xref: /openbmc/qemu/hw/i386/pc.c (revision 583b28d45bf978832be024276fda8fdfe571c447)
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 "sysemu/cpus.h"
34 #include "hw/block/fdc.h"
35 #include "hw/ide.h"
36 #include "hw/pci/pci.h"
37 #include "hw/pci/pci_bus.h"
38 #include "hw/nvram/fw_cfg.h"
39 #include "hw/timer/hpet.h"
40 #include "hw/firmware/smbios.h"
41 #include "hw/loader.h"
42 #include "elf.h"
43 #include "multiboot.h"
44 #include "hw/timer/mc146818rtc.h"
45 #include "hw/dma/i8257.h"
46 #include "hw/timer/i8254.h"
47 #include "hw/input/i8042.h"
48 #include "hw/audio/pcspk.h"
49 #include "hw/pci/msi.h"
50 #include "hw/sysbus.h"
51 #include "sysemu/sysemu.h"
52 #include "sysemu/numa.h"
53 #include "sysemu/kvm.h"
54 #include "sysemu/qtest.h"
55 #include "kvm_i386.h"
56 #include "hw/xen/xen.h"
57 #include "ui/qemu-spice.h"
58 #include "exec/memory.h"
59 #include "exec/address-spaces.h"
60 #include "sysemu/arch_init.h"
61 #include "qemu/bitmap.h"
62 #include "qemu/config-file.h"
63 #include "qemu/error-report.h"
64 #include "qemu/option.h"
65 #include "hw/acpi/acpi.h"
66 #include "hw/acpi/cpu_hotplug.h"
67 #include "hw/boards.h"
68 #include "acpi-build.h"
69 #include "hw/mem/pc-dimm.h"
70 #include "qapi/error.h"
71 #include "qapi/qapi-visit-common.h"
72 #include "qapi/visitor.h"
73 #include "qom/cpu.h"
74 #include "hw/nmi.h"
75 #include "hw/usb.h"
76 #include "hw/i386/intel_iommu.h"
77 #include "hw/net/ne2000-isa.h"
78 
79 /* debug PC/ISA interrupts */
80 //#define DEBUG_IRQ
81 
82 #ifdef DEBUG_IRQ
83 #define DPRINTF(fmt, ...)                                       \
84     do { printf("CPUIRQ: " fmt , ## __VA_ARGS__); } while (0)
85 #else
86 #define DPRINTF(fmt, ...)
87 #endif
88 
89 #define FW_CFG_ACPI_TABLES (FW_CFG_ARCH_LOCAL + 0)
90 #define FW_CFG_SMBIOS_ENTRIES (FW_CFG_ARCH_LOCAL + 1)
91 #define FW_CFG_IRQ0_OVERRIDE (FW_CFG_ARCH_LOCAL + 2)
92 #define FW_CFG_E820_TABLE (FW_CFG_ARCH_LOCAL + 3)
93 #define FW_CFG_HPET (FW_CFG_ARCH_LOCAL + 4)
94 
95 #define E820_NR_ENTRIES		16
96 
97 struct e820_entry {
98     uint64_t address;
99     uint64_t length;
100     uint32_t type;
101 } QEMU_PACKED __attribute((__aligned__(4)));
102 
103 struct e820_table {
104     uint32_t count;
105     struct e820_entry entry[E820_NR_ENTRIES];
106 } QEMU_PACKED __attribute((__aligned__(4)));
107 
108 static struct e820_table e820_reserve;
109 static struct e820_entry *e820_table;
110 static unsigned e820_entries;
111 struct hpet_fw_config hpet_cfg = {.count = UINT8_MAX};
112 
113 GlobalProperty pc_compat_3_1[] = {
114     { "intel-iommu", "dma-drain", "off" },
115     { "Opteron_G3" "-" TYPE_X86_CPU, "rdtscp", "off" },
116     { "Opteron_G4" "-" TYPE_X86_CPU, "rdtscp", "off" },
117     { "Opteron_G4" "-" TYPE_X86_CPU, "npt", "off" },
118     { "Opteron_G4" "-" TYPE_X86_CPU, "nrip-save", "off" },
119     { "Opteron_G5" "-" TYPE_X86_CPU, "rdtscp", "off" },
120     { "Opteron_G5" "-" TYPE_X86_CPU, "npt", "off" },
121     { "Opteron_G5" "-" TYPE_X86_CPU, "nrip-save", "off" },
122     { "EPYC" "-" TYPE_X86_CPU, "npt", "off" },
123     { "EPYC" "-" TYPE_X86_CPU, "nrip-save", "off" },
124     { "EPYC-IBPB" "-" TYPE_X86_CPU, "npt", "off" },
125     { "EPYC-IBPB" "-" TYPE_X86_CPU, "nrip-save", "off" },
126     { "Skylake-Client" "-" TYPE_X86_CPU,      "mpx", "on" },
127     { "Skylake-Client-IBRS" "-" TYPE_X86_CPU, "mpx", "on" },
128     { "Skylake-Server" "-" TYPE_X86_CPU,      "mpx", "on" },
129     { "Skylake-Server-IBRS" "-" TYPE_X86_CPU, "mpx", "on" },
130     { "Cascadelake-Server" "-" TYPE_X86_CPU,  "mpx", "on" },
131     { "Icelake-Client" "-" TYPE_X86_CPU,      "mpx", "on" },
132     { "Icelake-Server" "-" TYPE_X86_CPU,      "mpx", "on" },
133     { "Cascadelake-Server" "-" TYPE_X86_CPU, "stepping", "5" },
134 };
135 const size_t pc_compat_3_1_len = G_N_ELEMENTS(pc_compat_3_1);
136 
137 GlobalProperty pc_compat_3_0[] = {
138     { TYPE_X86_CPU, "x-hv-synic-kvm-only", "on" },
139     { "Skylake-Server" "-" TYPE_X86_CPU, "pku", "off" },
140     { "Skylake-Server-IBRS" "-" TYPE_X86_CPU, "pku", "off" },
141 };
142 const size_t pc_compat_3_0_len = G_N_ELEMENTS(pc_compat_3_0);
143 
144 GlobalProperty pc_compat_2_12[] = {
145     { TYPE_X86_CPU, "legacy-cache", "on" },
146     { TYPE_X86_CPU, "topoext", "off" },
147     { "EPYC-" TYPE_X86_CPU, "xlevel", "0x8000000a" },
148     { "EPYC-IBPB-" TYPE_X86_CPU, "xlevel", "0x8000000a" },
149 };
150 const size_t pc_compat_2_12_len = G_N_ELEMENTS(pc_compat_2_12);
151 
152 GlobalProperty pc_compat_2_11[] = {
153     { TYPE_X86_CPU, "x-migrate-smi-count", "off" },
154     { "Skylake-Server" "-" TYPE_X86_CPU, "clflushopt", "off" },
155 };
156 const size_t pc_compat_2_11_len = G_N_ELEMENTS(pc_compat_2_11);
157 
158 GlobalProperty pc_compat_2_10[] = {
159     { TYPE_X86_CPU, "x-hv-max-vps", "0x40" },
160     { "i440FX-pcihost", "x-pci-hole64-fix", "off" },
161     { "q35-pcihost", "x-pci-hole64-fix", "off" },
162 };
163 const size_t pc_compat_2_10_len = G_N_ELEMENTS(pc_compat_2_10);
164 
165 GlobalProperty pc_compat_2_9[] = {
166     { "mch", "extended-tseg-mbytes", "0" },
167 };
168 const size_t pc_compat_2_9_len = G_N_ELEMENTS(pc_compat_2_9);
169 
170 GlobalProperty pc_compat_2_8[] = {
171     { TYPE_X86_CPU, "tcg-cpuid", "off" },
172     { "kvmclock", "x-mach-use-reliable-get-clock", "off" },
173     { "ICH9-LPC", "x-smi-broadcast", "off" },
174     { TYPE_X86_CPU, "vmware-cpuid-freq", "off" },
175     { "Haswell-" TYPE_X86_CPU, "stepping", "1" },
176 };
177 const size_t pc_compat_2_8_len = G_N_ELEMENTS(pc_compat_2_8);
178 
179 GlobalProperty pc_compat_2_7[] = {
180     { TYPE_X86_CPU, "l3-cache", "off" },
181     { TYPE_X86_CPU, "full-cpuid-auto-level", "off" },
182     { "Opteron_G3" "-" TYPE_X86_CPU, "family", "15" },
183     { "Opteron_G3" "-" TYPE_X86_CPU, "model", "6" },
184     { "Opteron_G3" "-" TYPE_X86_CPU, "stepping", "1" },
185     { "isa-pcspk", "migrate", "off" },
186 };
187 const size_t pc_compat_2_7_len = G_N_ELEMENTS(pc_compat_2_7);
188 
189 GlobalProperty pc_compat_2_6[] = {
190     { TYPE_X86_CPU, "cpuid-0xb", "off" },
191     { "vmxnet3", "romfile", "" },
192     { TYPE_X86_CPU, "fill-mtrr-mask", "off" },
193     { "apic-common", "legacy-instance-id", "on", }
194 };
195 const size_t pc_compat_2_6_len = G_N_ELEMENTS(pc_compat_2_6);
196 
197 GlobalProperty pc_compat_2_5[] = {};
198 const size_t pc_compat_2_5_len = G_N_ELEMENTS(pc_compat_2_5);
199 
200 GlobalProperty pc_compat_2_4[] = {
201     PC_CPU_MODEL_IDS("2.4.0")
202     { "Haswell-" TYPE_X86_CPU, "abm", "off" },
203     { "Haswell-noTSX-" TYPE_X86_CPU, "abm", "off" },
204     { "Broadwell-" TYPE_X86_CPU, "abm", "off" },
205     { "Broadwell-noTSX-" TYPE_X86_CPU, "abm", "off" },
206     { "host" "-" TYPE_X86_CPU, "host-cache-info", "on" },
207     { TYPE_X86_CPU, "check", "off" },
208     { "qemu64" "-" TYPE_X86_CPU, "sse4a", "on" },
209     { "qemu64" "-" TYPE_X86_CPU, "abm", "on" },
210     { "qemu64" "-" TYPE_X86_CPU, "popcnt", "on" },
211     { "qemu32" "-" TYPE_X86_CPU, "popcnt", "on" },
212     { "Opteron_G2" "-" TYPE_X86_CPU, "rdtscp", "on" },
213     { "Opteron_G3" "-" TYPE_X86_CPU, "rdtscp", "on" },
214     { "Opteron_G4" "-" TYPE_X86_CPU, "rdtscp", "on" },
215     { "Opteron_G5" "-" TYPE_X86_CPU, "rdtscp", "on", }
216 };
217 const size_t pc_compat_2_4_len = G_N_ELEMENTS(pc_compat_2_4);
218 
219 GlobalProperty pc_compat_2_3[] = {
220     PC_CPU_MODEL_IDS("2.3.0")
221     { TYPE_X86_CPU, "arat", "off" },
222     { "qemu64" "-" TYPE_X86_CPU, "min-level", "4" },
223     { "kvm64" "-" TYPE_X86_CPU, "min-level", "5" },
224     { "pentium3" "-" TYPE_X86_CPU, "min-level", "2" },
225     { "n270" "-" TYPE_X86_CPU, "min-level", "5" },
226     { "Conroe" "-" TYPE_X86_CPU, "min-level", "4" },
227     { "Penryn" "-" TYPE_X86_CPU, "min-level", "4" },
228     { "Nehalem" "-" TYPE_X86_CPU, "min-level", "4" },
229     { "n270" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
230     { "Penryn" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
231     { "Conroe" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
232     { "Nehalem" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
233     { "Westmere" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
234     { "SandyBridge" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
235     { "IvyBridge" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
236     { "Haswell" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
237     { "Haswell-noTSX" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
238     { "Broadwell" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
239     { "Broadwell-noTSX" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
240     { TYPE_X86_CPU, "kvm-no-smi-migration", "on" },
241 };
242 const size_t pc_compat_2_3_len = G_N_ELEMENTS(pc_compat_2_3);
243 
244 GlobalProperty pc_compat_2_2[] = {
245     PC_CPU_MODEL_IDS("2.2.0")
246     { "kvm64" "-" TYPE_X86_CPU, "vme", "off" },
247     { "kvm32" "-" TYPE_X86_CPU, "vme", "off" },
248     { "Conroe" "-" TYPE_X86_CPU, "vme", "off" },
249     { "Penryn" "-" TYPE_X86_CPU, "vme", "off" },
250     { "Nehalem" "-" TYPE_X86_CPU, "vme", "off" },
251     { "Westmere" "-" TYPE_X86_CPU, "vme", "off" },
252     { "SandyBridge" "-" TYPE_X86_CPU, "vme", "off" },
253     { "Haswell" "-" TYPE_X86_CPU, "vme", "off" },
254     { "Broadwell" "-" TYPE_X86_CPU, "vme", "off" },
255     { "Opteron_G1" "-" TYPE_X86_CPU, "vme", "off" },
256     { "Opteron_G2" "-" TYPE_X86_CPU, "vme", "off" },
257     { "Opteron_G3" "-" TYPE_X86_CPU, "vme", "off" },
258     { "Opteron_G4" "-" TYPE_X86_CPU, "vme", "off" },
259     { "Opteron_G5" "-" TYPE_X86_CPU, "vme", "off" },
260     { "Haswell" "-" TYPE_X86_CPU, "f16c", "off" },
261     { "Haswell" "-" TYPE_X86_CPU, "rdrand", "off" },
262     { "Broadwell" "-" TYPE_X86_CPU, "f16c", "off" },
263     { "Broadwell" "-" TYPE_X86_CPU, "rdrand", "off" },
264 };
265 const size_t pc_compat_2_2_len = G_N_ELEMENTS(pc_compat_2_2);
266 
267 GlobalProperty pc_compat_2_1[] = {
268     PC_CPU_MODEL_IDS("2.1.0")
269     { "coreduo" "-" TYPE_X86_CPU, "vmx", "on" },
270     { "core2duo" "-" TYPE_X86_CPU, "vmx", "on" },
271 };
272 const size_t pc_compat_2_1_len = G_N_ELEMENTS(pc_compat_2_1);
273 
274 GlobalProperty pc_compat_2_0[] = {
275     PC_CPU_MODEL_IDS("2.0.0")
276     { "virtio-scsi-pci", "any_layout", "off" },
277     { "PIIX4_PM", "memory-hotplug-support", "off" },
278     { "apic", "version", "0x11" },
279     { "nec-usb-xhci", "superspeed-ports-first", "off" },
280     { "nec-usb-xhci", "force-pcie-endcap", "on" },
281     { "pci-serial", "prog_if", "0" },
282     { "pci-serial-2x", "prog_if", "0" },
283     { "pci-serial-4x", "prog_if", "0" },
284     { "virtio-net-pci", "guest_announce", "off" },
285     { "ICH9-LPC", "memory-hotplug-support", "off" },
286     { "xio3130-downstream", COMPAT_PROP_PCP, "off" },
287     { "ioh3420", COMPAT_PROP_PCP, "off" },
288 };
289 const size_t pc_compat_2_0_len = G_N_ELEMENTS(pc_compat_2_0);
290 
291 GlobalProperty pc_compat_1_7[] = {
292     PC_CPU_MODEL_IDS("1.7.0")
293     { TYPE_USB_DEVICE, "msos-desc", "no" },
294     { "PIIX4_PM", "acpi-pci-hotplug-with-bridge-support", "off" },
295     { "hpet", HPET_INTCAP, "4" },
296 };
297 const size_t pc_compat_1_7_len = G_N_ELEMENTS(pc_compat_1_7);
298 
299 GlobalProperty pc_compat_1_6[] = {
300     PC_CPU_MODEL_IDS("1.6.0")
301     { "e1000", "mitigation", "off" },
302     { "qemu64-" TYPE_X86_CPU, "model", "2" },
303     { "qemu32-" TYPE_X86_CPU, "model", "3" },
304     { "i440FX-pcihost", "short_root_bus", "1" },
305     { "q35-pcihost", "short_root_bus", "1" },
306 };
307 const size_t pc_compat_1_6_len = G_N_ELEMENTS(pc_compat_1_6);
308 
309 GlobalProperty pc_compat_1_5[] = {
310     PC_CPU_MODEL_IDS("1.5.0")
311     { "Conroe-" TYPE_X86_CPU, "model", "2" },
312     { "Conroe-" TYPE_X86_CPU, "min-level", "2" },
313     { "Penryn-" TYPE_X86_CPU, "model", "2" },
314     { "Penryn-" TYPE_X86_CPU, "min-level", "2" },
315     { "Nehalem-" TYPE_X86_CPU, "model", "2" },
316     { "Nehalem-" TYPE_X86_CPU, "min-level", "2" },
317     { "virtio-net-pci", "any_layout", "off" },
318     { TYPE_X86_CPU, "pmu", "on" },
319     { "i440FX-pcihost", "short_root_bus", "0" },
320     { "q35-pcihost", "short_root_bus", "0" },
321 };
322 const size_t pc_compat_1_5_len = G_N_ELEMENTS(pc_compat_1_5);
323 
324 GlobalProperty pc_compat_1_4[] = {
325     PC_CPU_MODEL_IDS("1.4.0")
326     { "scsi-hd", "discard_granularity", "0" },
327     { "scsi-cd", "discard_granularity", "0" },
328     { "scsi-disk", "discard_granularity", "0" },
329     { "ide-hd", "discard_granularity", "0" },
330     { "ide-cd", "discard_granularity", "0" },
331     { "ide-drive", "discard_granularity", "0" },
332     { "virtio-blk-pci", "discard_granularity", "0" },
333     /* DEV_NVECTORS_UNSPECIFIED as a uint32_t string: */
334     { "virtio-serial-pci", "vectors", "0xFFFFFFFF" },
335     { "virtio-net-pci", "ctrl_guest_offloads", "off" },
336     { "e1000", "romfile", "pxe-e1000.rom" },
337     { "ne2k_pci", "romfile", "pxe-ne2k_pci.rom" },
338     { "pcnet", "romfile", "pxe-pcnet.rom" },
339     { "rtl8139", "romfile", "pxe-rtl8139.rom" },
340     { "virtio-net-pci", "romfile", "pxe-virtio.rom" },
341     { "486-" TYPE_X86_CPU, "model", "0" },
342     { "n270" "-" TYPE_X86_CPU, "movbe", "off" },
343     { "Westmere" "-" TYPE_X86_CPU, "pclmulqdq", "off" },
344 };
345 const size_t pc_compat_1_4_len = G_N_ELEMENTS(pc_compat_1_4);
346 
347 void gsi_handler(void *opaque, int n, int level)
348 {
349     GSIState *s = opaque;
350 
351     DPRINTF("pc: %s GSI %d\n", level ? "raising" : "lowering", n);
352     if (n < ISA_NUM_IRQS) {
353         qemu_set_irq(s->i8259_irq[n], level);
354     }
355     qemu_set_irq(s->ioapic_irq[n], level);
356 }
357 
358 static void ioport80_write(void *opaque, hwaddr addr, uint64_t data,
359                            unsigned size)
360 {
361 }
362 
363 static uint64_t ioport80_read(void *opaque, hwaddr addr, unsigned size)
364 {
365     return 0xffffffffffffffffULL;
366 }
367 
368 /* MSDOS compatibility mode FPU exception support */
369 static qemu_irq ferr_irq;
370 
371 void pc_register_ferr_irq(qemu_irq irq)
372 {
373     ferr_irq = irq;
374 }
375 
376 /* XXX: add IGNNE support */
377 void cpu_set_ferr(CPUX86State *s)
378 {
379     qemu_irq_raise(ferr_irq);
380 }
381 
382 static void ioportF0_write(void *opaque, hwaddr addr, uint64_t data,
383                            unsigned size)
384 {
385     qemu_irq_lower(ferr_irq);
386 }
387 
388 static uint64_t ioportF0_read(void *opaque, hwaddr addr, unsigned size)
389 {
390     return 0xffffffffffffffffULL;
391 }
392 
393 /* TSC handling */
394 uint64_t cpu_get_tsc(CPUX86State *env)
395 {
396     return cpu_get_ticks();
397 }
398 
399 /* IRQ handling */
400 int cpu_get_pic_interrupt(CPUX86State *env)
401 {
402     X86CPU *cpu = x86_env_get_cpu(env);
403     int intno;
404 
405     if (!kvm_irqchip_in_kernel()) {
406         intno = apic_get_interrupt(cpu->apic_state);
407         if (intno >= 0) {
408             return intno;
409         }
410         /* read the irq from the PIC */
411         if (!apic_accept_pic_intr(cpu->apic_state)) {
412             return -1;
413         }
414     }
415 
416     intno = pic_read_irq(isa_pic);
417     return intno;
418 }
419 
420 static void pic_irq_request(void *opaque, int irq, int level)
421 {
422     CPUState *cs = first_cpu;
423     X86CPU *cpu = X86_CPU(cs);
424 
425     DPRINTF("pic_irqs: %s irq %d\n", level? "raise" : "lower", irq);
426     if (cpu->apic_state && !kvm_irqchip_in_kernel()) {
427         CPU_FOREACH(cs) {
428             cpu = X86_CPU(cs);
429             if (apic_accept_pic_intr(cpu->apic_state)) {
430                 apic_deliver_pic_intr(cpu->apic_state, level);
431             }
432         }
433     } else {
434         if (level) {
435             cpu_interrupt(cs, CPU_INTERRUPT_HARD);
436         } else {
437             cpu_reset_interrupt(cs, CPU_INTERRUPT_HARD);
438         }
439     }
440 }
441 
442 /* PC cmos mappings */
443 
444 #define REG_EQUIPMENT_BYTE          0x14
445 
446 int cmos_get_fd_drive_type(FloppyDriveType fd0)
447 {
448     int val;
449 
450     switch (fd0) {
451     case FLOPPY_DRIVE_TYPE_144:
452         /* 1.44 Mb 3"5 drive */
453         val = 4;
454         break;
455     case FLOPPY_DRIVE_TYPE_288:
456         /* 2.88 Mb 3"5 drive */
457         val = 5;
458         break;
459     case FLOPPY_DRIVE_TYPE_120:
460         /* 1.2 Mb 5"5 drive */
461         val = 2;
462         break;
463     case FLOPPY_DRIVE_TYPE_NONE:
464     default:
465         val = 0;
466         break;
467     }
468     return val;
469 }
470 
471 static void cmos_init_hd(ISADevice *s, int type_ofs, int info_ofs,
472                          int16_t cylinders, int8_t heads, int8_t sectors)
473 {
474     rtc_set_memory(s, type_ofs, 47);
475     rtc_set_memory(s, info_ofs, cylinders);
476     rtc_set_memory(s, info_ofs + 1, cylinders >> 8);
477     rtc_set_memory(s, info_ofs + 2, heads);
478     rtc_set_memory(s, info_ofs + 3, 0xff);
479     rtc_set_memory(s, info_ofs + 4, 0xff);
480     rtc_set_memory(s, info_ofs + 5, 0xc0 | ((heads > 8) << 3));
481     rtc_set_memory(s, info_ofs + 6, cylinders);
482     rtc_set_memory(s, info_ofs + 7, cylinders >> 8);
483     rtc_set_memory(s, info_ofs + 8, sectors);
484 }
485 
486 /* convert boot_device letter to something recognizable by the bios */
487 static int boot_device2nibble(char boot_device)
488 {
489     switch(boot_device) {
490     case 'a':
491     case 'b':
492         return 0x01; /* floppy boot */
493     case 'c':
494         return 0x02; /* hard drive boot */
495     case 'd':
496         return 0x03; /* CD-ROM boot */
497     case 'n':
498         return 0x04; /* Network boot */
499     }
500     return 0;
501 }
502 
503 static void set_boot_dev(ISADevice *s, const char *boot_device, Error **errp)
504 {
505 #define PC_MAX_BOOT_DEVICES 3
506     int nbds, bds[3] = { 0, };
507     int i;
508 
509     nbds = strlen(boot_device);
510     if (nbds > PC_MAX_BOOT_DEVICES) {
511         error_setg(errp, "Too many boot devices for PC");
512         return;
513     }
514     for (i = 0; i < nbds; i++) {
515         bds[i] = boot_device2nibble(boot_device[i]);
516         if (bds[i] == 0) {
517             error_setg(errp, "Invalid boot device for PC: '%c'",
518                        boot_device[i]);
519             return;
520         }
521     }
522     rtc_set_memory(s, 0x3d, (bds[1] << 4) | bds[0]);
523     rtc_set_memory(s, 0x38, (bds[2] << 4) | (fd_bootchk ? 0x0 : 0x1));
524 }
525 
526 static void pc_boot_set(void *opaque, const char *boot_device, Error **errp)
527 {
528     set_boot_dev(opaque, boot_device, errp);
529 }
530 
531 static void pc_cmos_init_floppy(ISADevice *rtc_state, ISADevice *floppy)
532 {
533     int val, nb, i;
534     FloppyDriveType fd_type[2] = { FLOPPY_DRIVE_TYPE_NONE,
535                                    FLOPPY_DRIVE_TYPE_NONE };
536 
537     /* floppy type */
538     if (floppy) {
539         for (i = 0; i < 2; i++) {
540             fd_type[i] = isa_fdc_get_drive_type(floppy, i);
541         }
542     }
543     val = (cmos_get_fd_drive_type(fd_type[0]) << 4) |
544         cmos_get_fd_drive_type(fd_type[1]);
545     rtc_set_memory(rtc_state, 0x10, val);
546 
547     val = rtc_get_memory(rtc_state, REG_EQUIPMENT_BYTE);
548     nb = 0;
549     if (fd_type[0] != FLOPPY_DRIVE_TYPE_NONE) {
550         nb++;
551     }
552     if (fd_type[1] != FLOPPY_DRIVE_TYPE_NONE) {
553         nb++;
554     }
555     switch (nb) {
556     case 0:
557         break;
558     case 1:
559         val |= 0x01; /* 1 drive, ready for boot */
560         break;
561     case 2:
562         val |= 0x41; /* 2 drives, ready for boot */
563         break;
564     }
565     rtc_set_memory(rtc_state, REG_EQUIPMENT_BYTE, val);
566 }
567 
568 typedef struct pc_cmos_init_late_arg {
569     ISADevice *rtc_state;
570     BusState *idebus[2];
571 } pc_cmos_init_late_arg;
572 
573 typedef struct check_fdc_state {
574     ISADevice *floppy;
575     bool multiple;
576 } CheckFdcState;
577 
578 static int check_fdc(Object *obj, void *opaque)
579 {
580     CheckFdcState *state = opaque;
581     Object *fdc;
582     uint32_t iobase;
583     Error *local_err = NULL;
584 
585     fdc = object_dynamic_cast(obj, TYPE_ISA_FDC);
586     if (!fdc) {
587         return 0;
588     }
589 
590     iobase = object_property_get_uint(obj, "iobase", &local_err);
591     if (local_err || iobase != 0x3f0) {
592         error_free(local_err);
593         return 0;
594     }
595 
596     if (state->floppy) {
597         state->multiple = true;
598     } else {
599         state->floppy = ISA_DEVICE(obj);
600     }
601     return 0;
602 }
603 
604 static const char * const fdc_container_path[] = {
605     "/unattached", "/peripheral", "/peripheral-anon"
606 };
607 
608 /*
609  * Locate the FDC at IO address 0x3f0, in order to configure the CMOS registers
610  * and ACPI objects.
611  */
612 ISADevice *pc_find_fdc0(void)
613 {
614     int i;
615     Object *container;
616     CheckFdcState state = { 0 };
617 
618     for (i = 0; i < ARRAY_SIZE(fdc_container_path); i++) {
619         container = container_get(qdev_get_machine(), fdc_container_path[i]);
620         object_child_foreach(container, check_fdc, &state);
621     }
622 
623     if (state.multiple) {
624         warn_report("multiple floppy disk controllers with "
625                     "iobase=0x3f0 have been found");
626         error_printf("the one being picked for CMOS setup might not reflect "
627                      "your intent");
628     }
629 
630     return state.floppy;
631 }
632 
633 static void pc_cmos_init_late(void *opaque)
634 {
635     pc_cmos_init_late_arg *arg = opaque;
636     ISADevice *s = arg->rtc_state;
637     int16_t cylinders;
638     int8_t heads, sectors;
639     int val;
640     int i, trans;
641 
642     val = 0;
643     if (arg->idebus[0] && ide_get_geometry(arg->idebus[0], 0,
644                                            &cylinders, &heads, &sectors) >= 0) {
645         cmos_init_hd(s, 0x19, 0x1b, cylinders, heads, sectors);
646         val |= 0xf0;
647     }
648     if (arg->idebus[0] && ide_get_geometry(arg->idebus[0], 1,
649                                            &cylinders, &heads, &sectors) >= 0) {
650         cmos_init_hd(s, 0x1a, 0x24, cylinders, heads, sectors);
651         val |= 0x0f;
652     }
653     rtc_set_memory(s, 0x12, val);
654 
655     val = 0;
656     for (i = 0; i < 4; i++) {
657         /* NOTE: ide_get_geometry() returns the physical
658            geometry.  It is always such that: 1 <= sects <= 63, 1
659            <= heads <= 16, 1 <= cylinders <= 16383. The BIOS
660            geometry can be different if a translation is done. */
661         if (arg->idebus[i / 2] &&
662             ide_get_geometry(arg->idebus[i / 2], i % 2,
663                              &cylinders, &heads, &sectors) >= 0) {
664             trans = ide_get_bios_chs_trans(arg->idebus[i / 2], i % 2) - 1;
665             assert((trans & ~3) == 0);
666             val |= trans << (i * 2);
667         }
668     }
669     rtc_set_memory(s, 0x39, val);
670 
671     pc_cmos_init_floppy(s, pc_find_fdc0());
672 
673     qemu_unregister_reset(pc_cmos_init_late, opaque);
674 }
675 
676 void pc_cmos_init(PCMachineState *pcms,
677                   BusState *idebus0, BusState *idebus1,
678                   ISADevice *s)
679 {
680     int val;
681     static pc_cmos_init_late_arg arg;
682 
683     /* various important CMOS locations needed by PC/Bochs bios */
684 
685     /* memory size */
686     /* base memory (first MiB) */
687     val = MIN(pcms->below_4g_mem_size / KiB, 640);
688     rtc_set_memory(s, 0x15, val);
689     rtc_set_memory(s, 0x16, val >> 8);
690     /* extended memory (next 64MiB) */
691     if (pcms->below_4g_mem_size > 1 * MiB) {
692         val = (pcms->below_4g_mem_size - 1 * MiB) / KiB;
693     } else {
694         val = 0;
695     }
696     if (val > 65535)
697         val = 65535;
698     rtc_set_memory(s, 0x17, val);
699     rtc_set_memory(s, 0x18, val >> 8);
700     rtc_set_memory(s, 0x30, val);
701     rtc_set_memory(s, 0x31, val >> 8);
702     /* memory between 16MiB and 4GiB */
703     if (pcms->below_4g_mem_size > 16 * MiB) {
704         val = (pcms->below_4g_mem_size - 16 * MiB) / (64 * KiB);
705     } else {
706         val = 0;
707     }
708     if (val > 65535)
709         val = 65535;
710     rtc_set_memory(s, 0x34, val);
711     rtc_set_memory(s, 0x35, val >> 8);
712     /* memory above 4GiB */
713     val = pcms->above_4g_mem_size / 65536;
714     rtc_set_memory(s, 0x5b, val);
715     rtc_set_memory(s, 0x5c, val >> 8);
716     rtc_set_memory(s, 0x5d, val >> 16);
717 
718     object_property_add_link(OBJECT(pcms), "rtc_state",
719                              TYPE_ISA_DEVICE,
720                              (Object **)&pcms->rtc,
721                              object_property_allow_set_link,
722                              OBJ_PROP_LINK_STRONG, &error_abort);
723     object_property_set_link(OBJECT(pcms), OBJECT(s),
724                              "rtc_state", &error_abort);
725 
726     set_boot_dev(s, MACHINE(pcms)->boot_order, &error_fatal);
727 
728     val = 0;
729     val |= 0x02; /* FPU is there */
730     val |= 0x04; /* PS/2 mouse installed */
731     rtc_set_memory(s, REG_EQUIPMENT_BYTE, val);
732 
733     /* hard drives and FDC */
734     arg.rtc_state = s;
735     arg.idebus[0] = idebus0;
736     arg.idebus[1] = idebus1;
737     qemu_register_reset(pc_cmos_init_late, &arg);
738 }
739 
740 #define TYPE_PORT92 "port92"
741 #define PORT92(obj) OBJECT_CHECK(Port92State, (obj), TYPE_PORT92)
742 
743 /* port 92 stuff: could be split off */
744 typedef struct Port92State {
745     ISADevice parent_obj;
746 
747     MemoryRegion io;
748     uint8_t outport;
749     qemu_irq a20_out;
750 } Port92State;
751 
752 static void port92_write(void *opaque, hwaddr addr, uint64_t val,
753                          unsigned size)
754 {
755     Port92State *s = opaque;
756     int oldval = s->outport;
757 
758     DPRINTF("port92: write 0x%02" PRIx64 "\n", val);
759     s->outport = val;
760     qemu_set_irq(s->a20_out, (val >> 1) & 1);
761     if ((val & 1) && !(oldval & 1)) {
762         qemu_system_reset_request(SHUTDOWN_CAUSE_GUEST_RESET);
763     }
764 }
765 
766 static uint64_t port92_read(void *opaque, hwaddr addr,
767                             unsigned size)
768 {
769     Port92State *s = opaque;
770     uint32_t ret;
771 
772     ret = s->outport;
773     DPRINTF("port92: read 0x%02x\n", ret);
774     return ret;
775 }
776 
777 static void port92_init(ISADevice *dev, qemu_irq a20_out)
778 {
779     qdev_connect_gpio_out_named(DEVICE(dev), PORT92_A20_LINE, 0, a20_out);
780 }
781 
782 static const VMStateDescription vmstate_port92_isa = {
783     .name = "port92",
784     .version_id = 1,
785     .minimum_version_id = 1,
786     .fields = (VMStateField[]) {
787         VMSTATE_UINT8(outport, Port92State),
788         VMSTATE_END_OF_LIST()
789     }
790 };
791 
792 static void port92_reset(DeviceState *d)
793 {
794     Port92State *s = PORT92(d);
795 
796     s->outport &= ~1;
797 }
798 
799 static const MemoryRegionOps port92_ops = {
800     .read = port92_read,
801     .write = port92_write,
802     .impl = {
803         .min_access_size = 1,
804         .max_access_size = 1,
805     },
806     .endianness = DEVICE_LITTLE_ENDIAN,
807 };
808 
809 static void port92_initfn(Object *obj)
810 {
811     Port92State *s = PORT92(obj);
812 
813     memory_region_init_io(&s->io, OBJECT(s), &port92_ops, s, "port92", 1);
814 
815     s->outport = 0;
816 
817     qdev_init_gpio_out_named(DEVICE(obj), &s->a20_out, PORT92_A20_LINE, 1);
818 }
819 
820 static void port92_realizefn(DeviceState *dev, Error **errp)
821 {
822     ISADevice *isadev = ISA_DEVICE(dev);
823     Port92State *s = PORT92(dev);
824 
825     isa_register_ioport(isadev, &s->io, 0x92);
826 }
827 
828 static void port92_class_initfn(ObjectClass *klass, void *data)
829 {
830     DeviceClass *dc = DEVICE_CLASS(klass);
831 
832     dc->realize = port92_realizefn;
833     dc->reset = port92_reset;
834     dc->vmsd = &vmstate_port92_isa;
835     /*
836      * Reason: unlike ordinary ISA devices, this one needs additional
837      * wiring: its A20 output line needs to be wired up by
838      * port92_init().
839      */
840     dc->user_creatable = false;
841 }
842 
843 static const TypeInfo port92_info = {
844     .name          = TYPE_PORT92,
845     .parent        = TYPE_ISA_DEVICE,
846     .instance_size = sizeof(Port92State),
847     .instance_init = port92_initfn,
848     .class_init    = port92_class_initfn,
849 };
850 
851 static void port92_register_types(void)
852 {
853     type_register_static(&port92_info);
854 }
855 
856 type_init(port92_register_types)
857 
858 static void handle_a20_line_change(void *opaque, int irq, int level)
859 {
860     X86CPU *cpu = opaque;
861 
862     /* XXX: send to all CPUs ? */
863     /* XXX: add logic to handle multiple A20 line sources */
864     x86_cpu_set_a20(cpu, level);
865 }
866 
867 int e820_add_entry(uint64_t address, uint64_t length, uint32_t type)
868 {
869     int index = le32_to_cpu(e820_reserve.count);
870     struct e820_entry *entry;
871 
872     if (type != E820_RAM) {
873         /* old FW_CFG_E820_TABLE entry -- reservations only */
874         if (index >= E820_NR_ENTRIES) {
875             return -EBUSY;
876         }
877         entry = &e820_reserve.entry[index++];
878 
879         entry->address = cpu_to_le64(address);
880         entry->length = cpu_to_le64(length);
881         entry->type = cpu_to_le32(type);
882 
883         e820_reserve.count = cpu_to_le32(index);
884     }
885 
886     /* new "etc/e820" file -- include ram too */
887     e820_table = g_renew(struct e820_entry, e820_table, e820_entries + 1);
888     e820_table[e820_entries].address = cpu_to_le64(address);
889     e820_table[e820_entries].length = cpu_to_le64(length);
890     e820_table[e820_entries].type = cpu_to_le32(type);
891     e820_entries++;
892 
893     return e820_entries;
894 }
895 
896 int e820_get_num_entries(void)
897 {
898     return e820_entries;
899 }
900 
901 bool e820_get_entry(int idx, uint32_t type, uint64_t *address, uint64_t *length)
902 {
903     if (idx < e820_entries && e820_table[idx].type == cpu_to_le32(type)) {
904         *address = le64_to_cpu(e820_table[idx].address);
905         *length = le64_to_cpu(e820_table[idx].length);
906         return true;
907     }
908     return false;
909 }
910 
911 /* Enables contiguous-apic-ID mode, for compatibility */
912 static bool compat_apic_id_mode;
913 
914 void enable_compat_apic_id_mode(void)
915 {
916     compat_apic_id_mode = true;
917 }
918 
919 /* Calculates initial APIC ID for a specific CPU index
920  *
921  * Currently we need to be able to calculate the APIC ID from the CPU index
922  * alone (without requiring a CPU object), as the QEMU<->Seabios interfaces have
923  * no concept of "CPU index", and the NUMA tables on fw_cfg need the APIC ID of
924  * all CPUs up to max_cpus.
925  */
926 static uint32_t x86_cpu_apic_id_from_index(unsigned int cpu_index)
927 {
928     uint32_t correct_id;
929     static bool warned;
930 
931     correct_id = x86_apicid_from_cpu_idx(smp_cores, smp_threads, cpu_index);
932     if (compat_apic_id_mode) {
933         if (cpu_index != correct_id && !warned && !qtest_enabled()) {
934             error_report("APIC IDs set in compatibility mode, "
935                          "CPU topology won't match the configuration");
936             warned = true;
937         }
938         return cpu_index;
939     } else {
940         return correct_id;
941     }
942 }
943 
944 static void pc_build_smbios(PCMachineState *pcms)
945 {
946     uint8_t *smbios_tables, *smbios_anchor;
947     size_t smbios_tables_len, smbios_anchor_len;
948     struct smbios_phys_mem_area *mem_array;
949     unsigned i, array_count;
950     MachineState *ms = MACHINE(pcms);
951     X86CPU *cpu = X86_CPU(ms->possible_cpus->cpus[0].cpu);
952 
953     /* tell smbios about cpuid version and features */
954     smbios_set_cpuid(cpu->env.cpuid_version, cpu->env.features[FEAT_1_EDX]);
955 
956     smbios_tables = smbios_get_table_legacy(&smbios_tables_len);
957     if (smbios_tables) {
958         fw_cfg_add_bytes(pcms->fw_cfg, FW_CFG_SMBIOS_ENTRIES,
959                          smbios_tables, smbios_tables_len);
960     }
961 
962     /* build the array of physical mem area from e820 table */
963     mem_array = g_malloc0(sizeof(*mem_array) * e820_get_num_entries());
964     for (i = 0, array_count = 0; i < e820_get_num_entries(); i++) {
965         uint64_t addr, len;
966 
967         if (e820_get_entry(i, E820_RAM, &addr, &len)) {
968             mem_array[array_count].address = addr;
969             mem_array[array_count].length = len;
970             array_count++;
971         }
972     }
973     smbios_get_tables(mem_array, array_count,
974                       &smbios_tables, &smbios_tables_len,
975                       &smbios_anchor, &smbios_anchor_len);
976     g_free(mem_array);
977 
978     if (smbios_anchor) {
979         fw_cfg_add_file(pcms->fw_cfg, "etc/smbios/smbios-tables",
980                         smbios_tables, smbios_tables_len);
981         fw_cfg_add_file(pcms->fw_cfg, "etc/smbios/smbios-anchor",
982                         smbios_anchor, smbios_anchor_len);
983     }
984 }
985 
986 static FWCfgState *bochs_bios_init(AddressSpace *as, PCMachineState *pcms)
987 {
988     FWCfgState *fw_cfg;
989     uint64_t *numa_fw_cfg;
990     int i;
991     const CPUArchIdList *cpus;
992     MachineClass *mc = MACHINE_GET_CLASS(pcms);
993 
994     fw_cfg = fw_cfg_init_io_dma(FW_CFG_IO_BASE, FW_CFG_IO_BASE + 4, as);
995     fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
996 
997     /* FW_CFG_MAX_CPUS is a bit confusing/problematic on x86:
998      *
999      * For machine types prior to 1.8, SeaBIOS needs FW_CFG_MAX_CPUS for
1000      * building MPTable, ACPI MADT, ACPI CPU hotplug and ACPI SRAT table,
1001      * that tables are based on xAPIC ID and QEMU<->SeaBIOS interface
1002      * for CPU hotplug also uses APIC ID and not "CPU index".
1003      * This means that FW_CFG_MAX_CPUS is not the "maximum number of CPUs",
1004      * but the "limit to the APIC ID values SeaBIOS may see".
1005      *
1006      * So for compatibility reasons with old BIOSes we are stuck with
1007      * "etc/max-cpus" actually being apic_id_limit
1008      */
1009     fw_cfg_add_i16(fw_cfg, FW_CFG_MAX_CPUS, (uint16_t)pcms->apic_id_limit);
1010     fw_cfg_add_i64(fw_cfg, FW_CFG_RAM_SIZE, (uint64_t)ram_size);
1011     fw_cfg_add_bytes(fw_cfg, FW_CFG_ACPI_TABLES,
1012                      acpi_tables, acpi_tables_len);
1013     fw_cfg_add_i32(fw_cfg, FW_CFG_IRQ0_OVERRIDE, kvm_allows_irq0_override());
1014 
1015     fw_cfg_add_bytes(fw_cfg, FW_CFG_E820_TABLE,
1016                      &e820_reserve, sizeof(e820_reserve));
1017     fw_cfg_add_file(fw_cfg, "etc/e820", e820_table,
1018                     sizeof(struct e820_entry) * e820_entries);
1019 
1020     fw_cfg_add_bytes(fw_cfg, FW_CFG_HPET, &hpet_cfg, sizeof(hpet_cfg));
1021     /* allocate memory for the NUMA channel: one (64bit) word for the number
1022      * of nodes, one word for each VCPU->node and one word for each node to
1023      * hold the amount of memory.
1024      */
1025     numa_fw_cfg = g_new0(uint64_t, 1 + pcms->apic_id_limit + nb_numa_nodes);
1026     numa_fw_cfg[0] = cpu_to_le64(nb_numa_nodes);
1027     cpus = mc->possible_cpu_arch_ids(MACHINE(pcms));
1028     for (i = 0; i < cpus->len; i++) {
1029         unsigned int apic_id = cpus->cpus[i].arch_id;
1030         assert(apic_id < pcms->apic_id_limit);
1031         numa_fw_cfg[apic_id + 1] = cpu_to_le64(cpus->cpus[i].props.node_id);
1032     }
1033     for (i = 0; i < nb_numa_nodes; i++) {
1034         numa_fw_cfg[pcms->apic_id_limit + 1 + i] =
1035             cpu_to_le64(numa_info[i].node_mem);
1036     }
1037     fw_cfg_add_bytes(fw_cfg, FW_CFG_NUMA, numa_fw_cfg,
1038                      (1 + pcms->apic_id_limit + nb_numa_nodes) *
1039                      sizeof(*numa_fw_cfg));
1040 
1041     return fw_cfg;
1042 }
1043 
1044 static long get_file_size(FILE *f)
1045 {
1046     long where, size;
1047 
1048     /* XXX: on Unix systems, using fstat() probably makes more sense */
1049 
1050     where = ftell(f);
1051     fseek(f, 0, SEEK_END);
1052     size = ftell(f);
1053     fseek(f, where, SEEK_SET);
1054 
1055     return size;
1056 }
1057 
1058 /* setup_data types */
1059 #define SETUP_NONE     0
1060 #define SETUP_E820_EXT 1
1061 #define SETUP_DTB      2
1062 #define SETUP_PCI      3
1063 #define SETUP_EFI      4
1064 
1065 struct setup_data {
1066     uint64_t next;
1067     uint32_t type;
1068     uint32_t len;
1069     uint8_t data[0];
1070 } __attribute__((packed));
1071 
1072 static void load_linux(PCMachineState *pcms,
1073                        FWCfgState *fw_cfg)
1074 {
1075     uint16_t protocol;
1076     int setup_size, kernel_size, cmdline_size;
1077     int dtb_size, setup_data_offset;
1078     uint32_t initrd_max;
1079     uint8_t header[8192], *setup, *kernel;
1080     hwaddr real_addr, prot_addr, cmdline_addr, initrd_addr = 0;
1081     FILE *f;
1082     char *vmode;
1083     MachineState *machine = MACHINE(pcms);
1084     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1085     struct setup_data *setup_data;
1086     const char *kernel_filename = machine->kernel_filename;
1087     const char *initrd_filename = machine->initrd_filename;
1088     const char *dtb_filename = machine->dtb;
1089     const char *kernel_cmdline = machine->kernel_cmdline;
1090 
1091     /* Align to 16 bytes as a paranoia measure */
1092     cmdline_size = (strlen(kernel_cmdline)+16) & ~15;
1093 
1094     /* load the kernel header */
1095     f = fopen(kernel_filename, "rb");
1096     if (!f || !(kernel_size = get_file_size(f)) ||
1097         fread(header, 1, MIN(ARRAY_SIZE(header), kernel_size), f) !=
1098         MIN(ARRAY_SIZE(header), kernel_size)) {
1099         fprintf(stderr, "qemu: could not load kernel '%s': %s\n",
1100                 kernel_filename, strerror(errno));
1101         exit(1);
1102     }
1103 
1104     /* kernel protocol version */
1105 #if 0
1106     fprintf(stderr, "header magic: %#x\n", ldl_p(header+0x202));
1107 #endif
1108     if (ldl_p(header+0x202) == 0x53726448) {
1109         protocol = lduw_p(header+0x206);
1110     } else {
1111         /* This looks like a multiboot kernel. If it is, let's stop
1112            treating it like a Linux kernel. */
1113         if (load_multiboot(fw_cfg, f, kernel_filename, initrd_filename,
1114                            kernel_cmdline, kernel_size, header)) {
1115             return;
1116         }
1117         protocol = 0;
1118     }
1119 
1120     if (protocol < 0x200 || !(header[0x211] & 0x01)) {
1121         /* Low kernel */
1122         real_addr    = 0x90000;
1123         cmdline_addr = 0x9a000 - cmdline_size;
1124         prot_addr    = 0x10000;
1125     } else if (protocol < 0x202) {
1126         /* High but ancient kernel */
1127         real_addr    = 0x90000;
1128         cmdline_addr = 0x9a000 - cmdline_size;
1129         prot_addr    = 0x100000;
1130     } else {
1131         /* High and recent kernel */
1132         real_addr    = 0x10000;
1133         cmdline_addr = 0x20000;
1134         prot_addr    = 0x100000;
1135     }
1136 
1137 #if 0
1138     fprintf(stderr,
1139             "qemu: real_addr     = 0x" TARGET_FMT_plx "\n"
1140             "qemu: cmdline_addr  = 0x" TARGET_FMT_plx "\n"
1141             "qemu: prot_addr     = 0x" TARGET_FMT_plx "\n",
1142             real_addr,
1143             cmdline_addr,
1144             prot_addr);
1145 #endif
1146 
1147     /* highest address for loading the initrd */
1148     if (protocol >= 0x203) {
1149         initrd_max = ldl_p(header+0x22c);
1150     } else {
1151         initrd_max = 0x37ffffff;
1152     }
1153 
1154     if (initrd_max >= pcms->below_4g_mem_size - pcmc->acpi_data_size) {
1155         initrd_max = pcms->below_4g_mem_size - pcmc->acpi_data_size - 1;
1156     }
1157 
1158     fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_ADDR, cmdline_addr);
1159     fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_SIZE, strlen(kernel_cmdline)+1);
1160     fw_cfg_add_string(fw_cfg, FW_CFG_CMDLINE_DATA, kernel_cmdline);
1161 
1162     if (protocol >= 0x202) {
1163         stl_p(header+0x228, cmdline_addr);
1164     } else {
1165         stw_p(header+0x20, 0xA33F);
1166         stw_p(header+0x22, cmdline_addr-real_addr);
1167     }
1168 
1169     /* handle vga= parameter */
1170     vmode = strstr(kernel_cmdline, "vga=");
1171     if (vmode) {
1172         unsigned int video_mode;
1173         /* skip "vga=" */
1174         vmode += 4;
1175         if (!strncmp(vmode, "normal", 6)) {
1176             video_mode = 0xffff;
1177         } else if (!strncmp(vmode, "ext", 3)) {
1178             video_mode = 0xfffe;
1179         } else if (!strncmp(vmode, "ask", 3)) {
1180             video_mode = 0xfffd;
1181         } else {
1182             video_mode = strtol(vmode, NULL, 0);
1183         }
1184         stw_p(header+0x1fa, video_mode);
1185     }
1186 
1187     /* loader type */
1188     /* High nybble = B reserved for QEMU; low nybble is revision number.
1189        If this code is substantially changed, you may want to consider
1190        incrementing the revision. */
1191     if (protocol >= 0x200) {
1192         header[0x210] = 0xB0;
1193     }
1194     /* heap */
1195     if (protocol >= 0x201) {
1196         header[0x211] |= 0x80;	/* CAN_USE_HEAP */
1197         stw_p(header+0x224, cmdline_addr-real_addr-0x200);
1198     }
1199 
1200     /* load initrd */
1201     if (initrd_filename) {
1202         gsize initrd_size;
1203         gchar *initrd_data;
1204         GError *gerr = NULL;
1205 
1206         if (protocol < 0x200) {
1207             fprintf(stderr, "qemu: linux kernel too old to load a ram disk\n");
1208             exit(1);
1209         }
1210 
1211         if (!g_file_get_contents(initrd_filename, &initrd_data,
1212                                  &initrd_size, &gerr)) {
1213             fprintf(stderr, "qemu: error reading initrd %s: %s\n",
1214                     initrd_filename, gerr->message);
1215             exit(1);
1216         }
1217         if (initrd_size >= initrd_max) {
1218             fprintf(stderr, "qemu: initrd is too large, cannot support."
1219                     "(max: %"PRIu32", need %"PRId64")\n",
1220                     initrd_max, (uint64_t)initrd_size);
1221             exit(1);
1222         }
1223 
1224         initrd_addr = (initrd_max-initrd_size) & ~4095;
1225 
1226         fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_ADDR, initrd_addr);
1227         fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_SIZE, initrd_size);
1228         fw_cfg_add_bytes(fw_cfg, FW_CFG_INITRD_DATA, initrd_data, initrd_size);
1229 
1230         stl_p(header+0x218, initrd_addr);
1231         stl_p(header+0x21c, initrd_size);
1232     }
1233 
1234     /* load kernel and setup */
1235     setup_size = header[0x1f1];
1236     if (setup_size == 0) {
1237         setup_size = 4;
1238     }
1239     setup_size = (setup_size+1)*512;
1240     if (setup_size > kernel_size) {
1241         fprintf(stderr, "qemu: invalid kernel header\n");
1242         exit(1);
1243     }
1244     kernel_size -= setup_size;
1245 
1246     setup  = g_malloc(setup_size);
1247     kernel = g_malloc(kernel_size);
1248     fseek(f, 0, SEEK_SET);
1249     if (fread(setup, 1, setup_size, f) != setup_size) {
1250         fprintf(stderr, "fread() failed\n");
1251         exit(1);
1252     }
1253     if (fread(kernel, 1, kernel_size, f) != kernel_size) {
1254         fprintf(stderr, "fread() failed\n");
1255         exit(1);
1256     }
1257     fclose(f);
1258 
1259     /* append dtb to kernel */
1260     if (dtb_filename) {
1261         if (protocol < 0x209) {
1262             fprintf(stderr, "qemu: Linux kernel too old to load a dtb\n");
1263             exit(1);
1264         }
1265 
1266         dtb_size = get_image_size(dtb_filename);
1267         if (dtb_size <= 0) {
1268             fprintf(stderr, "qemu: error reading dtb %s: %s\n",
1269                     dtb_filename, strerror(errno));
1270             exit(1);
1271         }
1272 
1273         setup_data_offset = QEMU_ALIGN_UP(kernel_size, 16);
1274         kernel_size = setup_data_offset + sizeof(struct setup_data) + dtb_size;
1275         kernel = g_realloc(kernel, kernel_size);
1276 
1277         stq_p(header+0x250, prot_addr + setup_data_offset);
1278 
1279         setup_data = (struct setup_data *)(kernel + setup_data_offset);
1280         setup_data->next = 0;
1281         setup_data->type = cpu_to_le32(SETUP_DTB);
1282         setup_data->len = cpu_to_le32(dtb_size);
1283 
1284         load_image_size(dtb_filename, setup_data->data, dtb_size);
1285     }
1286 
1287     memcpy(setup, header, MIN(sizeof(header), setup_size));
1288 
1289     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ADDR, prot_addr);
1290     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_SIZE, kernel_size);
1291     fw_cfg_add_bytes(fw_cfg, FW_CFG_KERNEL_DATA, kernel, kernel_size);
1292 
1293     fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_ADDR, real_addr);
1294     fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_SIZE, setup_size);
1295     fw_cfg_add_bytes(fw_cfg, FW_CFG_SETUP_DATA, setup, setup_size);
1296 
1297     option_rom[nb_option_roms].bootindex = 0;
1298     option_rom[nb_option_roms].name = "linuxboot.bin";
1299     if (pcmc->linuxboot_dma_enabled && fw_cfg_dma_enabled(fw_cfg)) {
1300         option_rom[nb_option_roms].name = "linuxboot_dma.bin";
1301     }
1302     nb_option_roms++;
1303 }
1304 
1305 #define NE2000_NB_MAX 6
1306 
1307 static const int ne2000_io[NE2000_NB_MAX] = { 0x300, 0x320, 0x340, 0x360,
1308                                               0x280, 0x380 };
1309 static const int ne2000_irq[NE2000_NB_MAX] = { 9, 10, 11, 3, 4, 5 };
1310 
1311 void pc_init_ne2k_isa(ISABus *bus, NICInfo *nd)
1312 {
1313     static int nb_ne2k = 0;
1314 
1315     if (nb_ne2k == NE2000_NB_MAX)
1316         return;
1317     isa_ne2000_init(bus, ne2000_io[nb_ne2k],
1318                     ne2000_irq[nb_ne2k], nd);
1319     nb_ne2k++;
1320 }
1321 
1322 DeviceState *cpu_get_current_apic(void)
1323 {
1324     if (current_cpu) {
1325         X86CPU *cpu = X86_CPU(current_cpu);
1326         return cpu->apic_state;
1327     } else {
1328         return NULL;
1329     }
1330 }
1331 
1332 void pc_acpi_smi_interrupt(void *opaque, int irq, int level)
1333 {
1334     X86CPU *cpu = opaque;
1335 
1336     if (level) {
1337         cpu_interrupt(CPU(cpu), CPU_INTERRUPT_SMI);
1338     }
1339 }
1340 
1341 static void pc_new_cpu(const char *typename, int64_t apic_id, Error **errp)
1342 {
1343     Object *cpu = NULL;
1344     Error *local_err = NULL;
1345 
1346     cpu = object_new(typename);
1347 
1348     object_property_set_uint(cpu, apic_id, "apic-id", &local_err);
1349     object_property_set_bool(cpu, true, "realized", &local_err);
1350 
1351     object_unref(cpu);
1352     error_propagate(errp, local_err);
1353 }
1354 
1355 void pc_hot_add_cpu(const int64_t id, Error **errp)
1356 {
1357     MachineState *ms = MACHINE(qdev_get_machine());
1358     int64_t apic_id = x86_cpu_apic_id_from_index(id);
1359     Error *local_err = NULL;
1360 
1361     if (id < 0) {
1362         error_setg(errp, "Invalid CPU id: %" PRIi64, id);
1363         return;
1364     }
1365 
1366     if (apic_id >= ACPI_CPU_HOTPLUG_ID_LIMIT) {
1367         error_setg(errp, "Unable to add CPU: %" PRIi64
1368                    ", resulting APIC ID (%" PRIi64 ") is too large",
1369                    id, apic_id);
1370         return;
1371     }
1372 
1373     pc_new_cpu(ms->cpu_type, apic_id, &local_err);
1374     if (local_err) {
1375         error_propagate(errp, local_err);
1376         return;
1377     }
1378 }
1379 
1380 void pc_cpus_init(PCMachineState *pcms)
1381 {
1382     int i;
1383     const CPUArchIdList *possible_cpus;
1384     MachineState *ms = MACHINE(pcms);
1385     MachineClass *mc = MACHINE_GET_CLASS(pcms);
1386 
1387     /* Calculates the limit to CPU APIC ID values
1388      *
1389      * Limit for the APIC ID value, so that all
1390      * CPU APIC IDs are < pcms->apic_id_limit.
1391      *
1392      * This is used for FW_CFG_MAX_CPUS. See comments on bochs_bios_init().
1393      */
1394     pcms->apic_id_limit = x86_cpu_apic_id_from_index(max_cpus - 1) + 1;
1395     possible_cpus = mc->possible_cpu_arch_ids(ms);
1396     for (i = 0; i < smp_cpus; i++) {
1397         pc_new_cpu(possible_cpus->cpus[i].type, possible_cpus->cpus[i].arch_id,
1398                    &error_fatal);
1399     }
1400 }
1401 
1402 static void pc_build_feature_control_file(PCMachineState *pcms)
1403 {
1404     MachineState *ms = MACHINE(pcms);
1405     X86CPU *cpu = X86_CPU(ms->possible_cpus->cpus[0].cpu);
1406     CPUX86State *env = &cpu->env;
1407     uint32_t unused, ecx, edx;
1408     uint64_t feature_control_bits = 0;
1409     uint64_t *val;
1410 
1411     cpu_x86_cpuid(env, 1, 0, &unused, &unused, &ecx, &edx);
1412     if (ecx & CPUID_EXT_VMX) {
1413         feature_control_bits |= FEATURE_CONTROL_VMXON_ENABLED_OUTSIDE_SMX;
1414     }
1415 
1416     if ((edx & (CPUID_EXT2_MCE | CPUID_EXT2_MCA)) ==
1417         (CPUID_EXT2_MCE | CPUID_EXT2_MCA) &&
1418         (env->mcg_cap & MCG_LMCE_P)) {
1419         feature_control_bits |= FEATURE_CONTROL_LMCE;
1420     }
1421 
1422     if (!feature_control_bits) {
1423         return;
1424     }
1425 
1426     val = g_malloc(sizeof(*val));
1427     *val = cpu_to_le64(feature_control_bits | FEATURE_CONTROL_LOCKED);
1428     fw_cfg_add_file(pcms->fw_cfg, "etc/msr_feature_control", val, sizeof(*val));
1429 }
1430 
1431 static void rtc_set_cpus_count(ISADevice *rtc, uint16_t cpus_count)
1432 {
1433     if (cpus_count > 0xff) {
1434         /* If the number of CPUs can't be represented in 8 bits, the
1435          * BIOS must use "FW_CFG_NB_CPUS". Set RTC field to 0 just
1436          * to make old BIOSes fail more predictably.
1437          */
1438         rtc_set_memory(rtc, 0x5f, 0);
1439     } else {
1440         rtc_set_memory(rtc, 0x5f, cpus_count - 1);
1441     }
1442 }
1443 
1444 static
1445 void pc_machine_done(Notifier *notifier, void *data)
1446 {
1447     PCMachineState *pcms = container_of(notifier,
1448                                         PCMachineState, machine_done);
1449     PCIBus *bus = pcms->bus;
1450 
1451     /* set the number of CPUs */
1452     rtc_set_cpus_count(pcms->rtc, pcms->boot_cpus);
1453 
1454     if (bus) {
1455         int extra_hosts = 0;
1456 
1457         QLIST_FOREACH(bus, &bus->child, sibling) {
1458             /* look for expander root buses */
1459             if (pci_bus_is_root(bus)) {
1460                 extra_hosts++;
1461             }
1462         }
1463         if (extra_hosts && pcms->fw_cfg) {
1464             uint64_t *val = g_malloc(sizeof(*val));
1465             *val = cpu_to_le64(extra_hosts);
1466             fw_cfg_add_file(pcms->fw_cfg,
1467                     "etc/extra-pci-roots", val, sizeof(*val));
1468         }
1469     }
1470 
1471     acpi_setup();
1472     if (pcms->fw_cfg) {
1473         pc_build_smbios(pcms);
1474         pc_build_feature_control_file(pcms);
1475         /* update FW_CFG_NB_CPUS to account for -device added CPUs */
1476         fw_cfg_modify_i16(pcms->fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
1477     }
1478 
1479     if (pcms->apic_id_limit > 255 && !xen_enabled()) {
1480         IntelIOMMUState *iommu = INTEL_IOMMU_DEVICE(x86_iommu_get_default());
1481 
1482         if (!iommu || !x86_iommu_ir_supported(X86_IOMMU_DEVICE(iommu)) ||
1483             iommu->intr_eim != ON_OFF_AUTO_ON) {
1484             error_report("current -smp configuration requires "
1485                          "Extended Interrupt Mode enabled. "
1486                          "You can add an IOMMU using: "
1487                          "-device intel-iommu,intremap=on,eim=on");
1488             exit(EXIT_FAILURE);
1489         }
1490     }
1491 }
1492 
1493 void pc_guest_info_init(PCMachineState *pcms)
1494 {
1495     int i;
1496 
1497     pcms->apic_xrupt_override = kvm_allows_irq0_override();
1498     pcms->numa_nodes = nb_numa_nodes;
1499     pcms->node_mem = g_malloc0(pcms->numa_nodes *
1500                                     sizeof *pcms->node_mem);
1501     for (i = 0; i < nb_numa_nodes; i++) {
1502         pcms->node_mem[i] = numa_info[i].node_mem;
1503     }
1504 
1505     pcms->machine_done.notify = pc_machine_done;
1506     qemu_add_machine_init_done_notifier(&pcms->machine_done);
1507 }
1508 
1509 /* setup pci memory address space mapping into system address space */
1510 void pc_pci_as_mapping_init(Object *owner, MemoryRegion *system_memory,
1511                             MemoryRegion *pci_address_space)
1512 {
1513     /* Set to lower priority than RAM */
1514     memory_region_add_subregion_overlap(system_memory, 0x0,
1515                                         pci_address_space, -1);
1516 }
1517 
1518 void pc_acpi_init(const char *default_dsdt)
1519 {
1520     char *filename;
1521 
1522     if (acpi_tables != NULL) {
1523         /* manually set via -acpitable, leave it alone */
1524         return;
1525     }
1526 
1527     filename = qemu_find_file(QEMU_FILE_TYPE_BIOS, default_dsdt);
1528     if (filename == NULL) {
1529         warn_report("failed to find %s", default_dsdt);
1530     } else {
1531         QemuOpts *opts = qemu_opts_create(qemu_find_opts("acpi"), NULL, 0,
1532                                           &error_abort);
1533         Error *err = NULL;
1534 
1535         qemu_opt_set(opts, "file", filename, &error_abort);
1536 
1537         acpi_table_add_builtin(opts, &err);
1538         if (err) {
1539             warn_reportf_err(err, "failed to load %s: ", filename);
1540         }
1541         g_free(filename);
1542     }
1543 }
1544 
1545 void xen_load_linux(PCMachineState *pcms)
1546 {
1547     int i;
1548     FWCfgState *fw_cfg;
1549 
1550     assert(MACHINE(pcms)->kernel_filename != NULL);
1551 
1552     fw_cfg = fw_cfg_init_io(FW_CFG_IO_BASE);
1553     fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
1554     rom_set_fw(fw_cfg);
1555 
1556     load_linux(pcms, fw_cfg);
1557     for (i = 0; i < nb_option_roms; i++) {
1558         assert(!strcmp(option_rom[i].name, "linuxboot.bin") ||
1559                !strcmp(option_rom[i].name, "linuxboot_dma.bin") ||
1560                !strcmp(option_rom[i].name, "multiboot.bin"));
1561         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1562     }
1563     pcms->fw_cfg = fw_cfg;
1564 }
1565 
1566 void pc_memory_init(PCMachineState *pcms,
1567                     MemoryRegion *system_memory,
1568                     MemoryRegion *rom_memory,
1569                     MemoryRegion **ram_memory)
1570 {
1571     int linux_boot, i;
1572     MemoryRegion *ram, *option_rom_mr;
1573     MemoryRegion *ram_below_4g, *ram_above_4g;
1574     FWCfgState *fw_cfg;
1575     MachineState *machine = MACHINE(pcms);
1576     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1577 
1578     assert(machine->ram_size == pcms->below_4g_mem_size +
1579                                 pcms->above_4g_mem_size);
1580 
1581     linux_boot = (machine->kernel_filename != NULL);
1582 
1583     /* Allocate RAM.  We allocate it as a single memory region and use
1584      * aliases to address portions of it, mostly for backwards compatibility
1585      * with older qemus that used qemu_ram_alloc().
1586      */
1587     ram = g_malloc(sizeof(*ram));
1588     memory_region_allocate_system_memory(ram, NULL, "pc.ram",
1589                                          machine->ram_size);
1590     *ram_memory = ram;
1591     ram_below_4g = g_malloc(sizeof(*ram_below_4g));
1592     memory_region_init_alias(ram_below_4g, NULL, "ram-below-4g", ram,
1593                              0, pcms->below_4g_mem_size);
1594     memory_region_add_subregion(system_memory, 0, ram_below_4g);
1595     e820_add_entry(0, pcms->below_4g_mem_size, E820_RAM);
1596     if (pcms->above_4g_mem_size > 0) {
1597         ram_above_4g = g_malloc(sizeof(*ram_above_4g));
1598         memory_region_init_alias(ram_above_4g, NULL, "ram-above-4g", ram,
1599                                  pcms->below_4g_mem_size,
1600                                  pcms->above_4g_mem_size);
1601         memory_region_add_subregion(system_memory, 0x100000000ULL,
1602                                     ram_above_4g);
1603         e820_add_entry(0x100000000ULL, pcms->above_4g_mem_size, E820_RAM);
1604     }
1605 
1606     if (!pcmc->has_reserved_memory &&
1607         (machine->ram_slots ||
1608          (machine->maxram_size > machine->ram_size))) {
1609         MachineClass *mc = MACHINE_GET_CLASS(machine);
1610 
1611         error_report("\"-memory 'slots|maxmem'\" is not supported by: %s",
1612                      mc->name);
1613         exit(EXIT_FAILURE);
1614     }
1615 
1616     /* always allocate the device memory information */
1617     machine->device_memory = g_malloc0(sizeof(*machine->device_memory));
1618 
1619     /* initialize device memory address space */
1620     if (pcmc->has_reserved_memory &&
1621         (machine->ram_size < machine->maxram_size)) {
1622         ram_addr_t device_mem_size = machine->maxram_size - machine->ram_size;
1623 
1624         if (machine->ram_slots > ACPI_MAX_RAM_SLOTS) {
1625             error_report("unsupported amount of memory slots: %"PRIu64,
1626                          machine->ram_slots);
1627             exit(EXIT_FAILURE);
1628         }
1629 
1630         if (QEMU_ALIGN_UP(machine->maxram_size,
1631                           TARGET_PAGE_SIZE) != machine->maxram_size) {
1632             error_report("maximum memory size must by aligned to multiple of "
1633                          "%d bytes", TARGET_PAGE_SIZE);
1634             exit(EXIT_FAILURE);
1635         }
1636 
1637         machine->device_memory->base =
1638             ROUND_UP(0x100000000ULL + pcms->above_4g_mem_size, 1 * GiB);
1639 
1640         if (pcmc->enforce_aligned_dimm) {
1641             /* size device region assuming 1G page max alignment per slot */
1642             device_mem_size += (1 * GiB) * machine->ram_slots;
1643         }
1644 
1645         if ((machine->device_memory->base + device_mem_size) <
1646             device_mem_size) {
1647             error_report("unsupported amount of maximum memory: " RAM_ADDR_FMT,
1648                          machine->maxram_size);
1649             exit(EXIT_FAILURE);
1650         }
1651 
1652         memory_region_init(&machine->device_memory->mr, OBJECT(pcms),
1653                            "device-memory", device_mem_size);
1654         memory_region_add_subregion(system_memory, machine->device_memory->base,
1655                                     &machine->device_memory->mr);
1656     }
1657 
1658     /* Initialize PC system firmware */
1659     pc_system_firmware_init(rom_memory, !pcmc->pci_enabled);
1660 
1661     option_rom_mr = g_malloc(sizeof(*option_rom_mr));
1662     memory_region_init_ram(option_rom_mr, NULL, "pc.rom", PC_ROM_SIZE,
1663                            &error_fatal);
1664     if (pcmc->pci_enabled) {
1665         memory_region_set_readonly(option_rom_mr, true);
1666     }
1667     memory_region_add_subregion_overlap(rom_memory,
1668                                         PC_ROM_MIN_VGA,
1669                                         option_rom_mr,
1670                                         1);
1671 
1672     fw_cfg = bochs_bios_init(&address_space_memory, pcms);
1673 
1674     rom_set_fw(fw_cfg);
1675 
1676     if (pcmc->has_reserved_memory && machine->device_memory->base) {
1677         uint64_t *val = g_malloc(sizeof(*val));
1678         PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1679         uint64_t res_mem_end = machine->device_memory->base;
1680 
1681         if (!pcmc->broken_reserved_end) {
1682             res_mem_end += memory_region_size(&machine->device_memory->mr);
1683         }
1684         *val = cpu_to_le64(ROUND_UP(res_mem_end, 1 * GiB));
1685         fw_cfg_add_file(fw_cfg, "etc/reserved-memory-end", val, sizeof(*val));
1686     }
1687 
1688     if (linux_boot) {
1689         load_linux(pcms, fw_cfg);
1690     }
1691 
1692     for (i = 0; i < nb_option_roms; i++) {
1693         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1694     }
1695     pcms->fw_cfg = fw_cfg;
1696 
1697     /* Init default IOAPIC address space */
1698     pcms->ioapic_as = &address_space_memory;
1699 }
1700 
1701 /*
1702  * The 64bit pci hole starts after "above 4G RAM" and
1703  * potentially the space reserved for memory hotplug.
1704  */
1705 uint64_t pc_pci_hole64_start(void)
1706 {
1707     PCMachineState *pcms = PC_MACHINE(qdev_get_machine());
1708     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1709     MachineState *ms = MACHINE(pcms);
1710     uint64_t hole64_start = 0;
1711 
1712     if (pcmc->has_reserved_memory && ms->device_memory->base) {
1713         hole64_start = ms->device_memory->base;
1714         if (!pcmc->broken_reserved_end) {
1715             hole64_start += memory_region_size(&ms->device_memory->mr);
1716         }
1717     } else {
1718         hole64_start = 0x100000000ULL + pcms->above_4g_mem_size;
1719     }
1720 
1721     return ROUND_UP(hole64_start, 1 * GiB);
1722 }
1723 
1724 qemu_irq pc_allocate_cpu_irq(void)
1725 {
1726     return qemu_allocate_irq(pic_irq_request, NULL, 0);
1727 }
1728 
1729 DeviceState *pc_vga_init(ISABus *isa_bus, PCIBus *pci_bus)
1730 {
1731     DeviceState *dev = NULL;
1732 
1733     rom_set_order_override(FW_CFG_ORDER_OVERRIDE_VGA);
1734     if (pci_bus) {
1735         PCIDevice *pcidev = pci_vga_init(pci_bus);
1736         dev = pcidev ? &pcidev->qdev : NULL;
1737     } else if (isa_bus) {
1738         ISADevice *isadev = isa_vga_init(isa_bus);
1739         dev = isadev ? DEVICE(isadev) : NULL;
1740     }
1741     rom_reset_order_override();
1742     return dev;
1743 }
1744 
1745 static const MemoryRegionOps ioport80_io_ops = {
1746     .write = ioport80_write,
1747     .read = ioport80_read,
1748     .endianness = DEVICE_NATIVE_ENDIAN,
1749     .impl = {
1750         .min_access_size = 1,
1751         .max_access_size = 1,
1752     },
1753 };
1754 
1755 static const MemoryRegionOps ioportF0_io_ops = {
1756     .write = ioportF0_write,
1757     .read = ioportF0_read,
1758     .endianness = DEVICE_NATIVE_ENDIAN,
1759     .impl = {
1760         .min_access_size = 1,
1761         .max_access_size = 1,
1762     },
1763 };
1764 
1765 static void pc_superio_init(ISABus *isa_bus, bool create_fdctrl, bool no_vmport)
1766 {
1767     int i;
1768     DriveInfo *fd[MAX_FD];
1769     qemu_irq *a20_line;
1770     ISADevice *i8042, *port92, *vmmouse;
1771 
1772     serial_hds_isa_init(isa_bus, 0, MAX_ISA_SERIAL_PORTS);
1773     parallel_hds_isa_init(isa_bus, MAX_PARALLEL_PORTS);
1774 
1775     for (i = 0; i < MAX_FD; i++) {
1776         fd[i] = drive_get(IF_FLOPPY, 0, i);
1777         create_fdctrl |= !!fd[i];
1778     }
1779     if (create_fdctrl) {
1780         fdctrl_init_isa(isa_bus, fd);
1781     }
1782 
1783     i8042 = isa_create_simple(isa_bus, "i8042");
1784     if (!no_vmport) {
1785         vmport_init(isa_bus);
1786         vmmouse = isa_try_create(isa_bus, "vmmouse");
1787     } else {
1788         vmmouse = NULL;
1789     }
1790     if (vmmouse) {
1791         DeviceState *dev = DEVICE(vmmouse);
1792         qdev_prop_set_ptr(dev, "ps2_mouse", i8042);
1793         qdev_init_nofail(dev);
1794     }
1795     port92 = isa_create_simple(isa_bus, "port92");
1796 
1797     a20_line = qemu_allocate_irqs(handle_a20_line_change, first_cpu, 2);
1798     i8042_setup_a20_line(i8042, a20_line[0]);
1799     port92_init(port92, a20_line[1]);
1800     g_free(a20_line);
1801 }
1802 
1803 void pc_basic_device_init(ISABus *isa_bus, qemu_irq *gsi,
1804                           ISADevice **rtc_state,
1805                           bool create_fdctrl,
1806                           bool no_vmport,
1807                           bool has_pit,
1808                           uint32_t hpet_irqs)
1809 {
1810     int i;
1811     DeviceState *hpet = NULL;
1812     int pit_isa_irq = 0;
1813     qemu_irq pit_alt_irq = NULL;
1814     qemu_irq rtc_irq = NULL;
1815     ISADevice *pit = NULL;
1816     MemoryRegion *ioport80_io = g_new(MemoryRegion, 1);
1817     MemoryRegion *ioportF0_io = g_new(MemoryRegion, 1);
1818 
1819     memory_region_init_io(ioport80_io, NULL, &ioport80_io_ops, NULL, "ioport80", 1);
1820     memory_region_add_subregion(isa_bus->address_space_io, 0x80, ioport80_io);
1821 
1822     memory_region_init_io(ioportF0_io, NULL, &ioportF0_io_ops, NULL, "ioportF0", 1);
1823     memory_region_add_subregion(isa_bus->address_space_io, 0xf0, ioportF0_io);
1824 
1825     /*
1826      * Check if an HPET shall be created.
1827      *
1828      * Without KVM_CAP_PIT_STATE2, we cannot switch off the in-kernel PIT
1829      * when the HPET wants to take over. Thus we have to disable the latter.
1830      */
1831     if (!no_hpet && (!kvm_irqchip_in_kernel() || kvm_has_pit_state2())) {
1832         /* In order to set property, here not using sysbus_try_create_simple */
1833         hpet = qdev_try_create(NULL, TYPE_HPET);
1834         if (hpet) {
1835             /* For pc-piix-*, hpet's intcap is always IRQ2. For pc-q35-1.7
1836              * and earlier, use IRQ2 for compat. Otherwise, use IRQ16~23,
1837              * IRQ8 and IRQ2.
1838              */
1839             uint8_t compat = object_property_get_uint(OBJECT(hpet),
1840                     HPET_INTCAP, NULL);
1841             if (!compat) {
1842                 qdev_prop_set_uint32(hpet, HPET_INTCAP, hpet_irqs);
1843             }
1844             qdev_init_nofail(hpet);
1845             sysbus_mmio_map(SYS_BUS_DEVICE(hpet), 0, HPET_BASE);
1846 
1847             for (i = 0; i < GSI_NUM_PINS; i++) {
1848                 sysbus_connect_irq(SYS_BUS_DEVICE(hpet), i, gsi[i]);
1849             }
1850             pit_isa_irq = -1;
1851             pit_alt_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_PIT_INT);
1852             rtc_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_RTC_INT);
1853         }
1854     }
1855     *rtc_state = mc146818_rtc_init(isa_bus, 2000, rtc_irq);
1856 
1857     qemu_register_boot_set(pc_boot_set, *rtc_state);
1858 
1859     if (!xen_enabled() && has_pit) {
1860         if (kvm_pit_in_kernel()) {
1861             pit = kvm_pit_init(isa_bus, 0x40);
1862         } else {
1863             pit = i8254_pit_init(isa_bus, 0x40, pit_isa_irq, pit_alt_irq);
1864         }
1865         if (hpet) {
1866             /* connect PIT to output control line of the HPET */
1867             qdev_connect_gpio_out(hpet, 0, qdev_get_gpio_in(DEVICE(pit), 0));
1868         }
1869         pcspk_init(isa_bus, pit);
1870     }
1871 
1872     i8257_dma_init(isa_bus, 0);
1873 
1874     /* Super I/O */
1875     pc_superio_init(isa_bus, create_fdctrl, no_vmport);
1876 }
1877 
1878 void pc_nic_init(PCMachineClass *pcmc, ISABus *isa_bus, PCIBus *pci_bus)
1879 {
1880     int i;
1881 
1882     rom_set_order_override(FW_CFG_ORDER_OVERRIDE_NIC);
1883     for (i = 0; i < nb_nics; i++) {
1884         NICInfo *nd = &nd_table[i];
1885         const char *model = nd->model ? nd->model : pcmc->default_nic_model;
1886 
1887         if (g_str_equal(model, "ne2k_isa")) {
1888             pc_init_ne2k_isa(isa_bus, nd);
1889         } else {
1890             pci_nic_init_nofail(nd, pci_bus, model, NULL);
1891         }
1892     }
1893     rom_reset_order_override();
1894 }
1895 
1896 void ioapic_init_gsi(GSIState *gsi_state, const char *parent_name)
1897 {
1898     DeviceState *dev;
1899     SysBusDevice *d;
1900     unsigned int i;
1901 
1902     if (kvm_ioapic_in_kernel()) {
1903         dev = qdev_create(NULL, TYPE_KVM_IOAPIC);
1904     } else {
1905         dev = qdev_create(NULL, TYPE_IOAPIC);
1906     }
1907     if (parent_name) {
1908         object_property_add_child(object_resolve_path(parent_name, NULL),
1909                                   "ioapic", OBJECT(dev), NULL);
1910     }
1911     qdev_init_nofail(dev);
1912     d = SYS_BUS_DEVICE(dev);
1913     sysbus_mmio_map(d, 0, IO_APIC_DEFAULT_ADDRESS);
1914 
1915     for (i = 0; i < IOAPIC_NUM_PINS; i++) {
1916         gsi_state->ioapic_irq[i] = qdev_get_gpio_in(dev, i);
1917     }
1918 }
1919 
1920 static void pc_memory_pre_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
1921                                Error **errp)
1922 {
1923     const PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1924     const PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1925     const bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
1926     const uint64_t legacy_align = TARGET_PAGE_SIZE;
1927 
1928     /*
1929      * When -no-acpi is used with Q35 machine type, no ACPI is built,
1930      * but pcms->acpi_dev is still created. Check !acpi_enabled in
1931      * addition to cover this case.
1932      */
1933     if (!pcms->acpi_dev || !acpi_enabled) {
1934         error_setg(errp,
1935                    "memory hotplug is not enabled: missing acpi device or acpi disabled");
1936         return;
1937     }
1938 
1939     if (is_nvdimm && !pcms->acpi_nvdimm_state.is_enabled) {
1940         error_setg(errp, "nvdimm is not enabled: missing 'nvdimm' in '-M'");
1941         return;
1942     }
1943 
1944     pc_dimm_pre_plug(PC_DIMM(dev), MACHINE(hotplug_dev),
1945                      pcmc->enforce_aligned_dimm ? NULL : &legacy_align, errp);
1946 }
1947 
1948 static void pc_memory_plug(HotplugHandler *hotplug_dev,
1949                            DeviceState *dev, Error **errp)
1950 {
1951     HotplugHandlerClass *hhc;
1952     Error *local_err = NULL;
1953     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1954     bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
1955 
1956     pc_dimm_plug(PC_DIMM(dev), MACHINE(pcms), &local_err);
1957     if (local_err) {
1958         goto out;
1959     }
1960 
1961     if (is_nvdimm) {
1962         nvdimm_plug(&pcms->acpi_nvdimm_state);
1963     }
1964 
1965     hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
1966     hhc->plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &error_abort);
1967 out:
1968     error_propagate(errp, local_err);
1969 }
1970 
1971 static void pc_memory_unplug_request(HotplugHandler *hotplug_dev,
1972                                      DeviceState *dev, Error **errp)
1973 {
1974     HotplugHandlerClass *hhc;
1975     Error *local_err = NULL;
1976     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1977 
1978     /*
1979      * When -no-acpi is used with Q35 machine type, no ACPI is built,
1980      * but pcms->acpi_dev is still created. Check !acpi_enabled in
1981      * addition to cover this case.
1982      */
1983     if (!pcms->acpi_dev || !acpi_enabled) {
1984         error_setg(&local_err,
1985                    "memory hotplug is not enabled: missing acpi device or acpi disabled");
1986         goto out;
1987     }
1988 
1989     if (object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM)) {
1990         error_setg(&local_err,
1991                    "nvdimm device hot unplug is not supported yet.");
1992         goto out;
1993     }
1994 
1995     hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
1996     hhc->unplug_request(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
1997 
1998 out:
1999     error_propagate(errp, local_err);
2000 }
2001 
2002 static void pc_memory_unplug(HotplugHandler *hotplug_dev,
2003                              DeviceState *dev, Error **errp)
2004 {
2005     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2006     HotplugHandlerClass *hhc;
2007     Error *local_err = NULL;
2008 
2009     hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
2010     hhc->unplug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
2011 
2012     if (local_err) {
2013         goto out;
2014     }
2015 
2016     pc_dimm_unplug(PC_DIMM(dev), MACHINE(pcms));
2017     object_unparent(OBJECT(dev));
2018 
2019  out:
2020     error_propagate(errp, local_err);
2021 }
2022 
2023 static int pc_apic_cmp(const void *a, const void *b)
2024 {
2025    CPUArchId *apic_a = (CPUArchId *)a;
2026    CPUArchId *apic_b = (CPUArchId *)b;
2027 
2028    return apic_a->arch_id - apic_b->arch_id;
2029 }
2030 
2031 /* returns pointer to CPUArchId descriptor that matches CPU's apic_id
2032  * in ms->possible_cpus->cpus, if ms->possible_cpus->cpus has no
2033  * entry corresponding to CPU's apic_id returns NULL.
2034  */
2035 static CPUArchId *pc_find_cpu_slot(MachineState *ms, uint32_t id, int *idx)
2036 {
2037     CPUArchId apic_id, *found_cpu;
2038 
2039     apic_id.arch_id = id;
2040     found_cpu = bsearch(&apic_id, ms->possible_cpus->cpus,
2041         ms->possible_cpus->len, sizeof(*ms->possible_cpus->cpus),
2042         pc_apic_cmp);
2043     if (found_cpu && idx) {
2044         *idx = found_cpu - ms->possible_cpus->cpus;
2045     }
2046     return found_cpu;
2047 }
2048 
2049 static void pc_cpu_plug(HotplugHandler *hotplug_dev,
2050                         DeviceState *dev, Error **errp)
2051 {
2052     CPUArchId *found_cpu;
2053     HotplugHandlerClass *hhc;
2054     Error *local_err = NULL;
2055     X86CPU *cpu = X86_CPU(dev);
2056     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2057 
2058     if (pcms->acpi_dev) {
2059         hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
2060         hhc->plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
2061         if (local_err) {
2062             goto out;
2063         }
2064     }
2065 
2066     /* increment the number of CPUs */
2067     pcms->boot_cpus++;
2068     if (pcms->rtc) {
2069         rtc_set_cpus_count(pcms->rtc, pcms->boot_cpus);
2070     }
2071     if (pcms->fw_cfg) {
2072         fw_cfg_modify_i16(pcms->fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
2073     }
2074 
2075     found_cpu = pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, NULL);
2076     found_cpu->cpu = OBJECT(dev);
2077 out:
2078     error_propagate(errp, local_err);
2079 }
2080 static void pc_cpu_unplug_request_cb(HotplugHandler *hotplug_dev,
2081                                      DeviceState *dev, Error **errp)
2082 {
2083     int idx = -1;
2084     HotplugHandlerClass *hhc;
2085     Error *local_err = NULL;
2086     X86CPU *cpu = X86_CPU(dev);
2087     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2088 
2089     if (!pcms->acpi_dev) {
2090         error_setg(&local_err, "CPU hot unplug not supported without ACPI");
2091         goto out;
2092     }
2093 
2094     pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, &idx);
2095     assert(idx != -1);
2096     if (idx == 0) {
2097         error_setg(&local_err, "Boot CPU is unpluggable");
2098         goto out;
2099     }
2100 
2101     hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
2102     hhc->unplug_request(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
2103 
2104     if (local_err) {
2105         goto out;
2106     }
2107 
2108  out:
2109     error_propagate(errp, local_err);
2110 
2111 }
2112 
2113 static void pc_cpu_unplug_cb(HotplugHandler *hotplug_dev,
2114                              DeviceState *dev, Error **errp)
2115 {
2116     CPUArchId *found_cpu;
2117     HotplugHandlerClass *hhc;
2118     Error *local_err = NULL;
2119     X86CPU *cpu = X86_CPU(dev);
2120     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2121 
2122     hhc = HOTPLUG_HANDLER_GET_CLASS(pcms->acpi_dev);
2123     hhc->unplug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
2124 
2125     if (local_err) {
2126         goto out;
2127     }
2128 
2129     found_cpu = pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, NULL);
2130     found_cpu->cpu = NULL;
2131     object_unparent(OBJECT(dev));
2132 
2133     /* decrement the number of CPUs */
2134     pcms->boot_cpus--;
2135     /* Update the number of CPUs in CMOS */
2136     rtc_set_cpus_count(pcms->rtc, pcms->boot_cpus);
2137     fw_cfg_modify_i16(pcms->fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
2138  out:
2139     error_propagate(errp, local_err);
2140 }
2141 
2142 static void pc_cpu_pre_plug(HotplugHandler *hotplug_dev,
2143                             DeviceState *dev, Error **errp)
2144 {
2145     int idx;
2146     CPUState *cs;
2147     CPUArchId *cpu_slot;
2148     X86CPUTopoInfo topo;
2149     X86CPU *cpu = X86_CPU(dev);
2150     MachineState *ms = MACHINE(hotplug_dev);
2151     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2152 
2153     if(!object_dynamic_cast(OBJECT(cpu), ms->cpu_type)) {
2154         error_setg(errp, "Invalid CPU type, expected cpu type: '%s'",
2155                    ms->cpu_type);
2156         return;
2157     }
2158 
2159     /* if APIC ID is not set, set it based on socket/core/thread properties */
2160     if (cpu->apic_id == UNASSIGNED_APIC_ID) {
2161         int max_socket = (max_cpus - 1) / smp_threads / smp_cores;
2162 
2163         if (cpu->socket_id < 0) {
2164             error_setg(errp, "CPU socket-id is not set");
2165             return;
2166         } else if (cpu->socket_id > max_socket) {
2167             error_setg(errp, "Invalid CPU socket-id: %u must be in range 0:%u",
2168                        cpu->socket_id, max_socket);
2169             return;
2170         }
2171         if (cpu->core_id < 0) {
2172             error_setg(errp, "CPU core-id is not set");
2173             return;
2174         } else if (cpu->core_id > (smp_cores - 1)) {
2175             error_setg(errp, "Invalid CPU core-id: %u must be in range 0:%u",
2176                        cpu->core_id, smp_cores - 1);
2177             return;
2178         }
2179         if (cpu->thread_id < 0) {
2180             error_setg(errp, "CPU thread-id is not set");
2181             return;
2182         } else if (cpu->thread_id > (smp_threads - 1)) {
2183             error_setg(errp, "Invalid CPU thread-id: %u must be in range 0:%u",
2184                        cpu->thread_id, smp_threads - 1);
2185             return;
2186         }
2187 
2188         topo.pkg_id = cpu->socket_id;
2189         topo.core_id = cpu->core_id;
2190         topo.smt_id = cpu->thread_id;
2191         cpu->apic_id = apicid_from_topo_ids(smp_cores, smp_threads, &topo);
2192     }
2193 
2194     cpu_slot = pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, &idx);
2195     if (!cpu_slot) {
2196         MachineState *ms = MACHINE(pcms);
2197 
2198         x86_topo_ids_from_apicid(cpu->apic_id, smp_cores, smp_threads, &topo);
2199         error_setg(errp, "Invalid CPU [socket: %u, core: %u, thread: %u] with"
2200                   " APIC ID %" PRIu32 ", valid index range 0:%d",
2201                    topo.pkg_id, topo.core_id, topo.smt_id, cpu->apic_id,
2202                    ms->possible_cpus->len - 1);
2203         return;
2204     }
2205 
2206     if (cpu_slot->cpu) {
2207         error_setg(errp, "CPU[%d] with APIC ID %" PRIu32 " exists",
2208                    idx, cpu->apic_id);
2209         return;
2210     }
2211 
2212     /* if 'address' properties socket-id/core-id/thread-id are not set, set them
2213      * so that machine_query_hotpluggable_cpus would show correct values
2214      */
2215     /* TODO: move socket_id/core_id/thread_id checks into x86_cpu_realizefn()
2216      * once -smp refactoring is complete and there will be CPU private
2217      * CPUState::nr_cores and CPUState::nr_threads fields instead of globals */
2218     x86_topo_ids_from_apicid(cpu->apic_id, smp_cores, smp_threads, &topo);
2219     if (cpu->socket_id != -1 && cpu->socket_id != topo.pkg_id) {
2220         error_setg(errp, "property socket-id: %u doesn't match set apic-id:"
2221             " 0x%x (socket-id: %u)", cpu->socket_id, cpu->apic_id, topo.pkg_id);
2222         return;
2223     }
2224     cpu->socket_id = topo.pkg_id;
2225 
2226     if (cpu->core_id != -1 && cpu->core_id != topo.core_id) {
2227         error_setg(errp, "property core-id: %u doesn't match set apic-id:"
2228             " 0x%x (core-id: %u)", cpu->core_id, cpu->apic_id, topo.core_id);
2229         return;
2230     }
2231     cpu->core_id = topo.core_id;
2232 
2233     if (cpu->thread_id != -1 && cpu->thread_id != topo.smt_id) {
2234         error_setg(errp, "property thread-id: %u doesn't match set apic-id:"
2235             " 0x%x (thread-id: %u)", cpu->thread_id, cpu->apic_id, topo.smt_id);
2236         return;
2237     }
2238     cpu->thread_id = topo.smt_id;
2239 
2240     if (cpu->hyperv_vpindex && !kvm_hv_vpindex_settable()) {
2241         error_setg(errp, "kernel doesn't allow setting HyperV VP_INDEX");
2242         return;
2243     }
2244 
2245     cs = CPU(cpu);
2246     cs->cpu_index = idx;
2247 
2248     numa_cpu_pre_plug(cpu_slot, dev, errp);
2249 }
2250 
2251 static void pc_machine_device_pre_plug_cb(HotplugHandler *hotplug_dev,
2252                                           DeviceState *dev, Error **errp)
2253 {
2254     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2255         pc_memory_pre_plug(hotplug_dev, dev, errp);
2256     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2257         pc_cpu_pre_plug(hotplug_dev, dev, errp);
2258     }
2259 }
2260 
2261 static void pc_machine_device_plug_cb(HotplugHandler *hotplug_dev,
2262                                       DeviceState *dev, Error **errp)
2263 {
2264     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2265         pc_memory_plug(hotplug_dev, dev, errp);
2266     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2267         pc_cpu_plug(hotplug_dev, dev, errp);
2268     }
2269 }
2270 
2271 static void pc_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
2272                                                 DeviceState *dev, Error **errp)
2273 {
2274     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2275         pc_memory_unplug_request(hotplug_dev, dev, errp);
2276     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2277         pc_cpu_unplug_request_cb(hotplug_dev, dev, errp);
2278     } else {
2279         error_setg(errp, "acpi: device unplug request for not supported device"
2280                    " type: %s", object_get_typename(OBJECT(dev)));
2281     }
2282 }
2283 
2284 static void pc_machine_device_unplug_cb(HotplugHandler *hotplug_dev,
2285                                         DeviceState *dev, Error **errp)
2286 {
2287     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2288         pc_memory_unplug(hotplug_dev, dev, errp);
2289     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2290         pc_cpu_unplug_cb(hotplug_dev, dev, errp);
2291     } else {
2292         error_setg(errp, "acpi: device unplug for not supported device"
2293                    " type: %s", object_get_typename(OBJECT(dev)));
2294     }
2295 }
2296 
2297 static HotplugHandler *pc_get_hotplug_handler(MachineState *machine,
2298                                              DeviceState *dev)
2299 {
2300     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) ||
2301         object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2302         return HOTPLUG_HANDLER(machine);
2303     }
2304 
2305     return NULL;
2306 }
2307 
2308 static void
2309 pc_machine_get_device_memory_region_size(Object *obj, Visitor *v,
2310                                          const char *name, void *opaque,
2311                                          Error **errp)
2312 {
2313     MachineState *ms = MACHINE(obj);
2314     int64_t value = memory_region_size(&ms->device_memory->mr);
2315 
2316     visit_type_int(v, name, &value, errp);
2317 }
2318 
2319 static void pc_machine_get_max_ram_below_4g(Object *obj, Visitor *v,
2320                                             const char *name, void *opaque,
2321                                             Error **errp)
2322 {
2323     PCMachineState *pcms = PC_MACHINE(obj);
2324     uint64_t value = pcms->max_ram_below_4g;
2325 
2326     visit_type_size(v, name, &value, errp);
2327 }
2328 
2329 static void pc_machine_set_max_ram_below_4g(Object *obj, Visitor *v,
2330                                             const char *name, void *opaque,
2331                                             Error **errp)
2332 {
2333     PCMachineState *pcms = PC_MACHINE(obj);
2334     Error *error = NULL;
2335     uint64_t value;
2336 
2337     visit_type_size(v, name, &value, &error);
2338     if (error) {
2339         error_propagate(errp, error);
2340         return;
2341     }
2342     if (value > 4 * GiB) {
2343         error_setg(&error,
2344                    "Machine option 'max-ram-below-4g=%"PRIu64
2345                    "' expects size less than or equal to 4G", value);
2346         error_propagate(errp, error);
2347         return;
2348     }
2349 
2350     if (value < 1 * MiB) {
2351         warn_report("Only %" PRIu64 " bytes of RAM below the 4GiB boundary,"
2352                     "BIOS may not work with less than 1MiB", value);
2353     }
2354 
2355     pcms->max_ram_below_4g = value;
2356 }
2357 
2358 static void pc_machine_get_vmport(Object *obj, Visitor *v, const char *name,
2359                                   void *opaque, Error **errp)
2360 {
2361     PCMachineState *pcms = PC_MACHINE(obj);
2362     OnOffAuto vmport = pcms->vmport;
2363 
2364     visit_type_OnOffAuto(v, name, &vmport, errp);
2365 }
2366 
2367 static void pc_machine_set_vmport(Object *obj, Visitor *v, const char *name,
2368                                   void *opaque, Error **errp)
2369 {
2370     PCMachineState *pcms = PC_MACHINE(obj);
2371 
2372     visit_type_OnOffAuto(v, name, &pcms->vmport, errp);
2373 }
2374 
2375 bool pc_machine_is_smm_enabled(PCMachineState *pcms)
2376 {
2377     bool smm_available = false;
2378 
2379     if (pcms->smm == ON_OFF_AUTO_OFF) {
2380         return false;
2381     }
2382 
2383     if (tcg_enabled() || qtest_enabled()) {
2384         smm_available = true;
2385     } else if (kvm_enabled()) {
2386         smm_available = kvm_has_smm();
2387     }
2388 
2389     if (smm_available) {
2390         return true;
2391     }
2392 
2393     if (pcms->smm == ON_OFF_AUTO_ON) {
2394         error_report("System Management Mode not supported by this hypervisor.");
2395         exit(1);
2396     }
2397     return false;
2398 }
2399 
2400 static void pc_machine_get_smm(Object *obj, Visitor *v, const char *name,
2401                                void *opaque, Error **errp)
2402 {
2403     PCMachineState *pcms = PC_MACHINE(obj);
2404     OnOffAuto smm = pcms->smm;
2405 
2406     visit_type_OnOffAuto(v, name, &smm, errp);
2407 }
2408 
2409 static void pc_machine_set_smm(Object *obj, Visitor *v, const char *name,
2410                                void *opaque, Error **errp)
2411 {
2412     PCMachineState *pcms = PC_MACHINE(obj);
2413 
2414     visit_type_OnOffAuto(v, name, &pcms->smm, errp);
2415 }
2416 
2417 static bool pc_machine_get_nvdimm(Object *obj, Error **errp)
2418 {
2419     PCMachineState *pcms = PC_MACHINE(obj);
2420 
2421     return pcms->acpi_nvdimm_state.is_enabled;
2422 }
2423 
2424 static void pc_machine_set_nvdimm(Object *obj, bool value, Error **errp)
2425 {
2426     PCMachineState *pcms = PC_MACHINE(obj);
2427 
2428     pcms->acpi_nvdimm_state.is_enabled = value;
2429 }
2430 
2431 static char *pc_machine_get_nvdimm_persistence(Object *obj, Error **errp)
2432 {
2433     PCMachineState *pcms = PC_MACHINE(obj);
2434 
2435     return g_strdup(pcms->acpi_nvdimm_state.persistence_string);
2436 }
2437 
2438 static void pc_machine_set_nvdimm_persistence(Object *obj, const char *value,
2439                                                Error **errp)
2440 {
2441     PCMachineState *pcms = PC_MACHINE(obj);
2442     AcpiNVDIMMState *nvdimm_state = &pcms->acpi_nvdimm_state;
2443 
2444     if (strcmp(value, "cpu") == 0)
2445         nvdimm_state->persistence = 3;
2446     else if (strcmp(value, "mem-ctrl") == 0)
2447         nvdimm_state->persistence = 2;
2448     else {
2449         error_setg(errp, "-machine nvdimm-persistence=%s: unsupported option",
2450                    value);
2451         return;
2452     }
2453 
2454     g_free(nvdimm_state->persistence_string);
2455     nvdimm_state->persistence_string = g_strdup(value);
2456 }
2457 
2458 static bool pc_machine_get_smbus(Object *obj, Error **errp)
2459 {
2460     PCMachineState *pcms = PC_MACHINE(obj);
2461 
2462     return pcms->smbus_enabled;
2463 }
2464 
2465 static void pc_machine_set_smbus(Object *obj, bool value, Error **errp)
2466 {
2467     PCMachineState *pcms = PC_MACHINE(obj);
2468 
2469     pcms->smbus_enabled = value;
2470 }
2471 
2472 static bool pc_machine_get_sata(Object *obj, Error **errp)
2473 {
2474     PCMachineState *pcms = PC_MACHINE(obj);
2475 
2476     return pcms->sata_enabled;
2477 }
2478 
2479 static void pc_machine_set_sata(Object *obj, bool value, Error **errp)
2480 {
2481     PCMachineState *pcms = PC_MACHINE(obj);
2482 
2483     pcms->sata_enabled = value;
2484 }
2485 
2486 static bool pc_machine_get_pit(Object *obj, Error **errp)
2487 {
2488     PCMachineState *pcms = PC_MACHINE(obj);
2489 
2490     return pcms->pit_enabled;
2491 }
2492 
2493 static void pc_machine_set_pit(Object *obj, bool value, Error **errp)
2494 {
2495     PCMachineState *pcms = PC_MACHINE(obj);
2496 
2497     pcms->pit_enabled = value;
2498 }
2499 
2500 static void pc_machine_initfn(Object *obj)
2501 {
2502     PCMachineState *pcms = PC_MACHINE(obj);
2503 
2504     pcms->max_ram_below_4g = 0; /* use default */
2505     pcms->smm = ON_OFF_AUTO_AUTO;
2506     pcms->vmport = ON_OFF_AUTO_AUTO;
2507     /* nvdimm is disabled on default. */
2508     pcms->acpi_nvdimm_state.is_enabled = false;
2509     /* acpi build is enabled by default if machine supports it */
2510     pcms->acpi_build_enabled = PC_MACHINE_GET_CLASS(pcms)->has_acpi_build;
2511     pcms->smbus_enabled = true;
2512     pcms->sata_enabled = true;
2513     pcms->pit_enabled = true;
2514 }
2515 
2516 static void pc_machine_reset(void)
2517 {
2518     CPUState *cs;
2519     X86CPU *cpu;
2520 
2521     qemu_devices_reset();
2522 
2523     /* Reset APIC after devices have been reset to cancel
2524      * any changes that qemu_devices_reset() might have done.
2525      */
2526     CPU_FOREACH(cs) {
2527         cpu = X86_CPU(cs);
2528 
2529         if (cpu->apic_state) {
2530             device_reset(cpu->apic_state);
2531         }
2532     }
2533 }
2534 
2535 static CpuInstanceProperties
2536 pc_cpu_index_to_props(MachineState *ms, unsigned cpu_index)
2537 {
2538     MachineClass *mc = MACHINE_GET_CLASS(ms);
2539     const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(ms);
2540 
2541     assert(cpu_index < possible_cpus->len);
2542     return possible_cpus->cpus[cpu_index].props;
2543 }
2544 
2545 static int64_t pc_get_default_cpu_node_id(const MachineState *ms, int idx)
2546 {
2547    X86CPUTopoInfo topo;
2548 
2549    assert(idx < ms->possible_cpus->len);
2550    x86_topo_ids_from_apicid(ms->possible_cpus->cpus[idx].arch_id,
2551                             smp_cores, smp_threads, &topo);
2552    return topo.pkg_id % nb_numa_nodes;
2553 }
2554 
2555 static const CPUArchIdList *pc_possible_cpu_arch_ids(MachineState *ms)
2556 {
2557     int i;
2558 
2559     if (ms->possible_cpus) {
2560         /*
2561          * make sure that max_cpus hasn't changed since the first use, i.e.
2562          * -smp hasn't been parsed after it
2563         */
2564         assert(ms->possible_cpus->len == max_cpus);
2565         return ms->possible_cpus;
2566     }
2567 
2568     ms->possible_cpus = g_malloc0(sizeof(CPUArchIdList) +
2569                                   sizeof(CPUArchId) * max_cpus);
2570     ms->possible_cpus->len = max_cpus;
2571     for (i = 0; i < ms->possible_cpus->len; i++) {
2572         X86CPUTopoInfo topo;
2573 
2574         ms->possible_cpus->cpus[i].type = ms->cpu_type;
2575         ms->possible_cpus->cpus[i].vcpus_count = 1;
2576         ms->possible_cpus->cpus[i].arch_id = x86_cpu_apic_id_from_index(i);
2577         x86_topo_ids_from_apicid(ms->possible_cpus->cpus[i].arch_id,
2578                                  smp_cores, smp_threads, &topo);
2579         ms->possible_cpus->cpus[i].props.has_socket_id = true;
2580         ms->possible_cpus->cpus[i].props.socket_id = topo.pkg_id;
2581         ms->possible_cpus->cpus[i].props.has_core_id = true;
2582         ms->possible_cpus->cpus[i].props.core_id = topo.core_id;
2583         ms->possible_cpus->cpus[i].props.has_thread_id = true;
2584         ms->possible_cpus->cpus[i].props.thread_id = topo.smt_id;
2585     }
2586     return ms->possible_cpus;
2587 }
2588 
2589 static void x86_nmi(NMIState *n, int cpu_index, Error **errp)
2590 {
2591     /* cpu index isn't used */
2592     CPUState *cs;
2593 
2594     CPU_FOREACH(cs) {
2595         X86CPU *cpu = X86_CPU(cs);
2596 
2597         if (!cpu->apic_state) {
2598             cpu_interrupt(cs, CPU_INTERRUPT_NMI);
2599         } else {
2600             apic_deliver_nmi(cpu->apic_state);
2601         }
2602     }
2603 }
2604 
2605 static void pc_machine_class_init(ObjectClass *oc, void *data)
2606 {
2607     MachineClass *mc = MACHINE_CLASS(oc);
2608     PCMachineClass *pcmc = PC_MACHINE_CLASS(oc);
2609     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
2610     NMIClass *nc = NMI_CLASS(oc);
2611 
2612     pcmc->pci_enabled = true;
2613     pcmc->has_acpi_build = true;
2614     pcmc->rsdp_in_ram = true;
2615     pcmc->smbios_defaults = true;
2616     pcmc->smbios_uuid_encoded = true;
2617     pcmc->gigabyte_align = true;
2618     pcmc->has_reserved_memory = true;
2619     pcmc->kvmclock_enabled = true;
2620     pcmc->enforce_aligned_dimm = true;
2621     /* BIOS ACPI tables: 128K. Other BIOS datastructures: less than 4K reported
2622      * to be used at the moment, 32K should be enough for a while.  */
2623     pcmc->acpi_data_size = 0x20000 + 0x8000;
2624     pcmc->save_tsc_khz = true;
2625     pcmc->linuxboot_dma_enabled = true;
2626     assert(!mc->get_hotplug_handler);
2627     mc->get_hotplug_handler = pc_get_hotplug_handler;
2628     mc->cpu_index_to_instance_props = pc_cpu_index_to_props;
2629     mc->get_default_cpu_node_id = pc_get_default_cpu_node_id;
2630     mc->possible_cpu_arch_ids = pc_possible_cpu_arch_ids;
2631     mc->auto_enable_numa_with_memhp = true;
2632     mc->has_hotpluggable_cpus = true;
2633     mc->default_boot_order = "cad";
2634     mc->hot_add_cpu = pc_hot_add_cpu;
2635     mc->block_default_type = IF_IDE;
2636     mc->max_cpus = 255;
2637     mc->reset = pc_machine_reset;
2638     hc->pre_plug = pc_machine_device_pre_plug_cb;
2639     hc->plug = pc_machine_device_plug_cb;
2640     hc->unplug_request = pc_machine_device_unplug_request_cb;
2641     hc->unplug = pc_machine_device_unplug_cb;
2642     nc->nmi_monitor_handler = x86_nmi;
2643     mc->default_cpu_type = TARGET_DEFAULT_CPU_TYPE;
2644 
2645     object_class_property_add(oc, PC_MACHINE_DEVMEM_REGION_SIZE, "int",
2646         pc_machine_get_device_memory_region_size, NULL,
2647         NULL, NULL, &error_abort);
2648 
2649     object_class_property_add(oc, PC_MACHINE_MAX_RAM_BELOW_4G, "size",
2650         pc_machine_get_max_ram_below_4g, pc_machine_set_max_ram_below_4g,
2651         NULL, NULL, &error_abort);
2652 
2653     object_class_property_set_description(oc, PC_MACHINE_MAX_RAM_BELOW_4G,
2654         "Maximum ram below the 4G boundary (32bit boundary)", &error_abort);
2655 
2656     object_class_property_add(oc, PC_MACHINE_SMM, "OnOffAuto",
2657         pc_machine_get_smm, pc_machine_set_smm,
2658         NULL, NULL, &error_abort);
2659     object_class_property_set_description(oc, PC_MACHINE_SMM,
2660         "Enable SMM (pc & q35)", &error_abort);
2661 
2662     object_class_property_add(oc, PC_MACHINE_VMPORT, "OnOffAuto",
2663         pc_machine_get_vmport, pc_machine_set_vmport,
2664         NULL, NULL, &error_abort);
2665     object_class_property_set_description(oc, PC_MACHINE_VMPORT,
2666         "Enable vmport (pc & q35)", &error_abort);
2667 
2668     object_class_property_add_bool(oc, PC_MACHINE_NVDIMM,
2669         pc_machine_get_nvdimm, pc_machine_set_nvdimm, &error_abort);
2670 
2671     object_class_property_add_str(oc, PC_MACHINE_NVDIMM_PERSIST,
2672         pc_machine_get_nvdimm_persistence,
2673         pc_machine_set_nvdimm_persistence, &error_abort);
2674 
2675     object_class_property_add_bool(oc, PC_MACHINE_SMBUS,
2676         pc_machine_get_smbus, pc_machine_set_smbus, &error_abort);
2677 
2678     object_class_property_add_bool(oc, PC_MACHINE_SATA,
2679         pc_machine_get_sata, pc_machine_set_sata, &error_abort);
2680 
2681     object_class_property_add_bool(oc, PC_MACHINE_PIT,
2682         pc_machine_get_pit, pc_machine_set_pit, &error_abort);
2683 }
2684 
2685 static const TypeInfo pc_machine_info = {
2686     .name = TYPE_PC_MACHINE,
2687     .parent = TYPE_MACHINE,
2688     .abstract = true,
2689     .instance_size = sizeof(PCMachineState),
2690     .instance_init = pc_machine_initfn,
2691     .class_size = sizeof(PCMachineClass),
2692     .class_init = pc_machine_class_init,
2693     .interfaces = (InterfaceInfo[]) {
2694          { TYPE_HOTPLUG_HANDLER },
2695          { TYPE_NMI },
2696          { }
2697     },
2698 };
2699 
2700 static void pc_machine_register_types(void)
2701 {
2702     type_register_static(&pc_machine_info);
2703 }
2704 
2705 type_init(pc_machine_register_types)
2706