xref: /qemu/hw/i386/pc.c (revision 4d3ad3c3ba1f1e9c217d0581e4913a59ef2ac15f)
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/pc.h"
28 #include "hw/char/serial-isa.h"
29 #include "hw/char/parallel.h"
30 #include "hw/hyperv/hv-balloon.h"
31 #include "hw/i386/fw_cfg.h"
32 #include "hw/i386/vmport.h"
33 #include "system/cpus.h"
34 #include "hw/ide/ide-bus.h"
35 #include "hw/timer/hpet.h"
36 #include "hw/loader.h"
37 #include "hw/rtc/mc146818rtc.h"
38 #include "hw/intc/i8259.h"
39 #include "hw/timer/i8254.h"
40 #include "hw/input/i8042.h"
41 #include "hw/audio/pcspk.h"
42 #include "system/system.h"
43 #include "system/xen.h"
44 #include "system/reset.h"
45 #include "kvm/kvm_i386.h"
46 #include "hw/xen/xen.h"
47 #include "qobject/qlist.h"
48 #include "qemu/error-report.h"
49 #include "hw/acpi/cpu_hotplug.h"
50 #include "acpi-build.h"
51 #include "hw/mem/nvdimm.h"
52 #include "hw/cxl/cxl_host.h"
53 #include "hw/usb.h"
54 #include "hw/i386/intel_iommu.h"
55 #include "hw/net/ne2000-isa.h"
56 #include "hw/virtio/virtio-iommu.h"
57 #include "hw/virtio/virtio-md-pci.h"
58 #include "hw/i386/kvm/xen_overlay.h"
59 #include "hw/i386/kvm/xen_evtchn.h"
60 #include "hw/i386/kvm/xen_gnttab.h"
61 #include "hw/i386/kvm/xen_xenstore.h"
62 #include "hw/mem/memory-device.h"
63 #include "e820_memory_layout.h"
64 #include "trace.h"
65 #include "sev.h"
66 #include CONFIG_DEVICES
67 
68 #ifdef CONFIG_XEN_EMU
69 #include "hw/xen/xen-legacy-backend.h"
70 #include "hw/xen/xen-bus.h"
71 #endif
72 
73 /*
74  * Helper for setting model-id for CPU models that changed model-id
75  * depending on QEMU versions up to QEMU 2.4.
76  */
77 #define PC_CPU_MODEL_IDS(v) \
78     { "qemu32-" TYPE_X86_CPU, "model-id", "QEMU Virtual CPU version " v, },\
79     { "qemu64-" TYPE_X86_CPU, "model-id", "QEMU Virtual CPU version " v, },\
80     { "athlon-" TYPE_X86_CPU, "model-id", "QEMU Virtual CPU version " v, },
81 
82 GlobalProperty pc_compat_10_0[] = {};
83 const size_t pc_compat_10_0_len = G_N_ELEMENTS(pc_compat_10_0);
84 
85 GlobalProperty pc_compat_9_2[] = {};
86 const size_t pc_compat_9_2_len = G_N_ELEMENTS(pc_compat_9_2);
87 
88 GlobalProperty pc_compat_9_1[] = {
89     { "ICH9-LPC", "x-smi-swsmi-timer", "off" },
90     { "ICH9-LPC", "x-smi-periodic-timer", "off" },
91     { TYPE_INTEL_IOMMU_DEVICE, "stale-tm", "on" },
92     { TYPE_INTEL_IOMMU_DEVICE, "aw-bits", "39" },
93 };
94 const size_t pc_compat_9_1_len = G_N_ELEMENTS(pc_compat_9_1);
95 
96 GlobalProperty pc_compat_9_0[] = {
97     { TYPE_X86_CPU, "x-amd-topoext-features-only", "false" },
98     { TYPE_X86_CPU, "x-l1-cache-per-thread", "false" },
99     { TYPE_X86_CPU, "guest-phys-bits", "0" },
100     { "sev-guest", "legacy-vm-type", "on" },
101     { TYPE_X86_CPU, "legacy-multi-node", "on" },
102 };
103 const size_t pc_compat_9_0_len = G_N_ELEMENTS(pc_compat_9_0);
104 
105 GlobalProperty pc_compat_8_2[] = {};
106 const size_t pc_compat_8_2_len = G_N_ELEMENTS(pc_compat_8_2);
107 
108 GlobalProperty pc_compat_8_1[] = {};
109 const size_t pc_compat_8_1_len = G_N_ELEMENTS(pc_compat_8_1);
110 
111 GlobalProperty pc_compat_8_0[] = {
112     { "virtio-mem", "unplugged-inaccessible", "auto" },
113 };
114 const size_t pc_compat_8_0_len = G_N_ELEMENTS(pc_compat_8_0);
115 
116 GlobalProperty pc_compat_7_2[] = {
117     { "ICH9-LPC", "noreboot", "true" },
118 };
119 const size_t pc_compat_7_2_len = G_N_ELEMENTS(pc_compat_7_2);
120 
121 GlobalProperty pc_compat_7_1[] = {};
122 const size_t pc_compat_7_1_len = G_N_ELEMENTS(pc_compat_7_1);
123 
124 GlobalProperty pc_compat_7_0[] = {};
125 const size_t pc_compat_7_0_len = G_N_ELEMENTS(pc_compat_7_0);
126 
127 GlobalProperty pc_compat_6_2[] = {
128     { "virtio-mem", "unplugged-inaccessible", "off" },
129 };
130 const size_t pc_compat_6_2_len = G_N_ELEMENTS(pc_compat_6_2);
131 
132 GlobalProperty pc_compat_6_1[] = {
133     { TYPE_X86_CPU, "hv-version-id-build", "0x1bbc" },
134     { TYPE_X86_CPU, "hv-version-id-major", "0x0006" },
135     { TYPE_X86_CPU, "hv-version-id-minor", "0x0001" },
136     { "ICH9-LPC", "x-keep-pci-slot-hpc", "false" },
137 };
138 const size_t pc_compat_6_1_len = G_N_ELEMENTS(pc_compat_6_1);
139 
140 GlobalProperty pc_compat_6_0[] = {
141     { "qemu64" "-" TYPE_X86_CPU, "family", "6" },
142     { "qemu64" "-" TYPE_X86_CPU, "model", "6" },
143     { "qemu64" "-" TYPE_X86_CPU, "stepping", "3" },
144     { TYPE_X86_CPU, "x-vendor-cpuid-only", "off" },
145     { "ICH9-LPC", ACPI_PM_PROP_ACPI_PCIHP_BRIDGE, "off" },
146     { "ICH9-LPC", "x-keep-pci-slot-hpc", "true" },
147 };
148 const size_t pc_compat_6_0_len = G_N_ELEMENTS(pc_compat_6_0);
149 
150 GlobalProperty pc_compat_5_2[] = {
151     { "ICH9-LPC", "x-smi-cpu-hotunplug", "off" },
152 };
153 const size_t pc_compat_5_2_len = G_N_ELEMENTS(pc_compat_5_2);
154 
155 GlobalProperty pc_compat_5_1[] = {
156     { "ICH9-LPC", "x-smi-cpu-hotplug", "off" },
157     { TYPE_X86_CPU, "kvm-msi-ext-dest-id", "off" },
158 };
159 const size_t pc_compat_5_1_len = G_N_ELEMENTS(pc_compat_5_1);
160 
161 GlobalProperty pc_compat_5_0[] = {
162 };
163 const size_t pc_compat_5_0_len = G_N_ELEMENTS(pc_compat_5_0);
164 
165 GlobalProperty pc_compat_4_2[] = {
166     { "mch", "smbase-smram", "off" },
167 };
168 const size_t pc_compat_4_2_len = G_N_ELEMENTS(pc_compat_4_2);
169 
170 GlobalProperty pc_compat_4_1[] = {};
171 const size_t pc_compat_4_1_len = G_N_ELEMENTS(pc_compat_4_1);
172 
173 GlobalProperty pc_compat_4_0[] = {};
174 const size_t pc_compat_4_0_len = G_N_ELEMENTS(pc_compat_4_0);
175 
176 GlobalProperty pc_compat_3_1[] = {
177     { "intel-iommu", "dma-drain", "off" },
178     { "Opteron_G3" "-" TYPE_X86_CPU, "rdtscp", "off" },
179     { "Opteron_G4" "-" TYPE_X86_CPU, "rdtscp", "off" },
180     { "Opteron_G4" "-" TYPE_X86_CPU, "npt", "off" },
181     { "Opteron_G4" "-" TYPE_X86_CPU, "nrip-save", "off" },
182     { "Opteron_G5" "-" TYPE_X86_CPU, "rdtscp", "off" },
183     { "Opteron_G5" "-" TYPE_X86_CPU, "npt", "off" },
184     { "Opteron_G5" "-" TYPE_X86_CPU, "nrip-save", "off" },
185     { "EPYC" "-" TYPE_X86_CPU, "npt", "off" },
186     { "EPYC" "-" TYPE_X86_CPU, "nrip-save", "off" },
187     { "EPYC-IBPB" "-" TYPE_X86_CPU, "npt", "off" },
188     { "EPYC-IBPB" "-" TYPE_X86_CPU, "nrip-save", "off" },
189     { "Skylake-Client" "-" TYPE_X86_CPU,      "mpx", "on" },
190     { "Skylake-Client-IBRS" "-" TYPE_X86_CPU, "mpx", "on" },
191     { "Skylake-Server" "-" TYPE_X86_CPU,      "mpx", "on" },
192     { "Skylake-Server-IBRS" "-" TYPE_X86_CPU, "mpx", "on" },
193     { "Cascadelake-Server" "-" TYPE_X86_CPU,  "mpx", "on" },
194     { "Icelake-Client" "-" TYPE_X86_CPU,      "mpx", "on" },
195     { "Icelake-Server" "-" TYPE_X86_CPU,      "mpx", "on" },
196     { "Cascadelake-Server" "-" TYPE_X86_CPU, "stepping", "5" },
197     { TYPE_X86_CPU, "x-intel-pt-auto-level", "off" },
198 };
199 const size_t pc_compat_3_1_len = G_N_ELEMENTS(pc_compat_3_1);
200 
201 GlobalProperty pc_compat_3_0[] = {
202     { TYPE_X86_CPU, "x-hv-synic-kvm-only", "on" },
203     { "Skylake-Server" "-" TYPE_X86_CPU, "pku", "off" },
204     { "Skylake-Server-IBRS" "-" TYPE_X86_CPU, "pku", "off" },
205 };
206 const size_t pc_compat_3_0_len = G_N_ELEMENTS(pc_compat_3_0);
207 
208 GlobalProperty pc_compat_2_12[] = {
209     { TYPE_X86_CPU, "legacy-cache", "on" },
210     { TYPE_X86_CPU, "topoext", "off" },
211     { "EPYC-" TYPE_X86_CPU, "xlevel", "0x8000000a" },
212     { "EPYC-IBPB-" TYPE_X86_CPU, "xlevel", "0x8000000a" },
213 };
214 const size_t pc_compat_2_12_len = G_N_ELEMENTS(pc_compat_2_12);
215 
216 GlobalProperty pc_compat_2_11[] = {
217     { TYPE_X86_CPU, "x-migrate-smi-count", "off" },
218     { "Skylake-Server" "-" TYPE_X86_CPU, "clflushopt", "off" },
219 };
220 const size_t pc_compat_2_11_len = G_N_ELEMENTS(pc_compat_2_11);
221 
222 GlobalProperty pc_compat_2_10[] = {
223     { TYPE_X86_CPU, "x-hv-max-vps", "0x40" },
224     { "i440FX-pcihost", "x-pci-hole64-fix", "off" },
225     { "q35-pcihost", "x-pci-hole64-fix", "off" },
226 };
227 const size_t pc_compat_2_10_len = G_N_ELEMENTS(pc_compat_2_10);
228 
229 GlobalProperty pc_compat_2_9[] = {
230     { "mch", "extended-tseg-mbytes", "0" },
231 };
232 const size_t pc_compat_2_9_len = G_N_ELEMENTS(pc_compat_2_9);
233 
234 GlobalProperty pc_compat_2_8[] = {
235     { TYPE_X86_CPU, "tcg-cpuid", "off" },
236     { "kvmclock", "x-mach-use-reliable-get-clock", "off" },
237     { "ICH9-LPC", "x-smi-broadcast", "off" },
238     { TYPE_X86_CPU, "vmware-cpuid-freq", "off" },
239     { "Haswell-" TYPE_X86_CPU, "stepping", "1" },
240 };
241 const size_t pc_compat_2_8_len = G_N_ELEMENTS(pc_compat_2_8);
242 
243 GlobalProperty pc_compat_2_7[] = {
244     { TYPE_X86_CPU, "l3-cache", "off" },
245     { TYPE_X86_CPU, "full-cpuid-auto-level", "off" },
246     { "Opteron_G3" "-" TYPE_X86_CPU, "family", "15" },
247     { "Opteron_G3" "-" TYPE_X86_CPU, "model", "6" },
248     { "Opteron_G3" "-" TYPE_X86_CPU, "stepping", "1" },
249     { "isa-pcspk", "migrate", "off" },
250 };
251 const size_t pc_compat_2_7_len = G_N_ELEMENTS(pc_compat_2_7);
252 
253 GlobalProperty pc_compat_2_6[] = {
254     { TYPE_X86_CPU, "cpuid-0xb", "off" },
255     { "vmxnet3", "romfile", "" },
256     { TYPE_X86_CPU, "fill-mtrr-mask", "off" },
257     { "apic-common", "legacy-instance-id", "on", }
258 };
259 const size_t pc_compat_2_6_len = G_N_ELEMENTS(pc_compat_2_6);
260 
261 GlobalProperty pc_compat_2_5[] = {};
262 const size_t pc_compat_2_5_len = G_N_ELEMENTS(pc_compat_2_5);
263 
264 GlobalProperty pc_compat_2_4[] = {
265     PC_CPU_MODEL_IDS("2.4.0")
266     { "Haswell-" TYPE_X86_CPU, "abm", "off" },
267     { "Haswell-noTSX-" TYPE_X86_CPU, "abm", "off" },
268     { "Broadwell-" TYPE_X86_CPU, "abm", "off" },
269     { "Broadwell-noTSX-" TYPE_X86_CPU, "abm", "off" },
270     { "host" "-" TYPE_X86_CPU, "host-cache-info", "on" },
271     { TYPE_X86_CPU, "check", "off" },
272     { "qemu64" "-" TYPE_X86_CPU, "sse4a", "on" },
273     { "qemu64" "-" TYPE_X86_CPU, "abm", "on" },
274     { "qemu64" "-" TYPE_X86_CPU, "popcnt", "on" },
275     { "qemu32" "-" TYPE_X86_CPU, "popcnt", "on" },
276     { "Opteron_G2" "-" TYPE_X86_CPU, "rdtscp", "on" },
277     { "Opteron_G3" "-" TYPE_X86_CPU, "rdtscp", "on" },
278     { "Opteron_G4" "-" TYPE_X86_CPU, "rdtscp", "on" },
279     { "Opteron_G5" "-" TYPE_X86_CPU, "rdtscp", "on", }
280 };
281 const size_t pc_compat_2_4_len = G_N_ELEMENTS(pc_compat_2_4);
282 
283 /*
284  * @PC_FW_DATA:
285  * Size of the chunk of memory at the top of RAM for the BIOS ACPI tables
286  * and other BIOS datastructures.
287  *
288  * BIOS ACPI tables: 128K. Other BIOS datastructures: less than 4K
289  * reported to be used at the moment, 32K should be enough for a while.
290  */
291 #define PC_FW_DATA (0x20000 + 0x8000)
292 
293 GSIState *pc_gsi_create(qemu_irq **irqs, bool pci_enabled)
294 {
295     GSIState *s;
296 
297     s = g_new0(GSIState, 1);
298     if (kvm_ioapic_in_kernel()) {
299         kvm_pc_setup_irq_routing(pci_enabled);
300     }
301     *irqs = qemu_allocate_irqs(gsi_handler, s, IOAPIC_NUM_PINS);
302 
303     return s;
304 }
305 
306 static void ioport80_write(void *opaque, hwaddr addr, uint64_t data,
307                            unsigned size)
308 {
309 }
310 
311 static uint64_t ioport80_read(void *opaque, hwaddr addr, unsigned size)
312 {
313     return 0xffffffffffffffffULL;
314 }
315 
316 /* MS-DOS compatibility mode FPU exception support */
317 static void ioportF0_write(void *opaque, hwaddr addr, uint64_t data,
318                            unsigned size)
319 {
320     if (tcg_enabled()) {
321         cpu_set_ignne();
322     }
323 }
324 
325 static uint64_t ioportF0_read(void *opaque, hwaddr addr, unsigned size)
326 {
327     return 0xffffffffffffffffULL;
328 }
329 
330 /* PC cmos mappings */
331 
332 #define REG_EQUIPMENT_BYTE          0x14
333 
334 static void cmos_init_hd(MC146818RtcState *s, int type_ofs, int info_ofs,
335                          int16_t cylinders, int8_t heads, int8_t sectors)
336 {
337     mc146818rtc_set_cmos_data(s, type_ofs, 47);
338     mc146818rtc_set_cmos_data(s, info_ofs, cylinders);
339     mc146818rtc_set_cmos_data(s, info_ofs + 1, cylinders >> 8);
340     mc146818rtc_set_cmos_data(s, info_ofs + 2, heads);
341     mc146818rtc_set_cmos_data(s, info_ofs + 3, 0xff);
342     mc146818rtc_set_cmos_data(s, info_ofs + 4, 0xff);
343     mc146818rtc_set_cmos_data(s, info_ofs + 5, 0xc0 | ((heads > 8) << 3));
344     mc146818rtc_set_cmos_data(s, info_ofs + 6, cylinders);
345     mc146818rtc_set_cmos_data(s, info_ofs + 7, cylinders >> 8);
346     mc146818rtc_set_cmos_data(s, info_ofs + 8, sectors);
347 }
348 
349 /* convert boot_device letter to something recognizable by the bios */
350 static int boot_device2nibble(char boot_device)
351 {
352     switch(boot_device) {
353     case 'a':
354     case 'b':
355         return 0x01; /* floppy boot */
356     case 'c':
357         return 0x02; /* hard drive boot */
358     case 'd':
359         return 0x03; /* CD-ROM boot */
360     case 'n':
361         return 0x04; /* Network boot */
362     }
363     return 0;
364 }
365 
366 static void set_boot_dev(PCMachineState *pcms, MC146818RtcState *s,
367                          const char *boot_device, Error **errp)
368 {
369 #define PC_MAX_BOOT_DEVICES 3
370     int nbds, bds[3] = { 0, };
371     int i;
372 
373     nbds = strlen(boot_device);
374     if (nbds > PC_MAX_BOOT_DEVICES) {
375         error_setg(errp, "Too many boot devices for PC");
376         return;
377     }
378     for (i = 0; i < nbds; i++) {
379         bds[i] = boot_device2nibble(boot_device[i]);
380         if (bds[i] == 0) {
381             error_setg(errp, "Invalid boot device for PC: '%c'",
382                        boot_device[i]);
383             return;
384         }
385     }
386     mc146818rtc_set_cmos_data(s, 0x3d, (bds[1] << 4) | bds[0]);
387     mc146818rtc_set_cmos_data(s, 0x38, (bds[2] << 4) | !pcms->fd_bootchk);
388 }
389 
390 static void pc_boot_set(void *opaque, const char *boot_device, Error **errp)
391 {
392     PCMachineState *pcms = opaque;
393     X86MachineState *x86ms = X86_MACHINE(pcms);
394 
395     set_boot_dev(pcms, MC146818_RTC(x86ms->rtc), boot_device, errp);
396 }
397 
398 static void pc_cmos_init_floppy(MC146818RtcState *rtc_state, ISADevice *floppy)
399 {
400     int val, nb;
401     FloppyDriveType fd_type[2] = { FLOPPY_DRIVE_TYPE_NONE,
402                                    FLOPPY_DRIVE_TYPE_NONE };
403 
404 #ifdef CONFIG_FDC_ISA
405     /* floppy type */
406     if (floppy) {
407         for (int i = 0; i < 2; i++) {
408             fd_type[i] = isa_fdc_get_drive_type(floppy, i);
409         }
410     }
411 #endif
412 
413     val = (cmos_get_fd_drive_type(fd_type[0]) << 4) |
414         cmos_get_fd_drive_type(fd_type[1]);
415     mc146818rtc_set_cmos_data(rtc_state, 0x10, val);
416 
417     val = mc146818rtc_get_cmos_data(rtc_state, REG_EQUIPMENT_BYTE);
418     nb = 0;
419     if (fd_type[0] != FLOPPY_DRIVE_TYPE_NONE) {
420         nb++;
421     }
422     if (fd_type[1] != FLOPPY_DRIVE_TYPE_NONE) {
423         nb++;
424     }
425     switch (nb) {
426     case 0:
427         break;
428     case 1:
429         val |= 0x01; /* 1 drive, ready for boot */
430         break;
431     case 2:
432         val |= 0x41; /* 2 drives, ready for boot */
433         break;
434     }
435     mc146818rtc_set_cmos_data(rtc_state, REG_EQUIPMENT_BYTE, val);
436 }
437 
438 typedef struct check_fdc_state {
439     ISADevice *floppy;
440     bool multiple;
441 } CheckFdcState;
442 
443 static int check_fdc(Object *obj, void *opaque)
444 {
445     CheckFdcState *state = opaque;
446     Object *fdc;
447     uint32_t iobase;
448     Error *local_err = NULL;
449 
450     fdc = object_dynamic_cast(obj, TYPE_ISA_FDC);
451     if (!fdc) {
452         return 0;
453     }
454 
455     iobase = object_property_get_uint(obj, "iobase", &local_err);
456     if (local_err || iobase != 0x3f0) {
457         error_free(local_err);
458         return 0;
459     }
460 
461     if (state->floppy) {
462         state->multiple = true;
463     } else {
464         state->floppy = ISA_DEVICE(obj);
465     }
466     return 0;
467 }
468 
469 static const char * const fdc_container_path[] = {
470     "unattached", "peripheral", "peripheral-anon"
471 };
472 
473 /*
474  * Locate the FDC at IO address 0x3f0, in order to configure the CMOS registers
475  * and ACPI objects.
476  */
477 static ISADevice *pc_find_fdc0(void)
478 {
479     int i;
480     Object *container;
481     CheckFdcState state = { 0 };
482 
483     for (i = 0; i < ARRAY_SIZE(fdc_container_path); i++) {
484         container = machine_get_container(fdc_container_path[i]);
485         object_child_foreach(container, check_fdc, &state);
486     }
487 
488     if (state.multiple) {
489         warn_report("multiple floppy disk controllers with "
490                     "iobase=0x3f0 have been found");
491         error_printf("the one being picked for CMOS setup might not reflect "
492                      "your intent");
493     }
494 
495     return state.floppy;
496 }
497 
498 static void pc_cmos_init_late(PCMachineState *pcms)
499 {
500     X86MachineState *x86ms = X86_MACHINE(pcms);
501     MC146818RtcState *s = MC146818_RTC(x86ms->rtc);
502     int16_t cylinders;
503     int8_t heads, sectors;
504     int val;
505     int i, trans;
506 
507     val = 0;
508     if (pcms->idebus[0] &&
509         ide_get_geometry(pcms->idebus[0], 0,
510                          &cylinders, &heads, &sectors) >= 0) {
511         cmos_init_hd(s, 0x19, 0x1b, cylinders, heads, sectors);
512         val |= 0xf0;
513     }
514     if (pcms->idebus[0] &&
515         ide_get_geometry(pcms->idebus[0], 1,
516                          &cylinders, &heads, &sectors) >= 0) {
517         cmos_init_hd(s, 0x1a, 0x24, cylinders, heads, sectors);
518         val |= 0x0f;
519     }
520     mc146818rtc_set_cmos_data(s, 0x12, val);
521 
522     val = 0;
523     for (i = 0; i < 4; i++) {
524         /* NOTE: ide_get_geometry() returns the physical
525            geometry.  It is always such that: 1 <= sects <= 63, 1
526            <= heads <= 16, 1 <= cylinders <= 16383. The BIOS
527            geometry can be different if a translation is done. */
528         BusState *idebus = pcms->idebus[i / 2];
529         if (idebus &&
530             ide_get_geometry(idebus, i % 2,
531                              &cylinders, &heads, &sectors) >= 0) {
532             trans = ide_get_bios_chs_trans(idebus, i % 2) - 1;
533             assert((trans & ~3) == 0);
534             val |= trans << (i * 2);
535         }
536     }
537     mc146818rtc_set_cmos_data(s, 0x39, val);
538 
539     pc_cmos_init_floppy(s, pc_find_fdc0());
540 
541     /* various important CMOS locations needed by PC/Bochs bios */
542 
543     /* memory size */
544     /* base memory (first MiB) */
545     val = MIN(x86ms->below_4g_mem_size / KiB, 640);
546     mc146818rtc_set_cmos_data(s, 0x15, val);
547     mc146818rtc_set_cmos_data(s, 0x16, val >> 8);
548     /* extended memory (next 64MiB) */
549     if (x86ms->below_4g_mem_size > 1 * MiB) {
550         val = (x86ms->below_4g_mem_size - 1 * MiB) / KiB;
551     } else {
552         val = 0;
553     }
554     if (val > 65535)
555         val = 65535;
556     mc146818rtc_set_cmos_data(s, 0x17, val);
557     mc146818rtc_set_cmos_data(s, 0x18, val >> 8);
558     mc146818rtc_set_cmos_data(s, 0x30, val);
559     mc146818rtc_set_cmos_data(s, 0x31, val >> 8);
560     /* memory between 16MiB and 4GiB */
561     if (x86ms->below_4g_mem_size > 16 * MiB) {
562         val = (x86ms->below_4g_mem_size - 16 * MiB) / (64 * KiB);
563     } else {
564         val = 0;
565     }
566     if (val > 65535)
567         val = 65535;
568     mc146818rtc_set_cmos_data(s, 0x34, val);
569     mc146818rtc_set_cmos_data(s, 0x35, val >> 8);
570     /* memory above 4GiB */
571     val = x86ms->above_4g_mem_size / 65536;
572     mc146818rtc_set_cmos_data(s, 0x5b, val);
573     mc146818rtc_set_cmos_data(s, 0x5c, val >> 8);
574     mc146818rtc_set_cmos_data(s, 0x5d, val >> 16);
575 
576     val = 0;
577     val |= 0x02; /* FPU is there */
578     val |= 0x04; /* PS/2 mouse installed */
579     mc146818rtc_set_cmos_data(s, REG_EQUIPMENT_BYTE, val);
580 }
581 
582 static void handle_a20_line_change(void *opaque, int irq, int level)
583 {
584     X86CPU *cpu = opaque;
585 
586     /* XXX: send to all CPUs ? */
587     /* XXX: add logic to handle multiple A20 line sources */
588     x86_cpu_set_a20(cpu, level);
589 }
590 
591 #define NE2000_NB_MAX 6
592 
593 static const int ne2000_io[NE2000_NB_MAX] = { 0x300, 0x320, 0x340, 0x360,
594                                               0x280, 0x380 };
595 static const int ne2000_irq[NE2000_NB_MAX] = { 9, 10, 11, 3, 4, 5 };
596 
597 static gboolean pc_init_ne2k_isa(ISABus *bus, NICInfo *nd, Error **errp)
598 {
599     static int nb_ne2k = 0;
600 
601     if (nb_ne2k == NE2000_NB_MAX) {
602         error_setg(errp,
603                    "maximum number of ISA NE2000 devices exceeded");
604         return false;
605     }
606     isa_ne2000_init(bus, ne2000_io[nb_ne2k],
607                     ne2000_irq[nb_ne2k], nd);
608     nb_ne2k++;
609     return true;
610 }
611 
612 void pc_acpi_smi_interrupt(void *opaque, int irq, int level)
613 {
614     X86CPU *cpu = opaque;
615 
616     if (level) {
617         cpu_interrupt(CPU(cpu), CPU_INTERRUPT_SMI);
618     }
619 }
620 
621 static
622 void pc_machine_done(Notifier *notifier, void *data)
623 {
624     PCMachineState *pcms = container_of(notifier,
625                                         PCMachineState, machine_done);
626     X86MachineState *x86ms = X86_MACHINE(pcms);
627 
628     cxl_hook_up_pxb_registers(pcms->pcibus, &pcms->cxl_devices_state,
629                               &error_fatal);
630 
631     if (pcms->cxl_devices_state.is_enabled) {
632         cxl_fmws_link_targets(&pcms->cxl_devices_state, &error_fatal);
633     }
634 
635     /* set the number of CPUs */
636     x86_rtc_set_cpus_count(x86ms->rtc, x86ms->boot_cpus);
637 
638     pci_bus_add_fw_cfg_extra_pci_roots(x86ms->fw_cfg, pcms->pcibus,
639                                        &error_abort);
640 
641     acpi_setup();
642     if (x86ms->fw_cfg) {
643         fw_cfg_build_smbios(pcms, x86ms->fw_cfg, pcms->smbios_entry_point_type);
644         fw_cfg_add_e820(x86ms->fw_cfg);
645         fw_cfg_build_feature_control(MACHINE(pcms), x86ms->fw_cfg);
646         /* update FW_CFG_NB_CPUS to account for -device added CPUs */
647         fw_cfg_modify_i16(x86ms->fw_cfg, FW_CFG_NB_CPUS, x86ms->boot_cpus);
648     }
649 
650     pc_cmos_init_late(pcms);
651 }
652 
653 /* setup pci memory address space mapping into system address space */
654 void pc_pci_as_mapping_init(MemoryRegion *system_memory,
655                             MemoryRegion *pci_address_space)
656 {
657     /* Set to lower priority than RAM */
658     memory_region_add_subregion_overlap(system_memory, 0x0,
659                                         pci_address_space, -1);
660 }
661 
662 void xen_load_linux(PCMachineState *pcms)
663 {
664     int i;
665     FWCfgState *fw_cfg;
666     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
667     X86MachineState *x86ms = X86_MACHINE(pcms);
668 
669     assert(MACHINE(pcms)->kernel_filename != NULL);
670 
671     fw_cfg = fw_cfg_init_io_dma(FW_CFG_IO_BASE, FW_CFG_IO_BASE + 4,
672                                 &address_space_memory);
673     fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, x86ms->boot_cpus);
674     rom_set_fw(fw_cfg);
675 
676     x86_load_linux(x86ms, fw_cfg, PC_FW_DATA, pcmc->pvh_enabled);
677     for (i = 0; i < nb_option_roms; i++) {
678         assert(!strcmp(option_rom[i].name, "linuxboot.bin") ||
679                !strcmp(option_rom[i].name, "linuxboot_dma.bin") ||
680                !strcmp(option_rom[i].name, "pvh.bin") ||
681                !strcmp(option_rom[i].name, "multiboot.bin") ||
682                !strcmp(option_rom[i].name, "multiboot_dma.bin"));
683         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
684     }
685     x86ms->fw_cfg = fw_cfg;
686 }
687 
688 #define PC_ROM_MIN_VGA     0xc0000
689 #define PC_ROM_MIN_OPTION  0xc8000
690 #define PC_ROM_MAX         0xe0000
691 #define PC_ROM_ALIGN       0x800
692 #define PC_ROM_SIZE        (PC_ROM_MAX - PC_ROM_MIN_VGA)
693 
694 static hwaddr pc_above_4g_end(PCMachineState *pcms)
695 {
696     X86MachineState *x86ms = X86_MACHINE(pcms);
697 
698     if (pcms->sgx_epc.size != 0) {
699         return sgx_epc_above_4g_end(&pcms->sgx_epc);
700     }
701 
702     return x86ms->above_4g_mem_start + x86ms->above_4g_mem_size;
703 }
704 
705 static void pc_get_device_memory_range(PCMachineState *pcms,
706                                        hwaddr *base,
707                                        ram_addr_t *device_mem_size)
708 {
709     MachineState *machine = MACHINE(pcms);
710     ram_addr_t size;
711     hwaddr addr;
712 
713     size = machine->maxram_size - machine->ram_size;
714     addr = ROUND_UP(pc_above_4g_end(pcms), 1 * GiB);
715 
716     /* size device region assuming 1G page max alignment per slot */
717     size += (1 * GiB) * machine->ram_slots;
718 
719     *base = addr;
720     *device_mem_size = size;
721 }
722 
723 static uint64_t pc_get_cxl_range_start(PCMachineState *pcms)
724 {
725     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
726     MachineState *ms = MACHINE(pcms);
727     hwaddr cxl_base;
728     ram_addr_t size;
729 
730     if (pcmc->has_reserved_memory &&
731         (ms->ram_size < ms->maxram_size)) {
732         pc_get_device_memory_range(pcms, &cxl_base, &size);
733         cxl_base += size;
734     } else {
735         cxl_base = pc_above_4g_end(pcms);
736     }
737 
738     return cxl_base;
739 }
740 
741 static uint64_t pc_get_cxl_range_end(PCMachineState *pcms)
742 {
743     uint64_t start = pc_get_cxl_range_start(pcms) + MiB;
744 
745     if (pcms->cxl_devices_state.fixed_windows) {
746         GList *it;
747 
748         start = ROUND_UP(start, 256 * MiB);
749         for (it = pcms->cxl_devices_state.fixed_windows; it; it = it->next) {
750             CXLFixedWindow *fw = it->data;
751             start += fw->size;
752         }
753     }
754 
755     return start;
756 }
757 
758 static hwaddr pc_max_used_gpa(PCMachineState *pcms, uint64_t pci_hole64_size)
759 {
760     X86CPU *cpu = X86_CPU(first_cpu);
761     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
762     MachineState *ms = MACHINE(pcms);
763 
764     if (cpu->env.features[FEAT_8000_0001_EDX] & CPUID_EXT2_LM) {
765         /* 64-bit systems */
766         return pc_pci_hole64_start() + pci_hole64_size - 1;
767     }
768 
769     /* 32-bit systems */
770     if (pcmc->broken_32bit_mem_addr_check) {
771         /* old value for compatibility reasons */
772         return ((hwaddr)1 << cpu->phys_bits) - 1;
773     }
774 
775     /*
776      * 32-bit systems don't have hole64 but they might have a region for
777      * memory devices. Even if additional hotplugged memory devices might
778      * not be usable by most guest OSes, we need to still consider them for
779      * calculating the highest possible GPA so that we can properly report
780      * if someone configures them on a CPU that cannot possibly address them.
781      */
782     if (pcmc->has_reserved_memory &&
783         (ms->ram_size < ms->maxram_size)) {
784         hwaddr devmem_start;
785         ram_addr_t devmem_size;
786 
787         pc_get_device_memory_range(pcms, &devmem_start, &devmem_size);
788         devmem_start += devmem_size;
789         return devmem_start - 1;
790     }
791 
792     /* configuration without any memory hotplug */
793     return pc_above_4g_end(pcms) - 1;
794 }
795 
796 /*
797  * AMD systems with an IOMMU have an additional hole close to the
798  * 1Tb, which are special GPAs that cannot be DMA mapped. Depending
799  * on kernel version, VFIO may or may not let you DMA map those ranges.
800  * Starting Linux v5.4 we validate it, and can't create guests on AMD machines
801  * with certain memory sizes. It's also wrong to use those IOVA ranges
802  * in detriment of leading to IOMMU INVALID_DEVICE_REQUEST or worse.
803  * The ranges reserved for Hyper-Transport are:
804  *
805  * FD_0000_0000h - FF_FFFF_FFFFh
806  *
807  * The ranges represent the following:
808  *
809  * Base Address   Top Address  Use
810  *
811  * FD_0000_0000h FD_F7FF_FFFFh Reserved interrupt address space
812  * FD_F800_0000h FD_F8FF_FFFFh Interrupt/EOI IntCtl
813  * FD_F900_0000h FD_F90F_FFFFh Legacy PIC IACK
814  * FD_F910_0000h FD_F91F_FFFFh System Management
815  * FD_F920_0000h FD_FAFF_FFFFh Reserved Page Tables
816  * FD_FB00_0000h FD_FBFF_FFFFh Address Translation
817  * FD_FC00_0000h FD_FDFF_FFFFh I/O Space
818  * FD_FE00_0000h FD_FFFF_FFFFh Configuration
819  * FE_0000_0000h FE_1FFF_FFFFh Extended Configuration/Device Messages
820  * FE_2000_0000h FF_FFFF_FFFFh Reserved
821  *
822  * See AMD IOMMU spec, section 2.1.2 "IOMMU Logical Topology",
823  * Table 3: Special Address Controls (GPA) for more information.
824  */
825 #define AMD_HT_START         0xfd00000000UL
826 #define AMD_HT_END           0xffffffffffUL
827 #define AMD_ABOVE_1TB_START  (AMD_HT_END + 1)
828 #define AMD_HT_SIZE          (AMD_ABOVE_1TB_START - AMD_HT_START)
829 
830 void pc_memory_init(PCMachineState *pcms,
831                     MemoryRegion *system_memory,
832                     MemoryRegion *rom_memory,
833                     uint64_t pci_hole64_size)
834 {
835     int linux_boot, i;
836     MemoryRegion *option_rom_mr;
837     MemoryRegion *ram_below_4g, *ram_above_4g;
838     FWCfgState *fw_cfg;
839     MachineState *machine = MACHINE(pcms);
840     MachineClass *mc = MACHINE_GET_CLASS(machine);
841     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
842     X86MachineState *x86ms = X86_MACHINE(pcms);
843     hwaddr maxphysaddr, maxusedaddr;
844     hwaddr cxl_base, cxl_resv_end = 0;
845     X86CPU *cpu = X86_CPU(first_cpu);
846 
847     assert(machine->ram_size == x86ms->below_4g_mem_size +
848                                 x86ms->above_4g_mem_size);
849 
850     linux_boot = (machine->kernel_filename != NULL);
851 
852     /*
853      * The HyperTransport range close to the 1T boundary is unique to AMD
854      * hosts with IOMMUs enabled. Restrict the ram-above-4g relocation
855      * to above 1T to AMD vCPUs only. @enforce_amd_1tb_hole is only false in
856      * older machine types (<= 7.0) for compatibility purposes.
857      */
858     if (IS_AMD_CPU(&cpu->env) && pcmc->enforce_amd_1tb_hole) {
859         /* Bail out if max possible address does not cross HT range */
860         if (pc_max_used_gpa(pcms, pci_hole64_size) >= AMD_HT_START) {
861             x86ms->above_4g_mem_start = AMD_ABOVE_1TB_START;
862         }
863 
864         /*
865          * Advertise the HT region if address space covers the reserved
866          * region or if we relocate.
867          */
868         if (cpu->phys_bits >= 40) {
869             e820_add_entry(AMD_HT_START, AMD_HT_SIZE, E820_RESERVED);
870         }
871     }
872 
873     /*
874      * phys-bits is required to be appropriately configured
875      * to make sure max used GPA is reachable.
876      */
877     maxusedaddr = pc_max_used_gpa(pcms, pci_hole64_size);
878     maxphysaddr = ((hwaddr)1 << cpu->phys_bits) - 1;
879     if (maxphysaddr < maxusedaddr) {
880         error_report("Address space limit 0x%"PRIx64" < 0x%"PRIx64
881                      " phys-bits too low (%u)",
882                      maxphysaddr, maxusedaddr, cpu->phys_bits);
883         exit(EXIT_FAILURE);
884     }
885 
886     /*
887      * Split single memory region and use aliases to address portions of it,
888      * done for backwards compatibility with older qemus.
889      */
890     ram_below_4g = g_malloc(sizeof(*ram_below_4g));
891     memory_region_init_alias(ram_below_4g, NULL, "ram-below-4g", machine->ram,
892                              0, x86ms->below_4g_mem_size);
893     memory_region_add_subregion(system_memory, 0, ram_below_4g);
894     e820_add_entry(0, x86ms->below_4g_mem_size, E820_RAM);
895     if (x86ms->above_4g_mem_size > 0) {
896         ram_above_4g = g_malloc(sizeof(*ram_above_4g));
897         memory_region_init_alias(ram_above_4g, NULL, "ram-above-4g",
898                                  machine->ram,
899                                  x86ms->below_4g_mem_size,
900                                  x86ms->above_4g_mem_size);
901         memory_region_add_subregion(system_memory, x86ms->above_4g_mem_start,
902                                     ram_above_4g);
903         e820_add_entry(x86ms->above_4g_mem_start, x86ms->above_4g_mem_size,
904                        E820_RAM);
905     }
906 
907     if (pcms->sgx_epc.size != 0) {
908         e820_add_entry(pcms->sgx_epc.base, pcms->sgx_epc.size, E820_RESERVED);
909     }
910 
911     if (!pcmc->has_reserved_memory &&
912         (machine->ram_slots ||
913          (machine->maxram_size > machine->ram_size))) {
914 
915         error_report("\"-memory 'slots|maxmem'\" is not supported by: %s",
916                      mc->name);
917         exit(EXIT_FAILURE);
918     }
919 
920     /* initialize device memory address space */
921     if (pcmc->has_reserved_memory &&
922         (machine->ram_size < machine->maxram_size)) {
923         ram_addr_t device_mem_size;
924         hwaddr device_mem_base;
925 
926         if (machine->ram_slots > ACPI_MAX_RAM_SLOTS) {
927             error_report("unsupported amount of memory slots: %"PRIu64,
928                          machine->ram_slots);
929             exit(EXIT_FAILURE);
930         }
931 
932         if (QEMU_ALIGN_UP(machine->maxram_size,
933                           TARGET_PAGE_SIZE) != machine->maxram_size) {
934             error_report("maximum memory size must by aligned to multiple of "
935                          "%d bytes", TARGET_PAGE_SIZE);
936             exit(EXIT_FAILURE);
937         }
938 
939         pc_get_device_memory_range(pcms, &device_mem_base, &device_mem_size);
940 
941         if (device_mem_base + device_mem_size < device_mem_size) {
942             error_report("unsupported amount of maximum memory: " RAM_ADDR_FMT,
943                          machine->maxram_size);
944             exit(EXIT_FAILURE);
945         }
946         machine_memory_devices_init(machine, device_mem_base, device_mem_size);
947     }
948 
949     if (pcms->cxl_devices_state.is_enabled) {
950         MemoryRegion *mr = &pcms->cxl_devices_state.host_mr;
951         hwaddr cxl_size = MiB;
952 
953         cxl_base = pc_get_cxl_range_start(pcms);
954         memory_region_init(mr, OBJECT(machine), "cxl_host_reg", cxl_size);
955         memory_region_add_subregion(system_memory, cxl_base, mr);
956         cxl_resv_end = cxl_base + cxl_size;
957         if (pcms->cxl_devices_state.fixed_windows) {
958             hwaddr cxl_fmw_base;
959             GList *it;
960 
961             cxl_fmw_base = ROUND_UP(cxl_base + cxl_size, 256 * MiB);
962             for (it = pcms->cxl_devices_state.fixed_windows; it; it = it->next) {
963                 CXLFixedWindow *fw = it->data;
964 
965                 fw->base = cxl_fmw_base;
966                 memory_region_init_io(&fw->mr, OBJECT(machine), &cfmws_ops, fw,
967                                       "cxl-fixed-memory-region", fw->size);
968                 memory_region_add_subregion(system_memory, fw->base, &fw->mr);
969                 cxl_fmw_base += fw->size;
970                 cxl_resv_end = cxl_fmw_base;
971             }
972         }
973     }
974 
975     /* Initialize PC system firmware */
976     pc_system_firmware_init(pcms, rom_memory);
977 
978     option_rom_mr = g_malloc(sizeof(*option_rom_mr));
979     if (machine_require_guest_memfd(machine)) {
980         memory_region_init_ram_guest_memfd(option_rom_mr, NULL, "pc.rom",
981                                            PC_ROM_SIZE, &error_fatal);
982     } else {
983         memory_region_init_ram(option_rom_mr, NULL, "pc.rom", PC_ROM_SIZE,
984                                &error_fatal);
985         if (pcmc->pci_enabled) {
986             memory_region_set_readonly(option_rom_mr, true);
987         }
988     }
989     memory_region_add_subregion_overlap(rom_memory,
990                                         PC_ROM_MIN_VGA,
991                                         option_rom_mr,
992                                         1);
993 
994     fw_cfg = fw_cfg_arch_create(machine,
995                                 x86ms->boot_cpus, x86ms->apic_id_limit);
996 
997     rom_set_fw(fw_cfg);
998 
999     if (machine->device_memory) {
1000         uint64_t *val = g_malloc(sizeof(*val));
1001         uint64_t res_mem_end = machine->device_memory->base;
1002 
1003         if (!pcmc->broken_reserved_end) {
1004             res_mem_end += memory_region_size(&machine->device_memory->mr);
1005         }
1006 
1007         if (pcms->cxl_devices_state.is_enabled) {
1008             res_mem_end = cxl_resv_end;
1009         }
1010         *val = cpu_to_le64(ROUND_UP(res_mem_end, 1 * GiB));
1011         fw_cfg_add_file(fw_cfg, "etc/reserved-memory-end", val, sizeof(*val));
1012     }
1013 
1014     if (linux_boot) {
1015         x86_load_linux(x86ms, fw_cfg, PC_FW_DATA, pcmc->pvh_enabled);
1016     }
1017 
1018     for (i = 0; i < nb_option_roms; i++) {
1019         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1020     }
1021     x86ms->fw_cfg = fw_cfg;
1022 
1023     /* Init default IOAPIC address space */
1024     x86ms->ioapic_as = &address_space_memory;
1025 
1026     /* Init ACPI memory hotplug IO base address */
1027     pcms->memhp_io_base = ACPI_MEMORY_HOTPLUG_BASE;
1028 }
1029 
1030 /*
1031  * The 64bit pci hole starts after "above 4G RAM" and
1032  * potentially the space reserved for memory hotplug.
1033  */
1034 uint64_t pc_pci_hole64_start(void)
1035 {
1036     PCMachineState *pcms = PC_MACHINE(qdev_get_machine());
1037     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1038     MachineState *ms = MACHINE(pcms);
1039     uint64_t hole64_start = 0;
1040     ram_addr_t size = 0;
1041 
1042     if (pcms->cxl_devices_state.is_enabled) {
1043         hole64_start = pc_get_cxl_range_end(pcms);
1044     } else if (pcmc->has_reserved_memory && (ms->ram_size < ms->maxram_size)) {
1045         pc_get_device_memory_range(pcms, &hole64_start, &size);
1046         if (!pcmc->broken_reserved_end) {
1047             hole64_start += size;
1048         }
1049     } else {
1050         hole64_start = pc_above_4g_end(pcms);
1051     }
1052 
1053     return ROUND_UP(hole64_start, 1 * GiB);
1054 }
1055 
1056 DeviceState *pc_vga_init(ISABus *isa_bus, PCIBus *pci_bus)
1057 {
1058     DeviceState *dev = NULL;
1059 
1060     rom_set_order_override(FW_CFG_ORDER_OVERRIDE_VGA);
1061     if (pci_bus) {
1062         PCIDevice *pcidev = pci_vga_init(pci_bus);
1063         dev = pcidev ? &pcidev->qdev : NULL;
1064     } else if (isa_bus) {
1065         ISADevice *isadev = isa_vga_init(isa_bus);
1066         dev = isadev ? DEVICE(isadev) : NULL;
1067     }
1068     rom_reset_order_override();
1069     return dev;
1070 }
1071 
1072 static const MemoryRegionOps ioport80_io_ops = {
1073     .write = ioport80_write,
1074     .read = ioport80_read,
1075     .endianness = DEVICE_LITTLE_ENDIAN,
1076     .impl = {
1077         .min_access_size = 1,
1078         .max_access_size = 1,
1079     },
1080 };
1081 
1082 static const MemoryRegionOps ioportF0_io_ops = {
1083     .write = ioportF0_write,
1084     .read = ioportF0_read,
1085     .endianness = DEVICE_LITTLE_ENDIAN,
1086     .impl = {
1087         .min_access_size = 1,
1088         .max_access_size = 1,
1089     },
1090 };
1091 
1092 static void pc_superio_init(ISABus *isa_bus, bool create_fdctrl,
1093                             bool create_i8042, bool no_vmport, Error **errp)
1094 {
1095     int i;
1096     DriveInfo *fd[MAX_FD];
1097     qemu_irq *a20_line;
1098     ISADevice *i8042, *port92, *vmmouse;
1099 
1100     serial_hds_isa_init(isa_bus, 0, MAX_ISA_SERIAL_PORTS);
1101     parallel_hds_isa_init(isa_bus, MAX_PARALLEL_PORTS);
1102 
1103     for (i = 0; i < MAX_FD; i++) {
1104         fd[i] = drive_get(IF_FLOPPY, 0, i);
1105         create_fdctrl |= !!fd[i];
1106     }
1107     if (create_fdctrl) {
1108 #ifdef CONFIG_FDC_ISA
1109         ISADevice *fdc = isa_new(TYPE_ISA_FDC);
1110         if (fdc) {
1111             isa_realize_and_unref(fdc, isa_bus, &error_fatal);
1112             isa_fdc_init_drives(fdc, fd);
1113         }
1114 #endif
1115     }
1116 
1117     if (!create_i8042) {
1118         if (!no_vmport) {
1119             error_setg(errp,
1120                        "vmport requires the i8042 controller to be enabled");
1121         }
1122         return;
1123     }
1124 
1125     i8042 = isa_create_simple(isa_bus, TYPE_I8042);
1126     if (!no_vmport) {
1127         isa_create_simple(isa_bus, TYPE_VMPORT);
1128         vmmouse = isa_try_new("vmmouse");
1129     } else {
1130         vmmouse = NULL;
1131     }
1132     if (vmmouse) {
1133         object_property_set_link(OBJECT(vmmouse), TYPE_I8042, OBJECT(i8042),
1134                                  &error_abort);
1135         isa_realize_and_unref(vmmouse, isa_bus, &error_fatal);
1136     }
1137     port92 = isa_create_simple(isa_bus, TYPE_PORT92);
1138 
1139     a20_line = qemu_allocate_irqs(handle_a20_line_change, first_cpu, 2);
1140     qdev_connect_gpio_out_named(DEVICE(i8042),
1141                                 I8042_A20_LINE, 0, a20_line[0]);
1142     qdev_connect_gpio_out_named(DEVICE(port92),
1143                                 PORT92_A20_LINE, 0, a20_line[1]);
1144     g_free(a20_line);
1145 }
1146 
1147 void pc_basic_device_init(struct PCMachineState *pcms,
1148                           ISABus *isa_bus, qemu_irq *gsi,
1149                           ISADevice *rtc_state,
1150                           bool create_fdctrl,
1151                           uint32_t hpet_irqs)
1152 {
1153     int i;
1154     DeviceState *hpet = NULL;
1155     int pit_isa_irq = 0;
1156     qemu_irq pit_alt_irq = NULL;
1157     ISADevice *pit = NULL;
1158     MemoryRegion *ioport80_io = g_new(MemoryRegion, 1);
1159     MemoryRegion *ioportF0_io = g_new(MemoryRegion, 1);
1160     X86MachineState *x86ms = X86_MACHINE(pcms);
1161 
1162     memory_region_init_io(ioport80_io, NULL, &ioport80_io_ops, NULL, "ioport80", 1);
1163     memory_region_add_subregion(isa_bus->address_space_io, 0x80, ioport80_io);
1164 
1165     memory_region_init_io(ioportF0_io, NULL, &ioportF0_io_ops, NULL, "ioportF0", 1);
1166     memory_region_add_subregion(isa_bus->address_space_io, 0xf0, ioportF0_io);
1167 
1168     /*
1169      * Check if an HPET shall be created.
1170      */
1171     if (pcms->hpet_enabled) {
1172         qemu_irq rtc_irq;
1173 
1174         hpet = qdev_try_new(TYPE_HPET);
1175         if (!hpet) {
1176             error_report("couldn't create HPET device");
1177             exit(1);
1178         }
1179         /*
1180          * For pc-piix-*, hpet's intcap is always IRQ2. For pc-q35-*,
1181          * use IRQ16~23, IRQ8 and IRQ2.  If the user has already set
1182          * the property, use whatever mask they specified.
1183          */
1184         uint8_t compat = object_property_get_uint(OBJECT(hpet),
1185                 HPET_INTCAP, NULL);
1186         if (!compat) {
1187             qdev_prop_set_uint32(hpet, HPET_INTCAP, hpet_irqs);
1188         }
1189         sysbus_realize_and_unref(SYS_BUS_DEVICE(hpet), &error_fatal);
1190         sysbus_mmio_map(SYS_BUS_DEVICE(hpet), 0, HPET_BASE);
1191 
1192         for (i = 0; i < IOAPIC_NUM_PINS; i++) {
1193             sysbus_connect_irq(SYS_BUS_DEVICE(hpet), i, gsi[i]);
1194         }
1195         pit_isa_irq = -1;
1196         pit_alt_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_PIT_INT);
1197         rtc_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_RTC_INT);
1198 
1199         /* overwrite connection created by south bridge */
1200         qdev_connect_gpio_out(DEVICE(rtc_state), 0, rtc_irq);
1201     }
1202 
1203     object_property_add_alias(OBJECT(pcms), "rtc-time", OBJECT(rtc_state),
1204                               "date");
1205 
1206 #ifdef CONFIG_XEN_EMU
1207     if (xen_mode == XEN_EMULATE) {
1208         xen_overlay_create();
1209         xen_evtchn_create(IOAPIC_NUM_PINS, gsi);
1210         xen_gnttab_create();
1211         xen_xenstore_create();
1212         if (pcms->pcibus) {
1213             pci_create_simple(pcms->pcibus, -1, "xen-platform");
1214         }
1215         xen_bus_init();
1216     }
1217 #endif
1218 
1219     qemu_register_boot_set(pc_boot_set, pcms);
1220     set_boot_dev(pcms, MC146818_RTC(rtc_state),
1221                  MACHINE(pcms)->boot_config.order, &error_fatal);
1222 
1223     if (!xen_enabled() &&
1224         (x86ms->pit == ON_OFF_AUTO_AUTO || x86ms->pit == ON_OFF_AUTO_ON)) {
1225         if (kvm_pit_in_kernel()) {
1226             pit = kvm_pit_init(isa_bus, 0x40);
1227         } else {
1228             pit = i8254_pit_init(isa_bus, 0x40, pit_isa_irq, pit_alt_irq);
1229         }
1230         if (hpet) {
1231             /* connect PIT to output control line of the HPET */
1232             qdev_connect_gpio_out(hpet, 0, qdev_get_gpio_in(DEVICE(pit), 0));
1233         }
1234         object_property_set_link(OBJECT(pcms->pcspk), "pit",
1235                                  OBJECT(pit), &error_fatal);
1236         isa_realize_and_unref(pcms->pcspk, isa_bus, &error_fatal);
1237     }
1238 
1239     if (pcms->vmport == ON_OFF_AUTO_AUTO) {
1240         pcms->vmport = (xen_enabled() || !pcms->i8042_enabled)
1241             ? ON_OFF_AUTO_OFF : ON_OFF_AUTO_ON;
1242     }
1243 
1244     /* Super I/O */
1245     pc_superio_init(isa_bus, create_fdctrl, pcms->i8042_enabled,
1246                     pcms->vmport != ON_OFF_AUTO_ON, &error_fatal);
1247 
1248     pcms->machine_done.notify = pc_machine_done;
1249     qemu_add_machine_init_done_notifier(&pcms->machine_done);
1250 }
1251 
1252 void pc_nic_init(PCMachineClass *pcmc, ISABus *isa_bus, PCIBus *pci_bus)
1253 {
1254     MachineClass *mc = MACHINE_CLASS(pcmc);
1255     bool default_is_ne2k = g_str_equal(mc->default_nic, TYPE_ISA_NE2000);
1256     NICInfo *nd;
1257 
1258     rom_set_order_override(FW_CFG_ORDER_OVERRIDE_NIC);
1259 
1260     while ((nd = qemu_find_nic_info(TYPE_ISA_NE2000, default_is_ne2k, NULL))) {
1261         pc_init_ne2k_isa(isa_bus, nd, &error_fatal);
1262     }
1263 
1264     /* Anything remaining should be a PCI NIC */
1265     if (pci_bus) {
1266         pci_init_nic_devices(pci_bus, mc->default_nic);
1267     }
1268 
1269     rom_reset_order_override();
1270 }
1271 
1272 void pc_i8259_create(ISABus *isa_bus, qemu_irq *i8259_irqs)
1273 {
1274     qemu_irq *i8259;
1275 
1276     if (kvm_pic_in_kernel()) {
1277         i8259 = kvm_i8259_init(isa_bus);
1278     } else if (xen_enabled()) {
1279         i8259 = xen_interrupt_controller_init();
1280     } else {
1281         i8259 = i8259_init(isa_bus, x86_allocate_cpu_irq());
1282     }
1283 
1284     for (size_t i = 0; i < ISA_NUM_IRQS; i++) {
1285         i8259_irqs[i] = i8259[i];
1286     }
1287 
1288     g_free(i8259);
1289 }
1290 
1291 static void pc_memory_pre_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
1292                                Error **errp)
1293 {
1294     const X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
1295     const MachineState *ms = MACHINE(hotplug_dev);
1296     const bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
1297     Error *local_err = NULL;
1298 
1299     /*
1300      * When "acpi=off" is used with the Q35 machine type, no ACPI is built,
1301      * but pcms->acpi_dev is still created. Check !acpi_enabled in
1302      * addition to cover this case.
1303      */
1304     if (!x86ms->acpi_dev || !x86_machine_is_acpi_enabled(x86ms)) {
1305         error_setg(errp,
1306                    "memory hotplug is not enabled: missing acpi device or acpi disabled");
1307         return;
1308     }
1309 
1310     if (is_nvdimm && !ms->nvdimms_state->is_enabled) {
1311         error_setg(errp, "nvdimm is not enabled: missing 'nvdimm' in '-M'");
1312         return;
1313     }
1314 
1315     hotplug_handler_pre_plug(x86ms->acpi_dev, dev, &local_err);
1316     if (local_err) {
1317         error_propagate(errp, local_err);
1318         return;
1319     }
1320 
1321     pc_dimm_pre_plug(PC_DIMM(dev), MACHINE(hotplug_dev), errp);
1322 }
1323 
1324 static void pc_memory_plug(HotplugHandler *hotplug_dev,
1325                            DeviceState *dev, Error **errp)
1326 {
1327     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1328     X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
1329     MachineState *ms = MACHINE(hotplug_dev);
1330     bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
1331 
1332     pc_dimm_plug(PC_DIMM(dev), MACHINE(pcms));
1333 
1334     if (is_nvdimm) {
1335         nvdimm_plug(ms->nvdimms_state);
1336     }
1337 
1338     hotplug_handler_plug(x86ms->acpi_dev, dev, &error_abort);
1339 }
1340 
1341 static void pc_memory_unplug_request(HotplugHandler *hotplug_dev,
1342                                      DeviceState *dev, Error **errp)
1343 {
1344     X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
1345 
1346     /*
1347      * When "acpi=off" is used with the Q35 machine type, no ACPI is built,
1348      * but pcms->acpi_dev is still created. Check !acpi_enabled in
1349      * addition to cover this case.
1350      */
1351     if (!x86ms->acpi_dev || !x86_machine_is_acpi_enabled(x86ms)) {
1352         error_setg(errp,
1353                    "memory hotplug is not enabled: missing acpi device or acpi disabled");
1354         return;
1355     }
1356 
1357     if (object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM)) {
1358         error_setg(errp, "nvdimm device hot unplug is not supported yet.");
1359         return;
1360     }
1361 
1362     hotplug_handler_unplug_request(x86ms->acpi_dev, dev,
1363                                    errp);
1364 }
1365 
1366 static void pc_memory_unplug(HotplugHandler *hotplug_dev,
1367                              DeviceState *dev, Error **errp)
1368 {
1369     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1370     X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
1371     Error *local_err = NULL;
1372 
1373     hotplug_handler_unplug(x86ms->acpi_dev, dev, &local_err);
1374     if (local_err) {
1375         goto out;
1376     }
1377 
1378     pc_dimm_unplug(PC_DIMM(dev), MACHINE(pcms));
1379     qdev_unrealize(dev);
1380  out:
1381     error_propagate(errp, local_err);
1382 }
1383 
1384 static void pc_hv_balloon_pre_plug(HotplugHandler *hotplug_dev,
1385                                    DeviceState *dev, Error **errp)
1386 {
1387     /* The vmbus handler has no hotplug handler; we should never end up here. */
1388     g_assert(!dev->hotplugged);
1389     memory_device_pre_plug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev), errp);
1390 }
1391 
1392 static void pc_hv_balloon_plug(HotplugHandler *hotplug_dev,
1393                                DeviceState *dev, Error **errp)
1394 {
1395     memory_device_plug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev));
1396 }
1397 
1398 static void pc_machine_device_pre_plug_cb(HotplugHandler *hotplug_dev,
1399                                           DeviceState *dev, Error **errp)
1400 {
1401     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1402         pc_memory_pre_plug(hotplug_dev, dev, errp);
1403     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1404         x86_cpu_pre_plug(hotplug_dev, dev, errp);
1405     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI)) {
1406         virtio_md_pci_pre_plug(VIRTIO_MD_PCI(dev), MACHINE(hotplug_dev), errp);
1407     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
1408         /* Declare the APIC range as the reserved MSI region */
1409         char *resv_prop_str = g_strdup_printf("0xfee00000:0xfeefffff:%d",
1410                                               VIRTIO_IOMMU_RESV_MEM_T_MSI);
1411         QList *reserved_regions = qlist_new();
1412 
1413         qlist_append_str(reserved_regions, resv_prop_str);
1414         qdev_prop_set_array(dev, "reserved-regions", reserved_regions);
1415 
1416         g_free(resv_prop_str);
1417     }
1418 
1419     if (object_dynamic_cast(OBJECT(dev), TYPE_X86_IOMMU_DEVICE) ||
1420         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
1421         PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1422 
1423         if (pcms->iommu) {
1424             error_setg(errp, "QEMU does not support multiple vIOMMUs "
1425                        "for x86 yet.");
1426             return;
1427         }
1428         pcms->iommu = dev;
1429     } else if (object_dynamic_cast(OBJECT(dev), TYPE_HV_BALLOON)) {
1430         pc_hv_balloon_pre_plug(hotplug_dev, dev, errp);
1431     }
1432 }
1433 
1434 static void pc_machine_device_plug_cb(HotplugHandler *hotplug_dev,
1435                                       DeviceState *dev, Error **errp)
1436 {
1437     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1438         pc_memory_plug(hotplug_dev, dev, errp);
1439     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1440         x86_cpu_plug(hotplug_dev, dev, errp);
1441     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI)) {
1442         virtio_md_pci_plug(VIRTIO_MD_PCI(dev), MACHINE(hotplug_dev), errp);
1443     } else if (object_dynamic_cast(OBJECT(dev), TYPE_HV_BALLOON)) {
1444         pc_hv_balloon_plug(hotplug_dev, dev, errp);
1445     }
1446 }
1447 
1448 static void pc_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
1449                                                 DeviceState *dev, Error **errp)
1450 {
1451     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1452         pc_memory_unplug_request(hotplug_dev, dev, errp);
1453     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1454         x86_cpu_unplug_request_cb(hotplug_dev, dev, errp);
1455     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI)) {
1456         virtio_md_pci_unplug_request(VIRTIO_MD_PCI(dev), MACHINE(hotplug_dev),
1457                                      errp);
1458     } else {
1459         error_setg(errp, "acpi: device unplug request for not supported device"
1460                    " type: %s", object_get_typename(OBJECT(dev)));
1461     }
1462 }
1463 
1464 static void pc_machine_device_unplug_cb(HotplugHandler *hotplug_dev,
1465                                         DeviceState *dev, Error **errp)
1466 {
1467     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1468         pc_memory_unplug(hotplug_dev, dev, errp);
1469     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1470         x86_cpu_unplug_cb(hotplug_dev, dev, errp);
1471     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI)) {
1472         virtio_md_pci_unplug(VIRTIO_MD_PCI(dev), MACHINE(hotplug_dev), errp);
1473     } else {
1474         error_setg(errp, "acpi: device unplug for not supported device"
1475                    " type: %s", object_get_typename(OBJECT(dev)));
1476     }
1477 }
1478 
1479 static HotplugHandler *pc_get_hotplug_handler(MachineState *machine,
1480                                              DeviceState *dev)
1481 {
1482     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) ||
1483         object_dynamic_cast(OBJECT(dev), TYPE_CPU) ||
1484         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI) ||
1485         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI) ||
1486         object_dynamic_cast(OBJECT(dev), TYPE_HV_BALLOON) ||
1487         object_dynamic_cast(OBJECT(dev), TYPE_X86_IOMMU_DEVICE)) {
1488         return HOTPLUG_HANDLER(machine);
1489     }
1490 
1491     return NULL;
1492 }
1493 
1494 static void pc_machine_get_vmport(Object *obj, Visitor *v, const char *name,
1495                                   void *opaque, Error **errp)
1496 {
1497     PCMachineState *pcms = PC_MACHINE(obj);
1498     OnOffAuto vmport = pcms->vmport;
1499 
1500     visit_type_OnOffAuto(v, name, &vmport, errp);
1501 }
1502 
1503 static void pc_machine_set_vmport(Object *obj, Visitor *v, const char *name,
1504                                   void *opaque, Error **errp)
1505 {
1506     PCMachineState *pcms = PC_MACHINE(obj);
1507 
1508     visit_type_OnOffAuto(v, name, &pcms->vmport, errp);
1509 }
1510 
1511 static bool pc_machine_get_fd_bootchk(Object *obj, Error **errp)
1512 {
1513     PCMachineState *pcms = PC_MACHINE(obj);
1514 
1515     return pcms->fd_bootchk;
1516 }
1517 
1518 static void pc_machine_set_fd_bootchk(Object *obj, bool value, Error **errp)
1519 {
1520     PCMachineState *pcms = PC_MACHINE(obj);
1521 
1522     pcms->fd_bootchk = value;
1523 }
1524 
1525 static bool pc_machine_get_smbus(Object *obj, Error **errp)
1526 {
1527     PCMachineState *pcms = PC_MACHINE(obj);
1528 
1529     return pcms->smbus_enabled;
1530 }
1531 
1532 static void pc_machine_set_smbus(Object *obj, bool value, Error **errp)
1533 {
1534     PCMachineState *pcms = PC_MACHINE(obj);
1535 
1536     pcms->smbus_enabled = value;
1537 }
1538 
1539 static bool pc_machine_get_sata(Object *obj, Error **errp)
1540 {
1541     PCMachineState *pcms = PC_MACHINE(obj);
1542 
1543     return pcms->sata_enabled;
1544 }
1545 
1546 static void pc_machine_set_sata(Object *obj, bool value, Error **errp)
1547 {
1548     PCMachineState *pcms = PC_MACHINE(obj);
1549 
1550     pcms->sata_enabled = value;
1551 }
1552 
1553 static bool pc_machine_get_hpet(Object *obj, Error **errp)
1554 {
1555     PCMachineState *pcms = PC_MACHINE(obj);
1556 
1557     return pcms->hpet_enabled;
1558 }
1559 
1560 static void pc_machine_set_hpet(Object *obj, bool value, Error **errp)
1561 {
1562     PCMachineState *pcms = PC_MACHINE(obj);
1563 
1564     pcms->hpet_enabled = value;
1565 }
1566 
1567 static bool pc_machine_get_i8042(Object *obj, Error **errp)
1568 {
1569     PCMachineState *pcms = PC_MACHINE(obj);
1570 
1571     return pcms->i8042_enabled;
1572 }
1573 
1574 static void pc_machine_set_i8042(Object *obj, bool value, Error **errp)
1575 {
1576     PCMachineState *pcms = PC_MACHINE(obj);
1577 
1578     pcms->i8042_enabled = value;
1579 }
1580 
1581 static bool pc_machine_get_default_bus_bypass_iommu(Object *obj, Error **errp)
1582 {
1583     PCMachineState *pcms = PC_MACHINE(obj);
1584 
1585     return pcms->default_bus_bypass_iommu;
1586 }
1587 
1588 static void pc_machine_set_default_bus_bypass_iommu(Object *obj, bool value,
1589                                                     Error **errp)
1590 {
1591     PCMachineState *pcms = PC_MACHINE(obj);
1592 
1593     pcms->default_bus_bypass_iommu = value;
1594 }
1595 
1596 static void pc_machine_get_smbios_ep(Object *obj, Visitor *v, const char *name,
1597                                      void *opaque, Error **errp)
1598 {
1599     PCMachineState *pcms = PC_MACHINE(obj);
1600     SmbiosEntryPointType smbios_entry_point_type = pcms->smbios_entry_point_type;
1601 
1602     visit_type_SmbiosEntryPointType(v, name, &smbios_entry_point_type, errp);
1603 }
1604 
1605 static void pc_machine_set_smbios_ep(Object *obj, Visitor *v, const char *name,
1606                                      void *opaque, Error **errp)
1607 {
1608     PCMachineState *pcms = PC_MACHINE(obj);
1609 
1610     visit_type_SmbiosEntryPointType(v, name, &pcms->smbios_entry_point_type, errp);
1611 }
1612 
1613 static void pc_machine_get_max_ram_below_4g(Object *obj, Visitor *v,
1614                                             const char *name, void *opaque,
1615                                             Error **errp)
1616 {
1617     PCMachineState *pcms = PC_MACHINE(obj);
1618     uint64_t value = pcms->max_ram_below_4g;
1619 
1620     visit_type_size(v, name, &value, errp);
1621 }
1622 
1623 static void pc_machine_set_max_ram_below_4g(Object *obj, Visitor *v,
1624                                             const char *name, void *opaque,
1625                                             Error **errp)
1626 {
1627     PCMachineState *pcms = PC_MACHINE(obj);
1628     uint64_t value;
1629 
1630     if (!visit_type_size(v, name, &value, errp)) {
1631         return;
1632     }
1633     if (value > 4 * GiB) {
1634         error_setg(errp,
1635                    "Machine option 'max-ram-below-4g=%"PRIu64
1636                    "' expects size less than or equal to 4G", value);
1637         return;
1638     }
1639 
1640     if (value < 1 * MiB) {
1641         warn_report("Only %" PRIu64 " bytes of RAM below the 4GiB boundary,"
1642                     "BIOS may not work with less than 1MiB", value);
1643     }
1644 
1645     pcms->max_ram_below_4g = value;
1646 }
1647 
1648 static void pc_machine_get_max_fw_size(Object *obj, Visitor *v,
1649                                        const char *name, void *opaque,
1650                                        Error **errp)
1651 {
1652     PCMachineState *pcms = PC_MACHINE(obj);
1653     uint64_t value = pcms->max_fw_size;
1654 
1655     visit_type_size(v, name, &value, errp);
1656 }
1657 
1658 static void pc_machine_set_max_fw_size(Object *obj, Visitor *v,
1659                                        const char *name, void *opaque,
1660                                        Error **errp)
1661 {
1662     PCMachineState *pcms = PC_MACHINE(obj);
1663     uint64_t value;
1664 
1665     if (!visit_type_size(v, name, &value, errp)) {
1666         return;
1667     }
1668 
1669     /*
1670      * We don't have a theoretically justifiable exact lower bound on the base
1671      * address of any flash mapping. In practice, the IO-APIC MMIO range is
1672      * [0xFEE00000..0xFEE01000] -- see IO_APIC_DEFAULT_ADDRESS --, leaving free
1673      * only 18MiB-4KiB below 4GiB. For now, restrict the cumulative mapping to
1674      * 16MiB in size.
1675      */
1676     if (value > 16 * MiB) {
1677         error_setg(errp,
1678                    "User specified max allowed firmware size %" PRIu64 " is "
1679                    "greater than 16MiB. If combined firmware size exceeds "
1680                    "16MiB the system may not boot, or experience intermittent"
1681                    "stability issues.",
1682                    value);
1683         return;
1684     }
1685 
1686     pcms->max_fw_size = value;
1687 }
1688 
1689 
1690 static void pc_machine_initfn(Object *obj)
1691 {
1692     PCMachineState *pcms = PC_MACHINE(obj);
1693     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1694 
1695 #ifdef CONFIG_VMPORT
1696     pcms->vmport = ON_OFF_AUTO_AUTO;
1697 #else
1698     pcms->vmport = ON_OFF_AUTO_OFF;
1699 #endif /* CONFIG_VMPORT */
1700     pcms->max_ram_below_4g = 0; /* use default */
1701     pcms->smbios_entry_point_type = pcmc->default_smbios_ep_type;
1702     pcms->south_bridge = pcmc->default_south_bridge;
1703 
1704     /* acpi build is enabled by default if machine supports it */
1705     pcms->acpi_build_enabled = pcmc->has_acpi_build;
1706     pcms->smbus_enabled = true;
1707     pcms->sata_enabled = true;
1708     pcms->i8042_enabled = true;
1709     pcms->max_fw_size = 8 * MiB;
1710 #if defined(CONFIG_HPET)
1711     pcms->hpet_enabled = true;
1712 #endif
1713     pcms->fd_bootchk = true;
1714     pcms->default_bus_bypass_iommu = false;
1715 
1716     pc_system_flash_create(pcms);
1717     pcms->pcspk = isa_new(TYPE_PC_SPEAKER);
1718     object_property_add_alias(OBJECT(pcms), "pcspk-audiodev",
1719                               OBJECT(pcms->pcspk), "audiodev");
1720     if (pcmc->pci_enabled) {
1721         cxl_machine_init(obj, &pcms->cxl_devices_state);
1722     }
1723 }
1724 
1725 static void pc_machine_reset(MachineState *machine, ResetType type)
1726 {
1727     CPUState *cs;
1728     X86CPU *cpu;
1729 
1730     qemu_devices_reset(type);
1731 
1732     /* Reset APIC after devices have been reset to cancel
1733      * any changes that qemu_devices_reset() might have done.
1734      */
1735     CPU_FOREACH(cs) {
1736         cpu = X86_CPU(cs);
1737 
1738         x86_cpu_after_reset(cpu);
1739     }
1740 }
1741 
1742 static void pc_machine_wakeup(MachineState *machine)
1743 {
1744     cpu_synchronize_all_states();
1745     pc_machine_reset(machine, RESET_TYPE_WAKEUP);
1746     cpu_synchronize_all_post_reset();
1747 }
1748 
1749 static bool pc_hotplug_allowed(MachineState *ms, DeviceState *dev, Error **errp)
1750 {
1751     X86IOMMUState *iommu = x86_iommu_get_default();
1752     IntelIOMMUState *intel_iommu;
1753 
1754     if (iommu &&
1755         object_dynamic_cast((Object *)iommu, TYPE_INTEL_IOMMU_DEVICE) &&
1756         object_dynamic_cast((Object *)dev, "vfio-pci")) {
1757         intel_iommu = INTEL_IOMMU_DEVICE(iommu);
1758         if (!intel_iommu->caching_mode) {
1759             error_setg(errp, "Device assignment is not allowed without "
1760                        "enabling caching-mode=on for Intel IOMMU.");
1761             return false;
1762         }
1763     }
1764 
1765     return true;
1766 }
1767 
1768 static void pc_machine_class_init(ObjectClass *oc, void *data)
1769 {
1770     MachineClass *mc = MACHINE_CLASS(oc);
1771     X86MachineClass *x86mc = X86_MACHINE_CLASS(oc);
1772     PCMachineClass *pcmc = PC_MACHINE_CLASS(oc);
1773     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
1774 
1775     pcmc->pci_enabled = true;
1776     pcmc->has_acpi_build = true;
1777     pcmc->smbios_defaults = true;
1778     pcmc->gigabyte_align = true;
1779     pcmc->has_reserved_memory = true;
1780     pcmc->enforce_amd_1tb_hole = true;
1781     pcmc->isa_bios_alias = true;
1782     pcmc->pvh_enabled = true;
1783     pcmc->kvmclock_create_always = true;
1784     x86mc->apic_xrupt_override = true;
1785     assert(!mc->get_hotplug_handler);
1786     mc->get_hotplug_handler = pc_get_hotplug_handler;
1787     mc->hotplug_allowed = pc_hotplug_allowed;
1788     mc->auto_enable_numa_with_memhp = true;
1789     mc->auto_enable_numa_with_memdev = true;
1790     mc->has_hotpluggable_cpus = true;
1791     mc->default_boot_order = "cad";
1792     mc->block_default_type = IF_IDE;
1793     mc->max_cpus = 255;
1794     mc->reset = pc_machine_reset;
1795     mc->wakeup = pc_machine_wakeup;
1796     hc->pre_plug = pc_machine_device_pre_plug_cb;
1797     hc->plug = pc_machine_device_plug_cb;
1798     hc->unplug_request = pc_machine_device_unplug_request_cb;
1799     hc->unplug = pc_machine_device_unplug_cb;
1800     mc->default_cpu_type = TARGET_DEFAULT_CPU_TYPE;
1801     mc->nvdimm_supported = true;
1802     mc->smp_props.dies_supported = true;
1803     mc->smp_props.modules_supported = true;
1804     mc->smp_props.cache_supported[CACHE_LEVEL_AND_TYPE_L1D] = true;
1805     mc->smp_props.cache_supported[CACHE_LEVEL_AND_TYPE_L1I] = true;
1806     mc->smp_props.cache_supported[CACHE_LEVEL_AND_TYPE_L2] = true;
1807     mc->smp_props.cache_supported[CACHE_LEVEL_AND_TYPE_L3] = true;
1808     mc->default_ram_id = "pc.ram";
1809     pcmc->default_smbios_ep_type = SMBIOS_ENTRY_POINT_TYPE_AUTO;
1810 
1811     object_class_property_add(oc, PC_MACHINE_MAX_RAM_BELOW_4G, "size",
1812         pc_machine_get_max_ram_below_4g, pc_machine_set_max_ram_below_4g,
1813         NULL, NULL);
1814     object_class_property_set_description(oc, PC_MACHINE_MAX_RAM_BELOW_4G,
1815         "Maximum ram below the 4G boundary (32bit boundary)");
1816 
1817     object_class_property_add(oc, PC_MACHINE_VMPORT, "OnOffAuto",
1818         pc_machine_get_vmport, pc_machine_set_vmport,
1819         NULL, NULL);
1820     object_class_property_set_description(oc, PC_MACHINE_VMPORT,
1821         "Enable vmport (pc & q35)");
1822 
1823     object_class_property_add_bool(oc, PC_MACHINE_SMBUS,
1824         pc_machine_get_smbus, pc_machine_set_smbus);
1825     object_class_property_set_description(oc, PC_MACHINE_SMBUS,
1826         "Enable/disable system management bus");
1827 
1828     object_class_property_add_bool(oc, PC_MACHINE_SATA,
1829         pc_machine_get_sata, pc_machine_set_sata);
1830     object_class_property_set_description(oc, PC_MACHINE_SATA,
1831         "Enable/disable Serial ATA bus");
1832 
1833     object_class_property_add_bool(oc, "hpet",
1834         pc_machine_get_hpet, pc_machine_set_hpet);
1835     object_class_property_set_description(oc, "hpet",
1836         "Enable/disable high precision event timer emulation");
1837 
1838     object_class_property_add_bool(oc, PC_MACHINE_I8042,
1839         pc_machine_get_i8042, pc_machine_set_i8042);
1840     object_class_property_set_description(oc, PC_MACHINE_I8042,
1841         "Enable/disable Intel 8042 PS/2 controller emulation");
1842 
1843     object_class_property_add_bool(oc, "default-bus-bypass-iommu",
1844         pc_machine_get_default_bus_bypass_iommu,
1845         pc_machine_set_default_bus_bypass_iommu);
1846 
1847     object_class_property_add(oc, PC_MACHINE_MAX_FW_SIZE, "size",
1848         pc_machine_get_max_fw_size, pc_machine_set_max_fw_size,
1849         NULL, NULL);
1850     object_class_property_set_description(oc, PC_MACHINE_MAX_FW_SIZE,
1851         "Maximum combined firmware size");
1852 
1853     object_class_property_add(oc, PC_MACHINE_SMBIOS_EP, "str",
1854         pc_machine_get_smbios_ep, pc_machine_set_smbios_ep,
1855         NULL, NULL);
1856     object_class_property_set_description(oc, PC_MACHINE_SMBIOS_EP,
1857         "SMBIOS Entry Point type [32, 64]");
1858 
1859     object_class_property_add_bool(oc, "fd-bootchk",
1860         pc_machine_get_fd_bootchk,
1861         pc_machine_set_fd_bootchk);
1862 }
1863 
1864 static const TypeInfo pc_machine_info = {
1865     .name = TYPE_PC_MACHINE,
1866     .parent = TYPE_X86_MACHINE,
1867     .abstract = true,
1868     .instance_size = sizeof(PCMachineState),
1869     .instance_init = pc_machine_initfn,
1870     .class_size = sizeof(PCMachineClass),
1871     .class_init = pc_machine_class_init,
1872     .interfaces = (InterfaceInfo[]) {
1873          { TYPE_HOTPLUG_HANDLER },
1874          { }
1875     },
1876 };
1877 
1878 static void pc_machine_register_types(void)
1879 {
1880     type_register_static(&pc_machine_info);
1881 }
1882 
1883 type_init(pc_machine_register_types)
1884