYPE_X86_CPU, "pku", "off" },
147     { "Skylake-Server-IBRS" "-" TYPE_X86_CPU, "pku", "off" },
148 };
149 const size_t pc_compat_3_0_len = G_N_ELEMENTS(pc_compat_3_0);
150 
151 GlobalProperty pc_compat_2_12[] = {
152     { TYPE_X86_CPU, "legacy-cache", "on" },
153     { TYPE_X86_CPU, "topoext", "off" },
154     { "EPYC-" TYPE_X86_CPU, "xlevel", "0x8000000a" },
155     { "EPYC-IBPB-" TYPE_X86_CPU, "xlevel", "0x8000000a" },
156 };
157 const size_t pc_compat_2_12_len = G_N_ELEMENTS(pc_compat_2_12);
158 
159 GlobalProperty pc_compat_2_11[] = {
160     { TYPE_X86_CPU, "x-migrate-smi-count", "off" },
161     { "Skylake-Server" "-" TYPE_X86_CPU, "clflushopt", "off" },
162 };
163 const size_t pc_compat_2_11_len = G_N_ELEMENTS(pc_compat_2_11);
164 
165 GlobalProperty pc_compat_2_10[] = {
166     { TYPE_X86_CPU, "x-hv-max-vps", "0x40" },
167     { "i440FX-pcihost", "x-pci-hole64-fix", "off" },
168     { "q35-pcihost", "x-pci-hole64-fix", "off" },
169 };
170 const size_t pc_compat_2_10_len = G_N_ELEMENTS(pc_compat_2_10);
171 
172 GlobalProperty pc_compat_2_9[] = {
173     { "mch", "extended-tseg-mbytes", "0" },
174 };
175 const size_t pc_compat_2_9_len = G_N_ELEMENTS(pc_compat_2_9);
176 
177 GlobalProperty pc_compat_2_8[] = {
178     { TYPE_X86_CPU, "tcg-cpuid", "off" },
179     { "kvmclock", "x-mach-use-reliable-get-clock", "off" },
180     { "ICH9-LPC", "x-smi-broadcast", "off" },
181     { TYPE_X86_CPU, "vmware-cpuid-freq", "off" },
182     { "Haswell-" TYPE_X86_CPU, "stepping", "1" },
183 };
184 const size_t pc_compat_2_8_len = G_N_ELEMENTS(pc_compat_2_8);
185 
186 GlobalProperty pc_compat_2_7[] = {
187     { TYPE_X86_CPU, "l3-cache", "off" },
188     { TYPE_X86_CPU, "full-cpuid-auto-level", "off" },
189     { "Opteron_G3" "-" TYPE_X86_CPU, "family", "15" },
190     { "Opteron_G3" "-" TYPE_X86_CPU, "model", "6" },
191     { "Opteron_G3" "-" TYPE_X86_CPU, "stepping", "1" },
192     { "isa-pcspk", "migrate", "off" },
193 };
194 const size_t pc_compat_2_7_len = G_N_ELEMENTS(pc_compat_2_7);
195 
196 GlobalProperty pc_compat_2_6[] = {
197     { TYPE_X86_CPU, "cpuid-0xb", "off" },
198     { "vmxnet3", "romfile", "" },
199     { TYPE_X86_CPU, "fill-mtrr-mask", "off" },
200     { "apic-common", "legacy-instance-id", "on", }
201 };
202 const size_t pc_compat_2_6_len = G_N_ELEMENTS(pc_compat_2_6);
203 
204 GlobalProperty pc_compat_2_5[] = {};
205 const size_t pc_compat_2_5_len = G_N_ELEMENTS(pc_compat_2_5);
206 
207 GlobalProperty pc_compat_2_4[] = {
208     PC_CPU_MODEL_IDS("2.4.0")
209     { "Haswell-" TYPE_X86_CPU, "abm", "off" },
210     { "Haswell-noTSX-" TYPE_X86_CPU, "abm", "off" },
211     { "Broadwell-" TYPE_X86_CPU, "abm", "off" },
212     { "Broadwell-noTSX-" TYPE_X86_CPU, "abm", "off" },
213     { "host" "-" TYPE_X86_CPU, "host-cache-info", "on" },
214     { TYPE_X86_CPU, "check", "off" },
215     { "qemu64" "-" TYPE_X86_CPU, "sse4a", "on" },
216     { "qemu64" "-" TYPE_X86_CPU, "abm", "on" },
217     { "qemu64" "-" TYPE_X86_CPU, "popcnt", "on" },
218     { "qemu32" "-" TYPE_X86_CPU, "popcnt", "on" },
219     { "Opteron_G2" "-" TYPE_X86_CPU, "rdtscp", "on" },
220     { "Opteron_G3" "-" TYPE_X86_CPU, "rdtscp", "on" },
221     { "Opteron_G4" "-" TYPE_X86_CPU, "rdtscp", "on" },
222     { "Opteron_G5" "-" TYPE_X86_CPU, "rdtscp", "on", }
223 };
224 const size_t pc_compat_2_4_len = G_N_ELEMENTS(pc_compat_2_4);
225 
226 GlobalProperty pc_compat_2_3[] = {
227     PC_CPU_MODEL_IDS("2.3.0")
228     { TYPE_X86_CPU, "arat", "off" },
229     { "qemu64" "-" TYPE_X86_CPU, "min-level", "4" },
230     { "kvm64" "-" TYPE_X86_CPU, "min-level", "5" },
231     { "pentium3" "-" TYPE_X86_CPU, "min-level", "2" },
232     { "n270" "-" TYPE_X86_CPU, "min-level", "5" },
233     { "Conroe" "-" TYPE_X86_CPU, "min-level", "4" },
234     { "Penryn" "-" TYPE_X86_CPU, "min-level", "4" },
235     { "Nehalem" "-" TYPE_X86_CPU, "min-level", "4" },
236     { "n270" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
237     { "Penryn" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
238     { "Conroe" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
239     { "Nehalem" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
240     { "Westmere" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
241     { "SandyBridge" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
242     { "IvyBridge" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
243     { "Haswell" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
244     { "Haswell-noTSX" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
245     { "Broadwell" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
246     { "Broadwell-noTSX" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
247     { TYPE_X86_CPU, "kvm-no-smi-migration", "on" },
248 };
249 const size_t pc_compat_2_3_len = G_N_ELEMENTS(pc_compat_2_3);
250 
251 GlobalProperty pc_compat_2_2[] = {
252     PC_CPU_MODEL_IDS("2.2.0")
253     { "kvm64" "-" TYPE_X86_CPU, "vme", "off" },
254     { "kvm32" "-" TYPE_X86_CPU, "vme", "off" },
255     { "Conroe" "-" TYPE_X86_CPU, "vme", "off" },
256     { "Penryn" "-" TYPE_X86_CPU, "vme", "off" },
257     { "Nehalem" "-" TYPE_X86_CPU, "vme", "off" },
258     { "Westmere" "-" TYPE_X86_CPU, "vme", "off" },
259     { "SandyBridge" "-" TYPE_X86_CPU, "vme", "off" },
260     { "Haswell" "-" TYPE_X86_CPU, "vme", "off" },
261     { "Broadwell" "-" TYPE_X86_CPU, "vme", "off" },
262     { "Opteron_G1" "-" TYPE_X86_CPU, "vme", "off" },
263     { "Opteron_G2" "-" TYPE_X86_CPU, "vme", "off" },
264     { "Opteron_G3" "-" TYPE_X86_CPU, "vme", "off" },
265     { "Opteron_G4" "-" TYPE_X86_CPU, "vme", "off" },
266     { "Opteron_G5" "-" TYPE_X86_CPU, "vme", "off" },
267     { "Haswell" "-" TYPE_X86_CPU, "f16c", "off" },
268     { "Haswell" "-" TYPE_X86_CPU, "rdrand", "off" },
269     { "Broadwell" "-" TYPE_X86_CPU, "f16c", "off" },
270     { "Broadwell" "-" TYPE_X86_CPU, "rdrand", "off" },
271 };
272 const size_t pc_compat_2_2_len = G_N_ELEMENTS(pc_compat_2_2);
273 
274 GlobalProperty pc_compat_2_1[] = {
275     PC_CPU_MODEL_IDS("2.1.0")
276     { "coreduo" "-" TYPE_X86_CPU, "vmx", "on" },
277     { "core2duo" "-" TYPE_X86_CPU, "vmx", "on" },
278 };
279 const size_t pc_compat_2_1_len = G_N_ELEMENTS(pc_compat_2_1);
280 
281 GlobalProperty pc_compat_2_0[] = {
282     PC_CPU_MODEL_IDS("2.0.0")
283     { "virtio-scsi-pci", "any_layout", "off" },
284     { "PIIX4_PM", "memory-hotplug-support", "off" },
285     { "apic", "version", "0x11" },
286     { "nec-usb-xhci", "superspeed-ports-first", "off" },
287     { "nec-usb-xhci", "force-pcie-endcap", "on" },
288     { "pci-serial", "prog_if", "0" },
289     { "pci-serial-2x", "prog_if", "0" },
290     { "pci-serial-4x", "prog_if", "0" },
291     { "virtio-net-pci", "guest_announce", "off" },
292     { "ICH9-LPC", "memory-hotplug-support", "off" },
293     { "xio3130-downstream", COMPAT_PROP_PCP, "off" },
294     { "ioh3420", COMPAT_PROP_PCP, "off" },
295 };
296 const size_t pc_compat_2_0_len = G_N_ELEMENTS(pc_compat_2_0);
297 
298 GlobalProperty pc_compat_1_7[] = {
299     PC_CPU_MODEL_IDS("1.7.0")
300     { TYPE_USB_DEVICE, "msos-desc", "no" },
301     { "PIIX4_PM", "acpi-pci-hotplug-with-bridge-support", "off" },
302     { "hpet", HPET_INTCAP, "4" },
303 };
304 const size_t pc_compat_1_7_len = G_N_ELEMENTS(pc_compat_1_7);
305 
306 GlobalProperty pc_compat_1_6[] = {
307     PC_CPU_MODEL_IDS("1.6.0")
308     { "e1000", "mitigation", "off" },
309     { "qemu64-" TYPE_X86_CPU, "model", "2" },
310     { "qemu32-" TYPE_X86_CPU, "model", "3" },
311     { "i440FX-pcihost", "short_root_bus", "1" },
312     { "q35-pcihost", "short_root_bus", "1" },
313 };
314 const size_t pc_compat_1_6_len = G_N_ELEMENTS(pc_compat_1_6);
315 
316 GlobalProperty pc_compat_1_5[] = {
317     PC_CPU_MODEL_IDS("1.5.0")
318     { "Conroe-" TYPE_X86_CPU, "model", "2" },
319     { "Conroe-" TYPE_X86_CPU, "min-level", "2" },
320     { "Penryn-" TYPE_X86_CPU, "model", "2" },
321     { "Penryn-" TYPE_X86_CPU, "min-level", "2" },
322     { "Nehalem-" TYPE_X86_CPU, "model", "2" },
323     { "Nehalem-" TYPE_X86_CPU, "min-level", "2" },
324     { "virtio-net-pci", "any_layout", "off" },
325     { TYPE_X86_CPU, "pmu", "on" },
326     { "i440FX-pcihost", "short_root_bus", "0" },
327     { "q35-pcihost", "short_root_bus", "0" },
328 };
329 const size_t pc_compat_1_5_len = G_N_ELEMENTS(pc_compat_1_5);
330 
331 GlobalProperty pc_compat_1_4[] = {
332     PC_CPU_MODEL_IDS("1.4.0")
333     { "scsi-hd", "discard_granularity", "0" },
334     { "scsi-cd", "discard_granularity", "0" },
335     { "scsi-disk", "discard_granularity", "0" },
336     { "ide-hd", "discard_granularity", "0" },
337     { "ide-cd", "discard_granularity", "0" },
338     { "ide-drive", "discard_granularity", "0" },
339     { "virtio-blk-pci", "discard_granularity", "0" },
340     /* DEV_NVECTORS_UNSPECIFIED as a uint32_t string: */
341     { "virtio-serial-pci", "vectors", "0xFFFFFFFF" },
342     { "virtio-net-pci", "ctrl_guest_offloads", "off" },
343     { "e1000", "romfile", "pxe-e1000.rom" },
344     { "ne2k_pci", "romfile", "pxe-ne2k_pci.rom" },
345     { "pcnet", "romfile", "pxe-pcnet.rom" },
346     { "rtl8139", "romfile", "pxe-rtl8139.rom" },
347     { "virtio-net-pci", "romfile", "pxe-virtio.rom" },
348     { "486-" TYPE_X86_CPU, "model", "0" },
349     { "n270" "-" TYPE_X86_CPU, "movbe", "off" },
350     { "Westmere" "-" TYPE_X86_CPU, "pclmulqdq", "off" },
351 };
352 const size_t pc_compat_1_4_len = G_N_ELEMENTS(pc_compat_1_4);
353 
354 void gsi_handler(void *opaque, int n, int level)
355 {
356     GSIState *s = opaque;
357 
358     DPRINTF("pc: %s GSI %d\n", level ? "raising" : "lowering", n);
359     if (n < ISA_NUM_IRQS) {
360         qemu_set_irq(s->i8259_irq[n], level);
361     }
362     qemu_set_irq(s->ioapic_irq[n], level);
363 }
364 
365 static void ioport80_write(void *opaque, hwaddr addr, uint64_t data,
366                            unsigned size)
367 {
368 }
369 
370 static uint64_t ioport80_read(void *opaque, hwaddr addr, unsigned size)
371 {
372     return 0xffffffffffffffffULL;
373 }
374 
375 /* MSDOS compatibility mode FPU exception support */
376 static qemu_irq ferr_irq;
377 
378 void pc_register_ferr_irq(qemu_irq irq)
379 {
380     ferr_irq = irq;
381 }
382 
383 /* XXX: add IGNNE support */
384 void cpu_set_ferr(CPUX86State *s)
385 {
386     qemu_irq_raise(ferr_irq);
387 }
388 
389 static void ioportF0_write(void *opaque, hwaddr addr, uint64_t data,
390                            unsigned size)
391 {
392     qemu_irq_lower(ferr_irq);
393 }
394 
395 static uint64_t ioportF0_read(void *opaque, hwaddr addr, unsigned size)
396 {
397     return 0xffffffffffffffffULL;
398 }
399 
400 /* TSC handling */
401 uint64_t cpu_get_tsc(CPUX86State *env)
402 {
403     return cpu_get_ticks();
404 }
405 
406 /* IRQ handling */
407 int cpu_get_pic_interrupt(CPUX86State *env)
408 {
409     X86CPU *cpu = x86_env_get_cpu(env);
410     int intno;
411 
412     if (!kvm_irqchip_in_kernel()) {
413         intno = apic_get_interrupt(cpu->apic_state);
414         if (intno >= 0) {
415             return intno;
416         }
417         /* read the irq from the PIC */
418         if (!apic_accept_pic_intr(cpu->apic_state)) {
419             return -1;
420         }
421     }
422 
423     intno = pic_read_irq(isa_pic);
424     return intno;
425 }
426 
427 static void pic_irq_request(void *opaque, int irq, int level)
428 {
429     CPUState *cs = first_cpu;
430     X86CPU *cpu = X86_CPU(cs);
431 
432     DPRINTF("pic_irqs: %s irq %d\n", level? "raise" : "lower", irq);
433     if (cpu->apic_state && !kvm_irqchip_in_kernel()) {
434         CPU_FOREACH(cs) {
435             cpu = X86_CPU(cs);
436             if (apic_accept_pic_intr(cpu->apic_state)) {
437                 apic_deliver_pic_intr(cpu->apic_state, level);
438             }
439         }
440     } else {
441         if (level) {
442             cpu_interrupt(cs, CPU_INTERRUPT_HARD);
443         } else {
444             cpu_reset_interrupt(cs, CPU_INTERRUPT_HARD);
445         }
446     }
447 }
448 
449 /* PC cmos mappings */
450 
451 #define REG_EQUIPMENT_BYTE          0x14
452 
453 int cmos_get_fd_drive_type(FloppyDriveType fd0)
454 {
455     int val;
456 
457     switch (fd0) {
458     case FLOPPY_DRIVE_TYPE_144:
459         /* 1.44 Mb 3"5 drive */
460         val = 4;
461         break;
462     case FLOPPY_DRIVE_TYPE_288:
463         /* 2.88 Mb 3"5 drive */
464         val = 5;
465         break;
466     case FLOPPY_DRIVE_TYPE_120:
467         /* 1.2 Mb 5"5 drive */
468         val = 2;
469         break;
470     case FLOPPY_DRIVE_TYPE_NONE:
471     default:
472         val = 0;
473         break;
474     }
475     return val;
476 }
477 
478 static void cmos_init_hd(ISADevice *s, int type_ofs, int info_ofs,
479                          int16_t cylinders, int8_t heads, int8_t sectors)
480 {
481     rtc_set_memory(s, type_ofs, 47);
482     rtc_set_memory(s, info_ofs, cylinders);
483     rtc_set_memory(s, info_ofs + 1, cylinders >> 8);
484     rtc_set_memory(s, info_ofs + 2, heads);
485     rtc_set_memory(s, info_ofs + 3, 0xff);
486     rtc_set_memory(s, info_ofs + 4, 0xff);
487     rtc_set_memory(s, info_ofs + 5, 0xc0 | ((heads > 8) << 3));
488     rtc_set_memory(s, info_ofs + 6, cylinders);
489     rtc_set_memory(s, info_ofs + 7, cylinders >> 8);
490     rtc_set_memory(s, info_ofs + 8, sectors);
491 }
492 
493 /* convert boot_device letter to something recognizable by the bios */
494 static int boot_device2nibble(char boot_device)
495 {
496     switch(boot_device) {
497     case 'a':
498     case 'b':
499         return 0x01; /* floppy boot */
500     case 'c':
501         return 0x02; /* hard drive boot */
502     case 'd':
503         return 0x03; /* CD-ROM boot */
504     case 'n':
505         return 0x04; /* Network boot */
506     }
507     return 0;
508 }
509 
510 static void set_boot_dev(ISADevice *s, const char *boot_device, Error **errp)
511 {
512 #define PC_MAX_BOOT_DEVICES 3
513     int nbds, bds[3] = { 0, };
514     int i;
515 
516     nbds = strlen(boot_device);
517     if (nbds > PC_MAX_BOOT_DEVICES) {
518         error_setg(errp, "Too many boot devices for PC");
519         return;
520     }
521     for (i = 0; i < nbds; i++) {
522         bds[i] = boot_device2nibble(boot_device[i]);
523         if (bds[i] == 0) {
524             error_setg(errp, "Invalid boot device for PC: '%c'",
525                        boot_device[i]);
526             return;
527         }
528     }
529     rtc_set_memory(s, 0x3d, (bds[1] << 4) | bds[0]);
530     rtc_set_memory(s, 0x38, (bds[2] << 4) | (fd_bootchk ? 0x0 : 0x1));
531 }
532 
533 static void pc_boot_set(void *opaque, const char *boot_device, Error **errp)
534 {
535     set_boot_dev(opaque, boot_device, errp);
536 }
537 
538 static void pc_cmos_init_floppy(ISADevice *rtc_state, ISADevice *floppy)
539 {
540     int val, nb, i;
541     FloppyDriveType fd_type[2] = { FLOPPY_DRIVE_TYPE_NONE,
542                                    FLOPPY_DRIVE_TYPE_NONE };
543 
544     /* floppy type */
545     if (floppy) {
546         for (i = 0; i < 2; i++) {
547             fd_type[i] = isa_fdc_get_drive_type(floppy, i);
548         }
549     }
550     val = (cmos_get_fd_drive_type(fd_type[0]) << 4) |
551         cmos_get_fd_drive_type(fd_type[1]);
552     rtc_set_memory(rtc_state, 0x10, val);
553 
554     val = rtc_get_memory(rtc_state, REG_EQUIPMENT_BYTE);
555     nb = 0;
556     if (fd_type[0] != FLOPPY_DRIVE_TYPE_NONE) {
557         nb++;
558     }
559     if (fd_type[1] != FLOPPY_DRIVE_TYPE_NONE) {
560         nb++;
561     }
562     switch (nb) {
563     case 0:
564         break;
565     case 1:
566         val |= 0x01; /* 1 drive, ready for boot */
567         break;
568     case 2:
569         val |= 0x41; /* 2 drives, ready for boot */
570         break;
571     }
572     rtc_set_memory(rtc_state, REG_EQUIPMENT_BYTE, val);
573 }
574 
575 typedef struct pc_cmos_init_late_arg {
576     ISADevice *rtc_state;
577     BusState *idebus[2];
578 } pc_cmos_init_late_arg;
579 
580 typedef struct check_fdc_state {
581     ISADevice *floppy;
582     bool multiple;
583 } CheckFdcState;
584 
585 static int check_fdc(Object *obj, void *opaque)
586 {
587     CheckFdcState *state = opaque;
588     Object *fdc;
589     uint32_t iobase;
590     Error *local_err = NULL;
591 
592     fdc = object_dynamic_cast(obj, TYPE_ISA_FDC);
593     if (!fdc) {
594         return 0;
595     }
596 
597     iobase = object_property_get_uint(obj, "iobase", &local_err);
598     if (local_err || iobase != 0x3f0) {
599         error_free(local_err);
600         return 0;
601     }
602 
603     if (state->floppy) {
604         state->multiple = true;
605     } else {
606         state->floppy = ISA_DEVICE(obj);
607     }
608     return 0;
609 }
610 
611 static const char * const fdc_container_path[] = {
612     "/unattached", "/peripheral", "/peripheral-anon"
613 };
614 
615 /*
616  * Locate the FDC at IO address 0x3f0, in order to configure the CMOS registers
617  * and ACPI objects.
618  */
619 ISADevice *pc_find_fdc0(void)
620 {
621     int i;
622     Object *container;
623     CheckFdcState state = { 0 };
624 
625     for (i = 0; i < ARRAY_SIZE(fdc_container_path); i++) {
626         container = container_get(qdev_get_machine(), fdc_container_path[i]);
627         object_child_foreach(container, check_fdc, &state);
628     }
629 
630     if (state.multiple) {
631         warn_report("multiple floppy disk controllers with "
632                     "iobase=0x3f0 have been found");
633         error_printf("the one being picked for CMOS setup might not reflect "
634                      "your intent");
635     }
636 
637     return state.floppy;
638 }
639 
640 static void pc_cmos_init_late(void *opaque)
641 {
642     pc_cmos_init_late_arg *arg = opaque;
643     ISADevice *s = arg->rtc_state;
644     int16_t cylinders;
645     int8_t heads, sectors;
646     int val;
647     int i, trans;
648 
649     val = 0;
650     if (arg->idebus[0] && ide_get_geometry(arg->idebus[0], 0,
651                                            &cylinders, &heads, &sectors) >= 0) {
652         cmos_init_hd(s, 0x19, 0x1b, cylinders, heads, sectors);
653         val |= 0xf0;
654     }
655     if (arg->idebus[0] && ide_get_geometry(arg->idebus[0], 1,
656                                            &cylinders, &heads, &sectors) >= 0) {
657         cmos_init_hd(s, 0x1a, 0x24, cylinders, heads, sectors);
658         val |= 0x0f;
659     }
660     rtc_set_memory(s, 0x12, val);
661 
662     val = 0;
663     for (i = 0; i < 4; i++) {
664         /* NOTE: ide_get_geometry() returns the physical
665            geometry.  It is always such that: 1 <= sects <= 63, 1
666            <= heads <= 16, 1 <= cylinders <= 16383. The BIOS
667            geometry can be different if a translation is done. */
668         if (arg->idebus[i / 2] &&
669             ide_get_geometry(arg->idebus[i / 2], i % 2,
670                              &cylinders, &heads, &sectors) >= 0) {
671             trans = ide_get_bios_chs_trans(arg->idebus[i / 2], i % 2) - 1;
672             assert((trans & ~3) == 0);
673             val |= trans << (i * 2);
674         }
675     }
676     rtc_set_memory(s, 0x39, val);
677 
678     pc_cmos_init_floppy(s, pc_find_fdc0());
679 
680     qemu_unregister_reset(pc_cmos_init_late, opaque);
681 }
682 
683 void pc_cmos_init(PCMachineState *pcms,
684                   BusState *idebus0, BusState *idebus1,
685                   ISADevice *s)
686 {
687     int val;
688     static pc_cmos_init_late_arg arg;
689 
690     /* various important CMOS locations needed by PC/Bochs bios */
691 
692     /* memory size */
693     /* base memory (first MiB) */
694     val = MIN(pcms->below_4g_mem_size / KiB, 640);
695     rtc_set_memory(s, 0x15, val);
696     rtc_set_memory(s, 0x16, val >> 8);
697     /* extended memory (next 64MiB) */
698     if (pcms->below_4g_mem_size > 1 * MiB) {
699         val = (pcms->below_4g_mem_size - 1 * MiB) / KiB;
700     } else {
701         val = 0;
702     }
703     if (val > 65535)
704         val = 65535;
705     rtc_set_memory(s, 0x17, val);
706     rtc_set_memory(s, 0x18, val >> 8);
707     rtc_set_memory(s, 0x30, val);
708     rtc_set_memory(s, 0x31, val >> 8);
709     /* memory between 16MiB and 4GiB */
710     if (pcms->below_4g_mem_size > 16 * MiB) {
711         val = (pcms->below_4g_mem_size - 16 * MiB) / (64 * KiB);
712     } else {
713         val = 0;
714     }
715     if (val > 65535)
716         val = 65535;
717     rtc_set_memory(s, 0x34, val);
718     rtc_set_memory(s, 0x35, val >> 8);
719     /* memory above 4GiB */
720     val = pcms->above_4g_mem_size / 65536;
721     rtc_set_memory(s, 0x5b, val);
722     rtc_set_memory(s, 0x5c, val >> 8);
723     rtc_set_memory(s, 0x5d, val >> 16);
724 
725     object_property_add_link(OBJECT(pcms), "rtc_state",
726                              TYPE_ISA_DEVICE,
727                              (Object **)&pcms->rtc,
728                              object_property_allow_set_link,
729                              OBJ_PROP_LINK_STRONG, &error_abort);
730     object_property_set_link(OBJECT(pcms), OBJECT(s),
731                              "rtc_state", &error_abort);
732 
733     set_boot_dev(s, MACHINE(pcms)->boot_order, &error_fatal);
734 
735     val = 0;
736     val |= 0x02; /* FPU is there */
737     val |= 0x04; /* PS/2 mouse installed */
738     rtc_set_memory(s, REG_EQUIPMENT_BYTE, val);
739 
740     /* hard drives and FDC */
741     arg.rtc_state = s;
742     arg.idebus[0] = idebus0;
743     arg.idebus[1] = idebus1;
744     qemu_register_reset(pc_cmos_init_late, &arg);
745 }
746 
747 #define TYPE_PORT92 "port92"
748 #define PORT92(obj) OBJECT_CHECK(Port92State, (obj), TYPE_PORT92)
749 
750 /* port 92 stuff: could be split off */
751 typedef struct Port92State {
752     ISADevice parent_obj;
753 
754     MemoryRegion io;
755     uint8_t outport;
756     qemu_irq a20_out;
757 } Port92State;
758 
759 static void port92_write(void *opaque, hwaddr addr, uint64_t val,
760                          unsigned size)
761 {
762     Port92State *s = opaque;
763     int oldval = s->outport;
764 
765     DPRINTF("port92: write 0x%02" PRIx64 "\n", val);
766     s->outport = val;
767     qemu_set_irq(s->a20_out, (val >> 1) & 1);
768     if ((val & 1) && !(oldval & 1)) {
769         qemu_system_reset_request(SHUTDOWN_CAUSE_GUEST_RESET);
770     }
771 }
772 
773 static uint64_t port92_read(void *opaque, hwaddr addr,
774                             unsigned size)
775 {
776     Port92State *s = opaque;
777     uint32_t ret;
778 
779     ret = s->outport;
780     DPRINTF("port92: read 0x%02x\n", ret);
781     return ret;
782 }
783 
784 static void port92_init(ISADevice *dev, qemu_irq a20_out)
785 {
786     qdev_connect_gpio_out_named(DEVICE(dev), PORT92_A20_LINE, 0, a20_out);
787 }
788 
789 static const VMStateDescription vmstate_port92_isa = {
790     .name = "port92",
791     .version_id = 1,
792     .minimum_version_id = 1,
793     .fields = (VMStateField[]) {
794         VMSTATE_UINT8(outport, Port92State),
795         VMSTATE_END_OF_LIST()
796     }
797 };
798 
799 static void port92_reset(DeviceState *d)
800 {
801     Port92State *s = PORT92(d);
802 
803     s->outport &= ~1;
804 }
805 
806 static const MemoryRegionOps port92_ops = {
807     .read = port92_read,
808     .write = port92_write,
809     .impl = {
810         .min_access_size = 1,
811         .max_access_size = 1,
812     },
813     .endianness = DEVICE_LITTLE_ENDIAN,
814 };
815 
816 static void port92_initfn(Object *obj)
817 {
818     Port92State *s = PORT92(obj);
819 
820     memory_region_init_io(&s->io, OBJECT(s), &port92_ops, s, "port92", 1);
821 
822     s->outport = 0;
823 
824     qdev_init_gpio_out_named(DEVICE(obj), &s->a20_out, PORT92_A20_LINE, 1);
825 }
826 
827 static void port92_realizefn(DeviceState *dev, Error **errp)
828 {
829     ISADevice *isadev = ISA_DEVICE(dev);
830     Port92State *s = PORT92(dev);
831 
832     isa_register_ioport(isadev, &s->io, 0x92);
833 }
834 
835 static void port92_class_initfn(ObjectClass *klass, void *data)
836 {
837     DeviceClass *dc = DEVICE_CLASS(klass);
838 
839     dc->realize = port92_realizefn;
840     dc->reset = port92_reset;
841     dc->vmsd = &vmstate_port92_isa;
842     /*
843      * Reason: unlike ordinary ISA devices, this one needs additional
844      * wiring: its A20 output line needs to be wired up by
845      * port92_init().
846      */
847     dc->user_creatable = false;
848 }
849 
850 static const TypeInfo port92_info = {
851     .name          = TYPE_PORT92,
852     .parent        = TYPE_ISA_DEVICE,
853     .instance_size = sizeof(Port92State),
854     .instance_init = port92_initfn,
855     .class_init    = port92_class_initfn,
856 };
857 
858 static void port92_register_types(void)
859 {
860     type_register_static(&port92_info);
861 }
862 
863 type_init(port92_register_types)
864 
865 static void handle_a20_line_change(void *opaque, int irq, int level)
866 {
867     X86CPU *cpu = opaque;
868 
869     /* XXX: send to all CPUs ? */
870     /* XXX: add logic to handle multiple A20 line sources */
871     x86_cpu_set_a20(cpu, level);
872 }
873 
874 int e820_add_entry(uint64_t address, uint64_t length, uint32_t type)
875 {
876     int index = le32_to_cpu(e820_reserve.count);
877     struct e820_entry *entry;
878 
879     if (type != E820_RAM) {
880         /* old FW_CFG_E820_TABLE entry -- reservations only */
881         if (index >= E820_NR_ENTRIES) {
882             return -EBUSY;
883         }
884         entry = &e820_reserve.entry[index++];
885 
886         entry->address = cpu_to_le64(address);
887         entry->length = cpu_to_le64(length);
888         entry->type = cpu_to_le32(type);
889 
890         e820_reserve.count = cpu_to_le32(index);
891     }
892 
893     /* new "etc/e820" file -- include ram too */
894     e820_table = g_renew(struct e820_entry, e820_table, e820_entries + 1);
895     e820_table[e820_entries].address = cpu_to_le64(address);
896     e820_table[e820_entries].length = cpu_to_le64(length);
897     e820_table[e820_entries].type = cpu_to_le32(type);
898     e820_entries++;
899 
900     return e820_entries;
901 }
902 
903 int e820_get_num_entries(void)
904 {
905     return e820_entries;
906 }
907 
908 bool e820_get_entry(int idx, uint32_t type, uint64_t *address, uint64_t *length)
909 {
910     if (idx < e820_entries && e820_table[idx].type == cpu_to_le32(type)) {
911         *address = le64_to_cpu(e820_table[idx].address);
912         *length = le64_to_cpu(e820_table[idx].length);
913         return true;
914     }
915     return false;
916 }
917 
918 /* Enables contiguous-apic-ID mode, for compatibility */
919 static bool compat_apic_id_mode;
920 
921 void enable_compat_apic_id_mode(void)
922 {
923     compat_apic_id_mode = true;
924 }
925 
926 /* Calculates initial APIC ID for a specific CPU index
927  *
928  * Currently we need to be able to calculate the APIC ID from the CPU index
929  * alone (without requiring a CPU object), as the QEMU<->Seabios interfaces have
930  * no concept of "CPU index", and the NUMA tables on fw_cfg need the APIC ID of
931  * all CPUs up to max_cpus.
932  */
933 static uint32_t x86_cpu_apic_id_from_index(unsigned int cpu_index)
934 {
935     uint32_t correct_id;
936     static bool warned;
937 
938     correct_id = x86_apicid_from_cpu_idx(smp_cores, smp_threads, cpu_index);
939     if (compat_apic_id_mode) {
940         if (cpu_index != correct_id && !warned && !qtest_enabled()) {
941             error_report("APIC IDs set in compatibility mode, "
942                          "CPU topology won't match the configuration");
943             warned = true;
944         }
945         return cpu_index;
946     } else {
947         return correct_id;
948     }
949 }
950 
951 static void pc_build_smbios(PCMachineState *pcms)
952 {
953     uint8_t *smbios_tables, *smbios_anchor;
954     size_t smbios_tables_len, smbios_anchor_len;
955     struct smbios_phys_mem_area *mem_array;
956     unsigned i, array_count;
957     MachineState *ms = MACHINE(pcms);
958     X86CPU *cpu = X86_CPU(ms->possible_cpus->cpus[0].cpu);
959 
960     /* tell smbios about cpuid version and features */
961     smbios_set_cpuid(cpu->env.cpuid_version, cpu->env.features[FEAT_1_EDX]);
962 
963     smbios_tables = smbios_get_table_legacy(&smbios_tables_len);
964     if (smbios_tables) {
965         fw_cfg_add_bytes(pcms->fw_cfg, FW_CFG_SMBIOS_ENTRIES,
966                          smbios_tables, smbios_tables_len);
967     }
968 
969     /* build the array of physical mem area from e820 table */
970     mem_array = g_malloc0(sizeof(*mem_array) * e820_get_num_entries());
971     for (i = 0, array_count = 0; i < e820_get_num_entries(); i++) {
972         uint64_t addr, len;
973 
974         if (e820_get_entry(i, E820_RAM, &addr, &len)) {
975             mem_array[array_count].address = addr;
976             mem_array[array_count].length = len;
977             array_count++;
978         }
979     }
980     smbios_get_tables(mem_array, array_count,
981                       &smbios_tables, &smbios_tables_len,
982                       &smbios_anchor, &smbios_anchor_len);
983     g_free(mem_array);
984 
985     if (smbios_anchor) {
986         fw_cfg_add_file(pcms->fw_cfg, "etc/smbios/smbios-tables",
987                         smbios_tables, smbios_tables_len);
988         fw_cfg_add_file(pcms->fw_cfg, "etc/smbios/smbios-anchor",
989                         smbios_anchor, smbios_anchor_len);
990     }
991 }
992 
993 static FWCfgState *bochs_bios_init(AddressSpace *as, PCMachineState *pcms)
994 {
995     FWCfgState *fw_cfg;
996     uint64_t *numa_fw_cfg;
997     int i;
998     const CPUArchIdList *cpus;
999     MachineClass *mc = MACHINE_GET_CLASS(pcms);
1000 
1001     fw_cfg = fw_cfg_init_io_dma(FW_CFG_IO_BASE, FW_CFG_IO_BASE + 4, as);
1002     fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
1003 
1004     /* FW_CFG_MAX_CPUS is a bit confusing/problematic on x86:
1005      *
1006      * For machine types prior to 1.8, SeaBIOS needs FW_CFG_MAX_CPUS for
1007      * building MPTable, ACPI MADT, ACPI CPU hotplug and ACPI SRAT table,
1008      * that tables are based on xAPIC ID and QEMU<->SeaBIOS interface
1009      * for CPU hotplug also uses APIC ID and not "CPU index".
1010      * This means that FW_CFG_MAX_CPUS is not the "maximum number of CPUs",
1011      * but the "limit to the APIC ID values SeaBIOS may see".
1012      *
1013      * So for compatibility reasons with old BIOSes we are stuck with
1014      * "etc/max-cpus" actually being apic_id_limit
1015      */
1016     fw_cfg_add_i16(fw_cfg, FW_CFG_MAX_CPUS, (uint16_t)pcms->apic_id_limit);
1017     fw_cfg_add_i64(fw_cfg, FW_CFG_RAM_SIZE, (uint64_t)ram_size);
1018     fw_cfg_add_bytes(fw_cfg, FW_CFG_ACPI_TABLES,
1019                      acpi_tables, acpi_tables_len);
1020     fw_cfg_add_i32(fw_cfg, FW_CFG_IRQ0_OVERRIDE, kvm_allows_irq0_override());
1021 
1022     fw_cfg_add_bytes(fw_cfg, FW_CFG_E820_TABLE,
1023                      &e820_reserve, sizeof(e820_reserve));
1024     fw_cfg_add_file(fw_cfg, "etc/e820", e820_table,
1025                     sizeof(struct e820_entry) * e820_entries);
1026 
1027     fw_cfg_add_bytes(fw_cfg, FW_CFG_HPET, &hpet_cfg, sizeof(hpet_cfg));
1028     /* allocate memory for the NUMA channel: one (64bit) word for the number
1029      * of nodes, one word for each VCPU->node and one word for each node to
1030      * hold the amount of memory.
1031      */
1032     numa_fw_cfg = g_new0(uint64_t, 1 + pcms->apic_id_limit + nb_numa_nodes);
1033     numa_fw_cfg[0] = cpu_to_le64(nb_numa_nodes);
1034     cpus = mc->possible_cpu_arch_ids(MACHINE(pcms));
1035     for (i = 0; i < cpus->len; i++) {
1036         unsigned int apic_id = cpus->cpus[i].arch_id;
1037         assert(apic_id < pcms->apic_id_limit);
1038         numa_fw_cfg[apic_id + 1] = cpu_to_le64(cpus->cpus[i].props.node_id);
1039     }
1040     for (i = 0; i < nb_numa_nodes; i++) {
1041         numa_fw_cfg[pcms->apic_id_limit + 1 + i] =
1042             cpu_to_le64(numa_info[i].node_mem);
1043     }
1044     fw_cfg_add_bytes(fw_cfg, FW_CFG_NUMA, numa_fw_cfg,
1045                      (1 + pcms->apic_id_limit + nb_numa_nodes) *
1046                      sizeof(*numa_fw_cfg));
1047 
1048     return fw_cfg;
1049 }
1050 
1051 static long get_file_size(FILE *f)
1052 {
1053     long where, size;
1054 
1055     /* XXX: on Unix systems, using fstat() probably makes more sense */
1056 
1057     where = ftell(f);
1058     fseek(f, 0, SEEK_END);
1059     size = ftell(f);
1060     fseek(f, where, SEEK_SET);
1061 
1062     return size;
1063 }
1064 
1065 struct setup_data {
1066     uint64_t next;
1067     uint32_t type;
1068     uint32_t len;
1069     uint8_t data[0];
1070 } __attribute__((packed));
1071 
1072 
1073 /*
1074  * The entry point into the kernel for PVH boot is different from
1075  * the native entry point.  The PVH entry is defined by the x86/HVM
1076  * direct boot ABI and is available in an ELFNOTE in the kernel binary.
1077  *
1078  * This function is passed to load_elf() when it is called from
1079  * load_elfboot() which then additionally checks for an ELF Note of
1080  * type XEN_ELFNOTE_PHYS32_ENTRY and passes it to this function to
1081  * parse the PVH entry address from the ELF Note.
1082  *
1083  * Due to trickery in elf_opts.h, load_elf() is actually available as
1084  * load_elf32() or load_elf64() and this routine needs to be able
1085  * to deal with being called as 32 or 64 bit.
1086  *
1087  * The address of the PVH entry point is saved to the 'pvh_start_addr'
1088  * global variable.  (although the entry point is 32-bit, the kernel
1089  * binary can be either 32-bit or 64-bit).
1090  */
1091 static uint64_t read_pvh_start_addr(void *arg1, void *arg2, bool is64)
1092 {
1093     size_t *elf_note_data_addr;
1094 
1095     /* Check if ELF Note header passed in is valid */
1096     if (arg1 == NULL) {
1097         return 0;
1098     }
1099 
1100     if (is64) {
1101         struct elf64_note *nhdr64 = (struct elf64_note *)arg1;
1102         uint64_t nhdr_size64 = sizeof(struct elf64_note);
1103         uint64_t phdr_align = *(uint64_t *)arg2;
1104         uint64_t nhdr_namesz = nhdr64->n_namesz;
1105 
1106         elf_note_data_addr =
1107             ((void *)nhdr64) + nhdr_size64 +
1108             QEMU_ALIGN_UP(nhdr_namesz, phdr_align);
1109     } else {
1110         struct elf32_note *nhdr32 = (struct elf32_note *)arg1;
1111         uint32_t nhdr_size32 = sizeof(struct elf32_note);
1112         uint32_t phdr_align = *(uint32_t *)arg2;
1113         uint32_t nhdr_namesz = nhdr32->n_namesz;
1114 
1115         elf_note_data_addr =
1116             ((void *)nhdr32) + nhdr_size32 +
1117             QEMU_ALIGN_UP(nhdr_namesz, phdr_align);
1118     }
1119 
1120     pvh_start_addr = *elf_note_data_addr;
1121 
1122     return pvh_start_addr;
1123 }
1124 
1125 static bool load_elfboot(const char *kernel_filename,
1126                    int kernel_file_size,
1127                    uint8_t *header,
1128                    size_t pvh_xen_start_addr,
1129                    FWCfgState *fw_cfg)
1130 {
1131     uint32_t flags = 0;
1132     uint32_t mh_load_addr = 0;
1133     uint32_t elf_kernel_size = 0;
1134     uint64_t elf_entry;
1135     uint64_t elf_low, elf_high;
1136     int kernel_size;
1137 
1138     if (ldl_p(header) != 0x464c457f) {
1139         return false; /* no elfboot */
1140     }
1141 
1142     bool elf_is64 = header[EI_CLASS] == ELFCLASS64;
1143     flags = elf_is64 ?
1144         ((Elf64_Ehdr *)header)->e_flags : ((Elf32_Ehdr *)header)->e_flags;
1145 
1146     if (flags & 0x00010004) { /* LOAD_ELF_HEADER_HAS_ADDR */
1147         error_report("elfboot unsupported flags = %x", flags);
1148         exit(1);
1149     }
1150 
1151     uint64_t elf_note_type = XEN_ELFNOTE_PHYS32_ENTRY;
1152     kernel_size = load_elf(kernel_filename, read_pvh_start_addr,
1153                            NULL, &elf_note_type, &elf_entry,
1154                            &elf_low, &elf_high, 0, I386_ELF_MACHINE,
1155                            0, 0);
1156 
1157     if (kernel_size < 0) {
1158         error_report("Error while loading elf kernel");
1159         exit(1);
1160     }
1161     mh_load_addr = elf_low;
1162     elf_kernel_size = elf_high - elf_low;
1163 
1164     if (pvh_start_addr == 0) {
1165         error_report("Error loading uncompressed kernel without PVH ELF Note");
1166         exit(1);
1167     }
1168     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ENTRY, pvh_start_addr);
1169     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ADDR, mh_load_addr);
1170     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_SIZE, elf_kernel_size);
1171 
1172     return true;
1173 }
1174 
1175 static void load_linux(PCMachineState *pcms,
1176                        FWCfgState *fw_cfg)
1177 {
1178     uint16_t protocol;
1179     int setup_size, kernel_size, cmdline_size;
1180     int dtb_size, setup_data_offset;
1181     uint32_t initrd_max;
1182     uint8_t header[8192], *setup, *kernel;
1183     hwaddr real_addr, prot_addr, cmdline_addr, initrd_addr = 0;
1184     FILE *f;
1185     char *vmode;
1186     MachineState *machine = MACHINE(pcms);
1187     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1188     struct setup_data *setup_data;
1189     const char *kernel_filename = machine->kernel_filename;
1190     const char *initrd_filename = machine->initrd_filename;
1191     const char *dtb_filename = machine->dtb;
1192     const char *kernel_cmdline = machine->kernel_cmdline;
1193 
1194     /* Align to 16 bytes as a paranoia measure */
1195     cmdline_size = (strlen(kernel_cmdline)+16) & ~15;
1196 
1197     /* load the kernel header */
1198     f = fopen(kernel_filename, "rb");
1199     if (!f || !(kernel_size = get_file_size(f)) ||
1200         fread(header, 1, MIN(ARRAY_SIZE(header), kernel_size), f) !=
1201         MIN(ARRAY_SIZE(header), kernel_size)) {
1202         fprintf(stderr, "qemu: could not load kernel '%s': %s\n",
1203                 kernel_filename, strerror(errno));
1204         exit(1);
1205     }
1206 
1207     /* kernel protocol version */
1208 #if 0
1209     fprintf(stderr, "header magic: %#x\n", ldl_p(header+0x202));
1210 #endif
1211     if (ldl_p(header+0x202) == 0x53726448) {
1212         protocol = lduw_p(header+0x206);
1213     } else {
1214         /*
1215          * This could be a multiboot kernel. If it is, let's stop treating it
1216          * like a Linux kernel.
1217          * Note: some multiboot images could be in the ELF format (the same of
1218          * PVH), so we try multiboot first since we check the multiboot magic
1219          * header before to load it.
1220          */
1221         if (load_multiboot(fw_cfg, f, kernel_filename, initrd_filename,
1222                            kernel_cmdline, kernel_size, header)) {
1223             return;
1224         }
1225         /*
1226          * Check if the file is an uncompressed kernel file (ELF) and load it,
1227          * saving the PVH entry point used by the x86/HVM direct boot ABI.
1228          * If load_elfboot() is successful, populate the fw_cfg info.
1229          */
1230         if (pcmc->pvh_enabled &&
1231             load_elfboot(kernel_filename, kernel_size,
1232                          header, pvh_start_addr, fw_cfg)) {
1233             fclose(f);
1234 
1235             fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_SIZE,
1236                 strlen(kernel_cmdline) + 1);
1237             fw_cfg_add_string(fw_cfg, FW_CFG_CMDLINE_DATA, kernel_cmdline);
1238 
1239             fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_SIZE, sizeof(header));
1240             fw_cfg_add_bytes(fw_cfg, FW_CFG_SETUP_DATA,
1241                              header, sizeof(header));
1242 
1243             /* load initrd */
1244             if (initrd_filename) {
1245                 gsize initrd_size;
1246                 gchar *initrd_data;
1247                 GError *gerr = NULL;
1248 
1249                 if (!g_file_get_contents(initrd_filename, &initrd_data,
1250                             &initrd_size, &gerr)) {
1251                     fprintf(stderr, "qemu: error reading initrd %s: %s\n",
1252                             initrd_filename, gerr->message);
1253                     exit(1);
1254                 }
1255 
1256                 initrd_max = pcms->below_4g_mem_size - pcmc->acpi_data_size - 1;
1257                 if (initrd_size >= initrd_max) {
1258                     fprintf(stderr, "qemu: initrd is too large, cannot support."
1259                             "(max: %"PRIu32", need %"PRId64")\n",
1260                             initrd_max, (uint64_t)initrd_size);
1261                     exit(1);
1262                 }
1263 
1264                 initrd_addr = (initrd_max - initrd_size) & ~4095;
1265 
1266                 fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_ADDR, initrd_addr);
1267                 fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_SIZE, initrd_size);
1268                 fw_cfg_add_bytes(fw_cfg, FW_CFG_INITRD_DATA, initrd_data,
1269                                  initrd_size);
1270             }
1271 
1272             option_rom[nb_option_roms].bootindex = 0;
1273             option_rom[nb_option_roms].name = "pvh.bin";
1274             nb_option_roms++;
1275 
1276             return;
1277         }
1278         protocol = 0;
1279     }
1280 
1281     if (protocol < 0x200 || !(header[0x211] & 0x01)) {
1282         /* Low kernel */
1283         real_addr    = 0x90000;
1284         cmdline_addr = 0x9a000 - cmdline_size;
1285         prot_addr    = 0x10000;
1286     } else if (protocol < 0x202) {
1287         /* High but ancient kernel */
1288         real_addr    = 0x90000;
1289         cmdline_addr = 0x9a000 - cmdline_size;
1290         prot_addr    = 0x100000;
1291     } else {
1292         /* High and recent kernel */
1293         real_addr    = 0x10000;
1294         cmdline_addr = 0x20000;
1295         prot_addr    = 0x100000;
1296     }
1297 
1298 #if 0
1299     fprintf(stderr,
1300             "qemu: real_addr     = 0x" TARGET_FMT_plx "\n"
1301             "qemu: cmdline_addr  = 0x" TARGET_FMT_plx "\n"
1302             "qemu: prot_addr     = 0x" TARGET_FMT_plx "\n",
1303             real_addr,
1304             cmdline_addr,
1305             prot_addr);
1306 #endif
1307 
1308     /* highest address for loading the initrd */
1309     if (protocol >= 0x20c &&
1310         lduw_p(header+0x236) & XLF_CAN_BE_LOADED_ABOVE_4G) {
1311         /*
1312          * Linux has supported initrd up to 4 GB for a very long time (2007,
1313          * long before XLF_CAN_BE_LOADED_ABOVE_4G which was added in 2013),
1314          * though it only sets initrd_max to 2 GB to "work around bootloader
1315          * bugs". Luckily, QEMU firmware(which does something like bootloader)
1316          * has supported this.
1317          *
1318          * It's believed that if XLF_CAN_BE_LOADED_ABOVE_4G is set, initrd can
1319          * be loaded into any address.
1320          *
1321          * In addition, initrd_max is uint32_t simply because QEMU doesn't
1322          * support the 64-bit boot protocol (specifically the ext_ramdisk_image
1323          * field).
1324          *
1325          * Therefore here just limit initrd_max to UINT32_MAX simply as well.
1326          */
1327         initrd_max = UINT32_MAX;
1328     } else if (protocol >= 0x203) {
1329         initrd_max = ldl_p(header+0x22c);
1330     } else {
1331         initrd_max = 0x37ffffff;
1332     }
1333 
1334     if (initrd_max >= pcms->below_4g_mem_size - pcmc->acpi_data_size) {
1335         initrd_max = pcms->below_4g_mem_size - pcmc->acpi_data_size - 1;
1336     }
1337 
1338     fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_ADDR, cmdline_addr);
1339     fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_SIZE, strlen(kernel_cmdline)+1);
1340     fw_cfg_add_string(fw_cfg, FW_CFG_CMDLINE_DATA, kernel_cmdline);
1341 
1342     if (protocol >= 0x202) {
1343         stl_p(header+0x228, cmdline_addr);
1344     } else {
1345         stw_p(header+0x20, 0xA33F);
1346         stw_p(header+0x22, cmdline_addr-real_addr);
1347     }
1348 
1349     /* handle vga= parameter */
1350     vmode = strstr(kernel_cmdline, "vga=");
1351     if (vmode) {
1352         unsigned int video_mode;
1353         /* skip "vga=" */
1354         vmode += 4;
1355         if (!strncmp(vmode, "normal", 6)) {
1356             video_mode = 0xffff;
1357         } else if (!strncmp(vmode, "ext", 3)) {
1358             video_mode = 0xfffe;
1359         } else if (!strncmp(vmode, "ask", 3)) {
1360             video_mode = 0xfffd;
1361         } else {
1362             video_mode = strtol(vmode, NULL, 0);
1363         }
1364         stw_p(header+0x1fa, video_mode);
1365     }
1366 
1367     /* loader type */
1368     /* High nybble = B reserved for QEMU; low nybble is revision number.
1369        If this code is substantially changed, you may want to consider
1370        incrementing the revision. */
1371     if (protocol >= 0x200) {
1372         header[0x210] = 0xB0;
1373     }
1374     /* heap */
1375     if (protocol >= 0x201) {
1376         header[0x211] |= 0x80;	/* CAN_USE_HEAP */
1377         stw_p(header+0x224, cmdline_addr-real_addr-0x200);
1378     }
1379 
1380     /* load initrd */
1381     if (initrd_filename) {
1382         gsize initrd_size;
1383         gchar *initrd_data;
1384         GError *gerr = NULL;
1385 
1386         if (protocol < 0x200) {
1387             fprintf(stderr, "qemu: linux kernel too old to load a ram disk\n");
1388             exit(1);
1389         }
1390 
1391         if (!g_file_get_contents(initrd_filename, &initrd_data,
1392                                  &initrd_size, &gerr)) {
1393             fprintf(stderr, "qemu: error reading initrd %s: %s\n",
1394                     initrd_filename, gerr->message);
1395             exit(1);
1396         }
1397         if (initrd_size >= initrd_max) {
1398             fprintf(stderr, "qemu: initrd is too large, cannot support."
1399                     "(max: %"PRIu32", need %"PRId64")\n",
1400                     initrd_max, (uint64_t)initrd_size);
1401             exit(1);
1402         }
1403 
1404         initrd_addr = (initrd_max-initrd_size) & ~4095;
1405 
1406         fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_ADDR, initrd_addr);
1407         fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_SIZE, initrd_size);
1408         fw_cfg_add_bytes(fw_cfg, FW_CFG_INITRD_DATA, initrd_data, initrd_size);
1409 
1410         stl_p(header+0x218, initrd_addr);
1411         stl_p(header+0x21c, initrd_size);
1412     }
1413 
1414     /* load kernel and setup */
1415     setup_size = header[0x1f1];
1416     if (setup_size == 0) {
1417         setup_size = 4;
1418     }
1419     setup_size = (setup_size+1)*512;
1420     if (setup_size > kernel_size) {
1421         fprintf(stderr, "qemu: invalid kernel header\n");
1422         exit(1);
1423     }
1424     kernel_size -= setup_size;
1425 
1426     setup  = g_malloc(setup_size);
1427     kernel = g_malloc(kernel_size);
1428     fseek(f, 0, SEEK_SET);
1429     if (fread(setup, 1, setup_size, f) != setup_size) {
1430         fprintf(stderr, "fread() failed\n");
1431         exit(1);
1432     }
1433     if (fread(kernel, 1, kernel_size, f) != kernel_size) {
1434         fprintf(stderr, "fread() failed\n");
1435         exit(1);
1436     }
1437     fclose(f);
1438 
1439     /* append dtb to kernel */
1440     if (dtb_filename) {
1441         if (protocol < 0x209) {
1442             fprintf(stderr, "qemu: Linux kernel too old to load a dtb\n");
1443             exit(1);
1444         }
1445 
1446         dtb_size = get_image_size(dtb_filename);
1447         if (dtb_size <= 0) {
1448             fprintf(stderr, "qemu: error reading dtb %s: %s\n",
1449                     dtb_filename, strerror(errno));
1450             exit(1);
1451         }
1452 
1453         setup_data_offset = QEMU_ALIGN_UP(kernel_size, 16);
1454         kernel_size = setup_data_offset + sizeof(struct setup_data) + dtb_size;
1455         kernel = g_realloc(kernel, kernel_size);
1456 
1457         stq_p(header+0x250, prot_addr + setup_data_offset);
1458 
1459         setup_data = (struct setup_data *)(kernel + setup_data_offset);
1460         setup_data->next = 0;
1461         setup_data->type = cpu_to_le32(SETUP_DTB);
1462         setup_data->len = cpu_to_le32(dtb_size);
1463 
1464         load_image_size(dtb_filename, setup_data->data, dtb_size);
1465     }
1466 
1467     memcpy(setup, header, MIN(sizeof(header), setup_size));
1468 
1469     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ADDR, prot_addr);
1470     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_SIZE, kernel_size);
1471     fw_cfg_add_bytes(fw_cfg, FW_CFG_KERNEL_DATA, kernel, kernel_size);
1472 
1473     fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_ADDR, real_addr);
1474     fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_SIZE, setup_size);
1475     fw_cfg_add_bytes(fw_cfg, FW_CFG_SETUP_DATA, setup, setup_size);
1476 
1477     option_rom[nb_option_roms].bootindex = 0;
1478     option_rom[nb_option_roms].name = "linuxboot.bin";
1479     if (pcmc->linuxboot_dma_enabled && fw_cfg_dma_enabled(fw_cfg)) {
1480         option_rom[nb_option_roms].name = "linuxboot_dma.bin";
1481     }
1482     nb_option_roms++;
1483 }
1484 
1485 #define NE2000_NB_MAX 6
1486 
1487 static const int ne2000_io[NE2000_NB_MAX] = { 0x300, 0x320, 0x340, 0x360,
1488                                               0x280, 0x380 };
1489 static const int ne2000_irq[NE2000_NB_MAX] = { 9, 10, 11, 3, 4, 5 };
1490 
1491 void pc_init_ne2k_isa(ISABus *bus, NICInfo *nd)
1492 {
1493     static int nb_ne2k = 0;
1494 
1495     if (nb_ne2k == NE2000_NB_MAX)
1496         return;
1497     isa_ne2000_init(bus, ne2000_io[nb_ne2k],
1498                     ne2000_irq[nb_ne2k], nd);
1499     nb_ne2k++;
1500 }
1501 
1502 DeviceState *cpu_get_current_apic(void)
1503 {
1504     if (current_cpu) {
1505         X86CPU *cpu = X86_CPU(current_cpu);
1506         return cpu->apic_state;
1507     } else {
1508         return NULL;
1509     }
1510 }
1511 
1512 void pc_acpi_smi_interrupt(void *opaque, int irq, int level)
1513 {
1514     X86CPU *cpu = opaque;
1515 
1516     if (level) {
1517         cpu_interrupt(CPU(cpu), CPU_INTERRUPT_SMI);
1518     }
1519 }
1520 
1521 static void pc_new_cpu(const char *typename, int64_t apic_id, Error **errp)
1522 {
1523     Object *cpu = NULL;
1524     Error *local_err = NULL;
1525 
1526     cpu = object_new(typename);
1527 
1528     object_property_set_uint(cpu, apic_id, "apic-id", &local_err);
1529     object_property_set_bool(cpu, true, "realized", &local_err);
1530 
1531     object_unref(cpu);
1532     error_propagate(errp, local_err);
1533 }
1534 
1535 void pc_hot_add_cpu(const int64_t id, Error **errp)
1536 {
1537     MachineState *ms = MACHINE(qdev_get_machine());
1538     int64_t apic_id = x86_cpu_apic_id_from_index(id);
1539     Error *local_err = NULL;
1540 
1541     if (id < 0) {
1542         error_setg(errp, "Invalid CPU id: %" PRIi64, id);
1543         return;
1544     }
1545 
1546     if (apic_id >= ACPI_CPU_HOTPLUG_ID_LIMIT) {
1547         error_setg(errp, "Unable to add CPU: %" PRIi64
1548                    ", resulting APIC ID (%" PRIi64 ") is too large",
1549                    id, apic_id);
1550         return;
1551     }
1552 
1553     pc_new_cpu(ms->cpu_type, apic_id, &local_err);
1554     if (local_err) {
1555         error_propagate(errp, local_err);
1556         return;
1557     }
1558 }
1559 
1560 void pc_cpus_init(PCMachineState *pcms)
1561 {
1562     int i;
1563     const CPUArchIdList *possible_cpus;
1564     MachineState *ms = MACHINE(pcms);
1565     MachineClass *mc = MACHINE_GET_CLASS(pcms);
1566 
1567     /* Calculates the limit to CPU APIC ID values
1568      *
1569      * Limit for the APIC ID value, so that all
1570      * CPU APIC IDs are < pcms->apic_id_limit.
1571      *
1572      * This is used for FW_CFG_MAX_CPUS. See comments on bochs_bios_init().
1573      */
1574     pcms->apic_id_limit = x86_cpu_apic_id_from_index(max_cpus - 1) + 1;
1575     possible_cpus = mc->possible_cpu_arch_ids(ms);
1576     for (i = 0; i < smp_cpus; i++) {
1577         pc_new_cpu(possible_cpus->cpus[i].type, possible_cpus->cpus[i].arch_id,
1578                    &error_fatal);
1579     }
1580 }
1581 
1582 static void pc_build_feature_control_file(PCMachineState *pcms)
1583 {
1584     MachineState *ms = MACHINE(pcms);
1585     X86CPU *cpu = X86_CPU(ms->possible_cpus->cpus[0].cpu);
1586     CPUX86State *env = &cpu->env;
1587     uint32_t unused, ecx, edx;
1588     uint64_t feature_control_bits = 0;
1589     uint64_t *val;
1590 
1591     cpu_x86_cpuid(env, 1, 0, &unused, &unused, &ecx, &edx);
1592     if (ecx & CPUID_EXT_VMX) {
1593         feature_control_bits |= FEATURE_CONTROL_VMXON_ENABLED_OUTSIDE_SMX;
1594     }
1595 
1596     if ((edx & (CPUID_EXT2_MCE | CPUID_EXT2_MCA)) ==
1597         (CPUID_EXT2_MCE | CPUID_EXT2_MCA) &&
1598         (env->mcg_cap & MCG_LMCE_P)) {
1599         feature_control_bits |= FEATURE_CONTROL_LMCE;
1600     }
1601 
1602     if (!feature_control_bits) {
1603         return;
1604     }
1605 
1606     val = g_malloc(sizeof(*val));
1607     *val = cpu_to_le64(feature_control_bits | FEATURE_CONTROL_LOCKED);
1608     fw_cfg_add_file(pcms->fw_cfg, "etc/msr_feature_control", val, sizeof(*val));
1609 }
1610 
1611 static void rtc_set_cpus_count(ISADevice *rtc, uint16_t cpus_count)
1612 {
1613     if (cpus_count > 0xff) {
1614         /* If the number of CPUs can't be represented in 8 bits, the
1615          * BIOS must use "FW_CFG_NB_CPUS". Set RTC field to 0 just
1616          * to make old BIOSes fail more predictably.
1617          */
1618         rtc_set_memory(rtc, 0x5f, 0);
1619     } else {
1620         rtc_set_memory(rtc, 0x5f, cpus_count - 1);
1621     }
1622 }
1623 
1624 static
1625 void pc_machine_done(Notifier *notifier, void *data)
1626 {
1627     PCMachineState *pcms = container_of(notifier,
1628                                         PCMachineState, machine_done);
1629     PCIBus *bus = pcms->bus;
1630 
1631     /* set the number of CPUs */
1632     rtc_set_cpus_count(pcms->rtc, pcms->boot_cpus);
1633 
1634     if (bus) {
1635         int extra_hosts = 0;
1636 
1637         QLIST_FOREACH(bus, &bus->child, sibling) {
1638             /* look for expander root buses */
1639             if (pci_bus_is_root(bus)) {
1640                 extra_hosts++;
1641             }
1642         }
1643         if (extra_hosts && pcms->fw_cfg) {
1644             uint64_t *val = g_malloc(sizeof(*val));
1645             *val = cpu_to_le64(extra_hosts);
1646             fw_cfg_add_file(pcms->fw_cfg,
1647                     "etc/extra-pci-roots", val, sizeof(*val));
1648         }
1649     }
1650 
1651     acpi_setup();
1652     if (pcms->fw_cfg) {
1653         pc_build_smbios(pcms);
1654         pc_build_feature_control_file(pcms);
1655         /* update FW_CFG_NB_CPUS to account for -device added CPUs */
1656         fw_cfg_modify_i16(pcms->fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
1657     }
1658 
1659     if (pcms->apic_id_limit > 255 && !xen_enabled()) {
1660         IntelIOMMUState *iommu = INTEL_IOMMU_DEVICE(x86_iommu_get_default());
1661 
1662         if (!iommu || !x86_iommu_ir_supported(X86_IOMMU_DEVICE(iommu)) ||
1663             iommu->intr_eim != ON_OFF_AUTO_ON) {
1664             error_report("current -smp configuration requires "
1665                          "Extended Interrupt Mode enabled. "
1666                          "You can add an IOMMU using: "
1667                          "-device intel-iommu,intremap=on,eim=on");
1668             exit(EXIT_FAILURE);
1669         }
1670     }
1671 }
1672 
1673 void pc_guest_info_init(PCMachineState *pcms)
1674 {
1675     int i;
1676 
1677     pcms->apic_xrupt_override = kvm_allows_irq0_override();
1678     pcms->numa_nodes = nb_numa_nodes;
1679     pcms->node_mem = g_malloc0(pcms->numa_nodes *
1680                                     sizeof *pcms->node_mem);
1681     for (i = 0; i < nb_numa_nodes; i++) {
1682         pcms->node_mem[i] = numa_info[i].node_mem;
1683     }
1684 
1685     pcms->machine_done.notify = pc_machine_done;
1686     qemu_add_machine_init_done_notifier(&pcms->machine_done);
1687 }
1688 
1689 /* setup pci memory address space mapping into system address space */
1690 void pc_pci_as_mapping_init(Object *owner, MemoryRegion *system_memory,
1691                             MemoryRegion *pci_address_space)
1692 {
1693     /* Set to lower priority than RAM */
1694     memory_region_add_subregion_overlap(system_memory, 0x0,
1695                                         pci_address_space, -1);
1696 }
1697 
1698 void xen_load_linux(PCMachineState *pcms)
1699 {
1700     int i;
1701     FWCfgState *fw_cfg;
1702 
1703     assert(MACHINE(pcms)->kernel_filename != NULL);
1704 
1705     fw_cfg = fw_cfg_init_io(FW_CFG_IO_BASE);
1706     fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
1707     rom_set_fw(fw_cfg);
1708 
1709     load_linux(pcms, fw_cfg);
1710     for (i = 0; i < nb_option_roms; i++) {
1711         assert(!strcmp(option_rom[i].name, "linuxboot.bin") ||
1712                !strcmp(option_rom[i].name, "linuxboot_dma.bin") ||
1713                !strcmp(option_rom[i].name, "pvh.bin") ||
1714                !strcmp(option_rom[i].name, "multiboot.bin"));
1715         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1716     }
1717     pcms->fw_cfg = fw_cfg;
1718 }
1719 
1720 void pc_memory_init(PCMachineState *pcms,
1721                     MemoryRegion *system_memory,
1722                     MemoryRegion *rom_memory,
1723                     MemoryRegion **ram_memory)
1724 {
1725     int linux_boot, i;
1726     MemoryRegion *ram, *option_rom_mr;
1727     MemoryRegion *ram_below_4g, *ram_above_4g;
1728     FWCfgState *fw_cfg;
1729     MachineState *machine = MACHINE(pcms);
1730     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1731 
1732     assert(machine->ram_size == pcms->below_4g_mem_size +
1733                                 pcms->above_4g_mem_size);
1734 
1735     linux_boot = (machine->kernel_filename != NULL);
1736 
1737     /* Allocate RAM.  We allocate it as a single memory region and use
1738      * aliases to address portions of it, mostly for backwards compatibility
1739      * with older qemus that used qemu_ram_alloc().
1740      */
1741     ram = g_malloc(sizeof(*ram));
1742     memory_region_allocate_system_memory(ram, NULL, "pc.ram",
1743                                          machine->ram_size);
1744     *ram_memory = ram;
1745     ram_below_4g = g_malloc(sizeof(*ram_below_4g));
1746     memory_region_init_alias(ram_below_4g, NULL, "ram-below-4g", ram,
1747                              0, pcms->below_4g_mem_size);
1748     memory_region_add_subregion(system_memory, 0, ram_below_4g);
1749     e820_add_entry(0, pcms->below_4g_mem_size, E820_RAM);
1750     if (pcms->above_4g_mem_size > 0) {
1751         ram_above_4g = g_malloc(sizeof(*ram_above_4g));
1752         memory_region_init_alias(ram_above_4g, NULL, "ram-above-4g", ram,
1753                                  pcms->below_4g_mem_size,
1754                                  pcms->above_4g_mem_size);
1755         memory_region_add_subregion(system_memory, 0x100000000ULL,
1756                                     ram_above_4g);
1757         e820_add_entry(0x100000000ULL, pcms->above_4g_mem_size, E820_RAM);
1758     }
1759 
1760     if (!pcmc->has_reserved_memory &&
1761         (machine->ram_slots ||
1762          (machine->maxram_size > machine->ram_size))) {
1763         MachineClass *mc = MACHINE_GET_CLASS(machine);
1764 
1765         error_report("\"-memory 'slots|maxmem'\" is not supported by: %s",
1766                      mc->name);
1767         exit(EXIT_FAILURE);
1768     }
1769 
1770     /* always allocate the device memory information */
1771     machine->device_memory = g_malloc0(sizeof(*machine->device_memory));
1772 
1773     /* initialize device memory address space */
1774     if (pcmc->has_reserved_memory &&
1775         (machine->ram_size < machine->maxram_size)) {
1776         ram_addr_t device_mem_size = machine->maxram_size - machine->ram_size;
1777 
1778         if (machine->ram_slots > ACPI_MAX_RAM_SLOTS) {
1779             error_report("unsupported amount of memory slots: %"PRIu64,
1780                          machine->ram_slots);
1781             exit(EXIT_FAILURE);
1782         }
1783 
1784         if (QEMU_ALIGN_UP(machine->maxram_size,
1785                           TARGET_PAGE_SIZE) != machine->maxram_size) {
1786             error_report("maximum memory size must by aligned to multiple of "
1787                          "%d bytes", TARGET_PAGE_SIZE);
1788             exit(EXIT_FAILURE);
1789         }
1790 
1791         machine->device_memory->base =
1792             ROUND_UP(0x100000000ULL + pcms->above_4g_mem_size, 1 * GiB);
1793 
1794         if (pcmc->enforce_aligned_dimm) {
1795             /* size device region assuming 1G page max alignment per slot */
1796             device_mem_size += (1 * GiB) * machine->ram_slots;
1797         }
1798 
1799         if ((machine->device_memory->base + device_mem_size) <
1800             device_mem_size) {
1801             error_report("unsupported amount of maximum memory: " RAM_ADDR_FMT,
1802                          machine->maxram_size);
1803             exit(EXIT_FAILURE);
1804         }
1805 
1806         memory_region_init(&machine->device_memory->mr, OBJECT(pcms),
1807                            "device-memory", device_mem_size);
1808         memory_region_add_subregion(system_memory, machine->device_memory->base,
1809                                     &machine->device_memory->mr);
1810     }
1811 
1812     /* Initialize PC system firmware */
1813     pc_system_firmware_init(pcms, rom_memory);
1814 
1815     option_rom_mr = g_malloc(sizeof(*option_rom_mr));
1816     memory_region_init_ram(option_rom_mr, NULL, "pc.rom", PC_ROM_SIZE,
1817                            &error_fatal);
1818     if (pcmc->pci_enabled) {
1819         memory_region_set_readonly(option_rom_mr, true);
1820     }
1821     memory_region_add_subregion_overlap(rom_memory,
1822                                         PC_ROM_MIN_VGA,
1823                                         option_rom_mr,
1824                                         1);
1825 
1826     fw_cfg = bochs_bios_init(&address_space_memory, pcms);
1827 
1828     rom_set_fw(fw_cfg);
1829 
1830     if (pcmc->has_reserved_memory && machine->device_memory->base) {
1831         uint64_t *val = g_malloc(sizeof(*val));
1832         PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1833         uint64_t res_mem_end = machine->device_memory->base;
1834 
1835         if (!pcmc->broken_reserved_end) {
1836             res_mem_end += memory_region_size(&machine->device_memory->mr);
1837         }
1838         *val = cpu_to_le64(ROUND_UP(res_mem_end, 1 * GiB));
1839         fw_cfg_add_file(fw_cfg, "etc/reserved-memory-end", val, sizeof(*val));
1840     }
1841 
1842     if (linux_boot) {
1843         load_linux(pcms, fw_cfg);
1844     }
1845 
1846     for (i = 0; i < nb_option_roms; i++) {
1847         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1848     }
1849     pcms->fw_cfg = fw_cfg;
1850 
1851     /* Init default IOAPIC address space */
1852     pcms->ioapic_as = &address_space_memory;
1853 }
1854 
1855 /*
1856  * The 64bit pci hole starts after "above 4G RAM" and
1857  * potentially the space reserved for memory hotplug.
1858  */
1859 uint64_t pc_pci_hole64_start(void)
1860 {
1861     PCMachineState *pcms = PC_MACHINE(qdev_get_machine());
1862     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1863     MachineState *ms = MACHINE(pcms);
1864     uint64_t hole64_start = 0;
1865 
1866     if (pcmc->has_reserved_memory && ms->device_memory->base) {
1867         hole64_start = ms->device_memory->base;
1868         if (!pcmc->broken_reserved_end) {
1869             hole64_start += memory_region_size(&ms->device_memory->mr);
1870         }
1871     } else {
1872         hole64_start = 0x100000000ULL + pcms->above_4g_mem_size;
1873     }
1874 
1875     return ROUND_UP(hole64_start, 1 * GiB);
1876 }
1877 
1878 qemu_irq pc_allocate_cpu_irq(void)
1879 {
1880     return qemu_allocate_irq(pic_irq_request, NULL, 0);
1881 }
1882 
1883 DeviceState *pc_vga_init(ISABus *isa_bus, PCIBus *pci_bus)
1884 {
1885     DeviceState *dev = NULL;
1886 
1887     rom_set_order_override(FW_CFG_ORDER_OVERRIDE_VGA);
1888     if (pci_bus) {
1889         PCIDevice *pcidev = pci_vga_init(pci_bus);
1890         dev = pcidev ? &pcidev->qdev : NULL;
1891     } else if (isa_bus) {
1892         ISADevice *isadev = isa_vga_init(isa_bus);
1893         dev = isadev ? DEVICE(isadev) : NULL;
1894     }
1895     rom_reset_order_override();
1896     return dev;
1897 }
1898 
1899 static const MemoryRegionOps ioport80_io_ops = {
1900     .write = ioport80_write,
1901     .read = ioport80_read,
1902     .endianness = DEVICE_NATIVE_ENDIAN,
1903     .impl = {
1904         .min_access_size = 1,
1905         .max_access_size = 1,
1906     },
1907 };
1908 
1909 static const MemoryRegionOps ioportF0_io_ops = {
1910     .write = ioportF0_write,
1911     .read = ioportF0_read,
1912     .endianness = DEVICE_NATIVE_ENDIAN,
1913     .impl = {
1914         .min_access_size = 1,
1915         .max_access_size = 1,
1916     },
1917 };
1918 
1919 static void pc_superio_init(ISABus *isa_bus, bool create_fdctrl, bool no_vmport)
1920 {
1921     int i;
1922     DriveInfo *fd[MAX_FD];
1923     qemu_irq *a20_line;
1924     ISADevice *i8042, *port92, *vmmouse;
1925 
1926     serial_hds_isa_init(isa_bus, 0, MAX_ISA_SERIAL_PORTS);
1927     parallel_hds_isa_init(isa_bus, MAX_PARALLEL_PORTS);
1928 
1929     for (i = 0; i < MAX_FD; i++) {
1930         fd[i] = drive_get(IF_FLOPPY, 0, i);
1931         create_fdctrl |= !!fd[i];
1932     }
1933     if (create_fdctrl) {
1934         fdctrl_init_isa(isa_bus, fd);
1935     }
1936 
1937     i8042 = isa_create_simple(isa_bus, "i8042");
1938     if (!no_vmport) {
1939         vmport_init(isa_bus);
1940         vmmouse = isa_try_create(isa_bus, "vmmouse");
1941     } else {
1942         vmmouse = NULL;
1943     }
1944     if (vmmouse) {
1945         DeviceState *dev = DEVICE(vmmouse);
1946         qdev_prop_set_ptr(dev, "ps2_mouse", i8042);
1947         qdev_init_nofail(dev);
1948     }
1949     port92 = isa_create_simple(isa_bus, "port92");
1950 
1951     a20_line = qemu_allocate_irqs(handle_a20_line_change, first_cpu, 2);
1952     i8042_setup_a20_line(i8042, a20_line[0]);
1953     port92_init(port92, a20_line[1]);
1954     g_free(a20_line);
1955 }
1956 
1957 void pc_basic_device_init(ISABus *isa_bus, qemu_irq *gsi,
1958                           ISADevice **rtc_state,
1959                           bool create_fdctrl,
1960                           bool no_vmport,
1961                           bool has_pit,
1962                           uint32_t hpet_irqs)
1963 {
1964     int i;
1965     DeviceState *hpet = NULL;
1966     int pit_isa_irq = 0;
1967     qemu_irq pit_alt_irq = NULL;
1968     qemu_irq rtc_irq = NULL;
1969     ISADevice *pit = NULL;
1970     MemoryRegion *ioport80_io = g_new(MemoryRegion, 1);
1971     MemoryRegion *ioportF0_io = g_new(MemoryRegion, 1);
1972 
1973     memory_region_init_io(ioport80_io, NULL, &ioport80_io_ops, NULL, "ioport80", 1);
1974     memory_region_add_subregion(isa_bus->address_space_io, 0x80, ioport80_io);
1975 
1976     memory_region_init_io(ioportF0_io, NULL, &ioportF0_io_ops, NULL, "ioportF0", 1);
1977     memory_region_add_subregion(isa_bus->address_space_io, 0xf0, ioportF0_io);
1978 
1979     /*
1980      * Check if an HPET shall be created.
1981      *
1982      * Without KVM_CAP_PIT_STATE2, we cannot switch off the in-kernel PIT
1983      * when the HPET wants to take over. Thus we have to disable the latter.
1984      */
1985     if (!no_hpet && (!kvm_irqchip_in_kernel() || kvm_has_pit_state2())) {
1986         /* In order to set property, here not using sysbus_try_create_simple */
1987         hpet = qdev_try_create(NULL, TYPE_HPET);
1988         if (hpet) {
1989             /* For pc-piix-*, hpet's intcap is always IRQ2. For pc-q35-1.7
1990              * and earlier, use IRQ2 for compat. Otherwise, use IRQ16~23,
1991              * IRQ8 and IRQ2.
1992              */
1993             uint8_t compat = object_property_get_uint(OBJECT(hpet),
1994                     HPET_INTCAP, NULL);
1995             if (!compat) {
1996                 qdev_prop_set_uint32(hpet, HPET_INTCAP, hpet_irqs);
1997             }
1998             qdev_init_nofail(hpet);
1999             sysbus_mmio_map(SYS_BUS_DEVICE(hpet), 0, HPET_BASE);
2000 
2001             for (i = 0; i < GSI_NUM_PINS; i++) {
2002                 sysbus_connect_irq(SYS_BUS_DEVICE(hpet), i, gsi[i]);
2003             }
2004             pit_isa_irq = -1;
2005             pit_alt_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_PIT_INT);
2006             rtc_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_RTC_INT);
2007         }
2008     }
2009     *rtc_state = mc146818_rtc_init(isa_bus, 2000, rtc_irq);
2010 
2011     qemu_register_boot_set(pc_boot_set, *rtc_state);
2012 
2013     if (!xen_enabled() && has_pit) {
2014         if (kvm_pit_in_kernel()) {
2015             pit = kvm_pit_init(isa_bus, 0x40);
2016         } else {
2017             pit = i8254_pit_init(isa_bus, 0x40, pit_isa_irq, pit_alt_irq);
2018         }
2019         if (hpet) {
2020             /* connect PIT to output control line of the HPET */
2021             qdev_connect_gpio_out(hpet, 0, qdev_get_gpio_in(DEVICE(pit), 0));
2022         }
2023         pcspk_init(isa_bus, pit);
2024     }
2025 
2026     i8257_dma_init(isa_bus, 0);
2027 
2028     /* Super I/O */
2029     pc_superio_init(isa_bus, create_fdctrl, no_vmport);
2030 }
2031 
2032 void pc_nic_init(PCMachineClass *pcmc, ISABus *isa_bus, PCIBus *pci_bus)
2033 {
2034     int i;
2035 
2036     rom_set_order_override(FW_CFG_ORDER_OVERRIDE_NIC);
2037     for (i = 0; i < nb_nics; i++) {
2038         NICInfo *nd = &nd_table[i];
2039         const char *model = nd->model ? nd->model : pcmc->default_nic_model;
2040 
2041         if (g_str_equal(model, "ne2k_isa")) {
2042             pc_init_ne2k_isa(isa_bus, nd);
2043         } else {
2044             pci_nic_init_nofail(nd, pci_bus, model, NULL);
2045         }
2046     }
2047     rom_reset_order_override();
2048 }
2049 
2050 void ioapic_init_gsi(GSIState *gsi_state, const char *parent_name)
2051 {
2052     DeviceState *dev;
2053     SysBusDevice *d;
2054     unsigned int i;
2055 
2056     if (kvm_ioapic_in_kernel()) {
2057         dev = qdev_create(NULL, TYPE_KVM_IOAPIC);
2058     } else {
2059         dev = qdev_create(NULL, TYPE_IOAPIC);
2060     }
2061     if (parent_name) {
2062         object_property_add_child(object_resolve_path(parent_name, NULL),
2063                                   "ioapic", OBJECT(dev), NULL);
2064     }
2065     qdev_init_nofail(dev);
2066     d = SYS_BUS_DEVICE(dev);
2067     sysbus_mmio_map(d, 0, IO_APIC_DEFAULT_ADDRESS);
2068 
2069     for (i = 0; i < IOAPIC_NUM_PINS; i++) {
2070         gsi_state->ioapic_irq[i] = qdev_get_gpio_in(dev, i);
2071     }
2072 }
2073 
2074 static void pc_memory_pre_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
2075                                Error **errp)
2076 {
2077     const PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2078     const PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
2079     const MachineState *ms = MACHINE(hotplug_dev);
2080     const bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2081     const uint64_t legacy_align = TARGET_PAGE_SIZE;
2082     Error *local_err = NULL;
2083 
2084     /*
2085      * When -no-acpi is used with Q35 machine type, no ACPI is built,
2086      * but pcms->acpi_dev is still created. Check !acpi_enabled in
2087      * addition to cover this case.
2088      */
2089     if (!pcms->acpi_dev || !acpi_enabled) {
2090         error_setg(errp,
2091                    "memory hotplug is not enabled: missing acpi device or acpi disabled");
2092         return;
2093     }
2094 
2095     if (is_nvdimm && !ms->nvdimms_state->is_enabled) {
2096         error_setg(errp, "nvdimm is not enabled: missing 'nvdimm' in '-M'");
2097         return;
2098     }
2099 
2100     hotplug_handler_pre_plug(pcms->acpi_dev, dev, &local_err);
2101     if (local_err) {
2102         error_propagate(errp, local_err);
2103         return;
2104     }
2105 
2106     pc_dimm_pre_plug(PC_DIMM(dev), MACHINE(hotplug_dev),
2107                      pcmc->enforce_aligned_dimm ? NULL : &legacy_align, errp);
2108 }
2109 
2110 static void pc_memory_plug(HotplugHandler *hotplug_dev,
2111                            DeviceState *dev, Error **errp)
2112 {
2113     Error *local_err = NULL;
2114     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2115     MachineState *ms = MACHINE(hotplug_dev);
2116     bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2117 
2118     pc_dimm_plug(PC_DIMM(dev), MACHINE(pcms), &local_err);
2119     if (local_err) {
2120         goto out;
2121     }
2122 
2123     if (is_nvdimm) {
2124         nvdimm_plug(ms->nvdimms_state);
2125     }
2126 
2127     hotplug_handler_plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &error_abort);
2128 out:
2129     error_propagate(errp, local_err);
2130 }
2131 
2132 static void pc_memory_unplug_request(HotplugHandler *hotplug_dev,
2133                                      DeviceState *dev, Error **errp)
2134 {
2135     Error *local_err = NULL;
2136     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2137 
2138     /*
2139      * When -no-acpi is used with Q35 machine type, no ACPI is built,
2140      * but pcms->acpi_dev is still created. Check !acpi_enabled in
2141      * addition to cover this case.
2142      */
2143     if (!pcms->acpi_dev || !acpi_enabled) {
2144         error_setg(&local_err,
2145                    "memory hotplug is not enabled: missing acpi device or acpi disabled");
2146         goto out;
2147     }
2148 
2149     if (object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM)) {
2150         error_setg(&local_err,
2151                    "nvdimm device hot unplug is not supported yet.");
2152         goto out;
2153     }
2154 
2155     hotplug_handler_unplug_request(HOTPLUG_HANDLER(pcms->acpi_dev), dev,
2156                                    &local_err);
2157 out:
2158     error_propagate(errp, local_err);
2159 }
2160 
2161 static void pc_memory_unplug(HotplugHandler *hotplug_dev,
2162                              DeviceState *dev, Error **errp)
2163 {
2164     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2165     Error *local_err = NULL;
2166 
2167     hotplug_handler_unplug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
2168     if (local_err) {
2169         goto out;
2170     }
2171 
2172     pc_dimm_unplug(PC_DIMM(dev), MACHINE(pcms));
2173     object_property_set_bool(OBJECT(dev), false, "realized", NULL);
2174  out:
2175     error_propagate(errp, local_err);
2176 }
2177 
2178 static int pc_apic_cmp(const void *a, const void *b)
2179 {
2180    CPUArchId *apic_a = (CPUArchId *)a;
2181    CPUArchId *apic_b = (CPUArchId *)b;
2182 
2183    return apic_a->arch_id - apic_b->arch_id;
2184 }
2185 
2186 /* returns pointer to CPUArchId descriptor that matches CPU's apic_id
2187  * in ms->possible_cpus->cpus, if ms->possible_cpus->cpus has no
2188  * entry corresponding to CPU's apic_id returns NULL.
2189  */
2190 static CPUArchId *pc_find_cpu_slot(MachineState *ms, uint32_t id, int *idx)
2191 {
2192     CPUArchId apic_id, *found_cpu;
2193 
2194     apic_id.arch_id = id;
2195     found_cpu = bsearch(&apic_id, ms->possible_cpus->cpus,
2196         ms->possible_cpus->len, sizeof(*ms->possible_cpus->cpus),
2197         pc_apic_cmp);
2198     if (found_cpu && idx) {
2199         *idx = found_cpu - ms->possible_cpus->cpus;
2200     }
2201     return found_cpu;
2202 }
2203 
2204 static void pc_cpu_plug(HotplugHandler *hotplug_dev,
2205                         DeviceState *dev, Error **errp)
2206 {
2207     CPUArchId *found_cpu;
2208     Error *local_err = NULL;
2209     X86CPU *cpu = X86_CPU(dev);
2210     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2211 
2212     if (pcms->acpi_dev) {
2213         hotplug_handler_plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
2214         if (local_err) {
2215             goto out;
2216         }
2217     }
2218 
2219     /* increment the number of CPUs */
2220     pcms->boot_cpus++;
2221     if (pcms->rtc) {
2222         rtc_set_cpus_count(pcms->rtc, pcms->boot_cpus);
2223     }
2224     if (pcms->fw_cfg) {
2225         fw_cfg_modify_i16(pcms->fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
2226     }
2227 
2228     found_cpu = pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, NULL);
2229     found_cpu->cpu = OBJECT(dev);
2230 out:
2231     error_propagate(errp, local_err);
2232 }
2233 static void pc_cpu_unplug_request_cb(HotplugHandler *hotplug_dev,
2234                                      DeviceState *dev, Error **errp)
2235 {
2236     int idx = -1;
2237     Error *local_err = NULL;
2238     X86CPU *cpu = X86_CPU(dev);
2239     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2240 
2241     if (!pcms->acpi_dev) {
2242         error_setg(&local_err, "CPU hot unplug not supported without ACPI");
2243         goto out;
2244     }
2245 
2246     pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, &idx);
2247     assert(idx != -1);
2248     if (idx == 0) {
2249         error_setg(&local_err, "Boot CPU is unpluggable");
2250         goto out;
2251     }
2252 
2253     hotplug_handler_unplug_request(HOTPLUG_HANDLER(pcms->acpi_dev), dev,
2254                                    &local_err);
2255     if (local_err) {
2256         goto out;
2257     }
2258 
2259  out:
2260     error_propagate(errp, local_err);
2261 
2262 }
2263 
2264 static void pc_cpu_unplug_cb(HotplugHandler *hotplug_dev,
2265                              DeviceState *dev, Error **errp)
2266 {
2267     CPUArchId *found_cpu;
2268     Error *local_err = NULL;
2269     X86CPU *cpu = X86_CPU(dev);
2270     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2271 
2272     hotplug_handler_unplug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
2273     if (local_err) {
2274         goto out;
2275     }
2276 
2277     found_cpu = pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, NULL);
2278     found_cpu->cpu = NULL;
2279     object_property_set_bool(OBJECT(dev), false, "realized", NULL);
2280 
2281     /* decrement the number of CPUs */
2282     pcms->boot_cpus--;
2283     /* Update the number of CPUs in CMOS */
2284     rtc_set_cpus_count(pcms->rtc, pcms->boot_cpus);
2285     fw_cfg_modify_i16(pcms->fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
2286  out:
2287     error_propagate(errp, local_err);
2288 }
2289 
2290 static void pc_cpu_pre_plug(HotplugHandler *hotplug_dev,
2291                             DeviceState *dev, Error **errp)
2292 {
2293     int idx;
2294     CPUState *cs;
2295     CPUArchId *cpu_slot;
2296     X86CPUTopoInfo topo;
2297     X86CPU *cpu = X86_CPU(dev);
2298     MachineState *ms = MACHINE(hotplug_dev);
2299     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2300 
2301     if(!object_dynamic_cast(OBJECT(cpu), ms->cpu_type)) {
2302         error_setg(errp, "Invalid CPU type, expected cpu type: '%s'",
2303                    ms->cpu_type);
2304         return;
2305     }
2306 
2307     /* if APIC ID is not set, set it based on socket/core/thread properties */
2308     if (cpu->apic_id == UNASSIGNED_APIC_ID) {
2309         int max_socket = (max_cpus - 1) / smp_threads / smp_cores;
2310 
2311         if (cpu->socket_id < 0) {
2312             error_setg(errp, "CPU socket-id is not set");
2313             return;
2314         } else if (cpu->socket_id > max_socket) {
2315             error_setg(errp, "Invalid CPU socket-id: %u must be in range 0:%u",
2316                        cpu->socket_id, max_socket);
2317             return;
2318         }
2319         if (cpu->core_id < 0) {
2320             error_setg(errp, "CPU core-id is not set");
2321             return;
2322         } else if (cpu->core_id > (smp_cores - 1)) {
2323             error_setg(errp, "Invalid CPU core-id: %u must be in range 0:%u",
2324                        cpu->core_id, smp_cores - 1);
2325             return;
2326         }
2327         if (cpu->thread_id < 0) {
2328             error_setg(errp, "CPU thread-id is not set");
2329             return;
2330         } else if (cpu->thread_id > (smp_threads - 1)) {
2331             error_setg(errp, "Invalid CPU thread-id: %u must be in range 0:%u",
2332                        cpu->thread_id, smp_threads - 1);
2333             return;
2334         }
2335 
2336         topo.pkg_id = cpu->socket_id;
2337         topo.core_id = cpu->core_id;
2338         topo.smt_id = cpu->thread_id;
2339         cpu->apic_id = apicid_from_topo_ids(smp_cores, smp_threads, &topo);
2340     }
2341 
2342     cpu_slot = pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, &idx);
2343     if (!cpu_slot) {
2344         MachineState *ms = MACHINE(pcms);
2345 
2346         x86_topo_ids_from_apicid(cpu->apic_id, smp_cores, smp_threads, &topo);
2347         error_setg(errp, "Invalid CPU [socket: %u, core: %u, thread: %u] with"
2348                   " APIC ID %" PRIu32 ", valid index range 0:%d",
2349                    topo.pkg_id, topo.core_id, topo.smt_id, cpu->apic_id,
2350                    ms->possible_cpus->len - 1);
2351         return;
2352     }
2353 
2354     if (cpu_slot->cpu) {
2355         error_setg(errp, "CPU[%d] with APIC ID %" PRIu32 " exists",
2356                    idx, cpu->apic_id);
2357         return;
2358     }
2359 
2360     /* if 'address' properties socket-id/core-id/thread-id are not set, set them
2361      * so that machine_query_hotpluggable_cpus would show correct values
2362      */
2363     /* TODO: move socket_id/core_id/thread_id checks into x86_cpu_realizefn()
2364      * once -smp refactoring is complete and there will be CPU private
2365      * CPUState::nr_cores and CPUState::nr_threads fields instead of globals */
2366     x86_topo_ids_from_apicid(cpu->apic_id, smp_cores, smp_threads, &topo);
2367     if (cpu->socket_id != -1 && cpu->socket_id != topo.pkg_id) {
2368         error_setg(errp, "property socket-id: %u doesn't match set apic-id:"
2369             " 0x%x (socket-id: %u)", cpu->socket_id, cpu->apic_id, topo.pkg_id);
2370         return;
2371     }
2372     cpu->socket_id = topo.pkg_id;
2373 
2374     if (cpu->core_id != -1 && cpu->core_id != topo.core_id) {
2375         error_setg(errp, "property core-id: %u doesn't match set apic-id:"
2376             " 0x%x (core-id: %u)", cpu->core_id, cpu->apic_id, topo.core_id);
2377         return;
2378     }
2379     cpu->core_id = topo.core_id;
2380 
2381     if (cpu->thread_id != -1 && cpu->thread_id != topo.smt_id) {
2382         error_setg(errp, "property thread-id: %u doesn't match set apic-id:"
2383             " 0x%x (thread-id: %u)", cpu->thread_id, cpu->apic_id, topo.smt_id);
2384         return;
2385     }
2386     cpu->thread_id = topo.smt_id;
2387 
2388     if (cpu->hyperv_vpindex && !kvm_hv_vpindex_settable()) {
2389         error_setg(errp, "kernel doesn't allow setting HyperV VP_INDEX");
2390         return;
2391     }
2392 
2393     cs = CPU(cpu);
2394     cs->cpu_index = idx;
2395 
2396     numa_cpu_pre_plug(cpu_slot, dev, errp);
2397 }
2398 
2399 static void pc_machine_device_pre_plug_cb(HotplugHandler *hotplug_dev,
2400                                           DeviceState *dev, Error **errp)
2401 {
2402     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2403         pc_memory_pre_plug(hotplug_dev, dev, errp);
2404     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2405         pc_cpu_pre_plug(hotplug_dev, dev, errp);
2406     }
2407 }
2408 
2409 static void pc_machine_device_plug_cb(HotplugHandler *hotplug_dev,
2410                                       DeviceState *dev, Error **errp)
2411 {
2412     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2413         pc_memory_plug(hotplug_dev, dev, errp);
2414     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2415         pc_cpu_plug(hotplug_dev, dev, errp);
2416     }
2417 }
2418 
2419 static void pc_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
2420                                                 DeviceState *dev, Error **errp)
2421 {
2422     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2423         pc_memory_unplug_request(hotplug_dev, dev, errp);
2424     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2425         pc_cpu_unplug_request_cb(hotplug_dev, dev, errp);
2426     } else {
2427         error_setg(errp, "acpi: device unplug request for not supported device"
2428                    " type: %s", object_get_typename(OBJECT(dev)));
2429     }
2430 }
2431 
2432 static void pc_machine_device_unplug_cb(HotplugHandler *hotplug_dev,
2433                                         DeviceState *dev, Error **errp)
2434 {
2435     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2436         pc_memory_unplug(hotplug_dev, dev, errp);
2437     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2438         pc_cpu_unplug_cb(hotplug_dev, dev, errp);
2439     } else {
2440         error_setg(errp, "acpi: device unplug for not supported device"
2441                    " type: %s", object_get_typename(OBJECT(dev)));
2442     }
2443 }
2444 
2445 static HotplugHandler *pc_get_hotplug_handler(MachineState *machine,
2446                                              DeviceState *dev)
2447 {
2448     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) ||
2449         object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2450         return HOTPLUG_HANDLER(machine);
2451     }
2452 
2453     return NULL;
2454 }
2455 
2456 static void
2457 pc_machine_get_device_memory_region_size(Object *obj, Visitor *v,
2458                                          const char *name, void *opaque,
2459                                          Error **errp)
2460 {
2461     MachineState *ms = MACHINE(obj);
2462     int64_t value = memory_region_size(&ms->device_memory->mr);
2463 
2464     visit_type_int(v, name, &value, errp);
2465 }
2466 
2467 static void pc_machine_get_max_ram_below_4g(Object *obj, Visitor *v,
2468                                             const char *name, void *opaque,
2469                                             Error **errp)
2470 {
2471     PCMachineState *pcms = PC_MACHINE(obj);
2472     uint64_t value = pcms->max_ram_below_4g;
2473 
2474     visit_type_size(v, name, &value, errp);
2475 }
2476 
2477 static void pc_machine_set_max_ram_below_4g(Object *obj, Visitor *v,
2478                                             const char *name, void *opaque,
2479                                             Error **errp)
2480 {
2481     PCMachineState *pcms = PC_MACHINE(obj);
2482     Error *error = NULL;
2483     uint64_t value;
2484 
2485     visit_type_size(v, name, &value, &error);
2486     if (error) {
2487         error_propagate(errp, error);
2488         return;
2489     }
2490     if (value > 4 * GiB) {
2491         error_setg(&error,
2492                    "Machine option 'max-ram-below-4g=%"PRIu64
2493                    "' expects size less than or equal to 4G", value);
2494         error_propagate(errp, error);
2495         return;
2496     }
2497 
2498     if (value < 1 * MiB) {
2499         warn_report("Only %" PRIu64 " bytes of RAM below the 4GiB boundary,"
2500                     "BIOS may not work with less than 1MiB", value);
2501     }
2502 
2503     pcms->max_ram_below_4g = value;
2504 }
2505 
2506 static void pc_machine_get_vmport(Object *obj, Visitor *v, const char *name,
2507                                   void *opaque, Error **errp)
2508 {
2509     PCMachineState *pcms = PC_MACHINE(obj);
2510     OnOffAuto vmport = pcms->vmport;
2511 
2512     visit_type_OnOffAuto(v, name, &vmport, errp);
2513 }
2514 
2515 static void pc_machine_set_vmport(Object *obj, Visitor *v, const char *name,
2516                                   void *opaque, Error **errp)
2517 {
2518     PCMachineState *pcms = PC_MACHINE(obj);
2519 
2520     visit_type_OnOffAuto(v, name, &pcms->vmport, errp);
2521 }
2522 
2523 bool pc_machine_is_smm_enabled(PCMachineState *pcms)
2524 {
2525     bool smm_available = false;
2526 
2527     if (pcms->smm == ON_OFF_AUTO_OFF) {
2528         return false;
2529     }
2530 
2531     if (tcg_enabled() || qtest_enabled()) {
2532         smm_available = true;
2533     } else if (kvm_enabled()) {
2534         smm_available = kvm_has_smm();
2535     }
2536 
2537     if (smm_available) {
2538         return true;
2539     }
2540 
2541     if (pcms->smm == ON_OFF_AUTO_ON) {
2542         error_report("System Management Mode not supported by this hypervisor.");
2543         exit(1);
2544     }
2545     return false;
2546 }
2547 
2548 static void pc_machine_get_smm(Object *obj, Visitor *v, const char *name,
2549                                void *opaque, Error **errp)
2550 {
2551     PCMachineState *pcms = PC_MACHINE(obj);
2552     OnOffAuto smm = pcms->smm;
2553 
2554     visit_type_OnOffAuto(v, name, &smm, errp);
2555 }
2556 
2557 static void pc_machine_set_smm(Object *obj, Visitor *v, const char *name,
2558                                void *opaque, Error **errp)
2559 {
2560     PCMachineState *pcms = PC_MACHINE(obj);
2561 
2562     visit_type_OnOffAuto(v, name, &pcms->smm, errp);
2563 }
2564 
2565 static bool pc_machine_get_smbus(Object *obj, Error **errp)
2566 {
2567     PCMachineState *pcms = PC_MACHINE(obj);
2568 
2569     return pcms->smbus_enabled;
2570 }
2571 
2572 static void pc_machine_set_smbus(Object *obj, bool value, Error **errp)
2573 {
2574     PCMachineState *pcms = PC_MACHINE(obj);
2575 
2576     pcms->smbus_enabled = value;
2577 }
2578 
2579 static bool pc_machine_get_sata(Object *obj, Error **errp)
2580 {
2581     PCMachineState *pcms = PC_MACHINE(obj);
2582 
2583     return pcms->sata_enabled;
2584 }
2585 
2586 static void pc_machine_set_sata(Object *obj, bool value, Error **errp)
2587 {
2588     PCMachineState *pcms = PC_MACHINE(obj);
2589 
2590     pcms->sata_enabled = value;
2591 }
2592 
2593 static bool pc_machine_get_pit(Object *obj, Error **errp)
2594 {
2595     PCMachineState *pcms = PC_MACHINE(obj);
2596 
2597     return pcms->pit_enabled;
2598 }
2599 
2600 static void pc_machine_set_pit(Object *obj, bool value, Error **errp)
2601 {
2602     PCMachineState *pcms = PC_MACHINE(obj);
2603 
2604     pcms->pit_enabled = value;
2605 }
2606 
2607 static void pc_machine_initfn(Object *obj)
2608 {
2609     PCMachineState *pcms = PC_MACHINE(obj);
2610 
2611     pcms->max_ram_below_4g = 0; /* use default */
2612     pcms->smm = ON_OFF_AUTO_AUTO;
2613     pcms->vmport = ON_OFF_AUTO_AUTO;
2614     /* acpi build is enabled by default if machine supports it */
2615     pcms->acpi_build_enabled = PC_MACHINE_GET_CLASS(pcms)->has_acpi_build;
2616     pcms->smbus_enabled = true;
2617     pcms->sata_enabled = true;
2618     pcms->pit_enabled = true;
2619 
2620     pc_system_flash_create(pcms);
2621 }
2622 
2623 static void pc_machine_reset(void)
2624 {
2625     CPUState *cs;
2626     X86CPU *cpu;
2627 
2628     qemu_devices_reset();
2629 
2630     /* Reset APIC after devices have been reset to cancel
2631      * any changes that qemu_devices_reset() might have done.
2632      */
2633     CPU_FOREACH(cs) {
2634         cpu = X86_CPU(cs);
2635 
2636         if (cpu->apic_state) {
2637             device_reset(cpu->apic_state);
2638         }
2639     }
2640 }
2641 
2642 static CpuInstanceProperties
2643 pc_cpu_index_to_props(MachineState *ms, unsigned cpu_index)
2644 {
2645     MachineClass *mc = MACHINE_GET_CLASS(ms);
2646     const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(ms);
2647 
2648     assert(cpu_index < possible_cpus->len);
2649     return possible_cpus->cpus[cpu_index].props;
2650 }
2651 
2652 static int64_t pc_get_default_cpu_node_id(const MachineState *ms, int idx)
2653 {
2654    X86CPUTopoInfo topo;
2655 
2656    assert(idx < ms->possible_cpus->len);
2657    x86_topo_ids_from_apicid(ms->possible_cpus->cpus[idx].arch_id,
2658                             smp_cores, smp_threads, &topo);
2659    return topo.pkg_id % nb_numa_nodes;
2660 }
2661 
2662 static const CPUArchIdList *pc_possible_cpu_arch_ids(MachineState *ms)
2663 {
2664     int i;
2665 
2666     if (ms->possible_cpus) {
2667         /*
2668          * make sure that max_cpus hasn't changed since the first use, i.e.
2669          * -smp hasn't been parsed after it
2670         */
2671         assert(ms->possible_cpus->len == max_cpus);
2672         return ms->possible_cpus;
2673     }
2674 
2675     ms->possible_cpus = g_malloc0(sizeof(CPUArchIdList) +
2676                                   sizeof(CPUArchId) * max_cpus);
2677     ms->possible_cpus->len = max_cpus;
2678     for (i = 0; i < ms->possible_cpus->len; i++) {
2679         X86CPUTopoInfo topo;
2680 
2681         ms->possible_cpus->cpus[i].type = ms->cpu_type;
2682         ms->possible_cpus->cpus[i].vcpus_count = 1;
2683         ms->possible_cpus->cpus[i].arch_id = x86_cpu_apic_id_from_index(i);
2684         x86_topo_ids_from_apicid(ms->possible_cpus->cpus[i].arch_id,
2685                                  smp_cores, smp_threads, &topo);
2686         ms->possible_cpus->cpus[i].props.has_socket_id = true;
2687         ms->possible_cpus->cpus[i].props.socket_id = topo.pkg_id;
2688         ms->possible_cpus->cpus[i].props.has_core_id = true;
2689         ms->possible_cpus->cpus[i].props.core_id = topo.core_id;
2690         ms->possible_cpus->cpus[i].props.has_thread_id = true;
2691         ms->possible_cpus->cpus[i].props.thread_id = topo.smt_id;
2692     }
2693     return ms->possible_cpus;
2694 }
2695 
2696 static void x86_nmi(NMIState *n, int cpu_index, Error **errp)
2697 {
2698     /* cpu index isn't used */
2699     CPUState *cs;
2700 
2701     CPU_FOREACH(cs) {
2702         X86CPU *cpu = X86_CPU(cs);
2703 
2704         if (!cpu->apic_state) {
2705             cpu_interrupt(cs, CPU_INTERRUPT_NMI);
2706         } else {
2707             apic_deliver_nmi(cpu->apic_state);
2708         }
2709     }
2710 }
2711 
2712 static void pc_machine_class_init(ObjectClass *oc, void *data)
2713 {
2714     MachineClass *mc = MACHINE_CLASS(oc);
2715     PCMachineClass *pcmc = PC_MACHINE_CLASS(oc);
2716     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
2717     NMIClass *nc = NMI_CLASS(oc);
2718 
2719     pcmc->pci_enabled = true;
2720     pcmc->has_acpi_build = true;
2721     pcmc->rsdp_in_ram = true;
2722     pcmc->smbios_defaults = true;
2723     pcmc->smbios_uuid_encoded = true;
2724     pcmc->gigabyte_align = true;
2725     pcmc->has_reserved_memory = true;
2726     pcmc->kvmclock_enabled = true;
2727     pcmc->enforce_aligned_dimm = true;
2728     /* BIOS ACPI tables: 128K. Other BIOS datastructures: less than 4K reported
2729      * to be used at the moment, 32K should be enough for a while.  */
2730     pcmc->acpi_data_size = 0x20000 + 0x8000;
2731     pcmc->save_tsc_khz = true;
2732     pcmc->linuxboot_dma_enabled = true;
2733     pcmc->pvh_enabled = true;
2734     assert(!mc->get_hotplug_handler);
2735     mc->get_hotplug_handler = pc_get_hotplug_handler;
2736     mc->cpu_index_to_instance_props = pc_cpu_index_to_props;
2737     mc->get_default_cpu_node_id = pc_get_default_cpu_node_id;
2738     mc->possible_cpu_arch_ids = pc_possible_cpu_arch_ids;
2739     mc->auto_enable_numa_with_memhp = true;
2740     mc->has_hotpluggable_cpus = true;
2741     mc->default_boot_order = "cad";
2742     mc->hot_add_cpu = pc_hot_add_cpu;
2743     mc->block_default_type = IF_IDE;
2744     mc->max_cpus = 255;
2745     mc->reset = pc_machine_reset;
2746     hc->pre_plug = pc_machine_device_pre_plug_cb;
2747     hc->plug = pc_machine_device_plug_cb;
2748     hc->unplug_request = pc_machine_device_unplug_request_cb;
2749     hc->unplug = pc_machine_device_unplug_cb;
2750     nc->nmi_monitor_handler = x86_nmi;
2751     mc->default_cpu_type = TARGET_DEFAULT_CPU_TYPE;
2752     mc->nvdimm_supported = true;
2753 
2754     object_class_property_add(oc, PC_MACHINE_DEVMEM_REGION_SIZE, "int",
2755         pc_machine_get_device_memory_region_size, NULL,
2756         NULL, NULL, &error_abort);
2757 
2758     object_class_property_add(oc, PC_MACHINE_MAX_RAM_BELOW_4G, "size",
2759         pc_machine_get_max_ram_below_4g, pc_machine_set_max_ram_below_4g,
2760         NULL, NULL, &error_abort);
2761 
2762     object_class_property_set_description(oc, PC_MACHINE_MAX_RAM_BELOW_4G,
2763         "Maximum ram below the 4G boundary (32bit boundary)", &error_abort);
2764 
2765     object_class_property_add(oc, PC_MACHINE_SMM, "OnOffAuto",
2766         pc_machine_get_smm, pc_machine_set_smm,
2767         NULL, NULL, &error_abort);
2768     object_class_property_set_description(oc, PC_MACHINE_SMM,
2769         "Enable SMM (pc & q35)", &error_abort);
2770 
2771     object_class_property_add(oc, PC_MACHINE_VMPORT, "OnOffAuto",
2772         pc_machine_get_vmport, pc_machine_set_vmport,
2773         NULL, NULL, &error_abort);
2774     object_class_property_set_description(oc, PC_MACHINE_VMPORT,
2775         "Enable vmport (pc & q35)", &error_abort);
2776 
2777     object_class_property_add_bool(oc, PC_MACHINE_SMBUS,
2778         pc_machine_get_smbus, pc_machine_set_smbus, &error_abort);
2779 
2780     object_class_property_add_bool(oc, PC_MACHINE_SATA,
2781         pc_machine_get_sata, pc_machine_set_sata, &error_abort);
2782 
2783     object_class_property_add_bool(oc, PC_MACHINE_PIT,
2784         pc_machine_get_pit, pc_machine_set_pit, &error_abort);
2785 }
2786 
2787 static const TypeInfo pc_machine_info = {
2788     .name = TYPE_PC_MACHINE,
2789     .parent = TYPE_MACHINE,
2790     .abstract = true,
2791     .instance_size = sizeof(PCMachineState),
2792     .instance_init = pc_machine_initfn,
2793     .class_size = sizeof(PCMachineClass),
2794     .class_init = pc_machine_class_init,
2795     .interfaces = (InterfaceInfo[]) {
2796          { TYPE_HOTPLUG_HANDLER },
2797          { TYPE_NMI },
2798          { }
2799     },
2800 };
2801 
2802 static void pc_machine_register_types(void)
2803 {
2804     type_register_static(&pc_machine_info);
2805 }
2806 
2807 type_init(pc_machine_register_types)
2808