xref: /qemu/hw/i386/pc.c (revision abff1abf)
1 /*
2  * QEMU PC System Emulator
3  *
4  * Copyright (c) 2003-2004 Fabrice Bellard
5  *
6  * Permission is hereby granted, free of charge, to any person obtaining a copy
7  * of this software and associated documentation files (the "Software"), to deal
8  * in the Software without restriction, including without limitation the rights
9  * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
10  * copies of the Software, and to permit persons to whom the Software is
11  * furnished to do so, subject to the following conditions:
12  *
13  * The above copyright notice and this permission notice shall be included in
14  * all copies or substantial portions of the Software.
15  *
16  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
19  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
22  * THE SOFTWARE.
23  */
24 
25 #include "qemu/osdep.h"
26 #include "qemu/units.h"
27 #include "hw/i386/x86.h"
28 #include "hw/i386/pc.h"
29 #include "hw/char/serial.h"
30 #include "hw/char/parallel.h"
31 #include "hw/i386/apic.h"
32 #include "hw/i386/topology.h"
33 #include "hw/i386/fw_cfg.h"
34 #include "hw/i386/vmport.h"
35 #include "sysemu/cpus.h"
36 #include "hw/block/fdc.h"
37 #include "hw/ide.h"
38 #include "hw/pci/pci.h"
39 #include "hw/pci/pci_bus.h"
40 #include "hw/nvram/fw_cfg.h"
41 #include "hw/timer/hpet.h"
42 #include "hw/firmware/smbios.h"
43 #include "hw/loader.h"
44 #include "elf.h"
45 #include "migration/vmstate.h"
46 #include "multiboot.h"
47 #include "hw/rtc/mc146818rtc.h"
48 #include "hw/intc/i8259.h"
49 #include "hw/dma/i8257.h"
50 #include "hw/timer/i8254.h"
51 #include "hw/input/i8042.h"
52 #include "hw/irq.h"
53 #include "hw/audio/pcspk.h"
54 #include "hw/pci/msi.h"
55 #include "hw/sysbus.h"
56 #include "sysemu/sysemu.h"
57 #include "sysemu/tcg.h"
58 #include "sysemu/numa.h"
59 #include "sysemu/kvm.h"
60 #include "sysemu/xen.h"
61 #include "sysemu/qtest.h"
62 #include "sysemu/reset.h"
63 #include "sysemu/runstate.h"
64 #include "kvm_i386.h"
65 #include "hw/xen/xen.h"
66 #include "hw/xen/start_info.h"
67 #include "ui/qemu-spice.h"
68 #include "exec/memory.h"
69 #include "exec/address-spaces.h"
70 #include "sysemu/arch_init.h"
71 #include "qemu/bitmap.h"
72 #include "qemu/config-file.h"
73 #include "qemu/error-report.h"
74 #include "qemu/option.h"
75 #include "qemu/cutils.h"
76 #include "hw/acpi/acpi.h"
77 #include "hw/acpi/cpu_hotplug.h"
78 #include "hw/boards.h"
79 #include "acpi-build.h"
80 #include "hw/mem/pc-dimm.h"
81 #include "hw/mem/nvdimm.h"
82 #include "qapi/error.h"
83 #include "qapi/qapi-visit-common.h"
84 #include "qapi/visitor.h"
85 #include "hw/core/cpu.h"
86 #include "hw/usb.h"
87 #include "hw/i386/intel_iommu.h"
88 #include "hw/net/ne2000-isa.h"
89 #include "standard-headers/asm-x86/bootparam.h"
90 #include "hw/virtio/virtio-pmem-pci.h"
91 #include "hw/virtio/virtio-mem-pci.h"
92 #include "hw/mem/memory-device.h"
93 #include "sysemu/replay.h"
94 #include "qapi/qmp/qerror.h"
95 #include "e820_memory_layout.h"
96 #include "fw_cfg.h"
97 #include "trace.h"
98 #include CONFIG_DEVICES
99 
100 GlobalProperty pc_compat_5_0[] = {
101 };
102 const size_t pc_compat_5_0_len = G_N_ELEMENTS(pc_compat_5_0);
103 
104 GlobalProperty pc_compat_4_2[] = {
105     { "mch", "smbase-smram", "off" },
106 };
107 const size_t pc_compat_4_2_len = G_N_ELEMENTS(pc_compat_4_2);
108 
109 GlobalProperty pc_compat_4_1[] = {};
110 const size_t pc_compat_4_1_len = G_N_ELEMENTS(pc_compat_4_1);
111 
112 GlobalProperty pc_compat_4_0[] = {};
113 const size_t pc_compat_4_0_len = G_N_ELEMENTS(pc_compat_4_0);
114 
115 GlobalProperty pc_compat_3_1[] = {
116     { "intel-iommu", "dma-drain", "off" },
117     { "Opteron_G3" "-" TYPE_X86_CPU, "rdtscp", "off" },
118     { "Opteron_G4" "-" TYPE_X86_CPU, "rdtscp", "off" },
119     { "Opteron_G4" "-" TYPE_X86_CPU, "npt", "off" },
120     { "Opteron_G4" "-" TYPE_X86_CPU, "nrip-save", "off" },
121     { "Opteron_G5" "-" TYPE_X86_CPU, "rdtscp", "off" },
122     { "Opteron_G5" "-" TYPE_X86_CPU, "npt", "off" },
123     { "Opteron_G5" "-" TYPE_X86_CPU, "nrip-save", "off" },
124     { "EPYC" "-" TYPE_X86_CPU, "npt", "off" },
125     { "EPYC" "-" TYPE_X86_CPU, "nrip-save", "off" },
126     { "EPYC-IBPB" "-" TYPE_X86_CPU, "npt", "off" },
127     { "EPYC-IBPB" "-" TYPE_X86_CPU, "nrip-save", "off" },
128     { "Skylake-Client" "-" TYPE_X86_CPU,      "mpx", "on" },
129     { "Skylake-Client-IBRS" "-" TYPE_X86_CPU, "mpx", "on" },
130     { "Skylake-Server" "-" TYPE_X86_CPU,      "mpx", "on" },
131     { "Skylake-Server-IBRS" "-" TYPE_X86_CPU, "mpx", "on" },
132     { "Cascadelake-Server" "-" TYPE_X86_CPU,  "mpx", "on" },
133     { "Icelake-Client" "-" TYPE_X86_CPU,      "mpx", "on" },
134     { "Icelake-Server" "-" TYPE_X86_CPU,      "mpx", "on" },
135     { "Cascadelake-Server" "-" TYPE_X86_CPU, "stepping", "5" },
136     { TYPE_X86_CPU, "x-intel-pt-auto-level", "off" },
137 };
138 const size_t pc_compat_3_1_len = G_N_ELEMENTS(pc_compat_3_1);
139 
140 GlobalProperty pc_compat_3_0[] = {
141     { TYPE_X86_CPU, "x-hv-synic-kvm-only", "on" },
142     { "Skylake-Server" "-" TYPE_X86_CPU, "pku", "off" },
143     { "Skylake-Server-IBRS" "-" TYPE_X86_CPU, "pku", "off" },
144 };
145 const size_t pc_compat_3_0_len = G_N_ELEMENTS(pc_compat_3_0);
146 
147 GlobalProperty pc_compat_2_12[] = {
148     { TYPE_X86_CPU, "legacy-cache", "on" },
149     { TYPE_X86_CPU, "topoext", "off" },
150     { "EPYC-" TYPE_X86_CPU, "xlevel", "0x8000000a" },
151     { "EPYC-IBPB-" TYPE_X86_CPU, "xlevel", "0x8000000a" },
152 };
153 const size_t pc_compat_2_12_len = G_N_ELEMENTS(pc_compat_2_12);
154 
155 GlobalProperty pc_compat_2_11[] = {
156     { TYPE_X86_CPU, "x-migrate-smi-count", "off" },
157     { "Skylake-Server" "-" TYPE_X86_CPU, "clflushopt", "off" },
158 };
159 const size_t pc_compat_2_11_len = G_N_ELEMENTS(pc_compat_2_11);
160 
161 GlobalProperty pc_compat_2_10[] = {
162     { TYPE_X86_CPU, "x-hv-max-vps", "0x40" },
163     { "i440FX-pcihost", "x-pci-hole64-fix", "off" },
164     { "q35-pcihost", "x-pci-hole64-fix", "off" },
165 };
166 const size_t pc_compat_2_10_len = G_N_ELEMENTS(pc_compat_2_10);
167 
168 GlobalProperty pc_compat_2_9[] = {
169     { "mch", "extended-tseg-mbytes", "0" },
170 };
171 const size_t pc_compat_2_9_len = G_N_ELEMENTS(pc_compat_2_9);
172 
173 GlobalProperty pc_compat_2_8[] = {
174     { TYPE_X86_CPU, "tcg-cpuid", "off" },
175     { "kvmclock", "x-mach-use-reliable-get-clock", "off" },
176     { "ICH9-LPC", "x-smi-broadcast", "off" },
177     { TYPE_X86_CPU, "vmware-cpuid-freq", "off" },
178     { "Haswell-" TYPE_X86_CPU, "stepping", "1" },
179 };
180 const size_t pc_compat_2_8_len = G_N_ELEMENTS(pc_compat_2_8);
181 
182 GlobalProperty pc_compat_2_7[] = {
183     { TYPE_X86_CPU, "l3-cache", "off" },
184     { TYPE_X86_CPU, "full-cpuid-auto-level", "off" },
185     { "Opteron_G3" "-" TYPE_X86_CPU, "family", "15" },
186     { "Opteron_G3" "-" TYPE_X86_CPU, "model", "6" },
187     { "Opteron_G3" "-" TYPE_X86_CPU, "stepping", "1" },
188     { "isa-pcspk", "migrate", "off" },
189 };
190 const size_t pc_compat_2_7_len = G_N_ELEMENTS(pc_compat_2_7);
191 
192 GlobalProperty pc_compat_2_6[] = {
193     { TYPE_X86_CPU, "cpuid-0xb", "off" },
194     { "vmxnet3", "romfile", "" },
195     { TYPE_X86_CPU, "fill-mtrr-mask", "off" },
196     { "apic-common", "legacy-instance-id", "on", }
197 };
198 const size_t pc_compat_2_6_len = G_N_ELEMENTS(pc_compat_2_6);
199 
200 GlobalProperty pc_compat_2_5[] = {};
201 const size_t pc_compat_2_5_len = G_N_ELEMENTS(pc_compat_2_5);
202 
203 GlobalProperty pc_compat_2_4[] = {
204     PC_CPU_MODEL_IDS("2.4.0")
205     { "Haswell-" TYPE_X86_CPU, "abm", "off" },
206     { "Haswell-noTSX-" TYPE_X86_CPU, "abm", "off" },
207     { "Broadwell-" TYPE_X86_CPU, "abm", "off" },
208     { "Broadwell-noTSX-" TYPE_X86_CPU, "abm", "off" },
209     { "host" "-" TYPE_X86_CPU, "host-cache-info", "on" },
210     { TYPE_X86_CPU, "check", "off" },
211     { "qemu64" "-" TYPE_X86_CPU, "sse4a", "on" },
212     { "qemu64" "-" TYPE_X86_CPU, "abm", "on" },
213     { "qemu64" "-" TYPE_X86_CPU, "popcnt", "on" },
214     { "qemu32" "-" TYPE_X86_CPU, "popcnt", "on" },
215     { "Opteron_G2" "-" TYPE_X86_CPU, "rdtscp", "on" },
216     { "Opteron_G3" "-" TYPE_X86_CPU, "rdtscp", "on" },
217     { "Opteron_G4" "-" TYPE_X86_CPU, "rdtscp", "on" },
218     { "Opteron_G5" "-" TYPE_X86_CPU, "rdtscp", "on", }
219 };
220 const size_t pc_compat_2_4_len = G_N_ELEMENTS(pc_compat_2_4);
221 
222 GlobalProperty pc_compat_2_3[] = {
223     PC_CPU_MODEL_IDS("2.3.0")
224     { TYPE_X86_CPU, "arat", "off" },
225     { "qemu64" "-" TYPE_X86_CPU, "min-level", "4" },
226     { "kvm64" "-" TYPE_X86_CPU, "min-level", "5" },
227     { "pentium3" "-" TYPE_X86_CPU, "min-level", "2" },
228     { "n270" "-" TYPE_X86_CPU, "min-level", "5" },
229     { "Conroe" "-" TYPE_X86_CPU, "min-level", "4" },
230     { "Penryn" "-" TYPE_X86_CPU, "min-level", "4" },
231     { "Nehalem" "-" TYPE_X86_CPU, "min-level", "4" },
232     { "n270" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
233     { "Penryn" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
234     { "Conroe" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
235     { "Nehalem" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
236     { "Westmere" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
237     { "SandyBridge" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
238     { "IvyBridge" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
239     { "Haswell" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
240     { "Haswell-noTSX" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
241     { "Broadwell" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
242     { "Broadwell-noTSX" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
243     { TYPE_X86_CPU, "kvm-no-smi-migration", "on" },
244 };
245 const size_t pc_compat_2_3_len = G_N_ELEMENTS(pc_compat_2_3);
246 
247 GlobalProperty pc_compat_2_2[] = {
248     PC_CPU_MODEL_IDS("2.2.0")
249     { "kvm64" "-" TYPE_X86_CPU, "vme", "off" },
250     { "kvm32" "-" TYPE_X86_CPU, "vme", "off" },
251     { "Conroe" "-" TYPE_X86_CPU, "vme", "off" },
252     { "Penryn" "-" TYPE_X86_CPU, "vme", "off" },
253     { "Nehalem" "-" TYPE_X86_CPU, "vme", "off" },
254     { "Westmere" "-" TYPE_X86_CPU, "vme", "off" },
255     { "SandyBridge" "-" TYPE_X86_CPU, "vme", "off" },
256     { "Haswell" "-" TYPE_X86_CPU, "vme", "off" },
257     { "Broadwell" "-" TYPE_X86_CPU, "vme", "off" },
258     { "Opteron_G1" "-" TYPE_X86_CPU, "vme", "off" },
259     { "Opteron_G2" "-" TYPE_X86_CPU, "vme", "off" },
260     { "Opteron_G3" "-" TYPE_X86_CPU, "vme", "off" },
261     { "Opteron_G4" "-" TYPE_X86_CPU, "vme", "off" },
262     { "Opteron_G5" "-" TYPE_X86_CPU, "vme", "off" },
263     { "Haswell" "-" TYPE_X86_CPU, "f16c", "off" },
264     { "Haswell" "-" TYPE_X86_CPU, "rdrand", "off" },
265     { "Broadwell" "-" TYPE_X86_CPU, "f16c", "off" },
266     { "Broadwell" "-" TYPE_X86_CPU, "rdrand", "off" },
267 };
268 const size_t pc_compat_2_2_len = G_N_ELEMENTS(pc_compat_2_2);
269 
270 GlobalProperty pc_compat_2_1[] = {
271     PC_CPU_MODEL_IDS("2.1.0")
272     { "coreduo" "-" TYPE_X86_CPU, "vmx", "on" },
273     { "core2duo" "-" TYPE_X86_CPU, "vmx", "on" },
274 };
275 const size_t pc_compat_2_1_len = G_N_ELEMENTS(pc_compat_2_1);
276 
277 GlobalProperty pc_compat_2_0[] = {
278     PC_CPU_MODEL_IDS("2.0.0")
279     { "virtio-scsi-pci", "any_layout", "off" },
280     { "PIIX4_PM", "memory-hotplug-support", "off" },
281     { "apic", "version", "0x11" },
282     { "nec-usb-xhci", "superspeed-ports-first", "off" },
283     { "nec-usb-xhci", "force-pcie-endcap", "on" },
284     { "pci-serial", "prog_if", "0" },
285     { "pci-serial-2x", "prog_if", "0" },
286     { "pci-serial-4x", "prog_if", "0" },
287     { "virtio-net-pci", "guest_announce", "off" },
288     { "ICH9-LPC", "memory-hotplug-support", "off" },
289     { "xio3130-downstream", COMPAT_PROP_PCP, "off" },
290     { "ioh3420", COMPAT_PROP_PCP, "off" },
291 };
292 const size_t pc_compat_2_0_len = G_N_ELEMENTS(pc_compat_2_0);
293 
294 GlobalProperty pc_compat_1_7[] = {
295     PC_CPU_MODEL_IDS("1.7.0")
296     { TYPE_USB_DEVICE, "msos-desc", "no" },
297     { "PIIX4_PM", "acpi-pci-hotplug-with-bridge-support", "off" },
298     { "hpet", HPET_INTCAP, "4" },
299 };
300 const size_t pc_compat_1_7_len = G_N_ELEMENTS(pc_compat_1_7);
301 
302 GlobalProperty pc_compat_1_6[] = {
303     PC_CPU_MODEL_IDS("1.6.0")
304     { "e1000", "mitigation", "off" },
305     { "qemu64-" TYPE_X86_CPU, "model", "2" },
306     { "qemu32-" TYPE_X86_CPU, "model", "3" },
307     { "i440FX-pcihost", "short_root_bus", "1" },
308     { "q35-pcihost", "short_root_bus", "1" },
309 };
310 const size_t pc_compat_1_6_len = G_N_ELEMENTS(pc_compat_1_6);
311 
312 GlobalProperty pc_compat_1_5[] = {
313     PC_CPU_MODEL_IDS("1.5.0")
314     { "Conroe-" TYPE_X86_CPU, "model", "2" },
315     { "Conroe-" TYPE_X86_CPU, "min-level", "2" },
316     { "Penryn-" TYPE_X86_CPU, "model", "2" },
317     { "Penryn-" TYPE_X86_CPU, "min-level", "2" },
318     { "Nehalem-" TYPE_X86_CPU, "model", "2" },
319     { "Nehalem-" TYPE_X86_CPU, "min-level", "2" },
320     { "virtio-net-pci", "any_layout", "off" },
321     { TYPE_X86_CPU, "pmu", "on" },
322     { "i440FX-pcihost", "short_root_bus", "0" },
323     { "q35-pcihost", "short_root_bus", "0" },
324 };
325 const size_t pc_compat_1_5_len = G_N_ELEMENTS(pc_compat_1_5);
326 
327 GlobalProperty pc_compat_1_4[] = {
328     PC_CPU_MODEL_IDS("1.4.0")
329     { "scsi-hd", "discard_granularity", "0" },
330     { "scsi-cd", "discard_granularity", "0" },
331     { "scsi-disk", "discard_granularity", "0" },
332     { "ide-hd", "discard_granularity", "0" },
333     { "ide-cd", "discard_granularity", "0" },
334     { "ide-drive", "discard_granularity", "0" },
335     { "virtio-blk-pci", "discard_granularity", "0" },
336     /* DEV_NVECTORS_UNSPECIFIED as a uint32_t string: */
337     { "virtio-serial-pci", "vectors", "0xFFFFFFFF" },
338     { "virtio-net-pci", "ctrl_guest_offloads", "off" },
339     { "e1000", "romfile", "pxe-e1000.rom" },
340     { "ne2k_pci", "romfile", "pxe-ne2k_pci.rom" },
341     { "pcnet", "romfile", "pxe-pcnet.rom" },
342     { "rtl8139", "romfile", "pxe-rtl8139.rom" },
343     { "virtio-net-pci", "romfile", "pxe-virtio.rom" },
344     { "486-" TYPE_X86_CPU, "model", "0" },
345     { "n270" "-" TYPE_X86_CPU, "movbe", "off" },
346     { "Westmere" "-" TYPE_X86_CPU, "pclmulqdq", "off" },
347 };
348 const size_t pc_compat_1_4_len = G_N_ELEMENTS(pc_compat_1_4);
349 
350 GSIState *pc_gsi_create(qemu_irq **irqs, bool pci_enabled)
351 {
352     GSIState *s;
353 
354     s = g_new0(GSIState, 1);
355     if (kvm_ioapic_in_kernel()) {
356         kvm_pc_setup_irq_routing(pci_enabled);
357     }
358     *irqs = qemu_allocate_irqs(gsi_handler, s, GSI_NUM_PINS);
359 
360     return s;
361 }
362 
363 static void ioport80_write(void *opaque, hwaddr addr, uint64_t data,
364                            unsigned size)
365 {
366 }
367 
368 static uint64_t ioport80_read(void *opaque, hwaddr addr, unsigned size)
369 {
370     return 0xffffffffffffffffULL;
371 }
372 
373 /* MSDOS compatibility mode FPU exception support */
374 static void ioportF0_write(void *opaque, hwaddr addr, uint64_t data,
375                            unsigned size)
376 {
377     if (tcg_enabled()) {
378         cpu_set_ignne();
379     }
380 }
381 
382 static uint64_t ioportF0_read(void *opaque, hwaddr addr, unsigned size)
383 {
384     return 0xffffffffffffffffULL;
385 }
386 
387 /* PC cmos mappings */
388 
389 #define REG_EQUIPMENT_BYTE          0x14
390 
391 static void cmos_init_hd(ISADevice *s, int type_ofs, int info_ofs,
392                          int16_t cylinders, int8_t heads, int8_t sectors)
393 {
394     rtc_set_memory(s, type_ofs, 47);
395     rtc_set_memory(s, info_ofs, cylinders);
396     rtc_set_memory(s, info_ofs + 1, cylinders >> 8);
397     rtc_set_memory(s, info_ofs + 2, heads);
398     rtc_set_memory(s, info_ofs + 3, 0xff);
399     rtc_set_memory(s, info_ofs + 4, 0xff);
400     rtc_set_memory(s, info_ofs + 5, 0xc0 | ((heads > 8) << 3));
401     rtc_set_memory(s, info_ofs + 6, cylinders);
402     rtc_set_memory(s, info_ofs + 7, cylinders >> 8);
403     rtc_set_memory(s, info_ofs + 8, sectors);
404 }
405 
406 /* convert boot_device letter to something recognizable by the bios */
407 static int boot_device2nibble(char boot_device)
408 {
409     switch(boot_device) {
410     case 'a':
411     case 'b':
412         return 0x01; /* floppy boot */
413     case 'c':
414         return 0x02; /* hard drive boot */
415     case 'd':
416         return 0x03; /* CD-ROM boot */
417     case 'n':
418         return 0x04; /* Network boot */
419     }
420     return 0;
421 }
422 
423 static void set_boot_dev(ISADevice *s, const char *boot_device, Error **errp)
424 {
425 #define PC_MAX_BOOT_DEVICES 3
426     int nbds, bds[3] = { 0, };
427     int i;
428 
429     nbds = strlen(boot_device);
430     if (nbds > PC_MAX_BOOT_DEVICES) {
431         error_setg(errp, "Too many boot devices for PC");
432         return;
433     }
434     for (i = 0; i < nbds; i++) {
435         bds[i] = boot_device2nibble(boot_device[i]);
436         if (bds[i] == 0) {
437             error_setg(errp, "Invalid boot device for PC: '%c'",
438                        boot_device[i]);
439             return;
440         }
441     }
442     rtc_set_memory(s, 0x3d, (bds[1] << 4) | bds[0]);
443     rtc_set_memory(s, 0x38, (bds[2] << 4) | (fd_bootchk ? 0x0 : 0x1));
444 }
445 
446 static void pc_boot_set(void *opaque, const char *boot_device, Error **errp)
447 {
448     set_boot_dev(opaque, boot_device, errp);
449 }
450 
451 static void pc_cmos_init_floppy(ISADevice *rtc_state, ISADevice *floppy)
452 {
453     int val, nb, i;
454     FloppyDriveType fd_type[2] = { FLOPPY_DRIVE_TYPE_NONE,
455                                    FLOPPY_DRIVE_TYPE_NONE };
456 
457     /* floppy type */
458     if (floppy) {
459         for (i = 0; i < 2; i++) {
460             fd_type[i] = isa_fdc_get_drive_type(floppy, i);
461         }
462     }
463     val = (cmos_get_fd_drive_type(fd_type[0]) << 4) |
464         cmos_get_fd_drive_type(fd_type[1]);
465     rtc_set_memory(rtc_state, 0x10, val);
466 
467     val = rtc_get_memory(rtc_state, REG_EQUIPMENT_BYTE);
468     nb = 0;
469     if (fd_type[0] != FLOPPY_DRIVE_TYPE_NONE) {
470         nb++;
471     }
472     if (fd_type[1] != FLOPPY_DRIVE_TYPE_NONE) {
473         nb++;
474     }
475     switch (nb) {
476     case 0:
477         break;
478     case 1:
479         val |= 0x01; /* 1 drive, ready for boot */
480         break;
481     case 2:
482         val |= 0x41; /* 2 drives, ready for boot */
483         break;
484     }
485     rtc_set_memory(rtc_state, REG_EQUIPMENT_BYTE, val);
486 }
487 
488 typedef struct pc_cmos_init_late_arg {
489     ISADevice *rtc_state;
490     BusState *idebus[2];
491 } pc_cmos_init_late_arg;
492 
493 typedef struct check_fdc_state {
494     ISADevice *floppy;
495     bool multiple;
496 } CheckFdcState;
497 
498 static int check_fdc(Object *obj, void *opaque)
499 {
500     CheckFdcState *state = opaque;
501     Object *fdc;
502     uint32_t iobase;
503     Error *local_err = NULL;
504 
505     fdc = object_dynamic_cast(obj, TYPE_ISA_FDC);
506     if (!fdc) {
507         return 0;
508     }
509 
510     iobase = object_property_get_uint(obj, "iobase", &local_err);
511     if (local_err || iobase != 0x3f0) {
512         error_free(local_err);
513         return 0;
514     }
515 
516     if (state->floppy) {
517         state->multiple = true;
518     } else {
519         state->floppy = ISA_DEVICE(obj);
520     }
521     return 0;
522 }
523 
524 static const char * const fdc_container_path[] = {
525     "/unattached", "/peripheral", "/peripheral-anon"
526 };
527 
528 /*
529  * Locate the FDC at IO address 0x3f0, in order to configure the CMOS registers
530  * and ACPI objects.
531  */
532 ISADevice *pc_find_fdc0(void)
533 {
534     int i;
535     Object *container;
536     CheckFdcState state = { 0 };
537 
538     for (i = 0; i < ARRAY_SIZE(fdc_container_path); i++) {
539         container = container_get(qdev_get_machine(), fdc_container_path[i]);
540         object_child_foreach(container, check_fdc, &state);
541     }
542 
543     if (state.multiple) {
544         warn_report("multiple floppy disk controllers with "
545                     "iobase=0x3f0 have been found");
546         error_printf("the one being picked for CMOS setup might not reflect "
547                      "your intent");
548     }
549 
550     return state.floppy;
551 }
552 
553 static void pc_cmos_init_late(void *opaque)
554 {
555     pc_cmos_init_late_arg *arg = opaque;
556     ISADevice *s = arg->rtc_state;
557     int16_t cylinders;
558     int8_t heads, sectors;
559     int val;
560     int i, trans;
561 
562     val = 0;
563     if (arg->idebus[0] && ide_get_geometry(arg->idebus[0], 0,
564                                            &cylinders, &heads, &sectors) >= 0) {
565         cmos_init_hd(s, 0x19, 0x1b, cylinders, heads, sectors);
566         val |= 0xf0;
567     }
568     if (arg->idebus[0] && ide_get_geometry(arg->idebus[0], 1,
569                                            &cylinders, &heads, &sectors) >= 0) {
570         cmos_init_hd(s, 0x1a, 0x24, cylinders, heads, sectors);
571         val |= 0x0f;
572     }
573     rtc_set_memory(s, 0x12, val);
574 
575     val = 0;
576     for (i = 0; i < 4; i++) {
577         /* NOTE: ide_get_geometry() returns the physical
578            geometry.  It is always such that: 1 <= sects <= 63, 1
579            <= heads <= 16, 1 <= cylinders <= 16383. The BIOS
580            geometry can be different if a translation is done. */
581         if (arg->idebus[i / 2] &&
582             ide_get_geometry(arg->idebus[i / 2], i % 2,
583                              &cylinders, &heads, &sectors) >= 0) {
584             trans = ide_get_bios_chs_trans(arg->idebus[i / 2], i % 2) - 1;
585             assert((trans & ~3) == 0);
586             val |= trans << (i * 2);
587         }
588     }
589     rtc_set_memory(s, 0x39, val);
590 
591     pc_cmos_init_floppy(s, pc_find_fdc0());
592 
593     qemu_unregister_reset(pc_cmos_init_late, opaque);
594 }
595 
596 void pc_cmos_init(PCMachineState *pcms,
597                   BusState *idebus0, BusState *idebus1,
598                   ISADevice *s)
599 {
600     int val;
601     static pc_cmos_init_late_arg arg;
602     X86MachineState *x86ms = X86_MACHINE(pcms);
603 
604     /* various important CMOS locations needed by PC/Bochs bios */
605 
606     /* memory size */
607     /* base memory (first MiB) */
608     val = MIN(x86ms->below_4g_mem_size / KiB, 640);
609     rtc_set_memory(s, 0x15, val);
610     rtc_set_memory(s, 0x16, val >> 8);
611     /* extended memory (next 64MiB) */
612     if (x86ms->below_4g_mem_size > 1 * MiB) {
613         val = (x86ms->below_4g_mem_size - 1 * MiB) / KiB;
614     } else {
615         val = 0;
616     }
617     if (val > 65535)
618         val = 65535;
619     rtc_set_memory(s, 0x17, val);
620     rtc_set_memory(s, 0x18, val >> 8);
621     rtc_set_memory(s, 0x30, val);
622     rtc_set_memory(s, 0x31, val >> 8);
623     /* memory between 16MiB and 4GiB */
624     if (x86ms->below_4g_mem_size > 16 * MiB) {
625         val = (x86ms->below_4g_mem_size - 16 * MiB) / (64 * KiB);
626     } else {
627         val = 0;
628     }
629     if (val > 65535)
630         val = 65535;
631     rtc_set_memory(s, 0x34, val);
632     rtc_set_memory(s, 0x35, val >> 8);
633     /* memory above 4GiB */
634     val = x86ms->above_4g_mem_size / 65536;
635     rtc_set_memory(s, 0x5b, val);
636     rtc_set_memory(s, 0x5c, val >> 8);
637     rtc_set_memory(s, 0x5d, val >> 16);
638 
639     object_property_add_link(OBJECT(pcms), "rtc_state",
640                              TYPE_ISA_DEVICE,
641                              (Object **)&x86ms->rtc,
642                              object_property_allow_set_link,
643                              OBJ_PROP_LINK_STRONG);
644     object_property_set_link(OBJECT(pcms), "rtc_state", OBJECT(s),
645                              &error_abort);
646 
647     set_boot_dev(s, MACHINE(pcms)->boot_order, &error_fatal);
648 
649     val = 0;
650     val |= 0x02; /* FPU is there */
651     val |= 0x04; /* PS/2 mouse installed */
652     rtc_set_memory(s, REG_EQUIPMENT_BYTE, val);
653 
654     /* hard drives and FDC */
655     arg.rtc_state = s;
656     arg.idebus[0] = idebus0;
657     arg.idebus[1] = idebus1;
658     qemu_register_reset(pc_cmos_init_late, &arg);
659 }
660 
661 static void handle_a20_line_change(void *opaque, int irq, int level)
662 {
663     X86CPU *cpu = opaque;
664 
665     /* XXX: send to all CPUs ? */
666     /* XXX: add logic to handle multiple A20 line sources */
667     x86_cpu_set_a20(cpu, level);
668 }
669 
670 #define NE2000_NB_MAX 6
671 
672 static const int ne2000_io[NE2000_NB_MAX] = { 0x300, 0x320, 0x340, 0x360,
673                                               0x280, 0x380 };
674 static const int ne2000_irq[NE2000_NB_MAX] = { 9, 10, 11, 3, 4, 5 };
675 
676 void pc_init_ne2k_isa(ISABus *bus, NICInfo *nd)
677 {
678     static int nb_ne2k = 0;
679 
680     if (nb_ne2k == NE2000_NB_MAX)
681         return;
682     isa_ne2000_init(bus, ne2000_io[nb_ne2k],
683                     ne2000_irq[nb_ne2k], nd);
684     nb_ne2k++;
685 }
686 
687 void pc_acpi_smi_interrupt(void *opaque, int irq, int level)
688 {
689     X86CPU *cpu = opaque;
690 
691     if (level) {
692         cpu_interrupt(CPU(cpu), CPU_INTERRUPT_SMI);
693     }
694 }
695 
696 /*
697  * This function is very similar to smp_parse()
698  * in hw/core/machine.c but includes CPU die support.
699  */
700 void pc_smp_parse(MachineState *ms, QemuOpts *opts)
701 {
702     X86MachineState *x86ms = X86_MACHINE(ms);
703 
704     if (opts) {
705         unsigned cpus    = qemu_opt_get_number(opts, "cpus", 0);
706         unsigned sockets = qemu_opt_get_number(opts, "sockets", 0);
707         unsigned dies = qemu_opt_get_number(opts, "dies", 1);
708         unsigned cores   = qemu_opt_get_number(opts, "cores", 0);
709         unsigned threads = qemu_opt_get_number(opts, "threads", 0);
710 
711         /* compute missing values, prefer sockets over cores over threads */
712         if (cpus == 0 || sockets == 0) {
713             cores = cores > 0 ? cores : 1;
714             threads = threads > 0 ? threads : 1;
715             if (cpus == 0) {
716                 sockets = sockets > 0 ? sockets : 1;
717                 cpus = cores * threads * dies * sockets;
718             } else {
719                 ms->smp.max_cpus =
720                         qemu_opt_get_number(opts, "maxcpus", cpus);
721                 sockets = ms->smp.max_cpus / (cores * threads * dies);
722             }
723         } else if (cores == 0) {
724             threads = threads > 0 ? threads : 1;
725             cores = cpus / (sockets * dies * threads);
726             cores = cores > 0 ? cores : 1;
727         } else if (threads == 0) {
728             threads = cpus / (cores * dies * sockets);
729             threads = threads > 0 ? threads : 1;
730         } else if (sockets * dies * cores * threads < cpus) {
731             error_report("cpu topology: "
732                          "sockets (%u) * dies (%u) * cores (%u) * threads (%u) < "
733                          "smp_cpus (%u)",
734                          sockets, dies, cores, threads, cpus);
735             exit(1);
736         }
737 
738         ms->smp.max_cpus =
739                 qemu_opt_get_number(opts, "maxcpus", cpus);
740 
741         if (ms->smp.max_cpus < cpus) {
742             error_report("maxcpus must be equal to or greater than smp");
743             exit(1);
744         }
745 
746         if (sockets * dies * cores * threads > ms->smp.max_cpus) {
747             error_report("cpu topology: "
748                          "sockets (%u) * dies (%u) * cores (%u) * threads (%u) > "
749                          "maxcpus (%u)",
750                          sockets, dies, cores, threads,
751                          ms->smp.max_cpus);
752             exit(1);
753         }
754 
755         if (sockets * dies * cores * threads != ms->smp.max_cpus) {
756             warn_report("Invalid CPU topology deprecated: "
757                         "sockets (%u) * dies (%u) * cores (%u) * threads (%u) "
758                         "!= maxcpus (%u)",
759                         sockets, dies, cores, threads,
760                         ms->smp.max_cpus);
761         }
762 
763         ms->smp.cpus = cpus;
764         ms->smp.cores = cores;
765         ms->smp.threads = threads;
766         ms->smp.sockets = sockets;
767         x86ms->smp_dies = dies;
768     }
769 
770     if (ms->smp.cpus > 1) {
771         Error *blocker = NULL;
772         error_setg(&blocker, QERR_REPLAY_NOT_SUPPORTED, "smp");
773         replay_add_blocker(blocker);
774     }
775 }
776 
777 void pc_hot_add_cpu(MachineState *ms, const int64_t id, Error **errp)
778 {
779     X86MachineState *x86ms = X86_MACHINE(ms);
780     int64_t apic_id = x86_cpu_apic_id_from_index(x86ms, id);
781     Error *local_err = NULL;
782 
783     if (id < 0) {
784         error_setg(errp, "Invalid CPU id: %" PRIi64, id);
785         return;
786     }
787 
788     if (apic_id >= ACPI_CPU_HOTPLUG_ID_LIMIT) {
789         error_setg(errp, "Unable to add CPU: %" PRIi64
790                    ", resulting APIC ID (%" PRIi64 ") is too large",
791                    id, apic_id);
792         return;
793     }
794 
795 
796     x86_cpu_new(X86_MACHINE(ms), apic_id, &local_err);
797     if (local_err) {
798         error_propagate(errp, local_err);
799         return;
800     }
801 }
802 
803 static void rtc_set_cpus_count(ISADevice *rtc, uint16_t cpus_count)
804 {
805     if (cpus_count > 0xff) {
806         /* If the number of CPUs can't be represented in 8 bits, the
807          * BIOS must use "FW_CFG_NB_CPUS". Set RTC field to 0 just
808          * to make old BIOSes fail more predictably.
809          */
810         rtc_set_memory(rtc, 0x5f, 0);
811     } else {
812         rtc_set_memory(rtc, 0x5f, cpus_count - 1);
813     }
814 }
815 
816 static
817 void pc_machine_done(Notifier *notifier, void *data)
818 {
819     PCMachineState *pcms = container_of(notifier,
820                                         PCMachineState, machine_done);
821     X86MachineState *x86ms = X86_MACHINE(pcms);
822     PCIBus *bus = pcms->bus;
823 
824     /* set the number of CPUs */
825     rtc_set_cpus_count(x86ms->rtc, x86ms->boot_cpus);
826 
827     if (bus) {
828         int extra_hosts = 0;
829 
830         QLIST_FOREACH(bus, &bus->child, sibling) {
831             /* look for expander root buses */
832             if (pci_bus_is_root(bus)) {
833                 extra_hosts++;
834             }
835         }
836         if (extra_hosts && x86ms->fw_cfg) {
837             uint64_t *val = g_malloc(sizeof(*val));
838             *val = cpu_to_le64(extra_hosts);
839             fw_cfg_add_file(x86ms->fw_cfg,
840                     "etc/extra-pci-roots", val, sizeof(*val));
841         }
842     }
843 
844     acpi_setup();
845     if (x86ms->fw_cfg) {
846         fw_cfg_build_smbios(MACHINE(pcms), x86ms->fw_cfg);
847         fw_cfg_build_feature_control(MACHINE(pcms), x86ms->fw_cfg);
848         /* update FW_CFG_NB_CPUS to account for -device added CPUs */
849         fw_cfg_modify_i16(x86ms->fw_cfg, FW_CFG_NB_CPUS, x86ms->boot_cpus);
850     }
851 
852     if (x86ms->apic_id_limit > 255 && !xen_enabled()) {
853         IntelIOMMUState *iommu = INTEL_IOMMU_DEVICE(x86_iommu_get_default());
854 
855         if (!iommu || !x86_iommu_ir_supported(X86_IOMMU_DEVICE(iommu)) ||
856             iommu->intr_eim != ON_OFF_AUTO_ON) {
857             error_report("current -smp configuration requires "
858                          "Extended Interrupt Mode enabled. "
859                          "You can add an IOMMU using: "
860                          "-device intel-iommu,intremap=on,eim=on");
861             exit(EXIT_FAILURE);
862         }
863     }
864 }
865 
866 void pc_guest_info_init(PCMachineState *pcms)
867 {
868     int i;
869     MachineState *ms = MACHINE(pcms);
870     X86MachineState *x86ms = X86_MACHINE(pcms);
871 
872     x86ms->apic_xrupt_override = kvm_allows_irq0_override();
873     pcms->numa_nodes = ms->numa_state->num_nodes;
874     pcms->node_mem = g_malloc0(pcms->numa_nodes *
875                                     sizeof *pcms->node_mem);
876     for (i = 0; i < ms->numa_state->num_nodes; i++) {
877         pcms->node_mem[i] = ms->numa_state->nodes[i].node_mem;
878     }
879 
880     pcms->machine_done.notify = pc_machine_done;
881     qemu_add_machine_init_done_notifier(&pcms->machine_done);
882 }
883 
884 /* setup pci memory address space mapping into system address space */
885 void pc_pci_as_mapping_init(Object *owner, MemoryRegion *system_memory,
886                             MemoryRegion *pci_address_space)
887 {
888     /* Set to lower priority than RAM */
889     memory_region_add_subregion_overlap(system_memory, 0x0,
890                                         pci_address_space, -1);
891 }
892 
893 void xen_load_linux(PCMachineState *pcms)
894 {
895     int i;
896     FWCfgState *fw_cfg;
897     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
898     X86MachineState *x86ms = X86_MACHINE(pcms);
899 
900     assert(MACHINE(pcms)->kernel_filename != NULL);
901 
902     fw_cfg = fw_cfg_init_io(FW_CFG_IO_BASE);
903     fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, x86ms->boot_cpus);
904     rom_set_fw(fw_cfg);
905 
906     x86_load_linux(x86ms, fw_cfg, pcmc->acpi_data_size,
907                    pcmc->pvh_enabled, pcmc->linuxboot_dma_enabled);
908     for (i = 0; i < nb_option_roms; i++) {
909         assert(!strcmp(option_rom[i].name, "linuxboot.bin") ||
910                !strcmp(option_rom[i].name, "linuxboot_dma.bin") ||
911                !strcmp(option_rom[i].name, "pvh.bin") ||
912                !strcmp(option_rom[i].name, "multiboot.bin"));
913         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
914     }
915     x86ms->fw_cfg = fw_cfg;
916 }
917 
918 void pc_memory_init(PCMachineState *pcms,
919                     MemoryRegion *system_memory,
920                     MemoryRegion *rom_memory,
921                     MemoryRegion **ram_memory)
922 {
923     int linux_boot, i;
924     MemoryRegion *option_rom_mr;
925     MemoryRegion *ram_below_4g, *ram_above_4g;
926     FWCfgState *fw_cfg;
927     MachineState *machine = MACHINE(pcms);
928     MachineClass *mc = MACHINE_GET_CLASS(machine);
929     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
930     X86MachineState *x86ms = X86_MACHINE(pcms);
931 
932     assert(machine->ram_size == x86ms->below_4g_mem_size +
933                                 x86ms->above_4g_mem_size);
934 
935     linux_boot = (machine->kernel_filename != NULL);
936 
937     /*
938      * Split single memory region and use aliases to address portions of it,
939      * done for backwards compatibility with older qemus.
940      */
941     *ram_memory = machine->ram;
942     ram_below_4g = g_malloc(sizeof(*ram_below_4g));
943     memory_region_init_alias(ram_below_4g, NULL, "ram-below-4g", machine->ram,
944                              0, x86ms->below_4g_mem_size);
945     memory_region_add_subregion(system_memory, 0, ram_below_4g);
946     e820_add_entry(0, x86ms->below_4g_mem_size, E820_RAM);
947     if (x86ms->above_4g_mem_size > 0) {
948         ram_above_4g = g_malloc(sizeof(*ram_above_4g));
949         memory_region_init_alias(ram_above_4g, NULL, "ram-above-4g",
950                                  machine->ram,
951                                  x86ms->below_4g_mem_size,
952                                  x86ms->above_4g_mem_size);
953         memory_region_add_subregion(system_memory, 0x100000000ULL,
954                                     ram_above_4g);
955         e820_add_entry(0x100000000ULL, x86ms->above_4g_mem_size, E820_RAM);
956     }
957 
958     if (!pcmc->has_reserved_memory &&
959         (machine->ram_slots ||
960          (machine->maxram_size > machine->ram_size))) {
961 
962         error_report("\"-memory 'slots|maxmem'\" is not supported by: %s",
963                      mc->name);
964         exit(EXIT_FAILURE);
965     }
966 
967     /* always allocate the device memory information */
968     machine->device_memory = g_malloc0(sizeof(*machine->device_memory));
969 
970     /* initialize device memory address space */
971     if (pcmc->has_reserved_memory &&
972         (machine->ram_size < machine->maxram_size)) {
973         ram_addr_t device_mem_size = machine->maxram_size - machine->ram_size;
974 
975         if (machine->ram_slots > ACPI_MAX_RAM_SLOTS) {
976             error_report("unsupported amount of memory slots: %"PRIu64,
977                          machine->ram_slots);
978             exit(EXIT_FAILURE);
979         }
980 
981         if (QEMU_ALIGN_UP(machine->maxram_size,
982                           TARGET_PAGE_SIZE) != machine->maxram_size) {
983             error_report("maximum memory size must by aligned to multiple of "
984                          "%d bytes", TARGET_PAGE_SIZE);
985             exit(EXIT_FAILURE);
986         }
987 
988         machine->device_memory->base =
989             ROUND_UP(0x100000000ULL + x86ms->above_4g_mem_size, 1 * GiB);
990 
991         if (pcmc->enforce_aligned_dimm) {
992             /* size device region assuming 1G page max alignment per slot */
993             device_mem_size += (1 * GiB) * machine->ram_slots;
994         }
995 
996         if ((machine->device_memory->base + device_mem_size) <
997             device_mem_size) {
998             error_report("unsupported amount of maximum memory: " RAM_ADDR_FMT,
999                          machine->maxram_size);
1000             exit(EXIT_FAILURE);
1001         }
1002 
1003         memory_region_init(&machine->device_memory->mr, OBJECT(pcms),
1004                            "device-memory", device_mem_size);
1005         memory_region_add_subregion(system_memory, machine->device_memory->base,
1006                                     &machine->device_memory->mr);
1007     }
1008 
1009     /* Initialize PC system firmware */
1010     pc_system_firmware_init(pcms, rom_memory);
1011 
1012     option_rom_mr = g_malloc(sizeof(*option_rom_mr));
1013     memory_region_init_ram(option_rom_mr, NULL, "pc.rom", PC_ROM_SIZE,
1014                            &error_fatal);
1015     if (pcmc->pci_enabled) {
1016         memory_region_set_readonly(option_rom_mr, true);
1017     }
1018     memory_region_add_subregion_overlap(rom_memory,
1019                                         PC_ROM_MIN_VGA,
1020                                         option_rom_mr,
1021                                         1);
1022 
1023     fw_cfg = fw_cfg_arch_create(machine,
1024                                 x86ms->boot_cpus, x86ms->apic_id_limit);
1025 
1026     rom_set_fw(fw_cfg);
1027 
1028     if (pcmc->has_reserved_memory && machine->device_memory->base) {
1029         uint64_t *val = g_malloc(sizeof(*val));
1030         PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1031         uint64_t res_mem_end = machine->device_memory->base;
1032 
1033         if (!pcmc->broken_reserved_end) {
1034             res_mem_end += memory_region_size(&machine->device_memory->mr);
1035         }
1036         *val = cpu_to_le64(ROUND_UP(res_mem_end, 1 * GiB));
1037         fw_cfg_add_file(fw_cfg, "etc/reserved-memory-end", val, sizeof(*val));
1038     }
1039 
1040     if (linux_boot) {
1041         x86_load_linux(x86ms, fw_cfg, pcmc->acpi_data_size,
1042                        pcmc->pvh_enabled, pcmc->linuxboot_dma_enabled);
1043     }
1044 
1045     for (i = 0; i < nb_option_roms; i++) {
1046         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1047     }
1048     x86ms->fw_cfg = fw_cfg;
1049 
1050     /* Init default IOAPIC address space */
1051     x86ms->ioapic_as = &address_space_memory;
1052 
1053     /* Init ACPI memory hotplug IO base address */
1054     pcms->memhp_io_base = ACPI_MEMORY_HOTPLUG_BASE;
1055 }
1056 
1057 /*
1058  * The 64bit pci hole starts after "above 4G RAM" and
1059  * potentially the space reserved for memory hotplug.
1060  */
1061 uint64_t pc_pci_hole64_start(void)
1062 {
1063     PCMachineState *pcms = PC_MACHINE(qdev_get_machine());
1064     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1065     MachineState *ms = MACHINE(pcms);
1066     X86MachineState *x86ms = X86_MACHINE(pcms);
1067     uint64_t hole64_start = 0;
1068 
1069     if (pcmc->has_reserved_memory && ms->device_memory->base) {
1070         hole64_start = ms->device_memory->base;
1071         if (!pcmc->broken_reserved_end) {
1072             hole64_start += memory_region_size(&ms->device_memory->mr);
1073         }
1074     } else {
1075         hole64_start = 0x100000000ULL + x86ms->above_4g_mem_size;
1076     }
1077 
1078     return ROUND_UP(hole64_start, 1 * GiB);
1079 }
1080 
1081 DeviceState *pc_vga_init(ISABus *isa_bus, PCIBus *pci_bus)
1082 {
1083     DeviceState *dev = NULL;
1084 
1085     rom_set_order_override(FW_CFG_ORDER_OVERRIDE_VGA);
1086     if (pci_bus) {
1087         PCIDevice *pcidev = pci_vga_init(pci_bus);
1088         dev = pcidev ? &pcidev->qdev : NULL;
1089     } else if (isa_bus) {
1090         ISADevice *isadev = isa_vga_init(isa_bus);
1091         dev = isadev ? DEVICE(isadev) : NULL;
1092     }
1093     rom_reset_order_override();
1094     return dev;
1095 }
1096 
1097 static const MemoryRegionOps ioport80_io_ops = {
1098     .write = ioport80_write,
1099     .read = ioport80_read,
1100     .endianness = DEVICE_NATIVE_ENDIAN,
1101     .impl = {
1102         .min_access_size = 1,
1103         .max_access_size = 1,
1104     },
1105 };
1106 
1107 static const MemoryRegionOps ioportF0_io_ops = {
1108     .write = ioportF0_write,
1109     .read = ioportF0_read,
1110     .endianness = DEVICE_NATIVE_ENDIAN,
1111     .impl = {
1112         .min_access_size = 1,
1113         .max_access_size = 1,
1114     },
1115 };
1116 
1117 static void pc_superio_init(ISABus *isa_bus, bool create_fdctrl, bool no_vmport)
1118 {
1119     int i;
1120     DriveInfo *fd[MAX_FD];
1121     qemu_irq *a20_line;
1122     ISADevice *fdc, *i8042, *port92, *vmmouse;
1123 
1124     serial_hds_isa_init(isa_bus, 0, MAX_ISA_SERIAL_PORTS);
1125     parallel_hds_isa_init(isa_bus, MAX_PARALLEL_PORTS);
1126 
1127     for (i = 0; i < MAX_FD; i++) {
1128         fd[i] = drive_get(IF_FLOPPY, 0, i);
1129         create_fdctrl |= !!fd[i];
1130     }
1131     if (create_fdctrl) {
1132         fdc = isa_new(TYPE_ISA_FDC);
1133         if (fdc) {
1134             isa_realize_and_unref(fdc, isa_bus, &error_fatal);
1135             isa_fdc_init_drives(fdc, fd);
1136         }
1137     }
1138 
1139     i8042 = isa_create_simple(isa_bus, "i8042");
1140     if (!no_vmport) {
1141         isa_create_simple(isa_bus, TYPE_VMPORT);
1142         vmmouse = isa_try_new("vmmouse");
1143     } else {
1144         vmmouse = NULL;
1145     }
1146     if (vmmouse) {
1147         object_property_set_link(OBJECT(vmmouse), "i8042", OBJECT(i8042),
1148                                  &error_abort);
1149         isa_realize_and_unref(vmmouse, isa_bus, &error_fatal);
1150     }
1151     port92 = isa_create_simple(isa_bus, TYPE_PORT92);
1152 
1153     a20_line = qemu_allocate_irqs(handle_a20_line_change, first_cpu, 2);
1154     i8042_setup_a20_line(i8042, a20_line[0]);
1155     qdev_connect_gpio_out_named(DEVICE(port92),
1156                                 PORT92_A20_LINE, 0, a20_line[1]);
1157     g_free(a20_line);
1158 }
1159 
1160 void pc_basic_device_init(struct PCMachineState *pcms,
1161                           ISABus *isa_bus, qemu_irq *gsi,
1162                           ISADevice **rtc_state,
1163                           bool create_fdctrl,
1164                           uint32_t hpet_irqs)
1165 {
1166     int i;
1167     DeviceState *hpet = NULL;
1168     int pit_isa_irq = 0;
1169     qemu_irq pit_alt_irq = NULL;
1170     qemu_irq rtc_irq = NULL;
1171     ISADevice *pit = NULL;
1172     MemoryRegion *ioport80_io = g_new(MemoryRegion, 1);
1173     MemoryRegion *ioportF0_io = g_new(MemoryRegion, 1);
1174 
1175     memory_region_init_io(ioport80_io, NULL, &ioport80_io_ops, NULL, "ioport80", 1);
1176     memory_region_add_subregion(isa_bus->address_space_io, 0x80, ioport80_io);
1177 
1178     memory_region_init_io(ioportF0_io, NULL, &ioportF0_io_ops, NULL, "ioportF0", 1);
1179     memory_region_add_subregion(isa_bus->address_space_io, 0xf0, ioportF0_io);
1180 
1181     /*
1182      * Check if an HPET shall be created.
1183      *
1184      * Without KVM_CAP_PIT_STATE2, we cannot switch off the in-kernel PIT
1185      * when the HPET wants to take over. Thus we have to disable the latter.
1186      */
1187     if (!no_hpet && (!kvm_irqchip_in_kernel() || kvm_has_pit_state2())) {
1188         hpet = qdev_try_new(TYPE_HPET);
1189         if (hpet) {
1190             /* For pc-piix-*, hpet's intcap is always IRQ2. For pc-q35-1.7
1191              * and earlier, use IRQ2 for compat. Otherwise, use IRQ16~23,
1192              * IRQ8 and IRQ2.
1193              */
1194             uint8_t compat = object_property_get_uint(OBJECT(hpet),
1195                     HPET_INTCAP, NULL);
1196             if (!compat) {
1197                 qdev_prop_set_uint32(hpet, HPET_INTCAP, hpet_irqs);
1198             }
1199             sysbus_realize_and_unref(SYS_BUS_DEVICE(hpet), &error_fatal);
1200             sysbus_mmio_map(SYS_BUS_DEVICE(hpet), 0, HPET_BASE);
1201 
1202             for (i = 0; i < GSI_NUM_PINS; i++) {
1203                 sysbus_connect_irq(SYS_BUS_DEVICE(hpet), i, gsi[i]);
1204             }
1205             pit_isa_irq = -1;
1206             pit_alt_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_PIT_INT);
1207             rtc_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_RTC_INT);
1208         }
1209     }
1210     *rtc_state = mc146818_rtc_init(isa_bus, 2000, rtc_irq);
1211 
1212     qemu_register_boot_set(pc_boot_set, *rtc_state);
1213 
1214     if (!xen_enabled() && pcms->pit_enabled) {
1215         if (kvm_pit_in_kernel()) {
1216             pit = kvm_pit_init(isa_bus, 0x40);
1217         } else {
1218             pit = i8254_pit_init(isa_bus, 0x40, pit_isa_irq, pit_alt_irq);
1219         }
1220         if (hpet) {
1221             /* connect PIT to output control line of the HPET */
1222             qdev_connect_gpio_out(hpet, 0, qdev_get_gpio_in(DEVICE(pit), 0));
1223         }
1224         pcspk_init(pcms->pcspk, isa_bus, pit);
1225     }
1226 
1227     i8257_dma_init(isa_bus, 0);
1228 
1229     /* Super I/O */
1230     pc_superio_init(isa_bus, create_fdctrl, pcms->vmport != ON_OFF_AUTO_ON);
1231 }
1232 
1233 void pc_nic_init(PCMachineClass *pcmc, ISABus *isa_bus, PCIBus *pci_bus)
1234 {
1235     int i;
1236 
1237     rom_set_order_override(FW_CFG_ORDER_OVERRIDE_NIC);
1238     for (i = 0; i < nb_nics; i++) {
1239         NICInfo *nd = &nd_table[i];
1240         const char *model = nd->model ? nd->model : pcmc->default_nic_model;
1241 
1242         if (g_str_equal(model, "ne2k_isa")) {
1243             pc_init_ne2k_isa(isa_bus, nd);
1244         } else {
1245             pci_nic_init_nofail(nd, pci_bus, model, NULL);
1246         }
1247     }
1248     rom_reset_order_override();
1249 }
1250 
1251 void pc_i8259_create(ISABus *isa_bus, qemu_irq *i8259_irqs)
1252 {
1253     qemu_irq *i8259;
1254 
1255     if (kvm_pic_in_kernel()) {
1256         i8259 = kvm_i8259_init(isa_bus);
1257     } else if (xen_enabled()) {
1258         i8259 = xen_interrupt_controller_init();
1259     } else {
1260         i8259 = i8259_init(isa_bus, x86_allocate_cpu_irq());
1261     }
1262 
1263     for (size_t i = 0; i < ISA_NUM_IRQS; i++) {
1264         i8259_irqs[i] = i8259[i];
1265     }
1266 
1267     g_free(i8259);
1268 }
1269 
1270 static void pc_memory_pre_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
1271                                Error **errp)
1272 {
1273     const PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1274     const PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1275     const MachineState *ms = MACHINE(hotplug_dev);
1276     const bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
1277     const uint64_t legacy_align = TARGET_PAGE_SIZE;
1278     Error *local_err = NULL;
1279 
1280     /*
1281      * When -no-acpi is used with Q35 machine type, no ACPI is built,
1282      * but pcms->acpi_dev is still created. Check !acpi_enabled in
1283      * addition to cover this case.
1284      */
1285     if (!pcms->acpi_dev || !x86_machine_is_acpi_enabled(X86_MACHINE(pcms))) {
1286         error_setg(errp,
1287                    "memory hotplug is not enabled: missing acpi device or acpi disabled");
1288         return;
1289     }
1290 
1291     if (is_nvdimm && !ms->nvdimms_state->is_enabled) {
1292         error_setg(errp, "nvdimm is not enabled: missing 'nvdimm' in '-M'");
1293         return;
1294     }
1295 
1296     hotplug_handler_pre_plug(pcms->acpi_dev, dev, &local_err);
1297     if (local_err) {
1298         error_propagate(errp, local_err);
1299         return;
1300     }
1301 
1302     pc_dimm_pre_plug(PC_DIMM(dev), MACHINE(hotplug_dev),
1303                      pcmc->enforce_aligned_dimm ? NULL : &legacy_align, errp);
1304 }
1305 
1306 static void pc_memory_plug(HotplugHandler *hotplug_dev,
1307                            DeviceState *dev, Error **errp)
1308 {
1309     Error *local_err = NULL;
1310     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1311     MachineState *ms = MACHINE(hotplug_dev);
1312     bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
1313 
1314     pc_dimm_plug(PC_DIMM(dev), MACHINE(pcms), &local_err);
1315     if (local_err) {
1316         goto out;
1317     }
1318 
1319     if (is_nvdimm) {
1320         nvdimm_plug(ms->nvdimms_state);
1321     }
1322 
1323     hotplug_handler_plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &error_abort);
1324 out:
1325     error_propagate(errp, local_err);
1326 }
1327 
1328 static void pc_memory_unplug_request(HotplugHandler *hotplug_dev,
1329                                      DeviceState *dev, Error **errp)
1330 {
1331     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1332 
1333     /*
1334      * When -no-acpi is used with Q35 machine type, no ACPI is built,
1335      * but pcms->acpi_dev is still created. Check !acpi_enabled in
1336      * addition to cover this case.
1337      */
1338     if (!pcms->acpi_dev || !x86_machine_is_acpi_enabled(X86_MACHINE(pcms))) {
1339         error_setg(errp,
1340                    "memory hotplug is not enabled: missing acpi device or acpi disabled");
1341         return;
1342     }
1343 
1344     if (object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM)) {
1345         error_setg(errp, "nvdimm device hot unplug is not supported yet.");
1346         return;
1347     }
1348 
1349     hotplug_handler_unplug_request(HOTPLUG_HANDLER(pcms->acpi_dev), dev,
1350                                    errp);
1351 }
1352 
1353 static void pc_memory_unplug(HotplugHandler *hotplug_dev,
1354                              DeviceState *dev, Error **errp)
1355 {
1356     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1357     Error *local_err = NULL;
1358 
1359     hotplug_handler_unplug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
1360     if (local_err) {
1361         goto out;
1362     }
1363 
1364     pc_dimm_unplug(PC_DIMM(dev), MACHINE(pcms));
1365     qdev_unrealize(dev);
1366  out:
1367     error_propagate(errp, local_err);
1368 }
1369 
1370 static int pc_apic_cmp(const void *a, const void *b)
1371 {
1372    CPUArchId *apic_a = (CPUArchId *)a;
1373    CPUArchId *apic_b = (CPUArchId *)b;
1374 
1375    return apic_a->arch_id - apic_b->arch_id;
1376 }
1377 
1378 /* returns pointer to CPUArchId descriptor that matches CPU's apic_id
1379  * in ms->possible_cpus->cpus, if ms->possible_cpus->cpus has no
1380  * entry corresponding to CPU's apic_id returns NULL.
1381  */
1382 static CPUArchId *pc_find_cpu_slot(MachineState *ms, uint32_t id, int *idx)
1383 {
1384     CPUArchId apic_id, *found_cpu;
1385 
1386     apic_id.arch_id = id;
1387     found_cpu = bsearch(&apic_id, ms->possible_cpus->cpus,
1388         ms->possible_cpus->len, sizeof(*ms->possible_cpus->cpus),
1389         pc_apic_cmp);
1390     if (found_cpu && idx) {
1391         *idx = found_cpu - ms->possible_cpus->cpus;
1392     }
1393     return found_cpu;
1394 }
1395 
1396 static void pc_cpu_plug(HotplugHandler *hotplug_dev,
1397                         DeviceState *dev, Error **errp)
1398 {
1399     CPUArchId *found_cpu;
1400     Error *local_err = NULL;
1401     X86CPU *cpu = X86_CPU(dev);
1402     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1403     X86MachineState *x86ms = X86_MACHINE(pcms);
1404 
1405     if (pcms->acpi_dev) {
1406         hotplug_handler_plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
1407         if (local_err) {
1408             goto out;
1409         }
1410     }
1411 
1412     /* increment the number of CPUs */
1413     x86ms->boot_cpus++;
1414     if (x86ms->rtc) {
1415         rtc_set_cpus_count(x86ms->rtc, x86ms->boot_cpus);
1416     }
1417     if (x86ms->fw_cfg) {
1418         fw_cfg_modify_i16(x86ms->fw_cfg, FW_CFG_NB_CPUS, x86ms->boot_cpus);
1419     }
1420 
1421     found_cpu = pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, NULL);
1422     found_cpu->cpu = OBJECT(dev);
1423 out:
1424     error_propagate(errp, local_err);
1425 }
1426 static void pc_cpu_unplug_request_cb(HotplugHandler *hotplug_dev,
1427                                      DeviceState *dev, Error **errp)
1428 {
1429     int idx = -1;
1430     X86CPU *cpu = X86_CPU(dev);
1431     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1432 
1433     if (!pcms->acpi_dev) {
1434         error_setg(errp, "CPU hot unplug not supported without ACPI");
1435         return;
1436     }
1437 
1438     pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, &idx);
1439     assert(idx != -1);
1440     if (idx == 0) {
1441         error_setg(errp, "Boot CPU is unpluggable");
1442         return;
1443     }
1444 
1445     hotplug_handler_unplug_request(HOTPLUG_HANDLER(pcms->acpi_dev), dev,
1446                                    errp);
1447 }
1448 
1449 static void pc_cpu_unplug_cb(HotplugHandler *hotplug_dev,
1450                              DeviceState *dev, Error **errp)
1451 {
1452     CPUArchId *found_cpu;
1453     Error *local_err = NULL;
1454     X86CPU *cpu = X86_CPU(dev);
1455     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1456     X86MachineState *x86ms = X86_MACHINE(pcms);
1457 
1458     hotplug_handler_unplug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
1459     if (local_err) {
1460         goto out;
1461     }
1462 
1463     found_cpu = pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, NULL);
1464     found_cpu->cpu = NULL;
1465     qdev_unrealize(dev);
1466 
1467     /* decrement the number of CPUs */
1468     x86ms->boot_cpus--;
1469     /* Update the number of CPUs in CMOS */
1470     rtc_set_cpus_count(x86ms->rtc, x86ms->boot_cpus);
1471     fw_cfg_modify_i16(x86ms->fw_cfg, FW_CFG_NB_CPUS, x86ms->boot_cpus);
1472  out:
1473     error_propagate(errp, local_err);
1474 }
1475 
1476 static void pc_cpu_pre_plug(HotplugHandler *hotplug_dev,
1477                             DeviceState *dev, Error **errp)
1478 {
1479     int idx;
1480     CPUState *cs;
1481     CPUArchId *cpu_slot;
1482     X86CPUTopoIDs topo_ids;
1483     X86CPU *cpu = X86_CPU(dev);
1484     CPUX86State *env = &cpu->env;
1485     MachineState *ms = MACHINE(hotplug_dev);
1486     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1487     X86MachineState *x86ms = X86_MACHINE(pcms);
1488     unsigned int smp_cores = ms->smp.cores;
1489     unsigned int smp_threads = ms->smp.threads;
1490     X86CPUTopoInfo topo_info;
1491 
1492     if(!object_dynamic_cast(OBJECT(cpu), ms->cpu_type)) {
1493         error_setg(errp, "Invalid CPU type, expected cpu type: '%s'",
1494                    ms->cpu_type);
1495         return;
1496     }
1497 
1498     init_topo_info(&topo_info, x86ms);
1499 
1500     env->nr_dies = x86ms->smp_dies;
1501     env->nr_nodes = topo_info.nodes_per_pkg;
1502     env->pkg_offset = x86ms->apicid_pkg_offset(&topo_info);
1503 
1504     /*
1505      * If APIC ID is not set,
1506      * set it based on socket/die/core/thread properties.
1507      */
1508     if (cpu->apic_id == UNASSIGNED_APIC_ID) {
1509         int max_socket = (ms->smp.max_cpus - 1) /
1510                                 smp_threads / smp_cores / x86ms->smp_dies;
1511 
1512         /*
1513          * die-id was optional in QEMU 4.0 and older, so keep it optional
1514          * if there's only one die per socket.
1515          */
1516         if (cpu->die_id < 0 && x86ms->smp_dies == 1) {
1517             cpu->die_id = 0;
1518         }
1519 
1520         if (cpu->socket_id < 0) {
1521             error_setg(errp, "CPU socket-id is not set");
1522             return;
1523         } else if (cpu->socket_id > max_socket) {
1524             error_setg(errp, "Invalid CPU socket-id: %u must be in range 0:%u",
1525                        cpu->socket_id, max_socket);
1526             return;
1527         }
1528         if (cpu->die_id < 0) {
1529             error_setg(errp, "CPU die-id is not set");
1530             return;
1531         } else if (cpu->die_id > x86ms->smp_dies - 1) {
1532             error_setg(errp, "Invalid CPU die-id: %u must be in range 0:%u",
1533                        cpu->die_id, x86ms->smp_dies - 1);
1534             return;
1535         }
1536         if (cpu->core_id < 0) {
1537             error_setg(errp, "CPU core-id is not set");
1538             return;
1539         } else if (cpu->core_id > (smp_cores - 1)) {
1540             error_setg(errp, "Invalid CPU core-id: %u must be in range 0:%u",
1541                        cpu->core_id, smp_cores - 1);
1542             return;
1543         }
1544         if (cpu->thread_id < 0) {
1545             error_setg(errp, "CPU thread-id is not set");
1546             return;
1547         } else if (cpu->thread_id > (smp_threads - 1)) {
1548             error_setg(errp, "Invalid CPU thread-id: %u must be in range 0:%u",
1549                        cpu->thread_id, smp_threads - 1);
1550             return;
1551         }
1552 
1553         topo_ids.pkg_id = cpu->socket_id;
1554         topo_ids.die_id = cpu->die_id;
1555         topo_ids.core_id = cpu->core_id;
1556         topo_ids.smt_id = cpu->thread_id;
1557         cpu->apic_id = x86ms->apicid_from_topo_ids(&topo_info, &topo_ids);
1558     }
1559 
1560     cpu_slot = pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, &idx);
1561     if (!cpu_slot) {
1562         MachineState *ms = MACHINE(pcms);
1563 
1564         x86ms->topo_ids_from_apicid(cpu->apic_id, &topo_info, &topo_ids);
1565         error_setg(errp,
1566             "Invalid CPU [socket: %u, die: %u, core: %u, thread: %u] with"
1567             " APIC ID %" PRIu32 ", valid index range 0:%d",
1568             topo_ids.pkg_id, topo_ids.die_id, topo_ids.core_id, topo_ids.smt_id,
1569             cpu->apic_id, ms->possible_cpus->len - 1);
1570         return;
1571     }
1572 
1573     if (cpu_slot->cpu) {
1574         error_setg(errp, "CPU[%d] with APIC ID %" PRIu32 " exists",
1575                    idx, cpu->apic_id);
1576         return;
1577     }
1578 
1579     /* if 'address' properties socket-id/core-id/thread-id are not set, set them
1580      * so that machine_query_hotpluggable_cpus would show correct values
1581      */
1582     /* TODO: move socket_id/core_id/thread_id checks into x86_cpu_realizefn()
1583      * once -smp refactoring is complete and there will be CPU private
1584      * CPUState::nr_cores and CPUState::nr_threads fields instead of globals */
1585     x86ms->topo_ids_from_apicid(cpu->apic_id, &topo_info, &topo_ids);
1586     if (cpu->socket_id != -1 && cpu->socket_id != topo_ids.pkg_id) {
1587         error_setg(errp, "property socket-id: %u doesn't match set apic-id:"
1588             " 0x%x (socket-id: %u)", cpu->socket_id, cpu->apic_id,
1589             topo_ids.pkg_id);
1590         return;
1591     }
1592     cpu->socket_id = topo_ids.pkg_id;
1593 
1594     if (cpu->die_id != -1 && cpu->die_id != topo_ids.die_id) {
1595         error_setg(errp, "property die-id: %u doesn't match set apic-id:"
1596             " 0x%x (die-id: %u)", cpu->die_id, cpu->apic_id, topo_ids.die_id);
1597         return;
1598     }
1599     cpu->die_id = topo_ids.die_id;
1600 
1601     if (cpu->core_id != -1 && cpu->core_id != topo_ids.core_id) {
1602         error_setg(errp, "property core-id: %u doesn't match set apic-id:"
1603             " 0x%x (core-id: %u)", cpu->core_id, cpu->apic_id,
1604             topo_ids.core_id);
1605         return;
1606     }
1607     cpu->core_id = topo_ids.core_id;
1608 
1609     if (cpu->thread_id != -1 && cpu->thread_id != topo_ids.smt_id) {
1610         error_setg(errp, "property thread-id: %u doesn't match set apic-id:"
1611             " 0x%x (thread-id: %u)", cpu->thread_id, cpu->apic_id,
1612             topo_ids.smt_id);
1613         return;
1614     }
1615     cpu->thread_id = topo_ids.smt_id;
1616 
1617     if (hyperv_feat_enabled(cpu, HYPERV_FEAT_VPINDEX) &&
1618         !kvm_hv_vpindex_settable()) {
1619         error_setg(errp, "kernel doesn't allow setting HyperV VP_INDEX");
1620         return;
1621     }
1622 
1623     cs = CPU(cpu);
1624     cs->cpu_index = idx;
1625 
1626     numa_cpu_pre_plug(cpu_slot, dev, errp);
1627 }
1628 
1629 static void pc_virtio_md_pci_pre_plug(HotplugHandler *hotplug_dev,
1630                                       DeviceState *dev, Error **errp)
1631 {
1632     HotplugHandler *hotplug_dev2 = qdev_get_bus_hotplug_handler(dev);
1633     Error *local_err = NULL;
1634 
1635     if (!hotplug_dev2 && dev->hotplugged) {
1636         /*
1637          * Without a bus hotplug handler, we cannot control the plug/unplug
1638          * order. We should never reach this point when hotplugging on x86,
1639          * however, better add a safety net.
1640          */
1641         error_setg(errp, "hotplug of virtio based memory devices not supported"
1642                    " on this bus.");
1643         return;
1644     }
1645     /*
1646      * First, see if we can plug this memory device at all. If that
1647      * succeeds, branch of to the actual hotplug handler.
1648      */
1649     memory_device_pre_plug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev), NULL,
1650                            &local_err);
1651     if (!local_err && hotplug_dev2) {
1652         hotplug_handler_pre_plug(hotplug_dev2, dev, &local_err);
1653     }
1654     error_propagate(errp, local_err);
1655 }
1656 
1657 static void pc_virtio_md_pci_plug(HotplugHandler *hotplug_dev,
1658                                   DeviceState *dev, Error **errp)
1659 {
1660     HotplugHandler *hotplug_dev2 = qdev_get_bus_hotplug_handler(dev);
1661     Error *local_err = NULL;
1662 
1663     /*
1664      * Plug the memory device first and then branch off to the actual
1665      * hotplug handler. If that one fails, we can easily undo the memory
1666      * device bits.
1667      */
1668     memory_device_plug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev));
1669     if (hotplug_dev2) {
1670         hotplug_handler_plug(hotplug_dev2, dev, &local_err);
1671         if (local_err) {
1672             memory_device_unplug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev));
1673         }
1674     }
1675     error_propagate(errp, local_err);
1676 }
1677 
1678 static void pc_virtio_md_pci_unplug_request(HotplugHandler *hotplug_dev,
1679                                             DeviceState *dev, Error **errp)
1680 {
1681     /* We don't support hot unplug of virtio based memory devices */
1682     error_setg(errp, "virtio based memory devices cannot be unplugged.");
1683 }
1684 
1685 static void pc_virtio_md_pci_unplug(HotplugHandler *hotplug_dev,
1686                                     DeviceState *dev, Error **errp)
1687 {
1688     /* We don't support hot unplug of virtio based memory devices */
1689 }
1690 
1691 static void pc_machine_device_pre_plug_cb(HotplugHandler *hotplug_dev,
1692                                           DeviceState *dev, Error **errp)
1693 {
1694     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1695         pc_memory_pre_plug(hotplug_dev, dev, errp);
1696     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1697         pc_cpu_pre_plug(hotplug_dev, dev, errp);
1698     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI) ||
1699                object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI)) {
1700         pc_virtio_md_pci_pre_plug(hotplug_dev, dev, errp);
1701     }
1702 }
1703 
1704 static void pc_machine_device_plug_cb(HotplugHandler *hotplug_dev,
1705                                       DeviceState *dev, Error **errp)
1706 {
1707     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1708         pc_memory_plug(hotplug_dev, dev, errp);
1709     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1710         pc_cpu_plug(hotplug_dev, dev, errp);
1711     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI) ||
1712                object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI)) {
1713         pc_virtio_md_pci_plug(hotplug_dev, dev, errp);
1714     }
1715 }
1716 
1717 static void pc_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
1718                                                 DeviceState *dev, Error **errp)
1719 {
1720     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1721         pc_memory_unplug_request(hotplug_dev, dev, errp);
1722     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1723         pc_cpu_unplug_request_cb(hotplug_dev, dev, errp);
1724     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI) ||
1725                object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI)) {
1726         pc_virtio_md_pci_unplug_request(hotplug_dev, dev, errp);
1727     } else {
1728         error_setg(errp, "acpi: device unplug request for not supported device"
1729                    " type: %s", object_get_typename(OBJECT(dev)));
1730     }
1731 }
1732 
1733 static void pc_machine_device_unplug_cb(HotplugHandler *hotplug_dev,
1734                                         DeviceState *dev, Error **errp)
1735 {
1736     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1737         pc_memory_unplug(hotplug_dev, dev, errp);
1738     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1739         pc_cpu_unplug_cb(hotplug_dev, dev, errp);
1740     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI) ||
1741                object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI)) {
1742         pc_virtio_md_pci_unplug(hotplug_dev, dev, errp);
1743     } else {
1744         error_setg(errp, "acpi: device unplug for not supported device"
1745                    " type: %s", object_get_typename(OBJECT(dev)));
1746     }
1747 }
1748 
1749 static HotplugHandler *pc_get_hotplug_handler(MachineState *machine,
1750                                              DeviceState *dev)
1751 {
1752     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) ||
1753         object_dynamic_cast(OBJECT(dev), TYPE_CPU) ||
1754         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI) ||
1755         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI)) {
1756         return HOTPLUG_HANDLER(machine);
1757     }
1758 
1759     return NULL;
1760 }
1761 
1762 static void
1763 pc_machine_get_device_memory_region_size(Object *obj, Visitor *v,
1764                                          const char *name, void *opaque,
1765                                          Error **errp)
1766 {
1767     MachineState *ms = MACHINE(obj);
1768     int64_t value = 0;
1769 
1770     if (ms->device_memory) {
1771         value = memory_region_size(&ms->device_memory->mr);
1772     }
1773 
1774     visit_type_int(v, name, &value, errp);
1775 }
1776 
1777 static void pc_machine_get_vmport(Object *obj, Visitor *v, const char *name,
1778                                   void *opaque, Error **errp)
1779 {
1780     PCMachineState *pcms = PC_MACHINE(obj);
1781     OnOffAuto vmport = pcms->vmport;
1782 
1783     visit_type_OnOffAuto(v, name, &vmport, errp);
1784 }
1785 
1786 static void pc_machine_set_vmport(Object *obj, Visitor *v, const char *name,
1787                                   void *opaque, Error **errp)
1788 {
1789     PCMachineState *pcms = PC_MACHINE(obj);
1790 
1791     visit_type_OnOffAuto(v, name, &pcms->vmport, errp);
1792 }
1793 
1794 static bool pc_machine_get_smbus(Object *obj, Error **errp)
1795 {
1796     PCMachineState *pcms = PC_MACHINE(obj);
1797 
1798     return pcms->smbus_enabled;
1799 }
1800 
1801 static void pc_machine_set_smbus(Object *obj, bool value, Error **errp)
1802 {
1803     PCMachineState *pcms = PC_MACHINE(obj);
1804 
1805     pcms->smbus_enabled = value;
1806 }
1807 
1808 static bool pc_machine_get_sata(Object *obj, Error **errp)
1809 {
1810     PCMachineState *pcms = PC_MACHINE(obj);
1811 
1812     return pcms->sata_enabled;
1813 }
1814 
1815 static void pc_machine_set_sata(Object *obj, bool value, Error **errp)
1816 {
1817     PCMachineState *pcms = PC_MACHINE(obj);
1818 
1819     pcms->sata_enabled = value;
1820 }
1821 
1822 static bool pc_machine_get_pit(Object *obj, Error **errp)
1823 {
1824     PCMachineState *pcms = PC_MACHINE(obj);
1825 
1826     return pcms->pit_enabled;
1827 }
1828 
1829 static void pc_machine_set_pit(Object *obj, bool value, Error **errp)
1830 {
1831     PCMachineState *pcms = PC_MACHINE(obj);
1832 
1833     pcms->pit_enabled = value;
1834 }
1835 
1836 static void pc_machine_get_max_ram_below_4g(Object *obj, Visitor *v,
1837                                             const char *name, void *opaque,
1838                                             Error **errp)
1839 {
1840     PCMachineState *pcms = PC_MACHINE(obj);
1841     uint64_t value = pcms->max_ram_below_4g;
1842 
1843     visit_type_size(v, name, &value, errp);
1844 }
1845 
1846 static void pc_machine_set_max_ram_below_4g(Object *obj, Visitor *v,
1847                                             const char *name, void *opaque,
1848                                             Error **errp)
1849 {
1850     PCMachineState *pcms = PC_MACHINE(obj);
1851     uint64_t value;
1852 
1853     if (!visit_type_size(v, name, &value, errp)) {
1854         return;
1855     }
1856     if (value > 4 * GiB) {
1857         error_setg(errp,
1858                    "Machine option 'max-ram-below-4g=%"PRIu64
1859                    "' expects size less than or equal to 4G", value);
1860         return;
1861     }
1862 
1863     if (value < 1 * MiB) {
1864         warn_report("Only %" PRIu64 " bytes of RAM below the 4GiB boundary,"
1865                     "BIOS may not work with less than 1MiB", value);
1866     }
1867 
1868     pcms->max_ram_below_4g = value;
1869 }
1870 
1871 static void pc_machine_initfn(Object *obj)
1872 {
1873     PCMachineState *pcms = PC_MACHINE(obj);
1874 
1875 #ifdef CONFIG_VMPORT
1876     pcms->vmport = ON_OFF_AUTO_AUTO;
1877 #else
1878     pcms->vmport = ON_OFF_AUTO_OFF;
1879 #endif /* CONFIG_VMPORT */
1880     pcms->max_ram_below_4g = 0; /* use default */
1881     /* acpi build is enabled by default if machine supports it */
1882     pcms->acpi_build_enabled = PC_MACHINE_GET_CLASS(pcms)->has_acpi_build;
1883     pcms->smbus_enabled = true;
1884     pcms->sata_enabled = true;
1885     pcms->pit_enabled = true;
1886 
1887     pc_system_flash_create(pcms);
1888     pcms->pcspk = isa_new(TYPE_PC_SPEAKER);
1889     object_property_add_alias(OBJECT(pcms), "pcspk-audiodev",
1890                               OBJECT(pcms->pcspk), "audiodev");
1891 }
1892 
1893 static void pc_machine_reset(MachineState *machine)
1894 {
1895     CPUState *cs;
1896     X86CPU *cpu;
1897 
1898     qemu_devices_reset();
1899 
1900     /* Reset APIC after devices have been reset to cancel
1901      * any changes that qemu_devices_reset() might have done.
1902      */
1903     CPU_FOREACH(cs) {
1904         cpu = X86_CPU(cs);
1905 
1906         if (cpu->apic_state) {
1907             device_legacy_reset(cpu->apic_state);
1908         }
1909     }
1910 }
1911 
1912 static void pc_machine_wakeup(MachineState *machine)
1913 {
1914     cpu_synchronize_all_states();
1915     pc_machine_reset(machine);
1916     cpu_synchronize_all_post_reset();
1917 }
1918 
1919 static bool pc_hotplug_allowed(MachineState *ms, DeviceState *dev, Error **errp)
1920 {
1921     X86IOMMUState *iommu = x86_iommu_get_default();
1922     IntelIOMMUState *intel_iommu;
1923 
1924     if (iommu &&
1925         object_dynamic_cast((Object *)iommu, TYPE_INTEL_IOMMU_DEVICE) &&
1926         object_dynamic_cast((Object *)dev, "vfio-pci")) {
1927         intel_iommu = INTEL_IOMMU_DEVICE(iommu);
1928         if (!intel_iommu->caching_mode) {
1929             error_setg(errp, "Device assignment is not allowed without "
1930                        "enabling caching-mode=on for Intel IOMMU.");
1931             return false;
1932         }
1933     }
1934 
1935     return true;
1936 }
1937 
1938 static void pc_machine_class_init(ObjectClass *oc, void *data)
1939 {
1940     MachineClass *mc = MACHINE_CLASS(oc);
1941     PCMachineClass *pcmc = PC_MACHINE_CLASS(oc);
1942     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
1943 
1944     pcmc->pci_enabled = true;
1945     pcmc->has_acpi_build = true;
1946     pcmc->rsdp_in_ram = true;
1947     pcmc->smbios_defaults = true;
1948     pcmc->smbios_uuid_encoded = true;
1949     pcmc->gigabyte_align = true;
1950     pcmc->has_reserved_memory = true;
1951     pcmc->kvmclock_enabled = true;
1952     pcmc->enforce_aligned_dimm = true;
1953     /* BIOS ACPI tables: 128K. Other BIOS datastructures: less than 4K reported
1954      * to be used at the moment, 32K should be enough for a while.  */
1955     pcmc->acpi_data_size = 0x20000 + 0x8000;
1956     pcmc->linuxboot_dma_enabled = true;
1957     pcmc->pvh_enabled = true;
1958     assert(!mc->get_hotplug_handler);
1959     mc->get_hotplug_handler = pc_get_hotplug_handler;
1960     mc->hotplug_allowed = pc_hotplug_allowed;
1961     mc->cpu_index_to_instance_props = x86_cpu_index_to_props;
1962     mc->get_default_cpu_node_id = x86_get_default_cpu_node_id;
1963     mc->possible_cpu_arch_ids = x86_possible_cpu_arch_ids;
1964     mc->auto_enable_numa_with_memhp = true;
1965     mc->auto_enable_numa_with_memdev = true;
1966     mc->has_hotpluggable_cpus = true;
1967     mc->default_boot_order = "cad";
1968     mc->hot_add_cpu = pc_hot_add_cpu;
1969     mc->smp_parse = pc_smp_parse;
1970     mc->block_default_type = IF_IDE;
1971     mc->max_cpus = 255;
1972     mc->reset = pc_machine_reset;
1973     mc->wakeup = pc_machine_wakeup;
1974     hc->pre_plug = pc_machine_device_pre_plug_cb;
1975     hc->plug = pc_machine_device_plug_cb;
1976     hc->unplug_request = pc_machine_device_unplug_request_cb;
1977     hc->unplug = pc_machine_device_unplug_cb;
1978     mc->default_cpu_type = TARGET_DEFAULT_CPU_TYPE;
1979     mc->nvdimm_supported = true;
1980     mc->default_ram_id = "pc.ram";
1981 
1982     object_class_property_add(oc, PC_MACHINE_MAX_RAM_BELOW_4G, "size",
1983         pc_machine_get_max_ram_below_4g, pc_machine_set_max_ram_below_4g,
1984         NULL, NULL);
1985     object_class_property_set_description(oc, PC_MACHINE_MAX_RAM_BELOW_4G,
1986         "Maximum ram below the 4G boundary (32bit boundary)");
1987 
1988     object_class_property_add(oc, PC_MACHINE_DEVMEM_REGION_SIZE, "int",
1989         pc_machine_get_device_memory_region_size, NULL,
1990         NULL, NULL);
1991 
1992     object_class_property_add(oc, PC_MACHINE_VMPORT, "OnOffAuto",
1993         pc_machine_get_vmport, pc_machine_set_vmport,
1994         NULL, NULL);
1995     object_class_property_set_description(oc, PC_MACHINE_VMPORT,
1996         "Enable vmport (pc & q35)");
1997 
1998     object_class_property_add_bool(oc, PC_MACHINE_SMBUS,
1999         pc_machine_get_smbus, pc_machine_set_smbus);
2000 
2001     object_class_property_add_bool(oc, PC_MACHINE_SATA,
2002         pc_machine_get_sata, pc_machine_set_sata);
2003 
2004     object_class_property_add_bool(oc, PC_MACHINE_PIT,
2005         pc_machine_get_pit, pc_machine_set_pit);
2006 }
2007 
2008 static const TypeInfo pc_machine_info = {
2009     .name = TYPE_PC_MACHINE,
2010     .parent = TYPE_X86_MACHINE,
2011     .abstract = true,
2012     .instance_size = sizeof(PCMachineState),
2013     .instance_init = pc_machine_initfn,
2014     .class_size = sizeof(PCMachineClass),
2015     .class_init = pc_machine_class_init,
2016     .interfaces = (InterfaceInfo[]) {
2017          { TYPE_HOTPLUG_HANDLER },
2018          { }
2019     },
2020 };
2021 
2022 static void pc_machine_register_types(void)
2023 {
2024     type_register_static(&pc_machine_info);
2025 }
2026 
2027 type_init(pc_machine_register_types)
2028