xref: /qemu/hw/i386/pc.c (revision a1f728ecc90cf6c6db24df53cc951713fca8b94d)
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 "exec/target_page.h"
28 #include "hw/i386/pc.h"
29 #include "hw/char/serial-isa.h"
30 #include "hw/char/parallel.h"
31 #include "hw/hyperv/hv-balloon.h"
32 #include "hw/i386/fw_cfg.h"
33 #include "hw/i386/vmport.h"
34 #include "system/cpus.h"
35 #include "hw/ide/ide-bus.h"
36 #include "hw/timer/hpet.h"
37 #include "hw/loader.h"
38 #include "hw/rtc/mc146818rtc.h"
39 #include "hw/intc/i8259.h"
40 #include "hw/timer/i8254.h"
41 #include "hw/input/i8042.h"
42 #include "hw/audio/pcspk.h"
43 #include "system/system.h"
44 #include "system/xen.h"
45 #include "system/reset.h"
46 #include "kvm/kvm_i386.h"
47 #include "hw/xen/xen.h"
48 #include "qobject/qlist.h"
49 #include "qemu/error-report.h"
50 #include "hw/acpi/cpu_hotplug.h"
51 #include "acpi-build.h"
52 #include "hw/mem/nvdimm.h"
53 #include "hw/cxl/cxl_host.h"
54 #include "hw/usb.h"
55 #include "hw/i386/intel_iommu.h"
56 #include "hw/net/ne2000-isa.h"
57 #include "hw/virtio/virtio-iommu.h"
58 #include "hw/virtio/virtio-md-pci.h"
59 #include "hw/i386/kvm/xen_overlay.h"
60 #include "hw/i386/kvm/xen_evtchn.h"
61 #include "hw/i386/kvm/xen_gnttab.h"
62 #include "hw/i386/kvm/xen_xenstore.h"
63 #include "hw/mem/memory-device.h"
64 #include "e820_memory_layout.h"
65 #include "trace.h"
66 #include "sev.h"
67 #include CONFIG_DEVICES
68 
69 #ifdef CONFIG_XEN_EMU
70 #include "hw/xen/xen-legacy-backend.h"
71 #include "hw/xen/xen-bus.h"
72 #endif
73 
74 /*
75  * Helper for setting model-id for CPU models that changed model-id
76  * depending on QEMU versions up to QEMU 2.4.
77  */
78 #define PC_CPU_MODEL_IDS(v) \
79     { "qemu32-" TYPE_X86_CPU, "model-id", "QEMU Virtual CPU version " v, },\
80     { "qemu64-" TYPE_X86_CPU, "model-id", "QEMU Virtual CPU version " v, },\
81     { "athlon-" TYPE_X86_CPU, "model-id", "QEMU Virtual CPU version " v, },
82 
83 GlobalProperty pc_compat_10_0[] = {};
84 const size_t pc_compat_10_0_len = G_N_ELEMENTS(pc_compat_10_0);
85 
86 GlobalProperty pc_compat_9_2[] = {};
87 const size_t pc_compat_9_2_len = G_N_ELEMENTS(pc_compat_9_2);
88 
89 GlobalProperty pc_compat_9_1[] = {
90     { "ICH9-LPC", "x-smi-swsmi-timer", "off" },
91     { "ICH9-LPC", "x-smi-periodic-timer", "off" },
92     { TYPE_INTEL_IOMMU_DEVICE, "stale-tm", "on" },
93     { TYPE_INTEL_IOMMU_DEVICE, "aw-bits", "39" },
94 };
95 const size_t pc_compat_9_1_len = G_N_ELEMENTS(pc_compat_9_1);
96 
97 GlobalProperty pc_compat_9_0[] = {
98     { TYPE_X86_CPU, "x-amd-topoext-features-only", "false" },
99     { TYPE_X86_CPU, "x-l1-cache-per-thread", "false" },
100     { TYPE_X86_CPU, "guest-phys-bits", "0" },
101     { "sev-guest", "legacy-vm-type", "on" },
102     { TYPE_X86_CPU, "legacy-multi-node", "on" },
103 };
104 const size_t pc_compat_9_0_len = G_N_ELEMENTS(pc_compat_9_0);
105 
106 GlobalProperty pc_compat_8_2[] = {};
107 const size_t pc_compat_8_2_len = G_N_ELEMENTS(pc_compat_8_2);
108 
109 GlobalProperty pc_compat_8_1[] = {};
110 const size_t pc_compat_8_1_len = G_N_ELEMENTS(pc_compat_8_1);
111 
112 GlobalProperty pc_compat_8_0[] = {
113     { "virtio-mem", "unplugged-inaccessible", "auto" },
114 };
115 const size_t pc_compat_8_0_len = G_N_ELEMENTS(pc_compat_8_0);
116 
117 GlobalProperty pc_compat_7_2[] = {
118     { "ICH9-LPC", "noreboot", "true" },
119 };
120 const size_t pc_compat_7_2_len = G_N_ELEMENTS(pc_compat_7_2);
121 
122 GlobalProperty pc_compat_7_1[] = {};
123 const size_t pc_compat_7_1_len = G_N_ELEMENTS(pc_compat_7_1);
124 
125 GlobalProperty pc_compat_7_0[] = {};
126 const size_t pc_compat_7_0_len = G_N_ELEMENTS(pc_compat_7_0);
127 
128 GlobalProperty pc_compat_6_2[] = {
129     { "virtio-mem", "unplugged-inaccessible", "off" },
130 };
131 const size_t pc_compat_6_2_len = G_N_ELEMENTS(pc_compat_6_2);
132 
133 GlobalProperty pc_compat_6_1[] = {
134     { TYPE_X86_CPU, "hv-version-id-build", "0x1bbc" },
135     { TYPE_X86_CPU, "hv-version-id-major", "0x0006" },
136     { TYPE_X86_CPU, "hv-version-id-minor", "0x0001" },
137     { "ICH9-LPC", "x-keep-pci-slot-hpc", "false" },
138 };
139 const size_t pc_compat_6_1_len = G_N_ELEMENTS(pc_compat_6_1);
140 
141 GlobalProperty pc_compat_6_0[] = {
142     { "qemu64" "-" TYPE_X86_CPU, "family", "6" },
143     { "qemu64" "-" TYPE_X86_CPU, "model", "6" },
144     { "qemu64" "-" TYPE_X86_CPU, "stepping", "3" },
145     { TYPE_X86_CPU, "x-vendor-cpuid-only", "off" },
146     { "ICH9-LPC", ACPI_PM_PROP_ACPI_PCIHP_BRIDGE, "off" },
147     { "ICH9-LPC", "x-keep-pci-slot-hpc", "true" },
148 };
149 const size_t pc_compat_6_0_len = G_N_ELEMENTS(pc_compat_6_0);
150 
151 GlobalProperty pc_compat_5_2[] = {
152     { "ICH9-LPC", "x-smi-cpu-hotunplug", "off" },
153 };
154 const size_t pc_compat_5_2_len = G_N_ELEMENTS(pc_compat_5_2);
155 
156 GlobalProperty pc_compat_5_1[] = {
157     { "ICH9-LPC", "x-smi-cpu-hotplug", "off" },
158     { TYPE_X86_CPU, "kvm-msi-ext-dest-id", "off" },
159 };
160 const size_t pc_compat_5_1_len = G_N_ELEMENTS(pc_compat_5_1);
161 
162 GlobalProperty pc_compat_5_0[] = {
163 };
164 const size_t pc_compat_5_0_len = G_N_ELEMENTS(pc_compat_5_0);
165 
166 GlobalProperty pc_compat_4_2[] = {
167     { "mch", "smbase-smram", "off" },
168 };
169 const size_t pc_compat_4_2_len = G_N_ELEMENTS(pc_compat_4_2);
170 
171 GlobalProperty pc_compat_4_1[] = {};
172 const size_t pc_compat_4_1_len = G_N_ELEMENTS(pc_compat_4_1);
173 
174 GlobalProperty pc_compat_4_0[] = {};
175 const size_t pc_compat_4_0_len = G_N_ELEMENTS(pc_compat_4_0);
176 
177 GlobalProperty pc_compat_3_1[] = {
178     { "intel-iommu", "dma-drain", "off" },
179     { "Opteron_G3" "-" TYPE_X86_CPU, "rdtscp", "off" },
180     { "Opteron_G4" "-" TYPE_X86_CPU, "rdtscp", "off" },
181     { "Opteron_G4" "-" TYPE_X86_CPU, "npt", "off" },
182     { "Opteron_G4" "-" TYPE_X86_CPU, "nrip-save", "off" },
183     { "Opteron_G5" "-" TYPE_X86_CPU, "rdtscp", "off" },
184     { "Opteron_G5" "-" TYPE_X86_CPU, "npt", "off" },
185     { "Opteron_G5" "-" TYPE_X86_CPU, "nrip-save", "off" },
186     { "EPYC" "-" TYPE_X86_CPU, "npt", "off" },
187     { "EPYC" "-" TYPE_X86_CPU, "nrip-save", "off" },
188     { "EPYC-IBPB" "-" TYPE_X86_CPU, "npt", "off" },
189     { "EPYC-IBPB" "-" TYPE_X86_CPU, "nrip-save", "off" },
190     { "Skylake-Client" "-" TYPE_X86_CPU,      "mpx", "on" },
191     { "Skylake-Client-IBRS" "-" TYPE_X86_CPU, "mpx", "on" },
192     { "Skylake-Server" "-" TYPE_X86_CPU,      "mpx", "on" },
193     { "Skylake-Server-IBRS" "-" TYPE_X86_CPU, "mpx", "on" },
194     { "Cascadelake-Server" "-" TYPE_X86_CPU,  "mpx", "on" },
195     { "Icelake-Client" "-" TYPE_X86_CPU,      "mpx", "on" },
196     { "Icelake-Server" "-" TYPE_X86_CPU,      "mpx", "on" },
197     { "Cascadelake-Server" "-" TYPE_X86_CPU, "stepping", "5" },
198     { TYPE_X86_CPU, "x-intel-pt-auto-level", "off" },
199 };
200 const size_t pc_compat_3_1_len = G_N_ELEMENTS(pc_compat_3_1);
201 
202 GlobalProperty pc_compat_3_0[] = {
203     { TYPE_X86_CPU, "x-hv-synic-kvm-only", "on" },
204     { "Skylake-Server" "-" TYPE_X86_CPU, "pku", "off" },
205     { "Skylake-Server-IBRS" "-" TYPE_X86_CPU, "pku", "off" },
206 };
207 const size_t pc_compat_3_0_len = G_N_ELEMENTS(pc_compat_3_0);
208 
209 GlobalProperty pc_compat_2_12[] = {
210     { TYPE_X86_CPU, "legacy-cache", "on" },
211     { TYPE_X86_CPU, "topoext", "off" },
212     { "EPYC-" TYPE_X86_CPU, "xlevel", "0x8000000a" },
213     { "EPYC-IBPB-" TYPE_X86_CPU, "xlevel", "0x8000000a" },
214 };
215 const size_t pc_compat_2_12_len = G_N_ELEMENTS(pc_compat_2_12);
216 
217 GlobalProperty pc_compat_2_11[] = {
218     { TYPE_X86_CPU, "x-migrate-smi-count", "off" },
219     { "Skylake-Server" "-" TYPE_X86_CPU, "clflushopt", "off" },
220 };
221 const size_t pc_compat_2_11_len = G_N_ELEMENTS(pc_compat_2_11);
222 
223 GlobalProperty pc_compat_2_10[] = {
224     { TYPE_X86_CPU, "x-hv-max-vps", "0x40" },
225     { "i440FX-pcihost", "x-pci-hole64-fix", "off" },
226     { "q35-pcihost", "x-pci-hole64-fix", "off" },
227 };
228 const size_t pc_compat_2_10_len = G_N_ELEMENTS(pc_compat_2_10);
229 
230 GlobalProperty pc_compat_2_9[] = {
231     { "mch", "extended-tseg-mbytes", "0" },
232 };
233 const size_t pc_compat_2_9_len = G_N_ELEMENTS(pc_compat_2_9);
234 
235 GlobalProperty pc_compat_2_8[] = {
236     { TYPE_X86_CPU, "tcg-cpuid", "off" },
237     { "kvmclock", "x-mach-use-reliable-get-clock", "off" },
238     { "ICH9-LPC", "x-smi-broadcast", "off" },
239     { TYPE_X86_CPU, "vmware-cpuid-freq", "off" },
240     { "Haswell-" TYPE_X86_CPU, "stepping", "1" },
241 };
242 const size_t pc_compat_2_8_len = G_N_ELEMENTS(pc_compat_2_8);
243 
244 GlobalProperty pc_compat_2_7[] = {
245     { TYPE_X86_CPU, "l3-cache", "off" },
246     { TYPE_X86_CPU, "full-cpuid-auto-level", "off" },
247     { "Opteron_G3" "-" TYPE_X86_CPU, "family", "15" },
248     { "Opteron_G3" "-" TYPE_X86_CPU, "model", "6" },
249     { "Opteron_G3" "-" TYPE_X86_CPU, "stepping", "1" },
250     { "isa-pcspk", "migrate", "off" },
251 };
252 const size_t pc_compat_2_7_len = G_N_ELEMENTS(pc_compat_2_7);
253 
254 GlobalProperty pc_compat_2_6[] = {
255     { TYPE_X86_CPU, "cpuid-0xb", "off" },
256     { "vmxnet3", "romfile", "" },
257     { TYPE_X86_CPU, "fill-mtrr-mask", "off" },
258     { "apic-common", "legacy-instance-id", "on", }
259 };
260 const size_t pc_compat_2_6_len = G_N_ELEMENTS(pc_compat_2_6);
261 
262 GlobalProperty pc_compat_2_5[] = {};
263 const size_t pc_compat_2_5_len = G_N_ELEMENTS(pc_compat_2_5);
264 
265 GlobalProperty pc_compat_2_4[] = {
266     PC_CPU_MODEL_IDS("2.4.0")
267     { "Haswell-" TYPE_X86_CPU, "abm", "off" },
268     { "Haswell-noTSX-" TYPE_X86_CPU, "abm", "off" },
269     { "Broadwell-" TYPE_X86_CPU, "abm", "off" },
270     { "Broadwell-noTSX-" TYPE_X86_CPU, "abm", "off" },
271     { "host" "-" TYPE_X86_CPU, "host-cache-info", "on" },
272     { TYPE_X86_CPU, "check", "off" },
273     { "qemu64" "-" TYPE_X86_CPU, "sse4a", "on" },
274     { "qemu64" "-" TYPE_X86_CPU, "abm", "on" },
275     { "qemu64" "-" TYPE_X86_CPU, "popcnt", "on" },
276     { "qemu32" "-" TYPE_X86_CPU, "popcnt", "on" },
277     { "Opteron_G2" "-" TYPE_X86_CPU, "rdtscp", "on" },
278     { "Opteron_G3" "-" TYPE_X86_CPU, "rdtscp", "on" },
279     { "Opteron_G4" "-" TYPE_X86_CPU, "rdtscp", "on" },
280     { "Opteron_G5" "-" TYPE_X86_CPU, "rdtscp", "on", }
281 };
282 const size_t pc_compat_2_4_len = G_N_ELEMENTS(pc_compat_2_4);
283 
284 /*
285  * @PC_FW_DATA:
286  * Size of the chunk of memory at the top of RAM for the BIOS ACPI tables
287  * and other BIOS datastructures.
288  *
289  * BIOS ACPI tables: 128K. Other BIOS datastructures: less than 4K
290  * reported to be used at the moment, 32K should be enough for a while.
291  */
292 #define PC_FW_DATA (0x20000 + 0x8000)
293 
294 GSIState *pc_gsi_create(qemu_irq **irqs, bool pci_enabled)
295 {
296     GSIState *s;
297 
298     s = g_new0(GSIState, 1);
299     if (kvm_ioapic_in_kernel()) {
300         kvm_pc_setup_irq_routing(pci_enabled);
301     }
302     *irqs = qemu_allocate_irqs(gsi_handler, s, IOAPIC_NUM_PINS);
303 
304     return s;
305 }
306 
307 static void ioport80_write(void *opaque, hwaddr addr, uint64_t data,
308                            unsigned size)
309 {
310 }
311 
312 static uint64_t ioport80_read(void *opaque, hwaddr addr, unsigned size)
313 {
314     return 0xffffffffffffffffULL;
315 }
316 
317 /* MS-DOS compatibility mode FPU exception support */
318 static void ioportF0_write(void *opaque, hwaddr addr, uint64_t data,
319                            unsigned size)
320 {
321     if (tcg_enabled()) {
322         cpu_set_ignne();
323     }
324 }
325 
326 static uint64_t ioportF0_read(void *opaque, hwaddr addr, unsigned size)
327 {
328     return 0xffffffffffffffffULL;
329 }
330 
331 /* PC cmos mappings */
332 
333 #define REG_EQUIPMENT_BYTE          0x14
334 
335 static void cmos_init_hd(MC146818RtcState *s, int type_ofs, int info_ofs,
336                          int16_t cylinders, int8_t heads, int8_t sectors)
337 {
338     mc146818rtc_set_cmos_data(s, type_ofs, 47);
339     mc146818rtc_set_cmos_data(s, info_ofs, cylinders);
340     mc146818rtc_set_cmos_data(s, info_ofs + 1, cylinders >> 8);
341     mc146818rtc_set_cmos_data(s, info_ofs + 2, heads);
342     mc146818rtc_set_cmos_data(s, info_ofs + 3, 0xff);
343     mc146818rtc_set_cmos_data(s, info_ofs + 4, 0xff);
344     mc146818rtc_set_cmos_data(s, info_ofs + 5, 0xc0 | ((heads > 8) << 3));
345     mc146818rtc_set_cmos_data(s, info_ofs + 6, cylinders);
346     mc146818rtc_set_cmos_data(s, info_ofs + 7, cylinders >> 8);
347     mc146818rtc_set_cmos_data(s, info_ofs + 8, sectors);
348 }
349 
350 /* convert boot_device letter to something recognizable by the bios */
351 static int boot_device2nibble(char boot_device)
352 {
353     switch(boot_device) {
354     case 'a':
355     case 'b':
356         return 0x01; /* floppy boot */
357     case 'c':
358         return 0x02; /* hard drive boot */
359     case 'd':
360         return 0x03; /* CD-ROM boot */
361     case 'n':
362         return 0x04; /* Network boot */
363     }
364     return 0;
365 }
366 
367 static void set_boot_dev(PCMachineState *pcms, MC146818RtcState *s,
368                          const char *boot_device, Error **errp)
369 {
370 #define PC_MAX_BOOT_DEVICES 3
371     int nbds, bds[3] = { 0, };
372     int i;
373 
374     nbds = strlen(boot_device);
375     if (nbds > PC_MAX_BOOT_DEVICES) {
376         error_setg(errp, "Too many boot devices for PC");
377         return;
378     }
379     for (i = 0; i < nbds; i++) {
380         bds[i] = boot_device2nibble(boot_device[i]);
381         if (bds[i] == 0) {
382             error_setg(errp, "Invalid boot device for PC: '%c'",
383                        boot_device[i]);
384             return;
385         }
386     }
387     mc146818rtc_set_cmos_data(s, 0x3d, (bds[1] << 4) | bds[0]);
388     mc146818rtc_set_cmos_data(s, 0x38, (bds[2] << 4) | !pcms->fd_bootchk);
389 }
390 
391 static void pc_boot_set(void *opaque, const char *boot_device, Error **errp)
392 {
393     PCMachineState *pcms = opaque;
394     X86MachineState *x86ms = X86_MACHINE(pcms);
395 
396     set_boot_dev(pcms, MC146818_RTC(x86ms->rtc), boot_device, errp);
397 }
398 
399 static void pc_cmos_init_floppy(MC146818RtcState *rtc_state, ISADevice *floppy)
400 {
401     int val, nb;
402     FloppyDriveType fd_type[2] = { FLOPPY_DRIVE_TYPE_NONE,
403                                    FLOPPY_DRIVE_TYPE_NONE };
404 
405 #ifdef CONFIG_FDC_ISA
406     /* floppy type */
407     if (floppy) {
408         for (int i = 0; i < 2; i++) {
409             fd_type[i] = isa_fdc_get_drive_type(floppy, i);
410         }
411     }
412 #endif
413 
414     val = (cmos_get_fd_drive_type(fd_type[0]) << 4) |
415         cmos_get_fd_drive_type(fd_type[1]);
416     mc146818rtc_set_cmos_data(rtc_state, 0x10, val);
417 
418     val = mc146818rtc_get_cmos_data(rtc_state, REG_EQUIPMENT_BYTE);
419     nb = 0;
420     if (fd_type[0] != FLOPPY_DRIVE_TYPE_NONE) {
421         nb++;
422     }
423     if (fd_type[1] != FLOPPY_DRIVE_TYPE_NONE) {
424         nb++;
425     }
426     switch (nb) {
427     case 0:
428         break;
429     case 1:
430         val |= 0x01; /* 1 drive, ready for boot */
431         break;
432     case 2:
433         val |= 0x41; /* 2 drives, ready for boot */
434         break;
435     }
436     mc146818rtc_set_cmos_data(rtc_state, REG_EQUIPMENT_BYTE, val);
437 }
438 
439 typedef struct check_fdc_state {
440     ISADevice *floppy;
441     bool multiple;
442 } CheckFdcState;
443 
444 static int check_fdc(Object *obj, void *opaque)
445 {
446     CheckFdcState *state = opaque;
447     Object *fdc;
448     uint32_t iobase;
449     Error *local_err = NULL;
450 
451     fdc = object_dynamic_cast(obj, TYPE_ISA_FDC);
452     if (!fdc) {
453         return 0;
454     }
455 
456     iobase = object_property_get_uint(obj, "iobase", &local_err);
457     if (local_err || iobase != 0x3f0) {
458         error_free(local_err);
459         return 0;
460     }
461 
462     if (state->floppy) {
463         state->multiple = true;
464     } else {
465         state->floppy = ISA_DEVICE(obj);
466     }
467     return 0;
468 }
469 
470 static const char * const fdc_container_path[] = {
471     "unattached", "peripheral", "peripheral-anon"
472 };
473 
474 /*
475  * Locate the FDC at IO address 0x3f0, in order to configure the CMOS registers
476  * and ACPI objects.
477  */
478 static ISADevice *pc_find_fdc0(void)
479 {
480     int i;
481     Object *container;
482     CheckFdcState state = { 0 };
483 
484     for (i = 0; i < ARRAY_SIZE(fdc_container_path); i++) {
485         container = machine_get_container(fdc_container_path[i]);
486         object_child_foreach(container, check_fdc, &state);
487     }
488 
489     if (state.multiple) {
490         warn_report("multiple floppy disk controllers with "
491                     "iobase=0x3f0 have been found");
492         error_printf("the one being picked for CMOS setup might not reflect "
493                      "your intent");
494     }
495 
496     return state.floppy;
497 }
498 
499 static void pc_cmos_init_late(PCMachineState *pcms)
500 {
501     X86MachineState *x86ms = X86_MACHINE(pcms);
502     MC146818RtcState *s = MC146818_RTC(x86ms->rtc);
503     int16_t cylinders;
504     int8_t heads, sectors;
505     int val;
506     int i, trans;
507 
508     val = 0;
509     if (pcms->idebus[0] &&
510         ide_get_geometry(pcms->idebus[0], 0,
511                          &cylinders, &heads, &sectors) >= 0) {
512         cmos_init_hd(s, 0x19, 0x1b, cylinders, heads, sectors);
513         val |= 0xf0;
514     }
515     if (pcms->idebus[0] &&
516         ide_get_geometry(pcms->idebus[0], 1,
517                          &cylinders, &heads, &sectors) >= 0) {
518         cmos_init_hd(s, 0x1a, 0x24, cylinders, heads, sectors);
519         val |= 0x0f;
520     }
521     mc146818rtc_set_cmos_data(s, 0x12, val);
522 
523     val = 0;
524     for (i = 0; i < 4; i++) {
525         /* NOTE: ide_get_geometry() returns the physical
526            geometry.  It is always such that: 1 <= sects <= 63, 1
527            <= heads <= 16, 1 <= cylinders <= 16383. The BIOS
528            geometry can be different if a translation is done. */
529         BusState *idebus = pcms->idebus[i / 2];
530         if (idebus &&
531             ide_get_geometry(idebus, i % 2,
532                              &cylinders, &heads, &sectors) >= 0) {
533             trans = ide_get_bios_chs_trans(idebus, i % 2) - 1;
534             assert((trans & ~3) == 0);
535             val |= trans << (i * 2);
536         }
537     }
538     mc146818rtc_set_cmos_data(s, 0x39, val);
539 
540     pc_cmos_init_floppy(s, pc_find_fdc0());
541 
542     /* various important CMOS locations needed by PC/Bochs bios */
543 
544     /* memory size */
545     /* base memory (first MiB) */
546     val = MIN(x86ms->below_4g_mem_size / KiB, 640);
547     mc146818rtc_set_cmos_data(s, 0x15, val);
548     mc146818rtc_set_cmos_data(s, 0x16, val >> 8);
549     /* extended memory (next 64MiB) */
550     if (x86ms->below_4g_mem_size > 1 * MiB) {
551         val = (x86ms->below_4g_mem_size - 1 * MiB) / KiB;
552     } else {
553         val = 0;
554     }
555     if (val > 65535)
556         val = 65535;
557     mc146818rtc_set_cmos_data(s, 0x17, val);
558     mc146818rtc_set_cmos_data(s, 0x18, val >> 8);
559     mc146818rtc_set_cmos_data(s, 0x30, val);
560     mc146818rtc_set_cmos_data(s, 0x31, val >> 8);
561     /* memory between 16MiB and 4GiB */
562     if (x86ms->below_4g_mem_size > 16 * MiB) {
563         val = (x86ms->below_4g_mem_size - 16 * MiB) / (64 * KiB);
564     } else {
565         val = 0;
566     }
567     if (val > 65535)
568         val = 65535;
569     mc146818rtc_set_cmos_data(s, 0x34, val);
570     mc146818rtc_set_cmos_data(s, 0x35, val >> 8);
571     /* memory above 4GiB */
572     val = x86ms->above_4g_mem_size / 65536;
573     mc146818rtc_set_cmos_data(s, 0x5b, val);
574     mc146818rtc_set_cmos_data(s, 0x5c, val >> 8);
575     mc146818rtc_set_cmos_data(s, 0x5d, val >> 16);
576 
577     val = 0;
578     val |= 0x02; /* FPU is there */
579     val |= 0x04; /* PS/2 mouse installed */
580     mc146818rtc_set_cmos_data(s, REG_EQUIPMENT_BYTE, val);
581 }
582 
583 static void handle_a20_line_change(void *opaque, int irq, int level)
584 {
585     X86CPU *cpu = opaque;
586 
587     /* XXX: send to all CPUs ? */
588     /* XXX: add logic to handle multiple A20 line sources */
589     x86_cpu_set_a20(cpu, level);
590 }
591 
592 #define NE2000_NB_MAX 6
593 
594 static const int ne2000_io[NE2000_NB_MAX] = { 0x300, 0x320, 0x340, 0x360,
595                                               0x280, 0x380 };
596 static const int ne2000_irq[NE2000_NB_MAX] = { 9, 10, 11, 3, 4, 5 };
597 
598 static gboolean pc_init_ne2k_isa(ISABus *bus, NICInfo *nd, Error **errp)
599 {
600     static int nb_ne2k = 0;
601 
602     if (nb_ne2k == NE2000_NB_MAX) {
603         error_setg(errp,
604                    "maximum number of ISA NE2000 devices exceeded");
605         return false;
606     }
607     isa_ne2000_init(bus, ne2000_io[nb_ne2k],
608                     ne2000_irq[nb_ne2k], nd);
609     nb_ne2k++;
610     return true;
611 }
612 
613 void pc_acpi_smi_interrupt(void *opaque, int irq, int level)
614 {
615     X86CPU *cpu = opaque;
616 
617     if (level) {
618         cpu_interrupt(CPU(cpu), CPU_INTERRUPT_SMI);
619     }
620 }
621 
622 static
623 void pc_machine_done(Notifier *notifier, void *data)
624 {
625     PCMachineState *pcms = container_of(notifier,
626                                         PCMachineState, machine_done);
627     X86MachineState *x86ms = X86_MACHINE(pcms);
628 
629     cxl_hook_up_pxb_registers(pcms->pcibus, &pcms->cxl_devices_state,
630                               &error_fatal);
631 
632     if (pcms->cxl_devices_state.is_enabled) {
633         cxl_fmws_link_targets(&pcms->cxl_devices_state, &error_fatal);
634     }
635 
636     /* set the number of CPUs */
637     x86_rtc_set_cpus_count(x86ms->rtc, x86ms->boot_cpus);
638 
639     pci_bus_add_fw_cfg_extra_pci_roots(x86ms->fw_cfg, pcms->pcibus,
640                                        &error_abort);
641 
642     acpi_setup();
643     if (x86ms->fw_cfg) {
644         fw_cfg_build_smbios(pcms, x86ms->fw_cfg, pcms->smbios_entry_point_type);
645         fw_cfg_add_e820(x86ms->fw_cfg);
646         fw_cfg_build_feature_control(MACHINE(pcms), x86ms->fw_cfg);
647         /* update FW_CFG_NB_CPUS to account for -device added CPUs */
648         fw_cfg_modify_i16(x86ms->fw_cfg, FW_CFG_NB_CPUS, x86ms->boot_cpus);
649     }
650 
651     pc_cmos_init_late(pcms);
652 }
653 
654 /* setup pci memory address space mapping into system address space */
655 void pc_pci_as_mapping_init(MemoryRegion *system_memory,
656                             MemoryRegion *pci_address_space)
657 {
658     /* Set to lower priority than RAM */
659     memory_region_add_subregion_overlap(system_memory, 0x0,
660                                         pci_address_space, -1);
661 }
662 
663 void xen_load_linux(PCMachineState *pcms)
664 {
665     int i;
666     FWCfgState *fw_cfg;
667     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
668     X86MachineState *x86ms = X86_MACHINE(pcms);
669 
670     assert(MACHINE(pcms)->kernel_filename != NULL);
671 
672     fw_cfg = fw_cfg_init_io_dma(FW_CFG_IO_BASE, FW_CFG_IO_BASE + 4,
673                                 &address_space_memory);
674     fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, x86ms->boot_cpus);
675     rom_set_fw(fw_cfg);
676 
677     x86_load_linux(x86ms, fw_cfg, PC_FW_DATA, pcmc->pvh_enabled);
678     for (i = 0; i < nb_option_roms; i++) {
679         assert(!strcmp(option_rom[i].name, "linuxboot.bin") ||
680                !strcmp(option_rom[i].name, "linuxboot_dma.bin") ||
681                !strcmp(option_rom[i].name, "pvh.bin") ||
682                !strcmp(option_rom[i].name, "multiboot.bin") ||
683                !strcmp(option_rom[i].name, "multiboot_dma.bin"));
684         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
685     }
686     x86ms->fw_cfg = fw_cfg;
687 }
688 
689 #define PC_ROM_MIN_VGA     0xc0000
690 #define PC_ROM_MIN_OPTION  0xc8000
691 #define PC_ROM_MAX         0xe0000
692 #define PC_ROM_ALIGN       0x800
693 #define PC_ROM_SIZE        (PC_ROM_MAX - PC_ROM_MIN_VGA)
694 
695 static hwaddr pc_above_4g_end(PCMachineState *pcms)
696 {
697     X86MachineState *x86ms = X86_MACHINE(pcms);
698 
699     if (pcms->sgx_epc.size != 0) {
700         return sgx_epc_above_4g_end(&pcms->sgx_epc);
701     }
702 
703     return x86ms->above_4g_mem_start + x86ms->above_4g_mem_size;
704 }
705 
706 static void pc_get_device_memory_range(PCMachineState *pcms,
707                                        hwaddr *base,
708                                        ram_addr_t *device_mem_size)
709 {
710     MachineState *machine = MACHINE(pcms);
711     ram_addr_t size;
712     hwaddr addr;
713 
714     size = machine->maxram_size - machine->ram_size;
715     addr = ROUND_UP(pc_above_4g_end(pcms), 1 * GiB);
716 
717     /* size device region assuming 1G page max alignment per slot */
718     size += (1 * GiB) * machine->ram_slots;
719 
720     *base = addr;
721     *device_mem_size = size;
722 }
723 
724 static uint64_t pc_get_cxl_range_start(PCMachineState *pcms)
725 {
726     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
727     MachineState *ms = MACHINE(pcms);
728     hwaddr cxl_base;
729     ram_addr_t size;
730 
731     if (pcmc->has_reserved_memory &&
732         (ms->ram_size < ms->maxram_size)) {
733         pc_get_device_memory_range(pcms, &cxl_base, &size);
734         cxl_base += size;
735     } else {
736         cxl_base = pc_above_4g_end(pcms);
737     }
738 
739     return cxl_base;
740 }
741 
742 static uint64_t pc_get_cxl_range_end(PCMachineState *pcms)
743 {
744     uint64_t start = pc_get_cxl_range_start(pcms) + MiB;
745 
746     if (pcms->cxl_devices_state.fixed_windows) {
747         GList *it;
748 
749         start = ROUND_UP(start, 256 * MiB);
750         for (it = pcms->cxl_devices_state.fixed_windows; it; it = it->next) {
751             CXLFixedWindow *fw = it->data;
752             start += fw->size;
753         }
754     }
755 
756     return start;
757 }
758 
759 static hwaddr pc_max_used_gpa(PCMachineState *pcms, uint64_t pci_hole64_size)
760 {
761     X86CPU *cpu = X86_CPU(first_cpu);
762     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
763     MachineState *ms = MACHINE(pcms);
764 
765     if (cpu->env.features[FEAT_8000_0001_EDX] & CPUID_EXT2_LM) {
766         /* 64-bit systems */
767         return pc_pci_hole64_start() + pci_hole64_size - 1;
768     }
769 
770     /* 32-bit systems */
771     if (pcmc->broken_32bit_mem_addr_check) {
772         /* old value for compatibility reasons */
773         return ((hwaddr)1 << cpu->phys_bits) - 1;
774     }
775 
776     /*
777      * 32-bit systems don't have hole64 but they might have a region for
778      * memory devices. Even if additional hotplugged memory devices might
779      * not be usable by most guest OSes, we need to still consider them for
780      * calculating the highest possible GPA so that we can properly report
781      * if someone configures them on a CPU that cannot possibly address them.
782      */
783     if (pcmc->has_reserved_memory &&
784         (ms->ram_size < ms->maxram_size)) {
785         hwaddr devmem_start;
786         ram_addr_t devmem_size;
787 
788         pc_get_device_memory_range(pcms, &devmem_start, &devmem_size);
789         devmem_start += devmem_size;
790         return devmem_start - 1;
791     }
792 
793     /* configuration without any memory hotplug */
794     return pc_above_4g_end(pcms) - 1;
795 }
796 
797 /*
798  * AMD systems with an IOMMU have an additional hole close to the
799  * 1Tb, which are special GPAs that cannot be DMA mapped. Depending
800  * on kernel version, VFIO may or may not let you DMA map those ranges.
801  * Starting Linux v5.4 we validate it, and can't create guests on AMD machines
802  * with certain memory sizes. It's also wrong to use those IOVA ranges
803  * in detriment of leading to IOMMU INVALID_DEVICE_REQUEST or worse.
804  * The ranges reserved for Hyper-Transport are:
805  *
806  * FD_0000_0000h - FF_FFFF_FFFFh
807  *
808  * The ranges represent the following:
809  *
810  * Base Address   Top Address  Use
811  *
812  * FD_0000_0000h FD_F7FF_FFFFh Reserved interrupt address space
813  * FD_F800_0000h FD_F8FF_FFFFh Interrupt/EOI IntCtl
814  * FD_F900_0000h FD_F90F_FFFFh Legacy PIC IACK
815  * FD_F910_0000h FD_F91F_FFFFh System Management
816  * FD_F920_0000h FD_FAFF_FFFFh Reserved Page Tables
817  * FD_FB00_0000h FD_FBFF_FFFFh Address Translation
818  * FD_FC00_0000h FD_FDFF_FFFFh I/O Space
819  * FD_FE00_0000h FD_FFFF_FFFFh Configuration
820  * FE_0000_0000h FE_1FFF_FFFFh Extended Configuration/Device Messages
821  * FE_2000_0000h FF_FFFF_FFFFh Reserved
822  *
823  * See AMD IOMMU spec, section 2.1.2 "IOMMU Logical Topology",
824  * Table 3: Special Address Controls (GPA) for more information.
825  */
826 #define AMD_HT_START         0xfd00000000UL
827 #define AMD_HT_END           0xffffffffffUL
828 #define AMD_ABOVE_1TB_START  (AMD_HT_END + 1)
829 #define AMD_HT_SIZE          (AMD_ABOVE_1TB_START - AMD_HT_START)
830 
831 void pc_memory_init(PCMachineState *pcms,
832                     MemoryRegion *system_memory,
833                     MemoryRegion *rom_memory,
834                     uint64_t pci_hole64_size)
835 {
836     int linux_boot, i;
837     MemoryRegion *option_rom_mr;
838     MemoryRegion *ram_below_4g, *ram_above_4g;
839     FWCfgState *fw_cfg;
840     MachineState *machine = MACHINE(pcms);
841     MachineClass *mc = MACHINE_GET_CLASS(machine);
842     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
843     X86MachineState *x86ms = X86_MACHINE(pcms);
844     hwaddr maxphysaddr, maxusedaddr;
845     hwaddr cxl_base, cxl_resv_end = 0;
846     X86CPU *cpu = X86_CPU(first_cpu);
847 
848     assert(machine->ram_size == x86ms->below_4g_mem_size +
849                                 x86ms->above_4g_mem_size);
850 
851     linux_boot = (machine->kernel_filename != NULL);
852 
853     /*
854      * The HyperTransport range close to the 1T boundary is unique to AMD
855      * hosts with IOMMUs enabled. Restrict the ram-above-4g relocation
856      * to above 1T to AMD vCPUs only. @enforce_amd_1tb_hole is only false in
857      * older machine types (<= 7.0) for compatibility purposes.
858      */
859     if (IS_AMD_CPU(&cpu->env) && pcmc->enforce_amd_1tb_hole) {
860         /* Bail out if max possible address does not cross HT range */
861         if (pc_max_used_gpa(pcms, pci_hole64_size) >= AMD_HT_START) {
862             x86ms->above_4g_mem_start = AMD_ABOVE_1TB_START;
863         }
864 
865         /*
866          * Advertise the HT region if address space covers the reserved
867          * region or if we relocate.
868          */
869         if (cpu->phys_bits >= 40) {
870             e820_add_entry(AMD_HT_START, AMD_HT_SIZE, E820_RESERVED);
871         }
872     }
873 
874     /*
875      * phys-bits is required to be appropriately configured
876      * to make sure max used GPA is reachable.
877      */
878     maxusedaddr = pc_max_used_gpa(pcms, pci_hole64_size);
879     maxphysaddr = ((hwaddr)1 << cpu->phys_bits) - 1;
880     if (maxphysaddr < maxusedaddr) {
881         error_report("Address space limit 0x%"PRIx64" < 0x%"PRIx64
882                      " phys-bits too low (%u)",
883                      maxphysaddr, maxusedaddr, cpu->phys_bits);
884         exit(EXIT_FAILURE);
885     }
886 
887     /*
888      * Split single memory region and use aliases to address portions of it,
889      * done for backwards compatibility with older qemus.
890      */
891     ram_below_4g = g_malloc(sizeof(*ram_below_4g));
892     memory_region_init_alias(ram_below_4g, NULL, "ram-below-4g", machine->ram,
893                              0, x86ms->below_4g_mem_size);
894     memory_region_add_subregion(system_memory, 0, ram_below_4g);
895     e820_add_entry(0, x86ms->below_4g_mem_size, E820_RAM);
896     if (x86ms->above_4g_mem_size > 0) {
897         ram_above_4g = g_malloc(sizeof(*ram_above_4g));
898         memory_region_init_alias(ram_above_4g, NULL, "ram-above-4g",
899                                  machine->ram,
900                                  x86ms->below_4g_mem_size,
901                                  x86ms->above_4g_mem_size);
902         memory_region_add_subregion(system_memory, x86ms->above_4g_mem_start,
903                                     ram_above_4g);
904         e820_add_entry(x86ms->above_4g_mem_start, x86ms->above_4g_mem_size,
905                        E820_RAM);
906     }
907 
908     if (pcms->sgx_epc.size != 0) {
909         e820_add_entry(pcms->sgx_epc.base, pcms->sgx_epc.size, E820_RESERVED);
910     }
911 
912     if (!pcmc->has_reserved_memory &&
913         (machine->ram_slots ||
914          (machine->maxram_size > machine->ram_size))) {
915 
916         error_report("\"-memory 'slots|maxmem'\" is not supported by: %s",
917                      mc->name);
918         exit(EXIT_FAILURE);
919     }
920 
921     /* initialize device memory address space */
922     if (pcmc->has_reserved_memory &&
923         (machine->ram_size < machine->maxram_size)) {
924         ram_addr_t device_mem_size;
925         hwaddr device_mem_base;
926 
927         if (machine->ram_slots > ACPI_MAX_RAM_SLOTS) {
928             error_report("unsupported amount of memory slots: %"PRIu64,
929                          machine->ram_slots);
930             exit(EXIT_FAILURE);
931         }
932 
933         if (QEMU_ALIGN_UP(machine->maxram_size,
934                           TARGET_PAGE_SIZE) != machine->maxram_size) {
935             error_report("maximum memory size must by aligned to multiple of "
936                          "%d bytes", TARGET_PAGE_SIZE);
937             exit(EXIT_FAILURE);
938         }
939 
940         pc_get_device_memory_range(pcms, &device_mem_base, &device_mem_size);
941 
942         if (device_mem_base + device_mem_size < device_mem_size) {
943             error_report("unsupported amount of maximum memory: " RAM_ADDR_FMT,
944                          machine->maxram_size);
945             exit(EXIT_FAILURE);
946         }
947         machine_memory_devices_init(machine, device_mem_base, device_mem_size);
948     }
949 
950     if (pcms->cxl_devices_state.is_enabled) {
951         MemoryRegion *mr = &pcms->cxl_devices_state.host_mr;
952         hwaddr cxl_size = MiB;
953 
954         cxl_base = pc_get_cxl_range_start(pcms);
955         memory_region_init(mr, OBJECT(machine), "cxl_host_reg", cxl_size);
956         memory_region_add_subregion(system_memory, cxl_base, mr);
957         cxl_resv_end = cxl_base + cxl_size;
958         if (pcms->cxl_devices_state.fixed_windows) {
959             hwaddr cxl_fmw_base;
960             GList *it;
961 
962             cxl_fmw_base = ROUND_UP(cxl_base + cxl_size, 256 * MiB);
963             for (it = pcms->cxl_devices_state.fixed_windows; it; it = it->next) {
964                 CXLFixedWindow *fw = it->data;
965 
966                 fw->base = cxl_fmw_base;
967                 memory_region_init_io(&fw->mr, OBJECT(machine), &cfmws_ops, fw,
968                                       "cxl-fixed-memory-region", fw->size);
969                 memory_region_add_subregion(system_memory, fw->base, &fw->mr);
970                 cxl_fmw_base += fw->size;
971                 cxl_resv_end = cxl_fmw_base;
972             }
973         }
974     }
975 
976     /* Initialize PC system firmware */
977     pc_system_firmware_init(pcms, rom_memory);
978 
979     option_rom_mr = g_malloc(sizeof(*option_rom_mr));
980     if (machine_require_guest_memfd(machine)) {
981         memory_region_init_ram_guest_memfd(option_rom_mr, NULL, "pc.rom",
982                                            PC_ROM_SIZE, &error_fatal);
983     } else {
984         memory_region_init_ram(option_rom_mr, NULL, "pc.rom", PC_ROM_SIZE,
985                                &error_fatal);
986         if (pcmc->pci_enabled) {
987             memory_region_set_readonly(option_rom_mr, true);
988         }
989     }
990     memory_region_add_subregion_overlap(rom_memory,
991                                         PC_ROM_MIN_VGA,
992                                         option_rom_mr,
993                                         1);
994 
995     fw_cfg = fw_cfg_arch_create(machine,
996                                 x86ms->boot_cpus, x86ms->apic_id_limit);
997 
998     rom_set_fw(fw_cfg);
999 
1000     if (machine->device_memory) {
1001         uint64_t *val = g_malloc(sizeof(*val));
1002         uint64_t res_mem_end = machine->device_memory->base;
1003 
1004         if (!pcmc->broken_reserved_end) {
1005             res_mem_end += memory_region_size(&machine->device_memory->mr);
1006         }
1007 
1008         if (pcms->cxl_devices_state.is_enabled) {
1009             res_mem_end = cxl_resv_end;
1010         }
1011         *val = cpu_to_le64(ROUND_UP(res_mem_end, 1 * GiB));
1012         fw_cfg_add_file(fw_cfg, "etc/reserved-memory-end", val, sizeof(*val));
1013     }
1014 
1015     if (linux_boot) {
1016         x86_load_linux(x86ms, fw_cfg, PC_FW_DATA, pcmc->pvh_enabled);
1017     }
1018 
1019     for (i = 0; i < nb_option_roms; i++) {
1020         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1021     }
1022     x86ms->fw_cfg = fw_cfg;
1023 
1024     /* Init default IOAPIC address space */
1025     x86ms->ioapic_as = &address_space_memory;
1026 
1027     /* Init ACPI memory hotplug IO base address */
1028     pcms->memhp_io_base = ACPI_MEMORY_HOTPLUG_BASE;
1029 }
1030 
1031 /*
1032  * The 64bit pci hole starts after "above 4G RAM" and
1033  * potentially the space reserved for memory hotplug.
1034  */
1035 uint64_t pc_pci_hole64_start(void)
1036 {
1037     PCMachineState *pcms = PC_MACHINE(qdev_get_machine());
1038     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1039     MachineState *ms = MACHINE(pcms);
1040     uint64_t hole64_start = 0;
1041     ram_addr_t size = 0;
1042 
1043     if (pcms->cxl_devices_state.is_enabled) {
1044         hole64_start = pc_get_cxl_range_end(pcms);
1045     } else if (pcmc->has_reserved_memory && (ms->ram_size < ms->maxram_size)) {
1046         pc_get_device_memory_range(pcms, &hole64_start, &size);
1047         if (!pcmc->broken_reserved_end) {
1048             hole64_start += size;
1049         }
1050     } else {
1051         hole64_start = pc_above_4g_end(pcms);
1052     }
1053 
1054     return ROUND_UP(hole64_start, 1 * GiB);
1055 }
1056 
1057 DeviceState *pc_vga_init(ISABus *isa_bus, PCIBus *pci_bus)
1058 {
1059     DeviceState *dev = NULL;
1060 
1061     rom_set_order_override(FW_CFG_ORDER_OVERRIDE_VGA);
1062     if (pci_bus) {
1063         PCIDevice *pcidev = pci_vga_init(pci_bus);
1064         dev = pcidev ? &pcidev->qdev : NULL;
1065     } else if (isa_bus) {
1066         ISADevice *isadev = isa_vga_init(isa_bus);
1067         dev = isadev ? DEVICE(isadev) : NULL;
1068     }
1069     rom_reset_order_override();
1070     return dev;
1071 }
1072 
1073 static const MemoryRegionOps ioport80_io_ops = {
1074     .write = ioport80_write,
1075     .read = ioport80_read,
1076     .endianness = DEVICE_LITTLE_ENDIAN,
1077     .impl = {
1078         .min_access_size = 1,
1079         .max_access_size = 1,
1080     },
1081 };
1082 
1083 static const MemoryRegionOps ioportF0_io_ops = {
1084     .write = ioportF0_write,
1085     .read = ioportF0_read,
1086     .endianness = DEVICE_LITTLE_ENDIAN,
1087     .impl = {
1088         .min_access_size = 1,
1089         .max_access_size = 1,
1090     },
1091 };
1092 
1093 static void pc_superio_init(ISABus *isa_bus, bool create_fdctrl,
1094                             bool create_i8042, bool no_vmport, Error **errp)
1095 {
1096     int i;
1097     DriveInfo *fd[MAX_FD];
1098     qemu_irq *a20_line;
1099     ISADevice *i8042, *port92, *vmmouse;
1100 
1101     serial_hds_isa_init(isa_bus, 0, MAX_ISA_SERIAL_PORTS);
1102     parallel_hds_isa_init(isa_bus, MAX_PARALLEL_PORTS);
1103 
1104     for (i = 0; i < MAX_FD; i++) {
1105         fd[i] = drive_get(IF_FLOPPY, 0, i);
1106         create_fdctrl |= !!fd[i];
1107     }
1108     if (create_fdctrl) {
1109 #ifdef CONFIG_FDC_ISA
1110         ISADevice *fdc = isa_new(TYPE_ISA_FDC);
1111         if (fdc) {
1112             isa_realize_and_unref(fdc, isa_bus, &error_fatal);
1113             isa_fdc_init_drives(fdc, fd);
1114         }
1115 #endif
1116     }
1117 
1118     if (!create_i8042) {
1119         if (!no_vmport) {
1120             error_setg(errp,
1121                        "vmport requires the i8042 controller to be enabled");
1122         }
1123         return;
1124     }
1125 
1126     i8042 = isa_create_simple(isa_bus, TYPE_I8042);
1127     if (!no_vmport) {
1128         isa_create_simple(isa_bus, TYPE_VMPORT);
1129         vmmouse = isa_try_new("vmmouse");
1130     } else {
1131         vmmouse = NULL;
1132     }
1133     if (vmmouse) {
1134         object_property_set_link(OBJECT(vmmouse), TYPE_I8042, OBJECT(i8042),
1135                                  &error_abort);
1136         isa_realize_and_unref(vmmouse, isa_bus, &error_fatal);
1137     }
1138     port92 = isa_create_simple(isa_bus, TYPE_PORT92);
1139 
1140     a20_line = qemu_allocate_irqs(handle_a20_line_change, first_cpu, 2);
1141     qdev_connect_gpio_out_named(DEVICE(i8042),
1142                                 I8042_A20_LINE, 0, a20_line[0]);
1143     qdev_connect_gpio_out_named(DEVICE(port92),
1144                                 PORT92_A20_LINE, 0, a20_line[1]);
1145     g_free(a20_line);
1146 }
1147 
1148 void pc_basic_device_init(struct PCMachineState *pcms,
1149                           ISABus *isa_bus, qemu_irq *gsi,
1150                           ISADevice *rtc_state,
1151                           bool create_fdctrl,
1152                           uint32_t hpet_irqs)
1153 {
1154     int i;
1155     DeviceState *hpet = NULL;
1156     int pit_isa_irq = 0;
1157     qemu_irq pit_alt_irq = NULL;
1158     ISADevice *pit = NULL;
1159     MemoryRegion *ioport80_io = g_new(MemoryRegion, 1);
1160     MemoryRegion *ioportF0_io = g_new(MemoryRegion, 1);
1161     X86MachineState *x86ms = X86_MACHINE(pcms);
1162 
1163     memory_region_init_io(ioport80_io, NULL, &ioport80_io_ops, NULL, "ioport80", 1);
1164     memory_region_add_subregion(isa_bus->address_space_io, 0x80, ioport80_io);
1165 
1166     memory_region_init_io(ioportF0_io, NULL, &ioportF0_io_ops, NULL, "ioportF0", 1);
1167     memory_region_add_subregion(isa_bus->address_space_io, 0xf0, ioportF0_io);
1168 
1169     /*
1170      * Check if an HPET shall be created.
1171      */
1172     if (pcms->hpet_enabled) {
1173         qemu_irq rtc_irq;
1174 
1175         hpet = qdev_try_new(TYPE_HPET);
1176         if (!hpet) {
1177             error_report("couldn't create HPET device");
1178             exit(1);
1179         }
1180         /*
1181          * For pc-piix-*, hpet's intcap is always IRQ2. For pc-q35-*,
1182          * use IRQ16~23, IRQ8 and IRQ2.  If the user has already set
1183          * the property, use whatever mask they specified.
1184          */
1185         uint8_t compat = object_property_get_uint(OBJECT(hpet),
1186                 HPET_INTCAP, NULL);
1187         if (!compat) {
1188             qdev_prop_set_uint32(hpet, HPET_INTCAP, hpet_irqs);
1189         }
1190         sysbus_realize_and_unref(SYS_BUS_DEVICE(hpet), &error_fatal);
1191         sysbus_mmio_map(SYS_BUS_DEVICE(hpet), 0, HPET_BASE);
1192 
1193         for (i = 0; i < IOAPIC_NUM_PINS; i++) {
1194             sysbus_connect_irq(SYS_BUS_DEVICE(hpet), i, gsi[i]);
1195         }
1196         pit_isa_irq = -1;
1197         pit_alt_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_PIT_INT);
1198         rtc_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_RTC_INT);
1199 
1200         /* overwrite connection created by south bridge */
1201         qdev_connect_gpio_out(DEVICE(rtc_state), 0, rtc_irq);
1202     }
1203 
1204     object_property_add_alias(OBJECT(pcms), "rtc-time", OBJECT(rtc_state),
1205                               "date");
1206 
1207 #ifdef CONFIG_XEN_EMU
1208     if (xen_mode == XEN_EMULATE) {
1209         xen_overlay_create();
1210         xen_evtchn_create(IOAPIC_NUM_PINS, gsi);
1211         xen_gnttab_create();
1212         xen_xenstore_create();
1213         if (pcms->pcibus) {
1214             pci_create_simple(pcms->pcibus, -1, "xen-platform");
1215         }
1216         xen_bus_init();
1217     }
1218 #endif
1219 
1220     qemu_register_boot_set(pc_boot_set, pcms);
1221     set_boot_dev(pcms, MC146818_RTC(rtc_state),
1222                  MACHINE(pcms)->boot_config.order, &error_fatal);
1223 
1224     if (!xen_enabled() &&
1225         (x86ms->pit == ON_OFF_AUTO_AUTO || x86ms->pit == ON_OFF_AUTO_ON)) {
1226         if (kvm_pit_in_kernel()) {
1227             pit = kvm_pit_init(isa_bus, 0x40);
1228         } else {
1229             pit = i8254_pit_init(isa_bus, 0x40, pit_isa_irq, pit_alt_irq);
1230         }
1231         if (hpet) {
1232             /* connect PIT to output control line of the HPET */
1233             qdev_connect_gpio_out(hpet, 0, qdev_get_gpio_in(DEVICE(pit), 0));
1234         }
1235         object_property_set_link(OBJECT(pcms->pcspk), "pit",
1236                                  OBJECT(pit), &error_fatal);
1237         isa_realize_and_unref(pcms->pcspk, isa_bus, &error_fatal);
1238     }
1239 
1240     if (pcms->vmport == ON_OFF_AUTO_AUTO) {
1241         pcms->vmport = (xen_enabled() || !pcms->i8042_enabled)
1242             ? ON_OFF_AUTO_OFF : ON_OFF_AUTO_ON;
1243     }
1244 
1245     /* Super I/O */
1246     pc_superio_init(isa_bus, create_fdctrl, pcms->i8042_enabled,
1247                     pcms->vmport != ON_OFF_AUTO_ON, &error_fatal);
1248 
1249     pcms->machine_done.notify = pc_machine_done;
1250     qemu_add_machine_init_done_notifier(&pcms->machine_done);
1251 }
1252 
1253 void pc_nic_init(PCMachineClass *pcmc, ISABus *isa_bus, PCIBus *pci_bus)
1254 {
1255     MachineClass *mc = MACHINE_CLASS(pcmc);
1256     bool default_is_ne2k = g_str_equal(mc->default_nic, TYPE_ISA_NE2000);
1257     NICInfo *nd;
1258 
1259     rom_set_order_override(FW_CFG_ORDER_OVERRIDE_NIC);
1260 
1261     while ((nd = qemu_find_nic_info(TYPE_ISA_NE2000, default_is_ne2k, NULL))) {
1262         pc_init_ne2k_isa(isa_bus, nd, &error_fatal);
1263     }
1264 
1265     /* Anything remaining should be a PCI NIC */
1266     if (pci_bus) {
1267         pci_init_nic_devices(pci_bus, mc->default_nic);
1268     }
1269 
1270     rom_reset_order_override();
1271 }
1272 
1273 void pc_i8259_create(ISABus *isa_bus, qemu_irq *i8259_irqs)
1274 {
1275     qemu_irq *i8259;
1276 
1277     if (kvm_pic_in_kernel()) {
1278         i8259 = kvm_i8259_init(isa_bus);
1279     } else if (xen_enabled()) {
1280         i8259 = xen_interrupt_controller_init();
1281     } else {
1282         i8259 = i8259_init(isa_bus, x86_allocate_cpu_irq());
1283     }
1284 
1285     for (size_t i = 0; i < ISA_NUM_IRQS; i++) {
1286         i8259_irqs[i] = i8259[i];
1287     }
1288 
1289     g_free(i8259);
1290 }
1291 
1292 static void pc_memory_pre_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
1293                                Error **errp)
1294 {
1295     const X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
1296     const MachineState *ms = MACHINE(hotplug_dev);
1297     const bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
1298     Error *local_err = NULL;
1299 
1300     /*
1301      * When "acpi=off" is used with the Q35 machine type, no ACPI is built,
1302      * but pcms->acpi_dev is still created. Check !acpi_enabled in
1303      * addition to cover this case.
1304      */
1305     if (!x86ms->acpi_dev || !x86_machine_is_acpi_enabled(x86ms)) {
1306         error_setg(errp,
1307                    "memory hotplug is not enabled: missing acpi device or acpi disabled");
1308         return;
1309     }
1310 
1311     if (is_nvdimm && !ms->nvdimms_state->is_enabled) {
1312         error_setg(errp, "nvdimm is not enabled: missing 'nvdimm' in '-M'");
1313         return;
1314     }
1315 
1316     hotplug_handler_pre_plug(x86ms->acpi_dev, dev, &local_err);
1317     if (local_err) {
1318         error_propagate(errp, local_err);
1319         return;
1320     }
1321 
1322     pc_dimm_pre_plug(PC_DIMM(dev), MACHINE(hotplug_dev), errp);
1323 }
1324 
1325 static void pc_memory_plug(HotplugHandler *hotplug_dev,
1326                            DeviceState *dev, Error **errp)
1327 {
1328     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1329     X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
1330     MachineState *ms = MACHINE(hotplug_dev);
1331     bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
1332 
1333     pc_dimm_plug(PC_DIMM(dev), MACHINE(pcms));
1334 
1335     if (is_nvdimm) {
1336         nvdimm_plug(ms->nvdimms_state);
1337     }
1338 
1339     hotplug_handler_plug(x86ms->acpi_dev, dev, &error_abort);
1340 }
1341 
1342 static void pc_memory_unplug_request(HotplugHandler *hotplug_dev,
1343                                      DeviceState *dev, Error **errp)
1344 {
1345     X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
1346 
1347     /*
1348      * When "acpi=off" is used with the Q35 machine type, no ACPI is built,
1349      * but pcms->acpi_dev is still created. Check !acpi_enabled in
1350      * addition to cover this case.
1351      */
1352     if (!x86ms->acpi_dev || !x86_machine_is_acpi_enabled(x86ms)) {
1353         error_setg(errp,
1354                    "memory hotplug is not enabled: missing acpi device or acpi disabled");
1355         return;
1356     }
1357 
1358     if (object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM)) {
1359         error_setg(errp, "nvdimm device hot unplug is not supported yet.");
1360         return;
1361     }
1362 
1363     hotplug_handler_unplug_request(x86ms->acpi_dev, dev,
1364                                    errp);
1365 }
1366 
1367 static void pc_memory_unplug(HotplugHandler *hotplug_dev,
1368                              DeviceState *dev, Error **errp)
1369 {
1370     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1371     X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
1372     Error *local_err = NULL;
1373 
1374     hotplug_handler_unplug(x86ms->acpi_dev, dev, &local_err);
1375     if (local_err) {
1376         goto out;
1377     }
1378 
1379     pc_dimm_unplug(PC_DIMM(dev), MACHINE(pcms));
1380     qdev_unrealize(dev);
1381  out:
1382     error_propagate(errp, local_err);
1383 }
1384 
1385 static void pc_hv_balloon_pre_plug(HotplugHandler *hotplug_dev,
1386                                    DeviceState *dev, Error **errp)
1387 {
1388     /* The vmbus handler has no hotplug handler; we should never end up here. */
1389     g_assert(!dev->hotplugged);
1390     memory_device_pre_plug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev), errp);
1391 }
1392 
1393 static void pc_hv_balloon_plug(HotplugHandler *hotplug_dev,
1394                                DeviceState *dev, Error **errp)
1395 {
1396     memory_device_plug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev));
1397 }
1398 
1399 static void pc_machine_device_pre_plug_cb(HotplugHandler *hotplug_dev,
1400                                           DeviceState *dev, Error **errp)
1401 {
1402     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1403         pc_memory_pre_plug(hotplug_dev, dev, errp);
1404     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1405         x86_cpu_pre_plug(hotplug_dev, dev, errp);
1406     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI)) {
1407         virtio_md_pci_pre_plug(VIRTIO_MD_PCI(dev), MACHINE(hotplug_dev), errp);
1408     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
1409         /* Declare the APIC range as the reserved MSI region */
1410         char *resv_prop_str = g_strdup_printf("0xfee00000:0xfeefffff:%d",
1411                                               VIRTIO_IOMMU_RESV_MEM_T_MSI);
1412         QList *reserved_regions = qlist_new();
1413 
1414         qlist_append_str(reserved_regions, resv_prop_str);
1415         qdev_prop_set_array(dev, "reserved-regions", reserved_regions);
1416 
1417         g_free(resv_prop_str);
1418     }
1419 
1420     if (object_dynamic_cast(OBJECT(dev), TYPE_X86_IOMMU_DEVICE) ||
1421         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
1422         PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1423 
1424         if (pcms->iommu) {
1425             error_setg(errp, "QEMU does not support multiple vIOMMUs "
1426                        "for x86 yet.");
1427             return;
1428         }
1429         pcms->iommu = dev;
1430     } else if (object_dynamic_cast(OBJECT(dev), TYPE_HV_BALLOON)) {
1431         pc_hv_balloon_pre_plug(hotplug_dev, dev, errp);
1432     }
1433 }
1434 
1435 static void pc_machine_device_plug_cb(HotplugHandler *hotplug_dev,
1436                                       DeviceState *dev, Error **errp)
1437 {
1438     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1439         pc_memory_plug(hotplug_dev, dev, errp);
1440     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1441         x86_cpu_plug(hotplug_dev, dev, errp);
1442     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI)) {
1443         virtio_md_pci_plug(VIRTIO_MD_PCI(dev), MACHINE(hotplug_dev), errp);
1444     } else if (object_dynamic_cast(OBJECT(dev), TYPE_HV_BALLOON)) {
1445         pc_hv_balloon_plug(hotplug_dev, dev, errp);
1446     }
1447 }
1448 
1449 static void pc_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
1450                                                 DeviceState *dev, Error **errp)
1451 {
1452     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1453         pc_memory_unplug_request(hotplug_dev, dev, errp);
1454     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1455         x86_cpu_unplug_request_cb(hotplug_dev, dev, errp);
1456     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI)) {
1457         virtio_md_pci_unplug_request(VIRTIO_MD_PCI(dev), MACHINE(hotplug_dev),
1458                                      errp);
1459     } else {
1460         error_setg(errp, "acpi: device unplug request for not supported device"
1461                    " type: %s", object_get_typename(OBJECT(dev)));
1462     }
1463 }
1464 
1465 static void pc_machine_device_unplug_cb(HotplugHandler *hotplug_dev,
1466                                         DeviceState *dev, Error **errp)
1467 {
1468     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1469         pc_memory_unplug(hotplug_dev, dev, errp);
1470     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1471         x86_cpu_unplug_cb(hotplug_dev, dev, errp);
1472     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI)) {
1473         virtio_md_pci_unplug(VIRTIO_MD_PCI(dev), MACHINE(hotplug_dev), errp);
1474     } else {
1475         error_setg(errp, "acpi: device unplug for not supported device"
1476                    " type: %s", object_get_typename(OBJECT(dev)));
1477     }
1478 }
1479 
1480 static HotplugHandler *pc_get_hotplug_handler(MachineState *machine,
1481                                              DeviceState *dev)
1482 {
1483     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) ||
1484         object_dynamic_cast(OBJECT(dev), TYPE_CPU) ||
1485         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI) ||
1486         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI) ||
1487         object_dynamic_cast(OBJECT(dev), TYPE_HV_BALLOON) ||
1488         object_dynamic_cast(OBJECT(dev), TYPE_X86_IOMMU_DEVICE)) {
1489         return HOTPLUG_HANDLER(machine);
1490     }
1491 
1492     return NULL;
1493 }
1494 
1495 static void pc_machine_get_vmport(Object *obj, Visitor *v, const char *name,
1496                                   void *opaque, Error **errp)
1497 {
1498     PCMachineState *pcms = PC_MACHINE(obj);
1499     OnOffAuto vmport = pcms->vmport;
1500 
1501     visit_type_OnOffAuto(v, name, &vmport, errp);
1502 }
1503 
1504 static void pc_machine_set_vmport(Object *obj, Visitor *v, const char *name,
1505                                   void *opaque, Error **errp)
1506 {
1507     PCMachineState *pcms = PC_MACHINE(obj);
1508 
1509     visit_type_OnOffAuto(v, name, &pcms->vmport, errp);
1510 }
1511 
1512 static bool pc_machine_get_fd_bootchk(Object *obj, Error **errp)
1513 {
1514     PCMachineState *pcms = PC_MACHINE(obj);
1515 
1516     return pcms->fd_bootchk;
1517 }
1518 
1519 static void pc_machine_set_fd_bootchk(Object *obj, bool value, Error **errp)
1520 {
1521     PCMachineState *pcms = PC_MACHINE(obj);
1522 
1523     pcms->fd_bootchk = value;
1524 }
1525 
1526 static bool pc_machine_get_smbus(Object *obj, Error **errp)
1527 {
1528     PCMachineState *pcms = PC_MACHINE(obj);
1529 
1530     return pcms->smbus_enabled;
1531 }
1532 
1533 static void pc_machine_set_smbus(Object *obj, bool value, Error **errp)
1534 {
1535     PCMachineState *pcms = PC_MACHINE(obj);
1536 
1537     pcms->smbus_enabled = value;
1538 }
1539 
1540 static bool pc_machine_get_sata(Object *obj, Error **errp)
1541 {
1542     PCMachineState *pcms = PC_MACHINE(obj);
1543 
1544     return pcms->sata_enabled;
1545 }
1546 
1547 static void pc_machine_set_sata(Object *obj, bool value, Error **errp)
1548 {
1549     PCMachineState *pcms = PC_MACHINE(obj);
1550 
1551     pcms->sata_enabled = value;
1552 }
1553 
1554 static bool pc_machine_get_hpet(Object *obj, Error **errp)
1555 {
1556     PCMachineState *pcms = PC_MACHINE(obj);
1557 
1558     return pcms->hpet_enabled;
1559 }
1560 
1561 static void pc_machine_set_hpet(Object *obj, bool value, Error **errp)
1562 {
1563     PCMachineState *pcms = PC_MACHINE(obj);
1564 
1565     pcms->hpet_enabled = value;
1566 }
1567 
1568 static bool pc_machine_get_i8042(Object *obj, Error **errp)
1569 {
1570     PCMachineState *pcms = PC_MACHINE(obj);
1571 
1572     return pcms->i8042_enabled;
1573 }
1574 
1575 static void pc_machine_set_i8042(Object *obj, bool value, Error **errp)
1576 {
1577     PCMachineState *pcms = PC_MACHINE(obj);
1578 
1579     pcms->i8042_enabled = value;
1580 }
1581 
1582 static bool pc_machine_get_default_bus_bypass_iommu(Object *obj, Error **errp)
1583 {
1584     PCMachineState *pcms = PC_MACHINE(obj);
1585 
1586     return pcms->default_bus_bypass_iommu;
1587 }
1588 
1589 static void pc_machine_set_default_bus_bypass_iommu(Object *obj, bool value,
1590                                                     Error **errp)
1591 {
1592     PCMachineState *pcms = PC_MACHINE(obj);
1593 
1594     pcms->default_bus_bypass_iommu = value;
1595 }
1596 
1597 static void pc_machine_get_smbios_ep(Object *obj, Visitor *v, const char *name,
1598                                      void *opaque, Error **errp)
1599 {
1600     PCMachineState *pcms = PC_MACHINE(obj);
1601     SmbiosEntryPointType smbios_entry_point_type = pcms->smbios_entry_point_type;
1602 
1603     visit_type_SmbiosEntryPointType(v, name, &smbios_entry_point_type, errp);
1604 }
1605 
1606 static void pc_machine_set_smbios_ep(Object *obj, Visitor *v, const char *name,
1607                                      void *opaque, Error **errp)
1608 {
1609     PCMachineState *pcms = PC_MACHINE(obj);
1610 
1611     visit_type_SmbiosEntryPointType(v, name, &pcms->smbios_entry_point_type, errp);
1612 }
1613 
1614 static void pc_machine_get_max_ram_below_4g(Object *obj, Visitor *v,
1615                                             const char *name, void *opaque,
1616                                             Error **errp)
1617 {
1618     PCMachineState *pcms = PC_MACHINE(obj);
1619     uint64_t value = pcms->max_ram_below_4g;
1620 
1621     visit_type_size(v, name, &value, errp);
1622 }
1623 
1624 static void pc_machine_set_max_ram_below_4g(Object *obj, Visitor *v,
1625                                             const char *name, void *opaque,
1626                                             Error **errp)
1627 {
1628     PCMachineState *pcms = PC_MACHINE(obj);
1629     uint64_t value;
1630 
1631     if (!visit_type_size(v, name, &value, errp)) {
1632         return;
1633     }
1634     if (value > 4 * GiB) {
1635         error_setg(errp,
1636                    "Machine option 'max-ram-below-4g=%"PRIu64
1637                    "' expects size less than or equal to 4G", value);
1638         return;
1639     }
1640 
1641     if (value < 1 * MiB) {
1642         warn_report("Only %" PRIu64 " bytes of RAM below the 4GiB boundary,"
1643                     "BIOS may not work with less than 1MiB", value);
1644     }
1645 
1646     pcms->max_ram_below_4g = value;
1647 }
1648 
1649 static void pc_machine_get_max_fw_size(Object *obj, Visitor *v,
1650                                        const char *name, void *opaque,
1651                                        Error **errp)
1652 {
1653     PCMachineState *pcms = PC_MACHINE(obj);
1654     uint64_t value = pcms->max_fw_size;
1655 
1656     visit_type_size(v, name, &value, errp);
1657 }
1658 
1659 static void pc_machine_set_max_fw_size(Object *obj, Visitor *v,
1660                                        const char *name, void *opaque,
1661                                        Error **errp)
1662 {
1663     PCMachineState *pcms = PC_MACHINE(obj);
1664     uint64_t value;
1665 
1666     if (!visit_type_size(v, name, &value, errp)) {
1667         return;
1668     }
1669 
1670     /*
1671      * We don't have a theoretically justifiable exact lower bound on the base
1672      * address of any flash mapping. In practice, the IO-APIC MMIO range is
1673      * [0xFEE00000..0xFEE01000] -- see IO_APIC_DEFAULT_ADDRESS --, leaving free
1674      * only 18MiB-4KiB below 4GiB. For now, restrict the cumulative mapping to
1675      * 16MiB in size.
1676      */
1677     if (value > 16 * MiB) {
1678         error_setg(errp,
1679                    "User specified max allowed firmware size %" PRIu64 " is "
1680                    "greater than 16MiB. If combined firmware size exceeds "
1681                    "16MiB the system may not boot, or experience intermittent"
1682                    "stability issues.",
1683                    value);
1684         return;
1685     }
1686 
1687     pcms->max_fw_size = value;
1688 }
1689 
1690 
1691 static void pc_machine_initfn(Object *obj)
1692 {
1693     PCMachineState *pcms = PC_MACHINE(obj);
1694     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1695 
1696 #ifdef CONFIG_VMPORT
1697     pcms->vmport = ON_OFF_AUTO_AUTO;
1698 #else
1699     pcms->vmport = ON_OFF_AUTO_OFF;
1700 #endif /* CONFIG_VMPORT */
1701     pcms->max_ram_below_4g = 0; /* use default */
1702     pcms->smbios_entry_point_type = pcmc->default_smbios_ep_type;
1703     pcms->south_bridge = pcmc->default_south_bridge;
1704 
1705     /* acpi build is enabled by default if machine supports it */
1706     pcms->acpi_build_enabled = pcmc->has_acpi_build;
1707     pcms->smbus_enabled = true;
1708     pcms->sata_enabled = true;
1709     pcms->i8042_enabled = true;
1710     pcms->max_fw_size = 8 * MiB;
1711 #if defined(CONFIG_HPET)
1712     pcms->hpet_enabled = true;
1713 #endif
1714     pcms->fd_bootchk = true;
1715     pcms->default_bus_bypass_iommu = false;
1716 
1717     pc_system_flash_create(pcms);
1718     pcms->pcspk = isa_new(TYPE_PC_SPEAKER);
1719     object_property_add_alias(OBJECT(pcms), "pcspk-audiodev",
1720                               OBJECT(pcms->pcspk), "audiodev");
1721     if (pcmc->pci_enabled) {
1722         cxl_machine_init(obj, &pcms->cxl_devices_state);
1723     }
1724 }
1725 
1726 static void pc_machine_reset(MachineState *machine, ResetType type)
1727 {
1728     CPUState *cs;
1729     X86CPU *cpu;
1730 
1731     qemu_devices_reset(type);
1732 
1733     /* Reset APIC after devices have been reset to cancel
1734      * any changes that qemu_devices_reset() might have done.
1735      */
1736     CPU_FOREACH(cs) {
1737         cpu = X86_CPU(cs);
1738 
1739         x86_cpu_after_reset(cpu);
1740     }
1741 }
1742 
1743 static void pc_machine_wakeup(MachineState *machine)
1744 {
1745     cpu_synchronize_all_states();
1746     pc_machine_reset(machine, RESET_TYPE_WAKEUP);
1747     cpu_synchronize_all_post_reset();
1748 }
1749 
1750 static bool pc_hotplug_allowed(MachineState *ms, DeviceState *dev, Error **errp)
1751 {
1752     X86IOMMUState *iommu = x86_iommu_get_default();
1753     IntelIOMMUState *intel_iommu;
1754 
1755     if (iommu &&
1756         object_dynamic_cast((Object *)iommu, TYPE_INTEL_IOMMU_DEVICE) &&
1757         object_dynamic_cast((Object *)dev, "vfio-pci")) {
1758         intel_iommu = INTEL_IOMMU_DEVICE(iommu);
1759         if (!intel_iommu->caching_mode) {
1760             error_setg(errp, "Device assignment is not allowed without "
1761                        "enabling caching-mode=on for Intel IOMMU.");
1762             return false;
1763         }
1764     }
1765 
1766     return true;
1767 }
1768 
1769 static void pc_machine_class_init(ObjectClass *oc, const void *data)
1770 {
1771     MachineClass *mc = MACHINE_CLASS(oc);
1772     X86MachineClass *x86mc = X86_MACHINE_CLASS(oc);
1773     PCMachineClass *pcmc = PC_MACHINE_CLASS(oc);
1774     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
1775 
1776     pcmc->pci_enabled = true;
1777     pcmc->has_acpi_build = true;
1778     pcmc->smbios_defaults = true;
1779     pcmc->gigabyte_align = true;
1780     pcmc->has_reserved_memory = true;
1781     pcmc->enforce_amd_1tb_hole = true;
1782     pcmc->isa_bios_alias = true;
1783     pcmc->pvh_enabled = true;
1784     pcmc->kvmclock_create_always = true;
1785     x86mc->apic_xrupt_override = true;
1786     assert(!mc->get_hotplug_handler);
1787     mc->get_hotplug_handler = pc_get_hotplug_handler;
1788     mc->hotplug_allowed = pc_hotplug_allowed;
1789     mc->auto_enable_numa_with_memhp = true;
1790     mc->auto_enable_numa_with_memdev = true;
1791     mc->has_hotpluggable_cpus = true;
1792     mc->default_boot_order = "cad";
1793     mc->block_default_type = IF_IDE;
1794     mc->max_cpus = 255;
1795     mc->reset = pc_machine_reset;
1796     mc->wakeup = pc_machine_wakeup;
1797     hc->pre_plug = pc_machine_device_pre_plug_cb;
1798     hc->plug = pc_machine_device_plug_cb;
1799     hc->unplug_request = pc_machine_device_unplug_request_cb;
1800     hc->unplug = pc_machine_device_unplug_cb;
1801     mc->default_cpu_type = TARGET_DEFAULT_CPU_TYPE;
1802     mc->nvdimm_supported = true;
1803     mc->smp_props.dies_supported = true;
1804     mc->smp_props.modules_supported = true;
1805     mc->smp_props.cache_supported[CACHE_LEVEL_AND_TYPE_L1D] = true;
1806     mc->smp_props.cache_supported[CACHE_LEVEL_AND_TYPE_L1I] = true;
1807     mc->smp_props.cache_supported[CACHE_LEVEL_AND_TYPE_L2] = true;
1808     mc->smp_props.cache_supported[CACHE_LEVEL_AND_TYPE_L3] = true;
1809     mc->default_ram_id = "pc.ram";
1810     pcmc->default_smbios_ep_type = SMBIOS_ENTRY_POINT_TYPE_AUTO;
1811 
1812     object_class_property_add(oc, PC_MACHINE_MAX_RAM_BELOW_4G, "size",
1813         pc_machine_get_max_ram_below_4g, pc_machine_set_max_ram_below_4g,
1814         NULL, NULL);
1815     object_class_property_set_description(oc, PC_MACHINE_MAX_RAM_BELOW_4G,
1816         "Maximum ram below the 4G boundary (32bit boundary)");
1817 
1818     object_class_property_add(oc, PC_MACHINE_VMPORT, "OnOffAuto",
1819         pc_machine_get_vmport, pc_machine_set_vmport,
1820         NULL, NULL);
1821     object_class_property_set_description(oc, PC_MACHINE_VMPORT,
1822         "Enable vmport (pc & q35)");
1823 
1824     object_class_property_add_bool(oc, PC_MACHINE_SMBUS,
1825         pc_machine_get_smbus, pc_machine_set_smbus);
1826     object_class_property_set_description(oc, PC_MACHINE_SMBUS,
1827         "Enable/disable system management bus");
1828 
1829     object_class_property_add_bool(oc, PC_MACHINE_SATA,
1830         pc_machine_get_sata, pc_machine_set_sata);
1831     object_class_property_set_description(oc, PC_MACHINE_SATA,
1832         "Enable/disable Serial ATA bus");
1833 
1834     object_class_property_add_bool(oc, "hpet",
1835         pc_machine_get_hpet, pc_machine_set_hpet);
1836     object_class_property_set_description(oc, "hpet",
1837         "Enable/disable high precision event timer emulation");
1838 
1839     object_class_property_add_bool(oc, PC_MACHINE_I8042,
1840         pc_machine_get_i8042, pc_machine_set_i8042);
1841     object_class_property_set_description(oc, PC_MACHINE_I8042,
1842         "Enable/disable Intel 8042 PS/2 controller emulation");
1843 
1844     object_class_property_add_bool(oc, "default-bus-bypass-iommu",
1845         pc_machine_get_default_bus_bypass_iommu,
1846         pc_machine_set_default_bus_bypass_iommu);
1847 
1848     object_class_property_add(oc, PC_MACHINE_MAX_FW_SIZE, "size",
1849         pc_machine_get_max_fw_size, pc_machine_set_max_fw_size,
1850         NULL, NULL);
1851     object_class_property_set_description(oc, PC_MACHINE_MAX_FW_SIZE,
1852         "Maximum combined firmware size");
1853 
1854     object_class_property_add(oc, PC_MACHINE_SMBIOS_EP, "str",
1855         pc_machine_get_smbios_ep, pc_machine_set_smbios_ep,
1856         NULL, NULL);
1857     object_class_property_set_description(oc, PC_MACHINE_SMBIOS_EP,
1858         "SMBIOS Entry Point type [32, 64]");
1859 
1860     object_class_property_add_bool(oc, "fd-bootchk",
1861         pc_machine_get_fd_bootchk,
1862         pc_machine_set_fd_bootchk);
1863 }
1864 
1865 static const TypeInfo pc_machine_info = {
1866     .name = TYPE_PC_MACHINE,
1867     .parent = TYPE_X86_MACHINE,
1868     .abstract = true,
1869     .instance_size = sizeof(PCMachineState),
1870     .instance_init = pc_machine_initfn,
1871     .class_size = sizeof(PCMachineClass),
1872     .class_init = pc_machine_class_init,
1873     .interfaces = (const InterfaceInfo[]) {
1874          { TYPE_HOTPLUG_HANDLER },
1875          { }
1876     },
1877 };
1878 
1879 static void pc_machine_register_types(void)
1880 {
1881     type_register_static(&pc_machine_info);
1882 }
1883 
1884 type_init(pc_machine_register_types)
1885