xref: /openbmc/qemu/hw/i386/pc.c (revision 200dbf37)
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[] = {};
114 const size_t pc_compat_4_0_len = G_N_ELEMENTS(pc_compat_4_0);
115 
116 GlobalProperty pc_compat_3_1[] = {
117     { "intel-iommu", "dma-drain", "off" },
118     { "Opteron_G3" "-" TYPE_X86_CPU, "rdtscp", "off" },
119     { "Opteron_G4" "-" TYPE_X86_CPU, "rdtscp", "off" },
120     { "Opteron_G4" "-" TYPE_X86_CPU, "npt", "off" },
121     { "Opteron_G4" "-" TYPE_X86_CPU, "nrip-save", "off" },
122     { "Opteron_G5" "-" TYPE_X86_CPU, "rdtscp", "off" },
123     { "Opteron_G5" "-" TYPE_X86_CPU, "npt", "off" },
124     { "Opteron_G5" "-" TYPE_X86_CPU, "nrip-save", "off" },
125     { "EPYC" "-" TYPE_X86_CPU, "npt", "off" },
126     { "EPYC" "-" TYPE_X86_CPU, "nrip-save", "off" },
127     { "EPYC-IBPB" "-" TYPE_X86_CPU, "npt", "off" },
128     { "EPYC-IBPB" "-" TYPE_X86_CPU, "nrip-save", "off" },
129     { "Skylake-Client" "-" TYPE_X86_CPU,      "mpx", "on" },
130     { "Skylake-Client-IBRS" "-" TYPE_X86_CPU, "mpx", "on" },
131     { "Skylake-Server" "-" TYPE_X86_CPU,      "mpx", "on" },
132     { "Skylake-Server-IBRS" "-" TYPE_X86_CPU, "mpx", "on" },
133     { "Cascadelake-Server" "-" TYPE_X86_CPU,  "mpx", "on" },
134     { "Icelake-Client" "-" TYPE_X86_CPU,      "mpx", "on" },
135     { "Icelake-Server" "-" TYPE_X86_CPU,      "mpx", "on" },
136     { "Cascadelake-Server" "-" TYPE_X86_CPU, "stepping", "5" },
137     { TYPE_X86_CPU, "x-intel-pt-auto-level", "off" },
138 };
139 const size_t pc_compat_3_1_len = G_N_ELEMENTS(pc_compat_3_1);
140 
141 GlobalProperty pc_compat_3_0[] = {
142     { TYPE_X86_CPU, "x-hv-synic-kvm-only", "on" },
143     { "Skylake-Server" "-" TYPE_X86_CPU, "pku", "off" },
144     { "Skylake-Server-IBRS" "-" TYPE_X86_CPU, "pku", "off" },
145 };
146 const size_t pc_compat_3_0_len = G_N_ELEMENTS(pc_compat_3_0);
147 
148 GlobalProperty pc_compat_2_12[] = {
149     { TYPE_X86_CPU, "legacy-cache", "on" },
150     { TYPE_X86_CPU, "topoext", "off" },
151     { "EPYC-" TYPE_X86_CPU, "xlevel", "0x8000000a" },
152     { "EPYC-IBPB-" TYPE_X86_CPU, "xlevel", "0x8000000a" },
153 };
154 const size_t pc_compat_2_12_len = G_N_ELEMENTS(pc_compat_2_12);
155 
156 GlobalProperty pc_compat_2_11[] = {
157     { TYPE_X86_CPU, "x-migrate-smi-count", "off" },
158     { "Skylake-Server" "-" TYPE_X86_CPU, "clflushopt", "off" },
159 };
160 const size_t pc_compat_2_11_len = G_N_ELEMENTS(pc_compat_2_11);
161 
162 GlobalProperty pc_compat_2_10[] = {
163     { TYPE_X86_CPU, "x-hv-max-vps", "0x40" },
164     { "i440FX-pcihost", "x-pci-hole64-fix", "off" },
165     { "q35-pcihost", "x-pci-hole64-fix", "off" },
166 };
167 const size_t pc_compat_2_10_len = G_N_ELEMENTS(pc_compat_2_10);
168 
169 GlobalProperty pc_compat_2_9[] = {
170     { "mch", "extended-tseg-mbytes", "0" },
171 };
172 const size_t pc_compat_2_9_len = G_N_ELEMENTS(pc_compat_2_9);
173 
174 GlobalProperty pc_compat_2_8[] = {
175     { TYPE_X86_CPU, "tcg-cpuid", "off" },
176     { "kvmclock", "x-mach-use-reliable-get-clock", "off" },
177     { "ICH9-LPC", "x-smi-broadcast", "off" },
178     { TYPE_X86_CPU, "vmware-cpuid-freq", "off" },
179     { "Haswell-" TYPE_X86_CPU, "stepping", "1" },
180 };
181 const size_t pc_compat_2_8_len = G_N_ELEMENTS(pc_compat_2_8);
182 
183 GlobalProperty pc_compat_2_7[] = {
184     { TYPE_X86_CPU, "l3-cache", "off" },
185     { TYPE_X86_CPU, "full-cpuid-auto-level", "off" },
186     { "Opteron_G3" "-" TYPE_X86_CPU, "family", "15" },
187     { "Opteron_G3" "-" TYPE_X86_CPU, "model", "6" },
188     { "Opteron_G3" "-" TYPE_X86_CPU, "stepping", "1" },
189     { "isa-pcspk", "migrate", "off" },
190 };
191 const size_t pc_compat_2_7_len = G_N_ELEMENTS(pc_compat_2_7);
192 
193 GlobalProperty pc_compat_2_6[] = {
194     { TYPE_X86_CPU, "cpuid-0xb", "off" },
195     { "vmxnet3", "romfile", "" },
196     { TYPE_X86_CPU, "fill-mtrr-mask", "off" },
197     { "apic-common", "legacy-instance-id", "on", }
198 };
199 const size_t pc_compat_2_6_len = G_N_ELEMENTS(pc_compat_2_6);
200 
201 GlobalProperty pc_compat_2_5[] = {};
202 const size_t pc_compat_2_5_len = G_N_ELEMENTS(pc_compat_2_5);
203 
204 GlobalProperty pc_compat_2_4[] = {
205     PC_CPU_MODEL_IDS("2.4.0")
206     { "Haswell-" TYPE_X86_CPU, "abm", "off" },
207     { "Haswell-noTSX-" TYPE_X86_CPU, "abm", "off" },
208     { "Broadwell-" TYPE_X86_CPU, "abm", "off" },
209     { "Broadwell-noTSX-" TYPE_X86_CPU, "abm", "off" },
210     { "host" "-" TYPE_X86_CPU, "host-cache-info", "on" },
211     { TYPE_X86_CPU, "check", "off" },
212     { "qemu64" "-" TYPE_X86_CPU, "sse4a", "on" },
213     { "qemu64" "-" TYPE_X86_CPU, "abm", "on" },
214     { "qemu64" "-" TYPE_X86_CPU, "popcnt", "on" },
215     { "qemu32" "-" TYPE_X86_CPU, "popcnt", "on" },
216     { "Opteron_G2" "-" TYPE_X86_CPU, "rdtscp", "on" },
217     { "Opteron_G3" "-" TYPE_X86_CPU, "rdtscp", "on" },
218     { "Opteron_G4" "-" TYPE_X86_CPU, "rdtscp", "on" },
219     { "Opteron_G5" "-" TYPE_X86_CPU, "rdtscp", "on", }
220 };
221 const size_t pc_compat_2_4_len = G_N_ELEMENTS(pc_compat_2_4);
222 
223 GlobalProperty pc_compat_2_3[] = {
224     PC_CPU_MODEL_IDS("2.3.0")
225     { TYPE_X86_CPU, "arat", "off" },
226     { "qemu64" "-" TYPE_X86_CPU, "min-level", "4" },
227     { "kvm64" "-" TYPE_X86_CPU, "min-level", "5" },
228     { "pentium3" "-" TYPE_X86_CPU, "min-level", "2" },
229     { "n270" "-" TYPE_X86_CPU, "min-level", "5" },
230     { "Conroe" "-" TYPE_X86_CPU, "min-level", "4" },
231     { "Penryn" "-" TYPE_X86_CPU, "min-level", "4" },
232     { "Nehalem" "-" TYPE_X86_CPU, "min-level", "4" },
233     { "n270" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
234     { "Penryn" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
235     { "Conroe" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
236     { "Nehalem" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
237     { "Westmere" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
238     { "SandyBridge" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
239     { "IvyBridge" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
240     { "Haswell" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
241     { "Haswell-noTSX" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
242     { "Broadwell" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
243     { "Broadwell-noTSX" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
244     { TYPE_X86_CPU, "kvm-no-smi-migration", "on" },
245 };
246 const size_t pc_compat_2_3_len = G_N_ELEMENTS(pc_compat_2_3);
247 
248 GlobalProperty pc_compat_2_2[] = {
249     PC_CPU_MODEL_IDS("2.2.0")
250     { "kvm64" "-" TYPE_X86_CPU, "vme", "off" },
251     { "kvm32" "-" TYPE_X86_CPU, "vme", "off" },
252     { "Conroe" "-" TYPE_X86_CPU, "vme", "off" },
253     { "Penryn" "-" TYPE_X86_CPU, "vme", "off" },
254     { "Nehalem" "-" TYPE_X86_CPU, "vme", "off" },
255     { "Westmere" "-" TYPE_X86_CPU, "vme", "off" },
256     { "SandyBridge" "-" TYPE_X86_CPU, "vme", "off" },
257     { "Haswell" "-" TYPE_X86_CPU, "vme", "off" },
258     { "Broadwell" "-" TYPE_X86_CPU, "vme", "off" },
259     { "Opteron_G1" "-" TYPE_X86_CPU, "vme", "off" },
260     { "Opteron_G2" "-" TYPE_X86_CPU, "vme", "off" },
261     { "Opteron_G3" "-" TYPE_X86_CPU, "vme", "off" },
262     { "Opteron_G4" "-" TYPE_X86_CPU, "vme", "off" },
263     { "Opteron_G5" "-" TYPE_X86_CPU, "vme", "off" },
264     { "Haswell" "-" TYPE_X86_CPU, "f16c", "off" },
265     { "Haswell" "-" TYPE_X86_CPU, "rdrand", "off" },
266     { "Broadwell" "-" TYPE_X86_CPU, "f16c", "off" },
267     { "Broadwell" "-" TYPE_X86_CPU, "rdrand", "off" },
268 };
269 const size_t pc_compat_2_2_len = G_N_ELEMENTS(pc_compat_2_2);
270 
271 GlobalProperty pc_compat_2_1[] = {
272     PC_CPU_MODEL_IDS("2.1.0")
273     { "coreduo" "-" TYPE_X86_CPU, "vmx", "on" },
274     { "core2duo" "-" TYPE_X86_CPU, "vmx", "on" },
275 };
276 const size_t pc_compat_2_1_len = G_N_ELEMENTS(pc_compat_2_1);
277 
278 GlobalProperty pc_compat_2_0[] = {
279     PC_CPU_MODEL_IDS("2.0.0")
280     { "virtio-scsi-pci", "any_layout", "off" },
281     { "PIIX4_PM", "memory-hotplug-support", "off" },
282     { "apic", "version", "0x11" },
283     { "nec-usb-xhci", "superspeed-ports-first", "off" },
284     { "nec-usb-xhci", "force-pcie-endcap", "on" },
285     { "pci-serial", "prog_if", "0" },
286     { "pci-serial-2x", "prog_if", "0" },
287     { "pci-serial-4x", "prog_if", "0" },
288     { "virtio-net-pci", "guest_announce", "off" },
289     { "ICH9-LPC", "memory-hotplug-support", "off" },
290     { "xio3130-downstream", COMPAT_PROP_PCP, "off" },
291     { "ioh3420", COMPAT_PROP_PCP, "off" },
292 };
293 const size_t pc_compat_2_0_len = G_N_ELEMENTS(pc_compat_2_0);
294 
295 GlobalProperty pc_compat_1_7[] = {
296     PC_CPU_MODEL_IDS("1.7.0")
297     { TYPE_USB_DEVICE, "msos-desc", "no" },
298     { "PIIX4_PM", "acpi-pci-hotplug-with-bridge-support", "off" },
299     { "hpet", HPET_INTCAP, "4" },
300 };
301 const size_t pc_compat_1_7_len = G_N_ELEMENTS(pc_compat_1_7);
302 
303 GlobalProperty pc_compat_1_6[] = {
304     PC_CPU_MODEL_IDS("1.6.0")
305     { "e1000", "mitigation", "off" },
306     { "qemu64-" TYPE_X86_CPU, "model", "2" },
307     { "qemu32-" TYPE_X86_CPU, "model", "3" },
308     { "i440FX-pcihost", "short_root_bus", "1" },
309     { "q35-pcihost", "short_root_bus", "1" },
310 };
311 const size_t pc_compat_1_6_len = G_N_ELEMENTS(pc_compat_1_6);
312 
313 GlobalProperty pc_compat_1_5[] = {
314     PC_CPU_MODEL_IDS("1.5.0")
315     { "Conroe-" TYPE_X86_CPU, "model", "2" },
316     { "Conroe-" TYPE_X86_CPU, "min-level", "2" },
317     { "Penryn-" TYPE_X86_CPU, "model", "2" },
318     { "Penryn-" TYPE_X86_CPU, "min-level", "2" },
319     { "Nehalem-" TYPE_X86_CPU, "model", "2" },
320     { "Nehalem-" TYPE_X86_CPU, "min-level", "2" },
321     { "virtio-net-pci", "any_layout", "off" },
322     { TYPE_X86_CPU, "pmu", "on" },
323     { "i440FX-pcihost", "short_root_bus", "0" },
324     { "q35-pcihost", "short_root_bus", "0" },
325 };
326 const size_t pc_compat_1_5_len = G_N_ELEMENTS(pc_compat_1_5);
327 
328 GlobalProperty pc_compat_1_4[] = {
329     PC_CPU_MODEL_IDS("1.4.0")
330     { "scsi-hd", "discard_granularity", "0" },
331     { "scsi-cd", "discard_granularity", "0" },
332     { "scsi-disk", "discard_granularity", "0" },
333     { "ide-hd", "discard_granularity", "0" },
334     { "ide-cd", "discard_granularity", "0" },
335     { "ide-drive", "discard_granularity", "0" },
336     { "virtio-blk-pci", "discard_granularity", "0" },
337     /* DEV_NVECTORS_UNSPECIFIED as a uint32_t string: */
338     { "virtio-serial-pci", "vectors", "0xFFFFFFFF" },
339     { "virtio-net-pci", "ctrl_guest_offloads", "off" },
340     { "e1000", "romfile", "pxe-e1000.rom" },
341     { "ne2k_pci", "romfile", "pxe-ne2k_pci.rom" },
342     { "pcnet", "romfile", "pxe-pcnet.rom" },
343     { "rtl8139", "romfile", "pxe-rtl8139.rom" },
344     { "virtio-net-pci", "romfile", "pxe-virtio.rom" },
345     { "486-" TYPE_X86_CPU, "model", "0" },
346     { "n270" "-" TYPE_X86_CPU, "movbe", "off" },
347     { "Westmere" "-" TYPE_X86_CPU, "pclmulqdq", "off" },
348 };
349 const size_t pc_compat_1_4_len = G_N_ELEMENTS(pc_compat_1_4);
350 
351 void gsi_handler(void *opaque, int n, int level)
352 {
353     GSIState *s = opaque;
354 
355     DPRINTF("pc: %s GSI %d\n", level ? "raising" : "lowering", n);
356     if (n < ISA_NUM_IRQS) {
357         qemu_set_irq(s->i8259_irq[n], level);
358     }
359     qemu_set_irq(s->ioapic_irq[n], level);
360 }
361 
362 static void ioport80_write(void *opaque, hwaddr addr, uint64_t data,
363                            unsigned size)
364 {
365 }
366 
367 static uint64_t ioport80_read(void *opaque, hwaddr addr, unsigned size)
368 {
369     return 0xffffffffffffffffULL;
370 }
371 
372 /* MSDOS compatibility mode FPU exception support */
373 static qemu_irq ferr_irq;
374 
375 void pc_register_ferr_irq(qemu_irq irq)
376 {
377     ferr_irq = irq;
378 }
379 
380 /* XXX: add IGNNE support */
381 void cpu_set_ferr(CPUX86State *s)
382 {
383     qemu_irq_raise(ferr_irq);
384 }
385 
386 static void ioportF0_write(void *opaque, hwaddr addr, uint64_t data,
387                            unsigned size)
388 {
389     qemu_irq_lower(ferr_irq);
390 }
391 
392 static uint64_t ioportF0_read(void *opaque, hwaddr addr, unsigned size)
393 {
394     return 0xffffffffffffffffULL;
395 }
396 
397 /* TSC handling */
398 uint64_t cpu_get_tsc(CPUX86State *env)
399 {
400     return cpu_get_ticks();
401 }
402 
403 /* IRQ handling */
404 int cpu_get_pic_interrupt(CPUX86State *env)
405 {
406     X86CPU *cpu = x86_env_get_cpu(env);
407     int intno;
408 
409     if (!kvm_irqchip_in_kernel()) {
410         intno = apic_get_interrupt(cpu->apic_state);
411         if (intno >= 0) {
412             return intno;
413         }
414         /* read the irq from the PIC */
415         if (!apic_accept_pic_intr(cpu->apic_state)) {
416             return -1;
417         }
418     }
419 
420     intno = pic_read_irq(isa_pic);
421     return intno;
422 }
423 
424 static void pic_irq_request(void *opaque, int irq, int level)
425 {
426     CPUState *cs = first_cpu;
427     X86CPU *cpu = X86_CPU(cs);
428 
429     DPRINTF("pic_irqs: %s irq %d\n", level? "raise" : "lower", irq);
430     if (cpu->apic_state && !kvm_irqchip_in_kernel()) {
431         CPU_FOREACH(cs) {
432             cpu = X86_CPU(cs);
433             if (apic_accept_pic_intr(cpu->apic_state)) {
434                 apic_deliver_pic_intr(cpu->apic_state, level);
435             }
436         }
437     } else {
438         if (level) {
439             cpu_interrupt(cs, CPU_INTERRUPT_HARD);
440         } else {
441             cpu_reset_interrupt(cs, CPU_INTERRUPT_HARD);
442         }
443     }
444 }
445 
446 /* PC cmos mappings */
447 
448 #define REG_EQUIPMENT_BYTE          0x14
449 
450 int cmos_get_fd_drive_type(FloppyDriveType fd0)
451 {
452     int val;
453 
454     switch (fd0) {
455     case FLOPPY_DRIVE_TYPE_144:
456         /* 1.44 Mb 3"5 drive */
457         val = 4;
458         break;
459     case FLOPPY_DRIVE_TYPE_288:
460         /* 2.88 Mb 3"5 drive */
461         val = 5;
462         break;
463     case FLOPPY_DRIVE_TYPE_120:
464         /* 1.2 Mb 5"5 drive */
465         val = 2;
466         break;
467     case FLOPPY_DRIVE_TYPE_NONE:
468     default:
469         val = 0;
470         break;
471     }
472     return val;
473 }
474 
475 static void cmos_init_hd(ISADevice *s, int type_ofs, int info_ofs,
476                          int16_t cylinders, int8_t heads, int8_t sectors)
477 {
478     rtc_set_memory(s, type_ofs, 47);
479     rtc_set_memory(s, info_ofs, cylinders);
480     rtc_set_memory(s, info_ofs + 1, cylinders >> 8);
481     rtc_set_memory(s, info_ofs + 2, heads);
482     rtc_set_memory(s, info_ofs + 3, 0xff);
483     rtc_set_memory(s, info_ofs + 4, 0xff);
484     rtc_set_memory(s, info_ofs + 5, 0xc0 | ((heads > 8) << 3));
485     rtc_set_memory(s, info_ofs + 6, cylinders);
486     rtc_set_memory(s, info_ofs + 7, cylinders >> 8);
487     rtc_set_memory(s, info_ofs + 8, sectors);
488 }
489 
490 /* convert boot_device letter to something recognizable by the bios */
491 static int boot_device2nibble(char boot_device)
492 {
493     switch(boot_device) {
494     case 'a':
495     case 'b':
496         return 0x01; /* floppy boot */
497     case 'c':
498         return 0x02; /* hard drive boot */
499     case 'd':
500         return 0x03; /* CD-ROM boot */
501     case 'n':
502         return 0x04; /* Network boot */
503     }
504     return 0;
505 }
506 
507 static void set_boot_dev(ISADevice *s, const char *boot_device, Error **errp)
508 {
509 #define PC_MAX_BOOT_DEVICES 3
510     int nbds, bds[3] = { 0, };
511     int i;
512 
513     nbds = strlen(boot_device);
514     if (nbds > PC_MAX_BOOT_DEVICES) {
515         error_setg(errp, "Too many boot devices for PC");
516         return;
517     }
518     for (i = 0; i < nbds; i++) {
519         bds[i] = boot_device2nibble(boot_device[i]);
520         if (bds[i] == 0) {
521             error_setg(errp, "Invalid boot device for PC: '%c'",
522                        boot_device[i]);
523             return;
524         }
525     }
526     rtc_set_memory(s, 0x3d, (bds[1] << 4) | bds[0]);
527     rtc_set_memory(s, 0x38, (bds[2] << 4) | (fd_bootchk ? 0x0 : 0x1));
528 }
529 
530 static void pc_boot_set(void *opaque, const char *boot_device, Error **errp)
531 {
532     set_boot_dev(opaque, boot_device, errp);
533 }
534 
535 static void pc_cmos_init_floppy(ISADevice *rtc_state, ISADevice *floppy)
536 {
537     int val, nb, i;
538     FloppyDriveType fd_type[2] = { FLOPPY_DRIVE_TYPE_NONE,
539                                    FLOPPY_DRIVE_TYPE_NONE };
540 
541     /* floppy type */
542     if (floppy) {
543         for (i = 0; i < 2; i++) {
544             fd_type[i] = isa_fdc_get_drive_type(floppy, i);
545         }
546     }
547     val = (cmos_get_fd_drive_type(fd_type[0]) << 4) |
548         cmos_get_fd_drive_type(fd_type[1]);
549     rtc_set_memory(rtc_state, 0x10, val);
550 
551     val = rtc_get_memory(rtc_state, REG_EQUIPMENT_BYTE);
552     nb = 0;
553     if (fd_type[0] != FLOPPY_DRIVE_TYPE_NONE) {
554         nb++;
555     }
556     if (fd_type[1] != FLOPPY_DRIVE_TYPE_NONE) {
557         nb++;
558     }
559     switch (nb) {
560     case 0:
561         break;
562     case 1:
563         val |= 0x01; /* 1 drive, ready for boot */
564         break;
565     case 2:
566         val |= 0x41; /* 2 drives, ready for boot */
567         break;
568     }
569     rtc_set_memory(rtc_state, REG_EQUIPMENT_BYTE, val);
570 }
571 
572 typedef struct pc_cmos_init_late_arg {
573     ISADevice *rtc_state;
574     BusState *idebus[2];
575 } pc_cmos_init_late_arg;
576 
577 typedef struct check_fdc_state {
578     ISADevice *floppy;
579     bool multiple;
580 } CheckFdcState;
581 
582 static int check_fdc(Object *obj, void *opaque)
583 {
584     CheckFdcState *state = opaque;
585     Object *fdc;
586     uint32_t iobase;
587     Error *local_err = NULL;
588 
589     fdc = object_dynamic_cast(obj, TYPE_ISA_FDC);
590     if (!fdc) {
591         return 0;
592     }
593 
594     iobase = object_property_get_uint(obj, "iobase", &local_err);
595     if (local_err || iobase != 0x3f0) {
596         error_free(local_err);
597         return 0;
598     }
599 
600     if (state->floppy) {
601         state->multiple = true;
602     } else {
603         state->floppy = ISA_DEVICE(obj);
604     }
605     return 0;
606 }
607 
608 static const char * const fdc_container_path[] = {
609     "/unattached", "/peripheral", "/peripheral-anon"
610 };
611 
612 /*
613  * Locate the FDC at IO address 0x3f0, in order to configure the CMOS registers
614  * and ACPI objects.
615  */
616 ISADevice *pc_find_fdc0(void)
617 {
618     int i;
619     Object *container;
620     CheckFdcState state = { 0 };
621 
622     for (i = 0; i < ARRAY_SIZE(fdc_container_path); i++) {
623         container = container_get(qdev_get_machine(), fdc_container_path[i]);
624         object_child_foreach(container, check_fdc, &state);
625     }
626 
627     if (state.multiple) {
628         warn_report("multiple floppy disk controllers with "
629                     "iobase=0x3f0 have been found");
630         error_printf("the one being picked for CMOS setup might not reflect "
631                      "your intent");
632     }
633 
634     return state.floppy;
635 }
636 
637 static void pc_cmos_init_late(void *opaque)
638 {
639     pc_cmos_init_late_arg *arg = opaque;
640     ISADevice *s = arg->rtc_state;
641     int16_t cylinders;
642     int8_t heads, sectors;
643     int val;
644     int i, trans;
645 
646     val = 0;
647     if (arg->idebus[0] && ide_get_geometry(arg->idebus[0], 0,
648                                            &cylinders, &heads, &sectors) >= 0) {
649         cmos_init_hd(s, 0x19, 0x1b, cylinders, heads, sectors);
650         val |= 0xf0;
651     }
652     if (arg->idebus[0] && ide_get_geometry(arg->idebus[0], 1,
653                                            &cylinders, &heads, &sectors) >= 0) {
654         cmos_init_hd(s, 0x1a, 0x24, cylinders, heads, sectors);
655         val |= 0x0f;
656     }
657     rtc_set_memory(s, 0x12, val);
658 
659     val = 0;
660     for (i = 0; i < 4; i++) {
661         /* NOTE: ide_get_geometry() returns the physical
662            geometry.  It is always such that: 1 <= sects <= 63, 1
663            <= heads <= 16, 1 <= cylinders <= 16383. The BIOS
664            geometry can be different if a translation is done. */
665         if (arg->idebus[i / 2] &&
666             ide_get_geometry(arg->idebus[i / 2], i % 2,
667                              &cylinders, &heads, &sectors) >= 0) {
668             trans = ide_get_bios_chs_trans(arg->idebus[i / 2], i % 2) - 1;
669             assert((trans & ~3) == 0);
670             val |= trans << (i * 2);
671         }
672     }
673     rtc_set_memory(s, 0x39, val);
674 
675     pc_cmos_init_floppy(s, pc_find_fdc0());
676 
677     qemu_unregister_reset(pc_cmos_init_late, opaque);
678 }
679 
680 void pc_cmos_init(PCMachineState *pcms,
681                   BusState *idebus0, BusState *idebus1,
682                   ISADevice *s)
683 {
684     int val;
685     static pc_cmos_init_late_arg arg;
686 
687     /* various important CMOS locations needed by PC/Bochs bios */
688 
689     /* memory size */
690     /* base memory (first MiB) */
691     val = MIN(pcms->below_4g_mem_size / KiB, 640);
692     rtc_set_memory(s, 0x15, val);
693     rtc_set_memory(s, 0x16, val >> 8);
694     /* extended memory (next 64MiB) */
695     if (pcms->below_4g_mem_size > 1 * MiB) {
696         val = (pcms->below_4g_mem_size - 1 * MiB) / KiB;
697     } else {
698         val = 0;
699     }
700     if (val > 65535)
701         val = 65535;
702     rtc_set_memory(s, 0x17, val);
703     rtc_set_memory(s, 0x18, val >> 8);
704     rtc_set_memory(s, 0x30, val);
705     rtc_set_memory(s, 0x31, val >> 8);
706     /* memory between 16MiB and 4GiB */
707     if (pcms->below_4g_mem_size > 16 * MiB) {
708         val = (pcms->below_4g_mem_size - 16 * MiB) / (64 * KiB);
709     } else {
710         val = 0;
711     }
712     if (val > 65535)
713         val = 65535;
714     rtc_set_memory(s, 0x34, val);
715     rtc_set_memory(s, 0x35, val >> 8);
716     /* memory above 4GiB */
717     val = pcms->above_4g_mem_size / 65536;
718     rtc_set_memory(s, 0x5b, val);
719     rtc_set_memory(s, 0x5c, val >> 8);
720     rtc_set_memory(s, 0x5d, val >> 16);
721 
722     object_property_add_link(OBJECT(pcms), "rtc_state",
723                              TYPE_ISA_DEVICE,
724                              (Object **)&pcms->rtc,
725                              object_property_allow_set_link,
726                              OBJ_PROP_LINK_STRONG, &error_abort);
727     object_property_set_link(OBJECT(pcms), OBJECT(s),
728                              "rtc_state", &error_abort);
729 
730     set_boot_dev(s, MACHINE(pcms)->boot_order, &error_fatal);
731 
732     val = 0;
733     val |= 0x02; /* FPU is there */
734     val |= 0x04; /* PS/2 mouse installed */
735     rtc_set_memory(s, REG_EQUIPMENT_BYTE, val);
736 
737     /* hard drives and FDC */
738     arg.rtc_state = s;
739     arg.idebus[0] = idebus0;
740     arg.idebus[1] = idebus1;
741     qemu_register_reset(pc_cmos_init_late, &arg);
742 }
743 
744 #define TYPE_PORT92 "port92"
745 #define PORT92(obj) OBJECT_CHECK(Port92State, (obj), TYPE_PORT92)
746 
747 /* port 92 stuff: could be split off */
748 typedef struct Port92State {
749     ISADevice parent_obj;
750 
751     MemoryRegion io;
752     uint8_t outport;
753     qemu_irq a20_out;
754 } Port92State;
755 
756 static void port92_write(void *opaque, hwaddr addr, uint64_t val,
757                          unsigned size)
758 {
759     Port92State *s = opaque;
760     int oldval = s->outport;
761 
762     DPRINTF("port92: write 0x%02" PRIx64 "\n", val);
763     s->outport = val;
764     qemu_set_irq(s->a20_out, (val >> 1) & 1);
765     if ((val & 1) && !(oldval & 1)) {
766         qemu_system_reset_request(SHUTDOWN_CAUSE_GUEST_RESET);
767     }
768 }
769 
770 static uint64_t port92_read(void *opaque, hwaddr addr,
771                             unsigned size)
772 {
773     Port92State *s = opaque;
774     uint32_t ret;
775 
776     ret = s->outport;
777     DPRINTF("port92: read 0x%02x\n", ret);
778     return ret;
779 }
780 
781 static void port92_init(ISADevice *dev, qemu_irq a20_out)
782 {
783     qdev_connect_gpio_out_named(DEVICE(dev), PORT92_A20_LINE, 0, a20_out);
784 }
785 
786 static const VMStateDescription vmstate_port92_isa = {
787     .name = "port92",
788     .version_id = 1,
789     .minimum_version_id = 1,
790     .fields = (VMStateField[]) {
791         VMSTATE_UINT8(outport, Port92State),
792         VMSTATE_END_OF_LIST()
793     }
794 };
795 
796 static void port92_reset(DeviceState *d)
797 {
798     Port92State *s = PORT92(d);
799 
800     s->outport &= ~1;
801 }
802 
803 static const MemoryRegionOps port92_ops = {
804     .read = port92_read,
805     .write = port92_write,
806     .impl = {
807         .min_access_size = 1,
808         .max_access_size = 1,
809     },
810     .endianness = DEVICE_LITTLE_ENDIAN,
811 };
812 
813 static void port92_initfn(Object *obj)
814 {
815     Port92State *s = PORT92(obj);
816 
817     memory_region_init_io(&s->io, OBJECT(s), &port92_ops, s, "port92", 1);
818 
819     s->outport = 0;
820 
821     qdev_init_gpio_out_named(DEVICE(obj), &s->a20_out, PORT92_A20_LINE, 1);
822 }
823 
824 static void port92_realizefn(DeviceState *dev, Error **errp)
825 {
826     ISADevice *isadev = ISA_DEVICE(dev);
827     Port92State *s = PORT92(dev);
828 
829     isa_register_ioport(isadev, &s->io, 0x92);
830 }
831 
832 static void port92_class_initfn(ObjectClass *klass, void *data)
833 {
834     DeviceClass *dc = DEVICE_CLASS(klass);
835 
836     dc->realize = port92_realizefn;
837     dc->reset = port92_reset;
838     dc->vmsd = &vmstate_port92_isa;
839     /*
840      * Reason: unlike ordinary ISA devices, this one needs additional
841      * wiring: its A20 output line needs to be wired up by
842      * port92_init().
843      */
844     dc->user_creatable = false;
845 }
846 
847 static const TypeInfo port92_info = {
848     .name          = TYPE_PORT92,
849     .parent        = TYPE_ISA_DEVICE,
850     .instance_size = sizeof(Port92State),
851     .instance_init = port92_initfn,
852     .class_init    = port92_class_initfn,
853 };
854 
855 static void port92_register_types(void)
856 {
857     type_register_static(&port92_info);
858 }
859 
860 type_init(port92_register_types)
861 
862 static void handle_a20_line_change(void *opaque, int irq, int level)
863 {
864     X86CPU *cpu = opaque;
865 
866     /* XXX: send to all CPUs ? */
867     /* XXX: add logic to handle multiple A20 line sources */
868     x86_cpu_set_a20(cpu, level);
869 }
870 
871 int e820_add_entry(uint64_t address, uint64_t length, uint32_t type)
872 {
873     int index = le32_to_cpu(e820_reserve.count);
874     struct e820_entry *entry;
875 
876     if (type != E820_RAM) {
877         /* old FW_CFG_E820_TABLE entry -- reservations only */
878         if (index >= E820_NR_ENTRIES) {
879             return -EBUSY;
880         }
881         entry = &e820_reserve.entry[index++];
882 
883         entry->address = cpu_to_le64(address);
884         entry->length = cpu_to_le64(length);
885         entry->type = cpu_to_le32(type);
886 
887         e820_reserve.count = cpu_to_le32(index);
888     }
889 
890     /* new "etc/e820" file -- include ram too */
891     e820_table = g_renew(struct e820_entry, e820_table, e820_entries + 1);
892     e820_table[e820_entries].address = cpu_to_le64(address);
893     e820_table[e820_entries].length = cpu_to_le64(length);
894     e820_table[e820_entries].type = cpu_to_le32(type);
895     e820_entries++;
896 
897     return e820_entries;
898 }
899 
900 int e820_get_num_entries(void)
901 {
902     return e820_entries;
903 }
904 
905 bool e820_get_entry(int idx, uint32_t type, uint64_t *address, uint64_t *length)
906 {
907     if (idx < e820_entries && e820_table[idx].type == cpu_to_le32(type)) {
908         *address = le64_to_cpu(e820_table[idx].address);
909         *length = le64_to_cpu(e820_table[idx].length);
910         return true;
911     }
912     return false;
913 }
914 
915 /* Enables contiguous-apic-ID mode, for compatibility */
916 static bool compat_apic_id_mode;
917 
918 void enable_compat_apic_id_mode(void)
919 {
920     compat_apic_id_mode = true;
921 }
922 
923 /* Calculates initial APIC ID for a specific CPU index
924  *
925  * Currently we need to be able to calculate the APIC ID from the CPU index
926  * alone (without requiring a CPU object), as the QEMU<->Seabios interfaces have
927  * no concept of "CPU index", and the NUMA tables on fw_cfg need the APIC ID of
928  * all CPUs up to max_cpus.
929  */
930 static uint32_t x86_cpu_apic_id_from_index(unsigned int cpu_index)
931 {
932     uint32_t correct_id;
933     static bool warned;
934 
935     correct_id = x86_apicid_from_cpu_idx(smp_cores, smp_threads, cpu_index);
936     if (compat_apic_id_mode) {
937         if (cpu_index != correct_id && !warned && !qtest_enabled()) {
938             error_report("APIC IDs set in compatibility mode, "
939                          "CPU topology won't match the configuration");
940             warned = true;
941         }
942         return cpu_index;
943     } else {
944         return correct_id;
945     }
946 }
947 
948 static void pc_build_smbios(PCMachineState *pcms)
949 {
950     uint8_t *smbios_tables, *smbios_anchor;
951     size_t smbios_tables_len, smbios_anchor_len;
952     struct smbios_phys_mem_area *mem_array;
953     unsigned i, array_count;
954     MachineState *ms = MACHINE(pcms);
955     X86CPU *cpu = X86_CPU(ms->possible_cpus->cpus[0].cpu);
956 
957     /* tell smbios about cpuid version and features */
958     smbios_set_cpuid(cpu->env.cpuid_version, cpu->env.features[FEAT_1_EDX]);
959 
960     smbios_tables = smbios_get_table_legacy(&smbios_tables_len);
961     if (smbios_tables) {
962         fw_cfg_add_bytes(pcms->fw_cfg, FW_CFG_SMBIOS_ENTRIES,
963                          smbios_tables, smbios_tables_len);
964     }
965 
966     /* build the array of physical mem area from e820 table */
967     mem_array = g_malloc0(sizeof(*mem_array) * e820_get_num_entries());
968     for (i = 0, array_count = 0; i < e820_get_num_entries(); i++) {
969         uint64_t addr, len;
970 
971         if (e820_get_entry(i, E820_RAM, &addr, &len)) {
972             mem_array[array_count].address = addr;
973             mem_array[array_count].length = len;
974             array_count++;
975         }
976     }
977     smbios_get_tables(mem_array, array_count,
978                       &smbios_tables, &smbios_tables_len,
979                       &smbios_anchor, &smbios_anchor_len);
980     g_free(mem_array);
981 
982     if (smbios_anchor) {
983         fw_cfg_add_file(pcms->fw_cfg, "etc/smbios/smbios-tables",
984                         smbios_tables, smbios_tables_len);
985         fw_cfg_add_file(pcms->fw_cfg, "etc/smbios/smbios-anchor",
986                         smbios_anchor, smbios_anchor_len);
987     }
988 }
989 
990 static FWCfgState *bochs_bios_init(AddressSpace *as, PCMachineState *pcms)
991 {
992     FWCfgState *fw_cfg;
993     uint64_t *numa_fw_cfg;
994     int i;
995     const CPUArchIdList *cpus;
996     MachineClass *mc = MACHINE_GET_CLASS(pcms);
997 
998     fw_cfg = fw_cfg_init_io_dma(FW_CFG_IO_BASE, FW_CFG_IO_BASE + 4, as);
999     fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
1000 
1001     /* FW_CFG_MAX_CPUS is a bit confusing/problematic on x86:
1002      *
1003      * For machine types prior to 1.8, SeaBIOS needs FW_CFG_MAX_CPUS for
1004      * building MPTable, ACPI MADT, ACPI CPU hotplug and ACPI SRAT table,
1005      * that tables are based on xAPIC ID and QEMU<->SeaBIOS interface
1006      * for CPU hotplug also uses APIC ID and not "CPU index".
1007      * This means that FW_CFG_MAX_CPUS is not the "maximum number of CPUs",
1008      * but the "limit to the APIC ID values SeaBIOS may see".
1009      *
1010      * So for compatibility reasons with old BIOSes we are stuck with
1011      * "etc/max-cpus" actually being apic_id_limit
1012      */
1013     fw_cfg_add_i16(fw_cfg, FW_CFG_MAX_CPUS, (uint16_t)pcms->apic_id_limit);
1014     fw_cfg_add_i64(fw_cfg, FW_CFG_RAM_SIZE, (uint64_t)ram_size);
1015     fw_cfg_add_bytes(fw_cfg, FW_CFG_ACPI_TABLES,
1016                      acpi_tables, acpi_tables_len);
1017     fw_cfg_add_i32(fw_cfg, FW_CFG_IRQ0_OVERRIDE, kvm_allows_irq0_override());
1018 
1019     fw_cfg_add_bytes(fw_cfg, FW_CFG_E820_TABLE,
1020                      &e820_reserve, sizeof(e820_reserve));
1021     fw_cfg_add_file(fw_cfg, "etc/e820", e820_table,
1022                     sizeof(struct e820_entry) * e820_entries);
1023 
1024     fw_cfg_add_bytes(fw_cfg, FW_CFG_HPET, &hpet_cfg, sizeof(hpet_cfg));
1025     /* allocate memory for the NUMA channel: one (64bit) word for the number
1026      * of nodes, one word for each VCPU->node and one word for each node to
1027      * hold the amount of memory.
1028      */
1029     numa_fw_cfg = g_new0(uint64_t, 1 + pcms->apic_id_limit + nb_numa_nodes);
1030     numa_fw_cfg[0] = cpu_to_le64(nb_numa_nodes);
1031     cpus = mc->possible_cpu_arch_ids(MACHINE(pcms));
1032     for (i = 0; i < cpus->len; i++) {
1033         unsigned int apic_id = cpus->cpus[i].arch_id;
1034         assert(apic_id < pcms->apic_id_limit);
1035         numa_fw_cfg[apic_id + 1] = cpu_to_le64(cpus->cpus[i].props.node_id);
1036     }
1037     for (i = 0; i < nb_numa_nodes; i++) {
1038         numa_fw_cfg[pcms->apic_id_limit + 1 + i] =
1039             cpu_to_le64(numa_info[i].node_mem);
1040     }
1041     fw_cfg_add_bytes(fw_cfg, FW_CFG_NUMA, numa_fw_cfg,
1042                      (1 + pcms->apic_id_limit + nb_numa_nodes) *
1043                      sizeof(*numa_fw_cfg));
1044 
1045     return fw_cfg;
1046 }
1047 
1048 static long get_file_size(FILE *f)
1049 {
1050     long where, size;
1051 
1052     /* XXX: on Unix systems, using fstat() probably makes more sense */
1053 
1054     where = ftell(f);
1055     fseek(f, 0, SEEK_END);
1056     size = ftell(f);
1057     fseek(f, where, SEEK_SET);
1058 
1059     return size;
1060 }
1061 
1062 struct setup_data {
1063     uint64_t next;
1064     uint32_t type;
1065     uint32_t len;
1066     uint8_t data[0];
1067 } __attribute__((packed));
1068 
1069 
1070 /*
1071  * The entry point into the kernel for PVH boot is different from
1072  * the native entry point.  The PVH entry is defined by the x86/HVM
1073  * direct boot ABI and is available in an ELFNOTE in the kernel binary.
1074  *
1075  * This function is passed to load_elf() when it is called from
1076  * load_elfboot() which then additionally checks for an ELF Note of
1077  * type XEN_ELFNOTE_PHYS32_ENTRY and passes it to this function to
1078  * parse the PVH entry address from the ELF Note.
1079  *
1080  * Due to trickery in elf_opts.h, load_elf() is actually available as
1081  * load_elf32() or load_elf64() and this routine needs to be able
1082  * to deal with being called as 32 or 64 bit.
1083  *
1084  * The address of the PVH entry point is saved to the 'pvh_start_addr'
1085  * global variable.  (although the entry point is 32-bit, the kernel
1086  * binary can be either 32-bit or 64-bit).
1087  */
1088 static uint64_t read_pvh_start_addr(void *arg1, void *arg2, bool is64)
1089 {
1090     size_t *elf_note_data_addr;
1091 
1092     /* Check if ELF Note header passed in is valid */
1093     if (arg1 == NULL) {
1094         return 0;
1095     }
1096 
1097     if (is64) {
1098         struct elf64_note *nhdr64 = (struct elf64_note *)arg1;
1099         uint64_t nhdr_size64 = sizeof(struct elf64_note);
1100         uint64_t phdr_align = *(uint64_t *)arg2;
1101         uint64_t nhdr_namesz = nhdr64->n_namesz;
1102 
1103         elf_note_data_addr =
1104             ((void *)nhdr64) + nhdr_size64 +
1105             QEMU_ALIGN_UP(nhdr_namesz, phdr_align);
1106     } else {
1107         struct elf32_note *nhdr32 = (struct elf32_note *)arg1;
1108         uint32_t nhdr_size32 = sizeof(struct elf32_note);
1109         uint32_t phdr_align = *(uint32_t *)arg2;
1110         uint32_t nhdr_namesz = nhdr32->n_namesz;
1111 
1112         elf_note_data_addr =
1113             ((void *)nhdr32) + nhdr_size32 +
1114             QEMU_ALIGN_UP(nhdr_namesz, phdr_align);
1115     }
1116 
1117     pvh_start_addr = *elf_note_data_addr;
1118 
1119     return pvh_start_addr;
1120 }
1121 
1122 static bool load_elfboot(const char *kernel_filename,
1123                    int kernel_file_size,
1124                    uint8_t *header,
1125                    size_t pvh_xen_start_addr,
1126                    FWCfgState *fw_cfg)
1127 {
1128     uint32_t flags = 0;
1129     uint32_t mh_load_addr = 0;
1130     uint32_t elf_kernel_size = 0;
1131     uint64_t elf_entry;
1132     uint64_t elf_low, elf_high;
1133     int kernel_size;
1134 
1135     if (ldl_p(header) != 0x464c457f) {
1136         return false; /* no elfboot */
1137     }
1138 
1139     bool elf_is64 = header[EI_CLASS] == ELFCLASS64;
1140     flags = elf_is64 ?
1141         ((Elf64_Ehdr *)header)->e_flags : ((Elf32_Ehdr *)header)->e_flags;
1142 
1143     if (flags & 0x00010004) { /* LOAD_ELF_HEADER_HAS_ADDR */
1144         error_report("elfboot unsupported flags = %x", flags);
1145         exit(1);
1146     }
1147 
1148     uint64_t elf_note_type = XEN_ELFNOTE_PHYS32_ENTRY;
1149     kernel_size = load_elf(kernel_filename, read_pvh_start_addr,
1150                            NULL, &elf_note_type, &elf_entry,
1151                            &elf_low, &elf_high, 0, I386_ELF_MACHINE,
1152                            0, 0);
1153 
1154     if (kernel_size < 0) {
1155         error_report("Error while loading elf kernel");
1156         exit(1);
1157     }
1158     mh_load_addr = elf_low;
1159     elf_kernel_size = elf_high - elf_low;
1160 
1161     if (pvh_start_addr == 0) {
1162         error_report("Error loading uncompressed kernel without PVH ELF Note");
1163         exit(1);
1164     }
1165     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ENTRY, pvh_start_addr);
1166     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ADDR, mh_load_addr);
1167     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_SIZE, elf_kernel_size);
1168 
1169     return true;
1170 }
1171 
1172 static void load_linux(PCMachineState *pcms,
1173                        FWCfgState *fw_cfg)
1174 {
1175     uint16_t protocol;
1176     int setup_size, kernel_size, cmdline_size;
1177     int dtb_size, setup_data_offset;
1178     uint32_t initrd_max;
1179     uint8_t header[8192], *setup, *kernel;
1180     hwaddr real_addr, prot_addr, cmdline_addr, initrd_addr = 0;
1181     FILE *f;
1182     char *vmode;
1183     MachineState *machine = MACHINE(pcms);
1184     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1185     struct setup_data *setup_data;
1186     const char *kernel_filename = machine->kernel_filename;
1187     const char *initrd_filename = machine->initrd_filename;
1188     const char *dtb_filename = machine->dtb;
1189     const char *kernel_cmdline = machine->kernel_cmdline;
1190 
1191     /* Align to 16 bytes as a paranoia measure */
1192     cmdline_size = (strlen(kernel_cmdline)+16) & ~15;
1193 
1194     /* load the kernel header */
1195     f = fopen(kernel_filename, "rb");
1196     if (!f || !(kernel_size = get_file_size(f)) ||
1197         fread(header, 1, MIN(ARRAY_SIZE(header), kernel_size), f) !=
1198         MIN(ARRAY_SIZE(header), kernel_size)) {
1199         fprintf(stderr, "qemu: could not load kernel '%s': %s\n",
1200                 kernel_filename, strerror(errno));
1201         exit(1);
1202     }
1203 
1204     /* kernel protocol version */
1205 #if 0
1206     fprintf(stderr, "header magic: %#x\n", ldl_p(header+0x202));
1207 #endif
1208     if (ldl_p(header+0x202) == 0x53726448) {
1209         protocol = lduw_p(header+0x206);
1210     } else {
1211         /*
1212          * This could be a multiboot kernel. If it is, let's stop treating it
1213          * like a Linux kernel.
1214          * Note: some multiboot images could be in the ELF format (the same of
1215          * PVH), so we try multiboot first since we check the multiboot magic
1216          * header before to load it.
1217          */
1218         if (load_multiboot(fw_cfg, f, kernel_filename, initrd_filename,
1219                            kernel_cmdline, kernel_size, header)) {
1220             return;
1221         }
1222         /*
1223          * Check if the file is an uncompressed kernel file (ELF) and load it,
1224          * saving the PVH entry point used by the x86/HVM direct boot ABI.
1225          * If load_elfboot() is successful, populate the fw_cfg info.
1226          */
1227         if (pcmc->pvh_enabled &&
1228             load_elfboot(kernel_filename, kernel_size,
1229                          header, pvh_start_addr, fw_cfg)) {
1230             fclose(f);
1231 
1232             fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_SIZE,
1233                 strlen(kernel_cmdline) + 1);
1234             fw_cfg_add_string(fw_cfg, FW_CFG_CMDLINE_DATA, kernel_cmdline);
1235 
1236             fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_SIZE, sizeof(header));
1237             fw_cfg_add_bytes(fw_cfg, FW_CFG_SETUP_DATA,
1238                              header, sizeof(header));
1239 
1240             /* load initrd */
1241             if (initrd_filename) {
1242                 gsize initrd_size;
1243                 gchar *initrd_data;
1244                 GError *gerr = NULL;
1245 
1246                 if (!g_file_get_contents(initrd_filename, &initrd_data,
1247                             &initrd_size, &gerr)) {
1248                     fprintf(stderr, "qemu: error reading initrd %s: %s\n",
1249                             initrd_filename, gerr->message);
1250                     exit(1);
1251                 }
1252 
1253                 initrd_max = pcms->below_4g_mem_size - pcmc->acpi_data_size - 1;
1254                 if (initrd_size >= initrd_max) {
1255                     fprintf(stderr, "qemu: initrd is too large, cannot support."
1256                             "(max: %"PRIu32", need %"PRId64")\n",
1257                             initrd_max, (uint64_t)initrd_size);
1258                     exit(1);
1259                 }
1260 
1261                 initrd_addr = (initrd_max - initrd_size) & ~4095;
1262 
1263                 fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_ADDR, initrd_addr);
1264                 fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_SIZE, initrd_size);
1265                 fw_cfg_add_bytes(fw_cfg, FW_CFG_INITRD_DATA, initrd_data,
1266                                  initrd_size);
1267             }
1268 
1269             option_rom[nb_option_roms].bootindex = 0;
1270             option_rom[nb_option_roms].name = "pvh.bin";
1271             nb_option_roms++;
1272 
1273             return;
1274         }
1275         protocol = 0;
1276     }
1277 
1278     if (protocol < 0x200 || !(header[0x211] & 0x01)) {
1279         /* Low kernel */
1280         real_addr    = 0x90000;
1281         cmdline_addr = 0x9a000 - cmdline_size;
1282         prot_addr    = 0x10000;
1283     } else if (protocol < 0x202) {
1284         /* High but ancient kernel */
1285         real_addr    = 0x90000;
1286         cmdline_addr = 0x9a000 - cmdline_size;
1287         prot_addr    = 0x100000;
1288     } else {
1289         /* High and recent kernel */
1290         real_addr    = 0x10000;
1291         cmdline_addr = 0x20000;
1292         prot_addr    = 0x100000;
1293     }
1294 
1295 #if 0
1296     fprintf(stderr,
1297             "qemu: real_addr     = 0x" TARGET_FMT_plx "\n"
1298             "qemu: cmdline_addr  = 0x" TARGET_FMT_plx "\n"
1299             "qemu: prot_addr     = 0x" TARGET_FMT_plx "\n",
1300             real_addr,
1301             cmdline_addr,
1302             prot_addr);
1303 #endif
1304 
1305     /* highest address for loading the initrd */
1306     if (protocol >= 0x20c &&
1307         lduw_p(header+0x236) & XLF_CAN_BE_LOADED_ABOVE_4G) {
1308         /*
1309          * Linux has supported initrd up to 4 GB for a very long time (2007,
1310          * long before XLF_CAN_BE_LOADED_ABOVE_4G which was added in 2013),
1311          * though it only sets initrd_max to 2 GB to "work around bootloader
1312          * bugs". Luckily, QEMU firmware(which does something like bootloader)
1313          * has supported this.
1314          *
1315          * It's believed that if XLF_CAN_BE_LOADED_ABOVE_4G is set, initrd can
1316          * be loaded into any address.
1317          *
1318          * In addition, initrd_max is uint32_t simply because QEMU doesn't
1319          * support the 64-bit boot protocol (specifically the ext_ramdisk_image
1320          * field).
1321          *
1322          * Therefore here just limit initrd_max to UINT32_MAX simply as well.
1323          */
1324         initrd_max = UINT32_MAX;
1325     } else if (protocol >= 0x203) {
1326         initrd_max = ldl_p(header+0x22c);
1327     } else {
1328         initrd_max = 0x37ffffff;
1329     }
1330 
1331     if (initrd_max >= pcms->below_4g_mem_size - pcmc->acpi_data_size) {
1332         initrd_max = pcms->below_4g_mem_size - pcmc->acpi_data_size - 1;
1333     }
1334 
1335     fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_ADDR, cmdline_addr);
1336     fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_SIZE, strlen(kernel_cmdline)+1);
1337     fw_cfg_add_string(fw_cfg, FW_CFG_CMDLINE_DATA, kernel_cmdline);
1338 
1339     if (protocol >= 0x202) {
1340         stl_p(header+0x228, cmdline_addr);
1341     } else {
1342         stw_p(header+0x20, 0xA33F);
1343         stw_p(header+0x22, cmdline_addr-real_addr);
1344     }
1345 
1346     /* handle vga= parameter */
1347     vmode = strstr(kernel_cmdline, "vga=");
1348     if (vmode) {
1349         unsigned int video_mode;
1350         /* skip "vga=" */
1351         vmode += 4;
1352         if (!strncmp(vmode, "normal", 6)) {
1353             video_mode = 0xffff;
1354         } else if (!strncmp(vmode, "ext", 3)) {
1355             video_mode = 0xfffe;
1356         } else if (!strncmp(vmode, "ask", 3)) {
1357             video_mode = 0xfffd;
1358         } else {
1359             video_mode = strtol(vmode, NULL, 0);
1360         }
1361         stw_p(header+0x1fa, video_mode);
1362     }
1363 
1364     /* loader type */
1365     /* High nybble = B reserved for QEMU; low nybble is revision number.
1366        If this code is substantially changed, you may want to consider
1367        incrementing the revision. */
1368     if (protocol >= 0x200) {
1369         header[0x210] = 0xB0;
1370     }
1371     /* heap */
1372     if (protocol >= 0x201) {
1373         header[0x211] |= 0x80;	/* CAN_USE_HEAP */
1374         stw_p(header+0x224, cmdline_addr-real_addr-0x200);
1375     }
1376 
1377     /* load initrd */
1378     if (initrd_filename) {
1379         gsize initrd_size;
1380         gchar *initrd_data;
1381         GError *gerr = NULL;
1382 
1383         if (protocol < 0x200) {
1384             fprintf(stderr, "qemu: linux kernel too old to load a ram disk\n");
1385             exit(1);
1386         }
1387 
1388         if (!g_file_get_contents(initrd_filename, &initrd_data,
1389                                  &initrd_size, &gerr)) {
1390             fprintf(stderr, "qemu: error reading initrd %s: %s\n",
1391                     initrd_filename, gerr->message);
1392             exit(1);
1393         }
1394         if (initrd_size >= initrd_max) {
1395             fprintf(stderr, "qemu: initrd is too large, cannot support."
1396                     "(max: %"PRIu32", need %"PRId64")\n",
1397                     initrd_max, (uint64_t)initrd_size);
1398             exit(1);
1399         }
1400 
1401         initrd_addr = (initrd_max-initrd_size) & ~4095;
1402 
1403         fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_ADDR, initrd_addr);
1404         fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_SIZE, initrd_size);
1405         fw_cfg_add_bytes(fw_cfg, FW_CFG_INITRD_DATA, initrd_data, initrd_size);
1406 
1407         stl_p(header+0x218, initrd_addr);
1408         stl_p(header+0x21c, initrd_size);
1409     }
1410 
1411     /* load kernel and setup */
1412     setup_size = header[0x1f1];
1413     if (setup_size == 0) {
1414         setup_size = 4;
1415     }
1416     setup_size = (setup_size+1)*512;
1417     if (setup_size > kernel_size) {
1418         fprintf(stderr, "qemu: invalid kernel header\n");
1419         exit(1);
1420     }
1421     kernel_size -= setup_size;
1422 
1423     setup  = g_malloc(setup_size);
1424     kernel = g_malloc(kernel_size);
1425     fseek(f, 0, SEEK_SET);
1426     if (fread(setup, 1, setup_size, f) != setup_size) {
1427         fprintf(stderr, "fread() failed\n");
1428         exit(1);
1429     }
1430     if (fread(kernel, 1, kernel_size, f) != kernel_size) {
1431         fprintf(stderr, "fread() failed\n");
1432         exit(1);
1433     }
1434     fclose(f);
1435 
1436     /* append dtb to kernel */
1437     if (dtb_filename) {
1438         if (protocol < 0x209) {
1439             fprintf(stderr, "qemu: Linux kernel too old to load a dtb\n");
1440             exit(1);
1441         }
1442 
1443         dtb_size = get_image_size(dtb_filename);
1444         if (dtb_size <= 0) {
1445             fprintf(stderr, "qemu: error reading dtb %s: %s\n",
1446                     dtb_filename, strerror(errno));
1447             exit(1);
1448         }
1449 
1450         setup_data_offset = QEMU_ALIGN_UP(kernel_size, 16);
1451         kernel_size = setup_data_offset + sizeof(struct setup_data) + dtb_size;
1452         kernel = g_realloc(kernel, kernel_size);
1453 
1454         stq_p(header+0x250, prot_addr + setup_data_offset);
1455 
1456         setup_data = (struct setup_data *)(kernel + setup_data_offset);
1457         setup_data->next = 0;
1458         setup_data->type = cpu_to_le32(SETUP_DTB);
1459         setup_data->len = cpu_to_le32(dtb_size);
1460 
1461         load_image_size(dtb_filename, setup_data->data, dtb_size);
1462     }
1463 
1464     memcpy(setup, header, MIN(sizeof(header), setup_size));
1465 
1466     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ADDR, prot_addr);
1467     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_SIZE, kernel_size);
1468     fw_cfg_add_bytes(fw_cfg, FW_CFG_KERNEL_DATA, kernel, kernel_size);
1469 
1470     fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_ADDR, real_addr);
1471     fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_SIZE, setup_size);
1472     fw_cfg_add_bytes(fw_cfg, FW_CFG_SETUP_DATA, setup, setup_size);
1473 
1474     option_rom[nb_option_roms].bootindex = 0;
1475     option_rom[nb_option_roms].name = "linuxboot.bin";
1476     if (pcmc->linuxboot_dma_enabled && fw_cfg_dma_enabled(fw_cfg)) {
1477         option_rom[nb_option_roms].name = "linuxboot_dma.bin";
1478     }
1479     nb_option_roms++;
1480 }
1481 
1482 #define NE2000_NB_MAX 6
1483 
1484 static const int ne2000_io[NE2000_NB_MAX] = { 0x300, 0x320, 0x340, 0x360,
1485                                               0x280, 0x380 };
1486 static const int ne2000_irq[NE2000_NB_MAX] = { 9, 10, 11, 3, 4, 5 };
1487 
1488 void pc_init_ne2k_isa(ISABus *bus, NICInfo *nd)
1489 {
1490     static int nb_ne2k = 0;
1491 
1492     if (nb_ne2k == NE2000_NB_MAX)
1493         return;
1494     isa_ne2000_init(bus, ne2000_io[nb_ne2k],
1495                     ne2000_irq[nb_ne2k], nd);
1496     nb_ne2k++;
1497 }
1498 
1499 DeviceState *cpu_get_current_apic(void)
1500 {
1501     if (current_cpu) {
1502         X86CPU *cpu = X86_CPU(current_cpu);
1503         return cpu->apic_state;
1504     } else {
1505         return NULL;
1506     }
1507 }
1508 
1509 void pc_acpi_smi_interrupt(void *opaque, int irq, int level)
1510 {
1511     X86CPU *cpu = opaque;
1512 
1513     if (level) {
1514         cpu_interrupt(CPU(cpu), CPU_INTERRUPT_SMI);
1515     }
1516 }
1517 
1518 static void pc_new_cpu(const char *typename, int64_t apic_id, Error **errp)
1519 {
1520     Object *cpu = NULL;
1521     Error *local_err = NULL;
1522 
1523     cpu = object_new(typename);
1524 
1525     object_property_set_uint(cpu, apic_id, "apic-id", &local_err);
1526     object_property_set_bool(cpu, true, "realized", &local_err);
1527 
1528     object_unref(cpu);
1529     error_propagate(errp, local_err);
1530 }
1531 
1532 void pc_hot_add_cpu(const int64_t id, Error **errp)
1533 {
1534     MachineState *ms = MACHINE(qdev_get_machine());
1535     int64_t apic_id = x86_cpu_apic_id_from_index(id);
1536     Error *local_err = NULL;
1537 
1538     if (id < 0) {
1539         error_setg(errp, "Invalid CPU id: %" PRIi64, id);
1540         return;
1541     }
1542 
1543     if (apic_id >= ACPI_CPU_HOTPLUG_ID_LIMIT) {
1544         error_setg(errp, "Unable to add CPU: %" PRIi64
1545                    ", resulting APIC ID (%" PRIi64 ") is too large",
1546                    id, apic_id);
1547         return;
1548     }
1549 
1550     pc_new_cpu(ms->cpu_type, apic_id, &local_err);
1551     if (local_err) {
1552         error_propagate(errp, local_err);
1553         return;
1554     }
1555 }
1556 
1557 void pc_cpus_init(PCMachineState *pcms)
1558 {
1559     int i;
1560     const CPUArchIdList *possible_cpus;
1561     MachineState *ms = MACHINE(pcms);
1562     MachineClass *mc = MACHINE_GET_CLASS(pcms);
1563 
1564     /* Calculates the limit to CPU APIC ID values
1565      *
1566      * Limit for the APIC ID value, so that all
1567      * CPU APIC IDs are < pcms->apic_id_limit.
1568      *
1569      * This is used for FW_CFG_MAX_CPUS. See comments on bochs_bios_init().
1570      */
1571     pcms->apic_id_limit = x86_cpu_apic_id_from_index(max_cpus - 1) + 1;
1572     possible_cpus = mc->possible_cpu_arch_ids(ms);
1573     for (i = 0; i < smp_cpus; i++) {
1574         pc_new_cpu(possible_cpus->cpus[i].type, possible_cpus->cpus[i].arch_id,
1575                    &error_fatal);
1576     }
1577 }
1578 
1579 static void pc_build_feature_control_file(PCMachineState *pcms)
1580 {
1581     MachineState *ms = MACHINE(pcms);
1582     X86CPU *cpu = X86_CPU(ms->possible_cpus->cpus[0].cpu);
1583     CPUX86State *env = &cpu->env;
1584     uint32_t unused, ecx, edx;
1585     uint64_t feature_control_bits = 0;
1586     uint64_t *val;
1587 
1588     cpu_x86_cpuid(env, 1, 0, &unused, &unused, &ecx, &edx);
1589     if (ecx & CPUID_EXT_VMX) {
1590         feature_control_bits |= FEATURE_CONTROL_VMXON_ENABLED_OUTSIDE_SMX;
1591     }
1592 
1593     if ((edx & (CPUID_EXT2_MCE | CPUID_EXT2_MCA)) ==
1594         (CPUID_EXT2_MCE | CPUID_EXT2_MCA) &&
1595         (env->mcg_cap & MCG_LMCE_P)) {
1596         feature_control_bits |= FEATURE_CONTROL_LMCE;
1597     }
1598 
1599     if (!feature_control_bits) {
1600         return;
1601     }
1602 
1603     val = g_malloc(sizeof(*val));
1604     *val = cpu_to_le64(feature_control_bits | FEATURE_CONTROL_LOCKED);
1605     fw_cfg_add_file(pcms->fw_cfg, "etc/msr_feature_control", val, sizeof(*val));
1606 }
1607 
1608 static void rtc_set_cpus_count(ISADevice *rtc, uint16_t cpus_count)
1609 {
1610     if (cpus_count > 0xff) {
1611         /* If the number of CPUs can't be represented in 8 bits, the
1612          * BIOS must use "FW_CFG_NB_CPUS". Set RTC field to 0 just
1613          * to make old BIOSes fail more predictably.
1614          */
1615         rtc_set_memory(rtc, 0x5f, 0);
1616     } else {
1617         rtc_set_memory(rtc, 0x5f, cpus_count - 1);
1618     }
1619 }
1620 
1621 static
1622 void pc_machine_done(Notifier *notifier, void *data)
1623 {
1624     PCMachineState *pcms = container_of(notifier,
1625                                         PCMachineState, machine_done);
1626     PCIBus *bus = pcms->bus;
1627 
1628     /* set the number of CPUs */
1629     rtc_set_cpus_count(pcms->rtc, pcms->boot_cpus);
1630 
1631     if (bus) {
1632         int extra_hosts = 0;
1633 
1634         QLIST_FOREACH(bus, &bus->child, sibling) {
1635             /* look for expander root buses */
1636             if (pci_bus_is_root(bus)) {
1637                 extra_hosts++;
1638             }
1639         }
1640         if (extra_hosts && pcms->fw_cfg) {
1641             uint64_t *val = g_malloc(sizeof(*val));
1642             *val = cpu_to_le64(extra_hosts);
1643             fw_cfg_add_file(pcms->fw_cfg,
1644                     "etc/extra-pci-roots", val, sizeof(*val));
1645         }
1646     }
1647 
1648     acpi_setup();
1649     if (pcms->fw_cfg) {
1650         pc_build_smbios(pcms);
1651         pc_build_feature_control_file(pcms);
1652         /* update FW_CFG_NB_CPUS to account for -device added CPUs */
1653         fw_cfg_modify_i16(pcms->fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
1654     }
1655 
1656     if (pcms->apic_id_limit > 255 && !xen_enabled()) {
1657         IntelIOMMUState *iommu = INTEL_IOMMU_DEVICE(x86_iommu_get_default());
1658 
1659         if (!iommu || !x86_iommu_ir_supported(X86_IOMMU_DEVICE(iommu)) ||
1660             iommu->intr_eim != ON_OFF_AUTO_ON) {
1661             error_report("current -smp configuration requires "
1662                          "Extended Interrupt Mode enabled. "
1663                          "You can add an IOMMU using: "
1664                          "-device intel-iommu,intremap=on,eim=on");
1665             exit(EXIT_FAILURE);
1666         }
1667     }
1668 }
1669 
1670 void pc_guest_info_init(PCMachineState *pcms)
1671 {
1672     int i;
1673 
1674     pcms->apic_xrupt_override = kvm_allows_irq0_override();
1675     pcms->numa_nodes = nb_numa_nodes;
1676     pcms->node_mem = g_malloc0(pcms->numa_nodes *
1677                                     sizeof *pcms->node_mem);
1678     for (i = 0; i < nb_numa_nodes; i++) {
1679         pcms->node_mem[i] = numa_info[i].node_mem;
1680     }
1681 
1682     pcms->machine_done.notify = pc_machine_done;
1683     qemu_add_machine_init_done_notifier(&pcms->machine_done);
1684 }
1685 
1686 /* setup pci memory address space mapping into system address space */
1687 void pc_pci_as_mapping_init(Object *owner, MemoryRegion *system_memory,
1688                             MemoryRegion *pci_address_space)
1689 {
1690     /* Set to lower priority than RAM */
1691     memory_region_add_subregion_overlap(system_memory, 0x0,
1692                                         pci_address_space, -1);
1693 }
1694 
1695 void xen_load_linux(PCMachineState *pcms)
1696 {
1697     int i;
1698     FWCfgState *fw_cfg;
1699 
1700     assert(MACHINE(pcms)->kernel_filename != NULL);
1701 
1702     fw_cfg = fw_cfg_init_io(FW_CFG_IO_BASE);
1703     fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
1704     rom_set_fw(fw_cfg);
1705 
1706     load_linux(pcms, fw_cfg);
1707     for (i = 0; i < nb_option_roms; i++) {
1708         assert(!strcmp(option_rom[i].name, "linuxboot.bin") ||
1709                !strcmp(option_rom[i].name, "linuxboot_dma.bin") ||
1710                !strcmp(option_rom[i].name, "pvh.bin") ||
1711                !strcmp(option_rom[i].name, "multiboot.bin"));
1712         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1713     }
1714     pcms->fw_cfg = fw_cfg;
1715 }
1716 
1717 void pc_memory_init(PCMachineState *pcms,
1718                     MemoryRegion *system_memory,
1719                     MemoryRegion *rom_memory,
1720                     MemoryRegion **ram_memory)
1721 {
1722     int linux_boot, i;
1723     MemoryRegion *ram, *option_rom_mr;
1724     MemoryRegion *ram_below_4g, *ram_above_4g;
1725     FWCfgState *fw_cfg;
1726     MachineState *machine = MACHINE(pcms);
1727     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1728 
1729     assert(machine->ram_size == pcms->below_4g_mem_size +
1730                                 pcms->above_4g_mem_size);
1731 
1732     linux_boot = (machine->kernel_filename != NULL);
1733 
1734     /* Allocate RAM.  We allocate it as a single memory region and use
1735      * aliases to address portions of it, mostly for backwards compatibility
1736      * with older qemus that used qemu_ram_alloc().
1737      */
1738     ram = g_malloc(sizeof(*ram));
1739     memory_region_allocate_system_memory(ram, NULL, "pc.ram",
1740                                          machine->ram_size);
1741     *ram_memory = ram;
1742     ram_below_4g = g_malloc(sizeof(*ram_below_4g));
1743     memory_region_init_alias(ram_below_4g, NULL, "ram-below-4g", ram,
1744                              0, pcms->below_4g_mem_size);
1745     memory_region_add_subregion(system_memory, 0, ram_below_4g);
1746     e820_add_entry(0, pcms->below_4g_mem_size, E820_RAM);
1747     if (pcms->above_4g_mem_size > 0) {
1748         ram_above_4g = g_malloc(sizeof(*ram_above_4g));
1749         memory_region_init_alias(ram_above_4g, NULL, "ram-above-4g", ram,
1750                                  pcms->below_4g_mem_size,
1751                                  pcms->above_4g_mem_size);
1752         memory_region_add_subregion(system_memory, 0x100000000ULL,
1753                                     ram_above_4g);
1754         e820_add_entry(0x100000000ULL, pcms->above_4g_mem_size, E820_RAM);
1755     }
1756 
1757     if (!pcmc->has_reserved_memory &&
1758         (machine->ram_slots ||
1759          (machine->maxram_size > machine->ram_size))) {
1760         MachineClass *mc = MACHINE_GET_CLASS(machine);
1761 
1762         error_report("\"-memory 'slots|maxmem'\" is not supported by: %s",
1763                      mc->name);
1764         exit(EXIT_FAILURE);
1765     }
1766 
1767     /* always allocate the device memory information */
1768     machine->device_memory = g_malloc0(sizeof(*machine->device_memory));
1769 
1770     /* initialize device memory address space */
1771     if (pcmc->has_reserved_memory &&
1772         (machine->ram_size < machine->maxram_size)) {
1773         ram_addr_t device_mem_size = machine->maxram_size - machine->ram_size;
1774 
1775         if (machine->ram_slots > ACPI_MAX_RAM_SLOTS) {
1776             error_report("unsupported amount of memory slots: %"PRIu64,
1777                          machine->ram_slots);
1778             exit(EXIT_FAILURE);
1779         }
1780 
1781         if (QEMU_ALIGN_UP(machine->maxram_size,
1782                           TARGET_PAGE_SIZE) != machine->maxram_size) {
1783             error_report("maximum memory size must by aligned to multiple of "
1784                          "%d bytes", TARGET_PAGE_SIZE);
1785             exit(EXIT_FAILURE);
1786         }
1787 
1788         machine->device_memory->base =
1789             ROUND_UP(0x100000000ULL + pcms->above_4g_mem_size, 1 * GiB);
1790 
1791         if (pcmc->enforce_aligned_dimm) {
1792             /* size device region assuming 1G page max alignment per slot */
1793             device_mem_size += (1 * GiB) * machine->ram_slots;
1794         }
1795 
1796         if ((machine->device_memory->base + device_mem_size) <
1797             device_mem_size) {
1798             error_report("unsupported amount of maximum memory: " RAM_ADDR_FMT,
1799                          machine->maxram_size);
1800             exit(EXIT_FAILURE);
1801         }
1802 
1803         memory_region_init(&machine->device_memory->mr, OBJECT(pcms),
1804                            "device-memory", device_mem_size);
1805         memory_region_add_subregion(system_memory, machine->device_memory->base,
1806                                     &machine->device_memory->mr);
1807     }
1808 
1809     /* Initialize PC system firmware */
1810     pc_system_firmware_init(pcms, rom_memory);
1811 
1812     option_rom_mr = g_malloc(sizeof(*option_rom_mr));
1813     memory_region_init_ram(option_rom_mr, NULL, "pc.rom", PC_ROM_SIZE,
1814                            &error_fatal);
1815     if (pcmc->pci_enabled) {
1816         memory_region_set_readonly(option_rom_mr, true);
1817     }
1818     memory_region_add_subregion_overlap(rom_memory,
1819                                         PC_ROM_MIN_VGA,
1820                                         option_rom_mr,
1821                                         1);
1822 
1823     fw_cfg = bochs_bios_init(&address_space_memory, pcms);
1824 
1825     rom_set_fw(fw_cfg);
1826 
1827     if (pcmc->has_reserved_memory && machine->device_memory->base) {
1828         uint64_t *val = g_malloc(sizeof(*val));
1829         PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1830         uint64_t res_mem_end = machine->device_memory->base;
1831 
1832         if (!pcmc->broken_reserved_end) {
1833             res_mem_end += memory_region_size(&machine->device_memory->mr);
1834         }
1835         *val = cpu_to_le64(ROUND_UP(res_mem_end, 1 * GiB));
1836         fw_cfg_add_file(fw_cfg, "etc/reserved-memory-end", val, sizeof(*val));
1837     }
1838 
1839     if (linux_boot) {
1840         load_linux(pcms, fw_cfg);
1841     }
1842 
1843     for (i = 0; i < nb_option_roms; i++) {
1844         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1845     }
1846     pcms->fw_cfg = fw_cfg;
1847 
1848     /* Init default IOAPIC address space */
1849     pcms->ioapic_as = &address_space_memory;
1850 }
1851 
1852 /*
1853  * The 64bit pci hole starts after "above 4G RAM" and
1854  * potentially the space reserved for memory hotplug.
1855  */
1856 uint64_t pc_pci_hole64_start(void)
1857 {
1858     PCMachineState *pcms = PC_MACHINE(qdev_get_machine());
1859     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1860     MachineState *ms = MACHINE(pcms);
1861     uint64_t hole64_start = 0;
1862 
1863     if (pcmc->has_reserved_memory && ms->device_memory->base) {
1864         hole64_start = ms->device_memory->base;
1865         if (!pcmc->broken_reserved_end) {
1866             hole64_start += memory_region_size(&ms->device_memory->mr);
1867         }
1868     } else {
1869         hole64_start = 0x100000000ULL + pcms->above_4g_mem_size;
1870     }
1871 
1872     return ROUND_UP(hole64_start, 1 * GiB);
1873 }
1874 
1875 qemu_irq pc_allocate_cpu_irq(void)
1876 {
1877     return qemu_allocate_irq(pic_irq_request, NULL, 0);
1878 }
1879 
1880 DeviceState *pc_vga_init(ISABus *isa_bus, PCIBus *pci_bus)
1881 {
1882     DeviceState *dev = NULL;
1883 
1884     rom_set_order_override(FW_CFG_ORDER_OVERRIDE_VGA);
1885     if (pci_bus) {
1886         PCIDevice *pcidev = pci_vga_init(pci_bus);
1887         dev = pcidev ? &pcidev->qdev : NULL;
1888     } else if (isa_bus) {
1889         ISADevice *isadev = isa_vga_init(isa_bus);
1890         dev = isadev ? DEVICE(isadev) : NULL;
1891     }
1892     rom_reset_order_override();
1893     return dev;
1894 }
1895 
1896 static const MemoryRegionOps ioport80_io_ops = {
1897     .write = ioport80_write,
1898     .read = ioport80_read,
1899     .endianness = DEVICE_NATIVE_ENDIAN,
1900     .impl = {
1901         .min_access_size = 1,
1902         .max_access_size = 1,
1903     },
1904 };
1905 
1906 static const MemoryRegionOps ioportF0_io_ops = {
1907     .write = ioportF0_write,
1908     .read = ioportF0_read,
1909     .endianness = DEVICE_NATIVE_ENDIAN,
1910     .impl = {
1911         .min_access_size = 1,
1912         .max_access_size = 1,
1913     },
1914 };
1915 
1916 static void pc_superio_init(ISABus *isa_bus, bool create_fdctrl, bool no_vmport)
1917 {
1918     int i;
1919     DriveInfo *fd[MAX_FD];
1920     qemu_irq *a20_line;
1921     ISADevice *i8042, *port92, *vmmouse;
1922 
1923     serial_hds_isa_init(isa_bus, 0, MAX_ISA_SERIAL_PORTS);
1924     parallel_hds_isa_init(isa_bus, MAX_PARALLEL_PORTS);
1925 
1926     for (i = 0; i < MAX_FD; i++) {
1927         fd[i] = drive_get(IF_FLOPPY, 0, i);
1928         create_fdctrl |= !!fd[i];
1929     }
1930     if (create_fdctrl) {
1931         fdctrl_init_isa(isa_bus, fd);
1932     }
1933 
1934     i8042 = isa_create_simple(isa_bus, "i8042");
1935     if (!no_vmport) {
1936         vmport_init(isa_bus);
1937         vmmouse = isa_try_create(isa_bus, "vmmouse");
1938     } else {
1939         vmmouse = NULL;
1940     }
1941     if (vmmouse) {
1942         DeviceState *dev = DEVICE(vmmouse);
1943         qdev_prop_set_ptr(dev, "ps2_mouse", i8042);
1944         qdev_init_nofail(dev);
1945     }
1946     port92 = isa_create_simple(isa_bus, "port92");
1947 
1948     a20_line = qemu_allocate_irqs(handle_a20_line_change, first_cpu, 2);
1949     i8042_setup_a20_line(i8042, a20_line[0]);
1950     port92_init(port92, a20_line[1]);
1951     g_free(a20_line);
1952 }
1953 
1954 void pc_basic_device_init(ISABus *isa_bus, qemu_irq *gsi,
1955                           ISADevice **rtc_state,
1956                           bool create_fdctrl,
1957                           bool no_vmport,
1958                           bool has_pit,
1959                           uint32_t hpet_irqs)
1960 {
1961     int i;
1962     DeviceState *hpet = NULL;
1963     int pit_isa_irq = 0;
1964     qemu_irq pit_alt_irq = NULL;
1965     qemu_irq rtc_irq = NULL;
1966     ISADevice *pit = NULL;
1967     MemoryRegion *ioport80_io = g_new(MemoryRegion, 1);
1968     MemoryRegion *ioportF0_io = g_new(MemoryRegion, 1);
1969 
1970     memory_region_init_io(ioport80_io, NULL, &ioport80_io_ops, NULL, "ioport80", 1);
1971     memory_region_add_subregion(isa_bus->address_space_io, 0x80, ioport80_io);
1972 
1973     memory_region_init_io(ioportF0_io, NULL, &ioportF0_io_ops, NULL, "ioportF0", 1);
1974     memory_region_add_subregion(isa_bus->address_space_io, 0xf0, ioportF0_io);
1975 
1976     /*
1977      * Check if an HPET shall be created.
1978      *
1979      * Without KVM_CAP_PIT_STATE2, we cannot switch off the in-kernel PIT
1980      * when the HPET wants to take over. Thus we have to disable the latter.
1981      */
1982     if (!no_hpet && (!kvm_irqchip_in_kernel() || kvm_has_pit_state2())) {
1983         /* In order to set property, here not using sysbus_try_create_simple */
1984         hpet = qdev_try_create(NULL, TYPE_HPET);
1985         if (hpet) {
1986             /* For pc-piix-*, hpet's intcap is always IRQ2. For pc-q35-1.7
1987              * and earlier, use IRQ2 for compat. Otherwise, use IRQ16~23,
1988              * IRQ8 and IRQ2.
1989              */
1990             uint8_t compat = object_property_get_uint(OBJECT(hpet),
1991                     HPET_INTCAP, NULL);
1992             if (!compat) {
1993                 qdev_prop_set_uint32(hpet, HPET_INTCAP, hpet_irqs);
1994             }
1995             qdev_init_nofail(hpet);
1996             sysbus_mmio_map(SYS_BUS_DEVICE(hpet), 0, HPET_BASE);
1997 
1998             for (i = 0; i < GSI_NUM_PINS; i++) {
1999                 sysbus_connect_irq(SYS_BUS_DEVICE(hpet), i, gsi[i]);
2000             }
2001             pit_isa_irq = -1;
2002             pit_alt_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_PIT_INT);
2003             rtc_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_RTC_INT);
2004         }
2005     }
2006     *rtc_state = mc146818_rtc_init(isa_bus, 2000, rtc_irq);
2007 
2008     qemu_register_boot_set(pc_boot_set, *rtc_state);
2009 
2010     if (!xen_enabled() && has_pit) {
2011         if (kvm_pit_in_kernel()) {
2012             pit = kvm_pit_init(isa_bus, 0x40);
2013         } else {
2014             pit = i8254_pit_init(isa_bus, 0x40, pit_isa_irq, pit_alt_irq);
2015         }
2016         if (hpet) {
2017             /* connect PIT to output control line of the HPET */
2018             qdev_connect_gpio_out(hpet, 0, qdev_get_gpio_in(DEVICE(pit), 0));
2019         }
2020         pcspk_init(isa_bus, pit);
2021     }
2022 
2023     i8257_dma_init(isa_bus, 0);
2024 
2025     /* Super I/O */
2026     pc_superio_init(isa_bus, create_fdctrl, no_vmport);
2027 }
2028 
2029 void pc_nic_init(PCMachineClass *pcmc, ISABus *isa_bus, PCIBus *pci_bus)
2030 {
2031     int i;
2032 
2033     rom_set_order_override(FW_CFG_ORDER_OVERRIDE_NIC);
2034     for (i = 0; i < nb_nics; i++) {
2035         NICInfo *nd = &nd_table[i];
2036         const char *model = nd->model ? nd->model : pcmc->default_nic_model;
2037 
2038         if (g_str_equal(model, "ne2k_isa")) {
2039             pc_init_ne2k_isa(isa_bus, nd);
2040         } else {
2041             pci_nic_init_nofail(nd, pci_bus, model, NULL);
2042         }
2043     }
2044     rom_reset_order_override();
2045 }
2046 
2047 void ioapic_init_gsi(GSIState *gsi_state, const char *parent_name)
2048 {
2049     DeviceState *dev;
2050     SysBusDevice *d;
2051     unsigned int i;
2052 
2053     if (kvm_ioapic_in_kernel()) {
2054         dev = qdev_create(NULL, TYPE_KVM_IOAPIC);
2055     } else {
2056         dev = qdev_create(NULL, TYPE_IOAPIC);
2057     }
2058     if (parent_name) {
2059         object_property_add_child(object_resolve_path(parent_name, NULL),
2060                                   "ioapic", OBJECT(dev), NULL);
2061     }
2062     qdev_init_nofail(dev);
2063     d = SYS_BUS_DEVICE(dev);
2064     sysbus_mmio_map(d, 0, IO_APIC_DEFAULT_ADDRESS);
2065 
2066     for (i = 0; i < IOAPIC_NUM_PINS; i++) {
2067         gsi_state->ioapic_irq[i] = qdev_get_gpio_in(dev, i);
2068     }
2069 }
2070 
2071 static void pc_memory_pre_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
2072                                Error **errp)
2073 {
2074     const PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2075     const PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
2076     const MachineState *ms = MACHINE(hotplug_dev);
2077     const bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2078     const uint64_t legacy_align = TARGET_PAGE_SIZE;
2079     Error *local_err = NULL;
2080 
2081     /*
2082      * When -no-acpi is used with Q35 machine type, no ACPI is built,
2083      * but pcms->acpi_dev is still created. Check !acpi_enabled in
2084      * addition to cover this case.
2085      */
2086     if (!pcms->acpi_dev || !acpi_enabled) {
2087         error_setg(errp,
2088                    "memory hotplug is not enabled: missing acpi device or acpi disabled");
2089         return;
2090     }
2091 
2092     if (is_nvdimm && !ms->nvdimms_state->is_enabled) {
2093         error_setg(errp, "nvdimm is not enabled: missing 'nvdimm' in '-M'");
2094         return;
2095     }
2096 
2097     hotplug_handler_pre_plug(pcms->acpi_dev, dev, &local_err);
2098     if (local_err) {
2099         error_propagate(errp, local_err);
2100         return;
2101     }
2102 
2103     pc_dimm_pre_plug(PC_DIMM(dev), MACHINE(hotplug_dev),
2104                      pcmc->enforce_aligned_dimm ? NULL : &legacy_align, errp);
2105 }
2106 
2107 static void pc_memory_plug(HotplugHandler *hotplug_dev,
2108                            DeviceState *dev, Error **errp)
2109 {
2110     Error *local_err = NULL;
2111     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2112     MachineState *ms = MACHINE(hotplug_dev);
2113     bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2114 
2115     pc_dimm_plug(PC_DIMM(dev), MACHINE(pcms), &local_err);
2116     if (local_err) {
2117         goto out;
2118     }
2119 
2120     if (is_nvdimm) {
2121         nvdimm_plug(ms->nvdimms_state);
2122     }
2123 
2124     hotplug_handler_plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &error_abort);
2125 out:
2126     error_propagate(errp, local_err);
2127 }
2128 
2129 static void pc_memory_unplug_request(HotplugHandler *hotplug_dev,
2130                                      DeviceState *dev, Error **errp)
2131 {
2132     Error *local_err = NULL;
2133     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2134 
2135     /*
2136      * When -no-acpi is used with Q35 machine type, no ACPI is built,
2137      * but pcms->acpi_dev is still created. Check !acpi_enabled in
2138      * addition to cover this case.
2139      */
2140     if (!pcms->acpi_dev || !acpi_enabled) {
2141         error_setg(&local_err,
2142                    "memory hotplug is not enabled: missing acpi device or acpi disabled");
2143         goto out;
2144     }
2145 
2146     if (object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM)) {
2147         error_setg(&local_err,
2148                    "nvdimm device hot unplug is not supported yet.");
2149         goto out;
2150     }
2151 
2152     hotplug_handler_unplug_request(HOTPLUG_HANDLER(pcms->acpi_dev), dev,
2153                                    &local_err);
2154 out:
2155     error_propagate(errp, local_err);
2156 }
2157 
2158 static void pc_memory_unplug(HotplugHandler *hotplug_dev,
2159                              DeviceState *dev, Error **errp)
2160 {
2161     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2162     Error *local_err = NULL;
2163 
2164     hotplug_handler_unplug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
2165     if (local_err) {
2166         goto out;
2167     }
2168 
2169     pc_dimm_unplug(PC_DIMM(dev), MACHINE(pcms));
2170     object_property_set_bool(OBJECT(dev), false, "realized", NULL);
2171  out:
2172     error_propagate(errp, local_err);
2173 }
2174 
2175 static int pc_apic_cmp(const void *a, const void *b)
2176 {
2177    CPUArchId *apic_a = (CPUArchId *)a;
2178    CPUArchId *apic_b = (CPUArchId *)b;
2179 
2180    return apic_a->arch_id - apic_b->arch_id;
2181 }
2182 
2183 /* returns pointer to CPUArchId descriptor that matches CPU's apic_id
2184  * in ms->possible_cpus->cpus, if ms->possible_cpus->cpus has no
2185  * entry corresponding to CPU's apic_id returns NULL.
2186  */
2187 static CPUArchId *pc_find_cpu_slot(MachineState *ms, uint32_t id, int *idx)
2188 {
2189     CPUArchId apic_id, *found_cpu;
2190 
2191     apic_id.arch_id = id;
2192     found_cpu = bsearch(&apic_id, ms->possible_cpus->cpus,
2193         ms->possible_cpus->len, sizeof(*ms->possible_cpus->cpus),
2194         pc_apic_cmp);
2195     if (found_cpu && idx) {
2196         *idx = found_cpu - ms->possible_cpus->cpus;
2197     }
2198     return found_cpu;
2199 }
2200 
2201 static void pc_cpu_plug(HotplugHandler *hotplug_dev,
2202                         DeviceState *dev, Error **errp)
2203 {
2204     CPUArchId *found_cpu;
2205     Error *local_err = NULL;
2206     X86CPU *cpu = X86_CPU(dev);
2207     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2208 
2209     if (pcms->acpi_dev) {
2210         hotplug_handler_plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
2211         if (local_err) {
2212             goto out;
2213         }
2214     }
2215 
2216     /* increment the number of CPUs */
2217     pcms->boot_cpus++;
2218     if (pcms->rtc) {
2219         rtc_set_cpus_count(pcms->rtc, pcms->boot_cpus);
2220     }
2221     if (pcms->fw_cfg) {
2222         fw_cfg_modify_i16(pcms->fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
2223     }
2224 
2225     found_cpu = pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, NULL);
2226     found_cpu->cpu = OBJECT(dev);
2227 out:
2228     error_propagate(errp, local_err);
2229 }
2230 static void pc_cpu_unplug_request_cb(HotplugHandler *hotplug_dev,
2231                                      DeviceState *dev, Error **errp)
2232 {
2233     int idx = -1;
2234     Error *local_err = NULL;
2235     X86CPU *cpu = X86_CPU(dev);
2236     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2237 
2238     if (!pcms->acpi_dev) {
2239         error_setg(&local_err, "CPU hot unplug not supported without ACPI");
2240         goto out;
2241     }
2242 
2243     pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, &idx);
2244     assert(idx != -1);
2245     if (idx == 0) {
2246         error_setg(&local_err, "Boot CPU is unpluggable");
2247         goto out;
2248     }
2249 
2250     hotplug_handler_unplug_request(HOTPLUG_HANDLER(pcms->acpi_dev), dev,
2251                                    &local_err);
2252     if (local_err) {
2253         goto out;
2254     }
2255 
2256  out:
2257     error_propagate(errp, local_err);
2258 
2259 }
2260 
2261 static void pc_cpu_unplug_cb(HotplugHandler *hotplug_dev,
2262                              DeviceState *dev, Error **errp)
2263 {
2264     CPUArchId *found_cpu;
2265     Error *local_err = NULL;
2266     X86CPU *cpu = X86_CPU(dev);
2267     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2268 
2269     hotplug_handler_unplug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
2270     if (local_err) {
2271         goto out;
2272     }
2273 
2274     found_cpu = pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, NULL);
2275     found_cpu->cpu = NULL;
2276     object_property_set_bool(OBJECT(dev), false, "realized", NULL);
2277 
2278     /* decrement the number of CPUs */
2279     pcms->boot_cpus--;
2280     /* Update the number of CPUs in CMOS */
2281     rtc_set_cpus_count(pcms->rtc, pcms->boot_cpus);
2282     fw_cfg_modify_i16(pcms->fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
2283  out:
2284     error_propagate(errp, local_err);
2285 }
2286 
2287 static void pc_cpu_pre_plug(HotplugHandler *hotplug_dev,
2288                             DeviceState *dev, Error **errp)
2289 {
2290     int idx;
2291     CPUState *cs;
2292     CPUArchId *cpu_slot;
2293     X86CPUTopoInfo topo;
2294     X86CPU *cpu = X86_CPU(dev);
2295     MachineState *ms = MACHINE(hotplug_dev);
2296     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2297 
2298     if(!object_dynamic_cast(OBJECT(cpu), ms->cpu_type)) {
2299         error_setg(errp, "Invalid CPU type, expected cpu type: '%s'",
2300                    ms->cpu_type);
2301         return;
2302     }
2303 
2304     /* if APIC ID is not set, set it based on socket/core/thread properties */
2305     if (cpu->apic_id == UNASSIGNED_APIC_ID) {
2306         int max_socket = (max_cpus - 1) / smp_threads / smp_cores;
2307 
2308         if (cpu->socket_id < 0) {
2309             error_setg(errp, "CPU socket-id is not set");
2310             return;
2311         } else if (cpu->socket_id > max_socket) {
2312             error_setg(errp, "Invalid CPU socket-id: %u must be in range 0:%u",
2313                        cpu->socket_id, max_socket);
2314             return;
2315         }
2316         if (cpu->core_id < 0) {
2317             error_setg(errp, "CPU core-id is not set");
2318             return;
2319         } else if (cpu->core_id > (smp_cores - 1)) {
2320             error_setg(errp, "Invalid CPU core-id: %u must be in range 0:%u",
2321                        cpu->core_id, smp_cores - 1);
2322             return;
2323         }
2324         if (cpu->thread_id < 0) {
2325             error_setg(errp, "CPU thread-id is not set");
2326             return;
2327         } else if (cpu->thread_id > (smp_threads - 1)) {
2328             error_setg(errp, "Invalid CPU thread-id: %u must be in range 0:%u",
2329                        cpu->thread_id, smp_threads - 1);
2330             return;
2331         }
2332 
2333         topo.pkg_id = cpu->socket_id;
2334         topo.core_id = cpu->core_id;
2335         topo.smt_id = cpu->thread_id;
2336         cpu->apic_id = apicid_from_topo_ids(smp_cores, smp_threads, &topo);
2337     }
2338 
2339     cpu_slot = pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, &idx);
2340     if (!cpu_slot) {
2341         MachineState *ms = MACHINE(pcms);
2342 
2343         x86_topo_ids_from_apicid(cpu->apic_id, smp_cores, smp_threads, &topo);
2344         error_setg(errp, "Invalid CPU [socket: %u, core: %u, thread: %u] with"
2345                   " APIC ID %" PRIu32 ", valid index range 0:%d",
2346                    topo.pkg_id, topo.core_id, topo.smt_id, cpu->apic_id,
2347                    ms->possible_cpus->len - 1);
2348         return;
2349     }
2350 
2351     if (cpu_slot->cpu) {
2352         error_setg(errp, "CPU[%d] with APIC ID %" PRIu32 " exists",
2353                    idx, cpu->apic_id);
2354         return;
2355     }
2356 
2357     /* if 'address' properties socket-id/core-id/thread-id are not set, set them
2358      * so that machine_query_hotpluggable_cpus would show correct values
2359      */
2360     /* TODO: move socket_id/core_id/thread_id checks into x86_cpu_realizefn()
2361      * once -smp refactoring is complete and there will be CPU private
2362      * CPUState::nr_cores and CPUState::nr_threads fields instead of globals */
2363     x86_topo_ids_from_apicid(cpu->apic_id, smp_cores, smp_threads, &topo);
2364     if (cpu->socket_id != -1 && cpu->socket_id != topo.pkg_id) {
2365         error_setg(errp, "property socket-id: %u doesn't match set apic-id:"
2366             " 0x%x (socket-id: %u)", cpu->socket_id, cpu->apic_id, topo.pkg_id);
2367         return;
2368     }
2369     cpu->socket_id = topo.pkg_id;
2370 
2371     if (cpu->core_id != -1 && cpu->core_id != topo.core_id) {
2372         error_setg(errp, "property core-id: %u doesn't match set apic-id:"
2373             " 0x%x (core-id: %u)", cpu->core_id, cpu->apic_id, topo.core_id);
2374         return;
2375     }
2376     cpu->core_id = topo.core_id;
2377 
2378     if (cpu->thread_id != -1 && cpu->thread_id != topo.smt_id) {
2379         error_setg(errp, "property thread-id: %u doesn't match set apic-id:"
2380             " 0x%x (thread-id: %u)", cpu->thread_id, cpu->apic_id, topo.smt_id);
2381         return;
2382     }
2383     cpu->thread_id = topo.smt_id;
2384 
2385     if (cpu->hyperv_vpindex && !kvm_hv_vpindex_settable()) {
2386         error_setg(errp, "kernel doesn't allow setting HyperV VP_INDEX");
2387         return;
2388     }
2389 
2390     cs = CPU(cpu);
2391     cs->cpu_index = idx;
2392 
2393     numa_cpu_pre_plug(cpu_slot, dev, errp);
2394 }
2395 
2396 static void pc_machine_device_pre_plug_cb(HotplugHandler *hotplug_dev,
2397                                           DeviceState *dev, Error **errp)
2398 {
2399     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2400         pc_memory_pre_plug(hotplug_dev, dev, errp);
2401     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2402         pc_cpu_pre_plug(hotplug_dev, dev, errp);
2403     }
2404 }
2405 
2406 static void pc_machine_device_plug_cb(HotplugHandler *hotplug_dev,
2407                                       DeviceState *dev, Error **errp)
2408 {
2409     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2410         pc_memory_plug(hotplug_dev, dev, errp);
2411     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2412         pc_cpu_plug(hotplug_dev, dev, errp);
2413     }
2414 }
2415 
2416 static void pc_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
2417                                                 DeviceState *dev, Error **errp)
2418 {
2419     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2420         pc_memory_unplug_request(hotplug_dev, dev, errp);
2421     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2422         pc_cpu_unplug_request_cb(hotplug_dev, dev, errp);
2423     } else {
2424         error_setg(errp, "acpi: device unplug request for not supported device"
2425                    " type: %s", object_get_typename(OBJECT(dev)));
2426     }
2427 }
2428 
2429 static void pc_machine_device_unplug_cb(HotplugHandler *hotplug_dev,
2430                                         DeviceState *dev, Error **errp)
2431 {
2432     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2433         pc_memory_unplug(hotplug_dev, dev, errp);
2434     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2435         pc_cpu_unplug_cb(hotplug_dev, dev, errp);
2436     } else {
2437         error_setg(errp, "acpi: device unplug for not supported device"
2438                    " type: %s", object_get_typename(OBJECT(dev)));
2439     }
2440 }
2441 
2442 static HotplugHandler *pc_get_hotplug_handler(MachineState *machine,
2443                                              DeviceState *dev)
2444 {
2445     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) ||
2446         object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2447         return HOTPLUG_HANDLER(machine);
2448     }
2449 
2450     return NULL;
2451 }
2452 
2453 static void
2454 pc_machine_get_device_memory_region_size(Object *obj, Visitor *v,
2455                                          const char *name, void *opaque,
2456                                          Error **errp)
2457 {
2458     MachineState *ms = MACHINE(obj);
2459     int64_t value = memory_region_size(&ms->device_memory->mr);
2460 
2461     visit_type_int(v, name, &value, errp);
2462 }
2463 
2464 static void pc_machine_get_max_ram_below_4g(Object *obj, Visitor *v,
2465                                             const char *name, void *opaque,
2466                                             Error **errp)
2467 {
2468     PCMachineState *pcms = PC_MACHINE(obj);
2469     uint64_t value = pcms->max_ram_below_4g;
2470 
2471     visit_type_size(v, name, &value, errp);
2472 }
2473 
2474 static void pc_machine_set_max_ram_below_4g(Object *obj, Visitor *v,
2475                                             const char *name, void *opaque,
2476                                             Error **errp)
2477 {
2478     PCMachineState *pcms = PC_MACHINE(obj);
2479     Error *error = NULL;
2480     uint64_t value;
2481 
2482     visit_type_size(v, name, &value, &error);
2483     if (error) {
2484         error_propagate(errp, error);
2485         return;
2486     }
2487     if (value > 4 * GiB) {
2488         error_setg(&error,
2489                    "Machine option 'max-ram-below-4g=%"PRIu64
2490                    "' expects size less than or equal to 4G", value);
2491         error_propagate(errp, error);
2492         return;
2493     }
2494 
2495     if (value < 1 * MiB) {
2496         warn_report("Only %" PRIu64 " bytes of RAM below the 4GiB boundary,"
2497                     "BIOS may not work with less than 1MiB", value);
2498     }
2499 
2500     pcms->max_ram_below_4g = value;
2501 }
2502 
2503 static void pc_machine_get_vmport(Object *obj, Visitor *v, const char *name,
2504                                   void *opaque, Error **errp)
2505 {
2506     PCMachineState *pcms = PC_MACHINE(obj);
2507     OnOffAuto vmport = pcms->vmport;
2508 
2509     visit_type_OnOffAuto(v, name, &vmport, errp);
2510 }
2511 
2512 static void pc_machine_set_vmport(Object *obj, Visitor *v, const char *name,
2513                                   void *opaque, Error **errp)
2514 {
2515     PCMachineState *pcms = PC_MACHINE(obj);
2516 
2517     visit_type_OnOffAuto(v, name, &pcms->vmport, errp);
2518 }
2519 
2520 bool pc_machine_is_smm_enabled(PCMachineState *pcms)
2521 {
2522     bool smm_available = false;
2523 
2524     if (pcms->smm == ON_OFF_AUTO_OFF) {
2525         return false;
2526     }
2527 
2528     if (tcg_enabled() || qtest_enabled()) {
2529         smm_available = true;
2530     } else if (kvm_enabled()) {
2531         smm_available = kvm_has_smm();
2532     }
2533 
2534     if (smm_available) {
2535         return true;
2536     }
2537 
2538     if (pcms->smm == ON_OFF_AUTO_ON) {
2539         error_report("System Management Mode not supported by this hypervisor.");
2540         exit(1);
2541     }
2542     return false;
2543 }
2544 
2545 static void pc_machine_get_smm(Object *obj, Visitor *v, const char *name,
2546                                void *opaque, Error **errp)
2547 {
2548     PCMachineState *pcms = PC_MACHINE(obj);
2549     OnOffAuto smm = pcms->smm;
2550 
2551     visit_type_OnOffAuto(v, name, &smm, errp);
2552 }
2553 
2554 static void pc_machine_set_smm(Object *obj, Visitor *v, const char *name,
2555                                void *opaque, Error **errp)
2556 {
2557     PCMachineState *pcms = PC_MACHINE(obj);
2558 
2559     visit_type_OnOffAuto(v, name, &pcms->smm, errp);
2560 }
2561 
2562 static bool pc_machine_get_smbus(Object *obj, Error **errp)
2563 {
2564     PCMachineState *pcms = PC_MACHINE(obj);
2565 
2566     return pcms->smbus_enabled;
2567 }
2568 
2569 static void pc_machine_set_smbus(Object *obj, bool value, Error **errp)
2570 {
2571     PCMachineState *pcms = PC_MACHINE(obj);
2572 
2573     pcms->smbus_enabled = value;
2574 }
2575 
2576 static bool pc_machine_get_sata(Object *obj, Error **errp)
2577 {
2578     PCMachineState *pcms = PC_MACHINE(obj);
2579 
2580     return pcms->sata_enabled;
2581 }
2582 
2583 static void pc_machine_set_sata(Object *obj, bool value, Error **errp)
2584 {
2585     PCMachineState *pcms = PC_MACHINE(obj);
2586 
2587     pcms->sata_enabled = value;
2588 }
2589 
2590 static bool pc_machine_get_pit(Object *obj, Error **errp)
2591 {
2592     PCMachineState *pcms = PC_MACHINE(obj);
2593 
2594     return pcms->pit_enabled;
2595 }
2596 
2597 static void pc_machine_set_pit(Object *obj, bool value, Error **errp)
2598 {
2599     PCMachineState *pcms = PC_MACHINE(obj);
2600 
2601     pcms->pit_enabled = value;
2602 }
2603 
2604 static void pc_machine_initfn(Object *obj)
2605 {
2606     PCMachineState *pcms = PC_MACHINE(obj);
2607 
2608     pcms->max_ram_below_4g = 0; /* use default */
2609     pcms->smm = ON_OFF_AUTO_AUTO;
2610     pcms->vmport = ON_OFF_AUTO_AUTO;
2611     /* acpi build is enabled by default if machine supports it */
2612     pcms->acpi_build_enabled = PC_MACHINE_GET_CLASS(pcms)->has_acpi_build;
2613     pcms->smbus_enabled = true;
2614     pcms->sata_enabled = true;
2615     pcms->pit_enabled = true;
2616 
2617     pc_system_flash_create(pcms);
2618 }
2619 
2620 static void pc_machine_reset(void)
2621 {
2622     CPUState *cs;
2623     X86CPU *cpu;
2624 
2625     qemu_devices_reset();
2626 
2627     /* Reset APIC after devices have been reset to cancel
2628      * any changes that qemu_devices_reset() might have done.
2629      */
2630     CPU_FOREACH(cs) {
2631         cpu = X86_CPU(cs);
2632 
2633         if (cpu->apic_state) {
2634             device_reset(cpu->apic_state);
2635         }
2636     }
2637 }
2638 
2639 static CpuInstanceProperties
2640 pc_cpu_index_to_props(MachineState *ms, unsigned cpu_index)
2641 {
2642     MachineClass *mc = MACHINE_GET_CLASS(ms);
2643     const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(ms);
2644 
2645     assert(cpu_index < possible_cpus->len);
2646     return possible_cpus->cpus[cpu_index].props;
2647 }
2648 
2649 static int64_t pc_get_default_cpu_node_id(const MachineState *ms, int idx)
2650 {
2651    X86CPUTopoInfo topo;
2652 
2653    assert(idx < ms->possible_cpus->len);
2654    x86_topo_ids_from_apicid(ms->possible_cpus->cpus[idx].arch_id,
2655                             smp_cores, smp_threads, &topo);
2656    return topo.pkg_id % nb_numa_nodes;
2657 }
2658 
2659 static const CPUArchIdList *pc_possible_cpu_arch_ids(MachineState *ms)
2660 {
2661     int i;
2662 
2663     if (ms->possible_cpus) {
2664         /*
2665          * make sure that max_cpus hasn't changed since the first use, i.e.
2666          * -smp hasn't been parsed after it
2667         */
2668         assert(ms->possible_cpus->len == max_cpus);
2669         return ms->possible_cpus;
2670     }
2671 
2672     ms->possible_cpus = g_malloc0(sizeof(CPUArchIdList) +
2673                                   sizeof(CPUArchId) * max_cpus);
2674     ms->possible_cpus->len = max_cpus;
2675     for (i = 0; i < ms->possible_cpus->len; i++) {
2676         X86CPUTopoInfo topo;
2677 
2678         ms->possible_cpus->cpus[i].type = ms->cpu_type;
2679         ms->possible_cpus->cpus[i].vcpus_count = 1;
2680         ms->possible_cpus->cpus[i].arch_id = x86_cpu_apic_id_from_index(i);
2681         x86_topo_ids_from_apicid(ms->possible_cpus->cpus[i].arch_id,
2682                                  smp_cores, smp_threads, &topo);
2683         ms->possible_cpus->cpus[i].props.has_socket_id = true;
2684         ms->possible_cpus->cpus[i].props.socket_id = topo.pkg_id;
2685         ms->possible_cpus->cpus[i].props.has_core_id = true;
2686         ms->possible_cpus->cpus[i].props.core_id = topo.core_id;
2687         ms->possible_cpus->cpus[i].props.has_thread_id = true;
2688         ms->possible_cpus->cpus[i].props.thread_id = topo.smt_id;
2689     }
2690     return ms->possible_cpus;
2691 }
2692 
2693 static void x86_nmi(NMIState *n, int cpu_index, Error **errp)
2694 {
2695     /* cpu index isn't used */
2696     CPUState *cs;
2697 
2698     CPU_FOREACH(cs) {
2699         X86CPU *cpu = X86_CPU(cs);
2700 
2701         if (!cpu->apic_state) {
2702             cpu_interrupt(cs, CPU_INTERRUPT_NMI);
2703         } else {
2704             apic_deliver_nmi(cpu->apic_state);
2705         }
2706     }
2707 }
2708 
2709 static void pc_machine_class_init(ObjectClass *oc, void *data)
2710 {
2711     MachineClass *mc = MACHINE_CLASS(oc);
2712     PCMachineClass *pcmc = PC_MACHINE_CLASS(oc);
2713     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
2714     NMIClass *nc = NMI_CLASS(oc);
2715 
2716     pcmc->pci_enabled = true;
2717     pcmc->has_acpi_build = true;
2718     pcmc->rsdp_in_ram = true;
2719     pcmc->smbios_defaults = true;
2720     pcmc->smbios_uuid_encoded = true;
2721     pcmc->gigabyte_align = true;
2722     pcmc->has_reserved_memory = true;
2723     pcmc->kvmclock_enabled = true;
2724     pcmc->enforce_aligned_dimm = true;
2725     /* BIOS ACPI tables: 128K. Other BIOS datastructures: less than 4K reported
2726      * to be used at the moment, 32K should be enough for a while.  */
2727     pcmc->acpi_data_size = 0x20000 + 0x8000;
2728     pcmc->save_tsc_khz = true;
2729     pcmc->linuxboot_dma_enabled = true;
2730     pcmc->pvh_enabled = true;
2731     assert(!mc->get_hotplug_handler);
2732     mc->get_hotplug_handler = pc_get_hotplug_handler;
2733     mc->cpu_index_to_instance_props = pc_cpu_index_to_props;
2734     mc->get_default_cpu_node_id = pc_get_default_cpu_node_id;
2735     mc->possible_cpu_arch_ids = pc_possible_cpu_arch_ids;
2736     mc->auto_enable_numa_with_memhp = true;
2737     mc->has_hotpluggable_cpus = true;
2738     mc->default_boot_order = "cad";
2739     mc->hot_add_cpu = pc_hot_add_cpu;
2740     mc->block_default_type = IF_IDE;
2741     mc->max_cpus = 255;
2742     mc->reset = pc_machine_reset;
2743     hc->pre_plug = pc_machine_device_pre_plug_cb;
2744     hc->plug = pc_machine_device_plug_cb;
2745     hc->unplug_request = pc_machine_device_unplug_request_cb;
2746     hc->unplug = pc_machine_device_unplug_cb;
2747     nc->nmi_monitor_handler = x86_nmi;
2748     mc->default_cpu_type = TARGET_DEFAULT_CPU_TYPE;
2749     mc->nvdimm_supported = true;
2750 
2751     object_class_property_add(oc, PC_MACHINE_DEVMEM_REGION_SIZE, "int",
2752         pc_machine_get_device_memory_region_size, NULL,
2753         NULL, NULL, &error_abort);
2754 
2755     object_class_property_add(oc, PC_MACHINE_MAX_RAM_BELOW_4G, "size",
2756         pc_machine_get_max_ram_below_4g, pc_machine_set_max_ram_below_4g,
2757         NULL, NULL, &error_abort);
2758 
2759     object_class_property_set_description(oc, PC_MACHINE_MAX_RAM_BELOW_4G,
2760         "Maximum ram below the 4G boundary (32bit boundary)", &error_abort);
2761 
2762     object_class_property_add(oc, PC_MACHINE_SMM, "OnOffAuto",
2763         pc_machine_get_smm, pc_machine_set_smm,
2764         NULL, NULL, &error_abort);
2765     object_class_property_set_description(oc, PC_MACHINE_SMM,
2766         "Enable SMM (pc & q35)", &error_abort);
2767 
2768     object_class_property_add(oc, PC_MACHINE_VMPORT, "OnOffAuto",
2769         pc_machine_get_vmport, pc_machine_set_vmport,
2770         NULL, NULL, &error_abort);
2771     object_class_property_set_description(oc, PC_MACHINE_VMPORT,
2772         "Enable vmport (pc & q35)", &error_abort);
2773 
2774     object_class_property_add_bool(oc, PC_MACHINE_SMBUS,
2775         pc_machine_get_smbus, pc_machine_set_smbus, &error_abort);
2776 
2777     object_class_property_add_bool(oc, PC_MACHINE_SATA,
2778         pc_machine_get_sata, pc_machine_set_sata, &error_abort);
2779 
2780     object_class_property_add_bool(oc, PC_MACHINE_PIT,
2781         pc_machine_get_pit, pc_machine_set_pit, &error_abort);
2782 }
2783 
2784 static const TypeInfo pc_machine_info = {
2785     .name = TYPE_PC_MACHINE,
2786     .parent = TYPE_MACHINE,
2787     .abstract = true,
2788     .instance_size = sizeof(PCMachineState),
2789     .instance_init = pc_machine_initfn,
2790     .class_size = sizeof(PCMachineClass),
2791     .class_init = pc_machine_class_init,
2792     .interfaces = (InterfaceInfo[]) {
2793          { TYPE_HOTPLUG_HANDLER },
2794          { TYPE_NMI },
2795          { }
2796     },
2797 };
2798 
2799 static void pc_machine_register_types(void)
2800 {
2801     type_register_static(&pc_machine_info);
2802 }
2803 
2804 type_init(pc_machine_register_types)
2805