xref: /qemu/hw/i386/pc.c (revision db65ac5e258e75e9aec45626bf1071626094e057)
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 "qapi/qmp/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_9_2[] = {};
83 const size_t pc_compat_9_2_len = G_N_ELEMENTS(pc_compat_9_2);
84 
85 GlobalProperty pc_compat_9_1[] = {
86     { "ICH9-LPC", "x-smi-swsmi-timer", "off" },
87     { "ICH9-LPC", "x-smi-periodic-timer", "off" },
88     { TYPE_INTEL_IOMMU_DEVICE, "stale-tm", "on" },
89 };
90 const size_t pc_compat_9_1_len = G_N_ELEMENTS(pc_compat_9_1);
91 
92 GlobalProperty pc_compat_9_0[] = {
93     { TYPE_X86_CPU, "x-amd-topoext-features-only", "false" },
94     { TYPE_X86_CPU, "x-l1-cache-per-thread", "false" },
95     { TYPE_X86_CPU, "guest-phys-bits", "0" },
96     { "sev-guest", "legacy-vm-type", "on" },
97     { TYPE_X86_CPU, "legacy-multi-node", "on" },
98 };
99 const size_t pc_compat_9_0_len = G_N_ELEMENTS(pc_compat_9_0);
100 
101 GlobalProperty pc_compat_8_2[] = {};
102 const size_t pc_compat_8_2_len = G_N_ELEMENTS(pc_compat_8_2);
103 
104 GlobalProperty pc_compat_8_1[] = {};
105 const size_t pc_compat_8_1_len = G_N_ELEMENTS(pc_compat_8_1);
106 
107 GlobalProperty pc_compat_8_0[] = {
108     { "virtio-mem", "unplugged-inaccessible", "auto" },
109 };
110 const size_t pc_compat_8_0_len = G_N_ELEMENTS(pc_compat_8_0);
111 
112 GlobalProperty pc_compat_7_2[] = {
113     { "ICH9-LPC", "noreboot", "true" },
114 };
115 const size_t pc_compat_7_2_len = G_N_ELEMENTS(pc_compat_7_2);
116 
117 GlobalProperty pc_compat_7_1[] = {};
118 const size_t pc_compat_7_1_len = G_N_ELEMENTS(pc_compat_7_1);
119 
120 GlobalProperty pc_compat_7_0[] = {};
121 const size_t pc_compat_7_0_len = G_N_ELEMENTS(pc_compat_7_0);
122 
123 GlobalProperty pc_compat_6_2[] = {
124     { "virtio-mem", "unplugged-inaccessible", "off" },
125 };
126 const size_t pc_compat_6_2_len = G_N_ELEMENTS(pc_compat_6_2);
127 
128 GlobalProperty pc_compat_6_1[] = {
129     { TYPE_X86_CPU, "hv-version-id-build", "0x1bbc" },
130     { TYPE_X86_CPU, "hv-version-id-major", "0x0006" },
131     { TYPE_X86_CPU, "hv-version-id-minor", "0x0001" },
132     { "ICH9-LPC", "x-keep-pci-slot-hpc", "false" },
133 };
134 const size_t pc_compat_6_1_len = G_N_ELEMENTS(pc_compat_6_1);
135 
136 GlobalProperty pc_compat_6_0[] = {
137     { "qemu64" "-" TYPE_X86_CPU, "family", "6" },
138     { "qemu64" "-" TYPE_X86_CPU, "model", "6" },
139     { "qemu64" "-" TYPE_X86_CPU, "stepping", "3" },
140     { TYPE_X86_CPU, "x-vendor-cpuid-only", "off" },
141     { "ICH9-LPC", ACPI_PM_PROP_ACPI_PCIHP_BRIDGE, "off" },
142     { "ICH9-LPC", "x-keep-pci-slot-hpc", "true" },
143 };
144 const size_t pc_compat_6_0_len = G_N_ELEMENTS(pc_compat_6_0);
145 
146 GlobalProperty pc_compat_5_2[] = {
147     { "ICH9-LPC", "x-smi-cpu-hotunplug", "off" },
148 };
149 const size_t pc_compat_5_2_len = G_N_ELEMENTS(pc_compat_5_2);
150 
151 GlobalProperty pc_compat_5_1[] = {
152     { "ICH9-LPC", "x-smi-cpu-hotplug", "off" },
153     { TYPE_X86_CPU, "kvm-msi-ext-dest-id", "off" },
154 };
155 const size_t pc_compat_5_1_len = G_N_ELEMENTS(pc_compat_5_1);
156 
157 GlobalProperty pc_compat_5_0[] = {
158 };
159 const size_t pc_compat_5_0_len = G_N_ELEMENTS(pc_compat_5_0);
160 
161 GlobalProperty pc_compat_4_2[] = {
162     { "mch", "smbase-smram", "off" },
163 };
164 const size_t pc_compat_4_2_len = G_N_ELEMENTS(pc_compat_4_2);
165 
166 GlobalProperty pc_compat_4_1[] = {};
167 const size_t pc_compat_4_1_len = G_N_ELEMENTS(pc_compat_4_1);
168 
169 GlobalProperty pc_compat_4_0[] = {};
170 const size_t pc_compat_4_0_len = G_N_ELEMENTS(pc_compat_4_0);
171 
172 GlobalProperty pc_compat_3_1[] = {
173     { "intel-iommu", "dma-drain", "off" },
174     { "Opteron_G3" "-" TYPE_X86_CPU, "rdtscp", "off" },
175     { "Opteron_G4" "-" TYPE_X86_CPU, "rdtscp", "off" },
176     { "Opteron_G4" "-" TYPE_X86_CPU, "npt", "off" },
177     { "Opteron_G4" "-" TYPE_X86_CPU, "nrip-save", "off" },
178     { "Opteron_G5" "-" TYPE_X86_CPU, "rdtscp", "off" },
179     { "Opteron_G5" "-" TYPE_X86_CPU, "npt", "off" },
180     { "Opteron_G5" "-" TYPE_X86_CPU, "nrip-save", "off" },
181     { "EPYC" "-" TYPE_X86_CPU, "npt", "off" },
182     { "EPYC" "-" TYPE_X86_CPU, "nrip-save", "off" },
183     { "EPYC-IBPB" "-" TYPE_X86_CPU, "npt", "off" },
184     { "EPYC-IBPB" "-" TYPE_X86_CPU, "nrip-save", "off" },
185     { "Skylake-Client" "-" TYPE_X86_CPU,      "mpx", "on" },
186     { "Skylake-Client-IBRS" "-" TYPE_X86_CPU, "mpx", "on" },
187     { "Skylake-Server" "-" TYPE_X86_CPU,      "mpx", "on" },
188     { "Skylake-Server-IBRS" "-" TYPE_X86_CPU, "mpx", "on" },
189     { "Cascadelake-Server" "-" TYPE_X86_CPU,  "mpx", "on" },
190     { "Icelake-Client" "-" TYPE_X86_CPU,      "mpx", "on" },
191     { "Icelake-Server" "-" TYPE_X86_CPU,      "mpx", "on" },
192     { "Cascadelake-Server" "-" TYPE_X86_CPU, "stepping", "5" },
193     { TYPE_X86_CPU, "x-intel-pt-auto-level", "off" },
194 };
195 const size_t pc_compat_3_1_len = G_N_ELEMENTS(pc_compat_3_1);
196 
197 GlobalProperty pc_compat_3_0[] = {
198     { TYPE_X86_CPU, "x-hv-synic-kvm-only", "on" },
199     { "Skylake-Server" "-" TYPE_X86_CPU, "pku", "off" },
200     { "Skylake-Server-IBRS" "-" TYPE_X86_CPU, "pku", "off" },
201 };
202 const size_t pc_compat_3_0_len = G_N_ELEMENTS(pc_compat_3_0);
203 
204 GlobalProperty pc_compat_2_12[] = {
205     { TYPE_X86_CPU, "legacy-cache", "on" },
206     { TYPE_X86_CPU, "topoext", "off" },
207     { "EPYC-" TYPE_X86_CPU, "xlevel", "0x8000000a" },
208     { "EPYC-IBPB-" TYPE_X86_CPU, "xlevel", "0x8000000a" },
209 };
210 const size_t pc_compat_2_12_len = G_N_ELEMENTS(pc_compat_2_12);
211 
212 GlobalProperty pc_compat_2_11[] = {
213     { TYPE_X86_CPU, "x-migrate-smi-count", "off" },
214     { "Skylake-Server" "-" TYPE_X86_CPU, "clflushopt", "off" },
215 };
216 const size_t pc_compat_2_11_len = G_N_ELEMENTS(pc_compat_2_11);
217 
218 GlobalProperty pc_compat_2_10[] = {
219     { TYPE_X86_CPU, "x-hv-max-vps", "0x40" },
220     { "i440FX-pcihost", "x-pci-hole64-fix", "off" },
221     { "q35-pcihost", "x-pci-hole64-fix", "off" },
222 };
223 const size_t pc_compat_2_10_len = G_N_ELEMENTS(pc_compat_2_10);
224 
225 GlobalProperty pc_compat_2_9[] = {
226     { "mch", "extended-tseg-mbytes", "0" },
227 };
228 const size_t pc_compat_2_9_len = G_N_ELEMENTS(pc_compat_2_9);
229 
230 GlobalProperty pc_compat_2_8[] = {
231     { TYPE_X86_CPU, "tcg-cpuid", "off" },
232     { "kvmclock", "x-mach-use-reliable-get-clock", "off" },
233     { "ICH9-LPC", "x-smi-broadcast", "off" },
234     { TYPE_X86_CPU, "vmware-cpuid-freq", "off" },
235     { "Haswell-" TYPE_X86_CPU, "stepping", "1" },
236 };
237 const size_t pc_compat_2_8_len = G_N_ELEMENTS(pc_compat_2_8);
238 
239 GlobalProperty pc_compat_2_7[] = {
240     { TYPE_X86_CPU, "l3-cache", "off" },
241     { TYPE_X86_CPU, "full-cpuid-auto-level", "off" },
242     { "Opteron_G3" "-" TYPE_X86_CPU, "family", "15" },
243     { "Opteron_G3" "-" TYPE_X86_CPU, "model", "6" },
244     { "Opteron_G3" "-" TYPE_X86_CPU, "stepping", "1" },
245     { "isa-pcspk", "migrate", "off" },
246 };
247 const size_t pc_compat_2_7_len = G_N_ELEMENTS(pc_compat_2_7);
248 
249 GlobalProperty pc_compat_2_6[] = {
250     { TYPE_X86_CPU, "cpuid-0xb", "off" },
251     { "vmxnet3", "romfile", "" },
252     { TYPE_X86_CPU, "fill-mtrr-mask", "off" },
253     { "apic-common", "legacy-instance-id", "on", }
254 };
255 const size_t pc_compat_2_6_len = G_N_ELEMENTS(pc_compat_2_6);
256 
257 GlobalProperty pc_compat_2_5[] = {};
258 const size_t pc_compat_2_5_len = G_N_ELEMENTS(pc_compat_2_5);
259 
260 GlobalProperty pc_compat_2_4[] = {
261     PC_CPU_MODEL_IDS("2.4.0")
262     { "Haswell-" TYPE_X86_CPU, "abm", "off" },
263     { "Haswell-noTSX-" TYPE_X86_CPU, "abm", "off" },
264     { "Broadwell-" TYPE_X86_CPU, "abm", "off" },
265     { "Broadwell-noTSX-" TYPE_X86_CPU, "abm", "off" },
266     { "host" "-" TYPE_X86_CPU, "host-cache-info", "on" },
267     { TYPE_X86_CPU, "check", "off" },
268     { "qemu64" "-" TYPE_X86_CPU, "sse4a", "on" },
269     { "qemu64" "-" TYPE_X86_CPU, "abm", "on" },
270     { "qemu64" "-" TYPE_X86_CPU, "popcnt", "on" },
271     { "qemu32" "-" TYPE_X86_CPU, "popcnt", "on" },
272     { "Opteron_G2" "-" TYPE_X86_CPU, "rdtscp", "on" },
273     { "Opteron_G3" "-" TYPE_X86_CPU, "rdtscp", "on" },
274     { "Opteron_G4" "-" TYPE_X86_CPU, "rdtscp", "on" },
275     { "Opteron_G5" "-" TYPE_X86_CPU, "rdtscp", "on", }
276 };
277 const size_t pc_compat_2_4_len = G_N_ELEMENTS(pc_compat_2_4);
278 
279 /*
280  * @PC_FW_DATA:
281  * Size of the chunk of memory at the top of RAM for the BIOS ACPI tables
282  * and other BIOS datastructures.
283  *
284  * BIOS ACPI tables: 128K. Other BIOS datastructures: less than 4K
285  * reported to be used at the moment, 32K should be enough for a while.
286  */
287 #define PC_FW_DATA (0x20000 + 0x8000)
288 
289 GSIState *pc_gsi_create(qemu_irq **irqs, bool pci_enabled)
290 {
291     GSIState *s;
292 
293     s = g_new0(GSIState, 1);
294     if (kvm_ioapic_in_kernel()) {
295         kvm_pc_setup_irq_routing(pci_enabled);
296     }
297     *irqs = qemu_allocate_irqs(gsi_handler, s, IOAPIC_NUM_PINS);
298 
299     return s;
300 }
301 
302 static void ioport80_write(void *opaque, hwaddr addr, uint64_t data,
303                            unsigned size)
304 {
305 }
306 
307 static uint64_t ioport80_read(void *opaque, hwaddr addr, unsigned size)
308 {
309     return 0xffffffffffffffffULL;
310 }
311 
312 /* MS-DOS compatibility mode FPU exception support */
313 static void ioportF0_write(void *opaque, hwaddr addr, uint64_t data,
314                            unsigned size)
315 {
316     if (tcg_enabled()) {
317         cpu_set_ignne();
318     }
319 }
320 
321 static uint64_t ioportF0_read(void *opaque, hwaddr addr, unsigned size)
322 {
323     return 0xffffffffffffffffULL;
324 }
325 
326 /* PC cmos mappings */
327 
328 #define REG_EQUIPMENT_BYTE          0x14
329 
330 static void cmos_init_hd(MC146818RtcState *s, int type_ofs, int info_ofs,
331                          int16_t cylinders, int8_t heads, int8_t sectors)
332 {
333     mc146818rtc_set_cmos_data(s, type_ofs, 47);
334     mc146818rtc_set_cmos_data(s, info_ofs, cylinders);
335     mc146818rtc_set_cmos_data(s, info_ofs + 1, cylinders >> 8);
336     mc146818rtc_set_cmos_data(s, info_ofs + 2, heads);
337     mc146818rtc_set_cmos_data(s, info_ofs + 3, 0xff);
338     mc146818rtc_set_cmos_data(s, info_ofs + 4, 0xff);
339     mc146818rtc_set_cmos_data(s, info_ofs + 5, 0xc0 | ((heads > 8) << 3));
340     mc146818rtc_set_cmos_data(s, info_ofs + 6, cylinders);
341     mc146818rtc_set_cmos_data(s, info_ofs + 7, cylinders >> 8);
342     mc146818rtc_set_cmos_data(s, info_ofs + 8, sectors);
343 }
344 
345 /* convert boot_device letter to something recognizable by the bios */
346 static int boot_device2nibble(char boot_device)
347 {
348     switch(boot_device) {
349     case 'a':
350     case 'b':
351         return 0x01; /* floppy boot */
352     case 'c':
353         return 0x02; /* hard drive boot */
354     case 'd':
355         return 0x03; /* CD-ROM boot */
356     case 'n':
357         return 0x04; /* Network boot */
358     }
359     return 0;
360 }
361 
362 static void set_boot_dev(PCMachineState *pcms, MC146818RtcState *s,
363                          const char *boot_device, Error **errp)
364 {
365 #define PC_MAX_BOOT_DEVICES 3
366     int nbds, bds[3] = { 0, };
367     int i;
368 
369     nbds = strlen(boot_device);
370     if (nbds > PC_MAX_BOOT_DEVICES) {
371         error_setg(errp, "Too many boot devices for PC");
372         return;
373     }
374     for (i = 0; i < nbds; i++) {
375         bds[i] = boot_device2nibble(boot_device[i]);
376         if (bds[i] == 0) {
377             error_setg(errp, "Invalid boot device for PC: '%c'",
378                        boot_device[i]);
379             return;
380         }
381     }
382     mc146818rtc_set_cmos_data(s, 0x3d, (bds[1] << 4) | bds[0]);
383     mc146818rtc_set_cmos_data(s, 0x38, (bds[2] << 4) | !pcms->fd_bootchk);
384 }
385 
386 static void pc_boot_set(void *opaque, const char *boot_device, Error **errp)
387 {
388     PCMachineState *pcms = opaque;
389     X86MachineState *x86ms = X86_MACHINE(pcms);
390 
391     set_boot_dev(pcms, MC146818_RTC(x86ms->rtc), boot_device, errp);
392 }
393 
394 static void pc_cmos_init_floppy(MC146818RtcState *rtc_state, ISADevice *floppy)
395 {
396     int val, nb;
397     FloppyDriveType fd_type[2] = { FLOPPY_DRIVE_TYPE_NONE,
398                                    FLOPPY_DRIVE_TYPE_NONE };
399 
400 #ifdef CONFIG_FDC_ISA
401     /* floppy type */
402     if (floppy) {
403         for (int i = 0; i < 2; i++) {
404             fd_type[i] = isa_fdc_get_drive_type(floppy, i);
405         }
406     }
407 #endif
408 
409     val = (cmos_get_fd_drive_type(fd_type[0]) << 4) |
410         cmos_get_fd_drive_type(fd_type[1]);
411     mc146818rtc_set_cmos_data(rtc_state, 0x10, val);
412 
413     val = mc146818rtc_get_cmos_data(rtc_state, REG_EQUIPMENT_BYTE);
414     nb = 0;
415     if (fd_type[0] != FLOPPY_DRIVE_TYPE_NONE) {
416         nb++;
417     }
418     if (fd_type[1] != FLOPPY_DRIVE_TYPE_NONE) {
419         nb++;
420     }
421     switch (nb) {
422     case 0:
423         break;
424     case 1:
425         val |= 0x01; /* 1 drive, ready for boot */
426         break;
427     case 2:
428         val |= 0x41; /* 2 drives, ready for boot */
429         break;
430     }
431     mc146818rtc_set_cmos_data(rtc_state, REG_EQUIPMENT_BYTE, val);
432 }
433 
434 typedef struct check_fdc_state {
435     ISADevice *floppy;
436     bool multiple;
437 } CheckFdcState;
438 
439 static int check_fdc(Object *obj, void *opaque)
440 {
441     CheckFdcState *state = opaque;
442     Object *fdc;
443     uint32_t iobase;
444     Error *local_err = NULL;
445 
446     fdc = object_dynamic_cast(obj, TYPE_ISA_FDC);
447     if (!fdc) {
448         return 0;
449     }
450 
451     iobase = object_property_get_uint(obj, "iobase", &local_err);
452     if (local_err || iobase != 0x3f0) {
453         error_free(local_err);
454         return 0;
455     }
456 
457     if (state->floppy) {
458         state->multiple = true;
459     } else {
460         state->floppy = ISA_DEVICE(obj);
461     }
462     return 0;
463 }
464 
465 static const char * const fdc_container_path[] = {
466     "/unattached", "/peripheral", "/peripheral-anon"
467 };
468 
469 /*
470  * Locate the FDC at IO address 0x3f0, in order to configure the CMOS registers
471  * and ACPI objects.
472  */
473 static ISADevice *pc_find_fdc0(void)
474 {
475     int i;
476     Object *container;
477     CheckFdcState state = { 0 };
478 
479     for (i = 0; i < ARRAY_SIZE(fdc_container_path); i++) {
480         container = container_get(qdev_get_machine(), fdc_container_path[i]);
481         object_child_foreach(container, check_fdc, &state);
482     }
483 
484     if (state.multiple) {
485         warn_report("multiple floppy disk controllers with "
486                     "iobase=0x3f0 have been found");
487         error_printf("the one being picked for CMOS setup might not reflect "
488                      "your intent");
489     }
490 
491     return state.floppy;
492 }
493 
494 static void pc_cmos_init_late(PCMachineState *pcms)
495 {
496     X86MachineState *x86ms = X86_MACHINE(pcms);
497     MC146818RtcState *s = MC146818_RTC(x86ms->rtc);
498     int16_t cylinders;
499     int8_t heads, sectors;
500     int val;
501     int i, trans;
502 
503     val = 0;
504     if (pcms->idebus[0] &&
505         ide_get_geometry(pcms->idebus[0], 0,
506                          &cylinders, &heads, &sectors) >= 0) {
507         cmos_init_hd(s, 0x19, 0x1b, cylinders, heads, sectors);
508         val |= 0xf0;
509     }
510     if (pcms->idebus[0] &&
511         ide_get_geometry(pcms->idebus[0], 1,
512                          &cylinders, &heads, &sectors) >= 0) {
513         cmos_init_hd(s, 0x1a, 0x24, cylinders, heads, sectors);
514         val |= 0x0f;
515     }
516     mc146818rtc_set_cmos_data(s, 0x12, val);
517 
518     val = 0;
519     for (i = 0; i < 4; i++) {
520         /* NOTE: ide_get_geometry() returns the physical
521            geometry.  It is always such that: 1 <= sects <= 63, 1
522            <= heads <= 16, 1 <= cylinders <= 16383. The BIOS
523            geometry can be different if a translation is done. */
524         BusState *idebus = pcms->idebus[i / 2];
525         if (idebus &&
526             ide_get_geometry(idebus, i % 2,
527                              &cylinders, &heads, &sectors) >= 0) {
528             trans = ide_get_bios_chs_trans(idebus, i % 2) - 1;
529             assert((trans & ~3) == 0);
530             val |= trans << (i * 2);
531         }
532     }
533     mc146818rtc_set_cmos_data(s, 0x39, val);
534 
535     pc_cmos_init_floppy(s, pc_find_fdc0());
536 
537     /* various important CMOS locations needed by PC/Bochs bios */
538 
539     /* memory size */
540     /* base memory (first MiB) */
541     val = MIN(x86ms->below_4g_mem_size / KiB, 640);
542     mc146818rtc_set_cmos_data(s, 0x15, val);
543     mc146818rtc_set_cmos_data(s, 0x16, val >> 8);
544     /* extended memory (next 64MiB) */
545     if (x86ms->below_4g_mem_size > 1 * MiB) {
546         val = (x86ms->below_4g_mem_size - 1 * MiB) / KiB;
547     } else {
548         val = 0;
549     }
550     if (val > 65535)
551         val = 65535;
552     mc146818rtc_set_cmos_data(s, 0x17, val);
553     mc146818rtc_set_cmos_data(s, 0x18, val >> 8);
554     mc146818rtc_set_cmos_data(s, 0x30, val);
555     mc146818rtc_set_cmos_data(s, 0x31, val >> 8);
556     /* memory between 16MiB and 4GiB */
557     if (x86ms->below_4g_mem_size > 16 * MiB) {
558         val = (x86ms->below_4g_mem_size - 16 * MiB) / (64 * KiB);
559     } else {
560         val = 0;
561     }
562     if (val > 65535)
563         val = 65535;
564     mc146818rtc_set_cmos_data(s, 0x34, val);
565     mc146818rtc_set_cmos_data(s, 0x35, val >> 8);
566     /* memory above 4GiB */
567     val = x86ms->above_4g_mem_size / 65536;
568     mc146818rtc_set_cmos_data(s, 0x5b, val);
569     mc146818rtc_set_cmos_data(s, 0x5c, val >> 8);
570     mc146818rtc_set_cmos_data(s, 0x5d, val >> 16);
571 
572     val = 0;
573     val |= 0x02; /* FPU is there */
574     val |= 0x04; /* PS/2 mouse installed */
575     mc146818rtc_set_cmos_data(s, REG_EQUIPMENT_BYTE, val);
576 }
577 
578 static void handle_a20_line_change(void *opaque, int irq, int level)
579 {
580     X86CPU *cpu = opaque;
581 
582     /* XXX: send to all CPUs ? */
583     /* XXX: add logic to handle multiple A20 line sources */
584     x86_cpu_set_a20(cpu, level);
585 }
586 
587 #define NE2000_NB_MAX 6
588 
589 static const int ne2000_io[NE2000_NB_MAX] = { 0x300, 0x320, 0x340, 0x360,
590                                               0x280, 0x380 };
591 static const int ne2000_irq[NE2000_NB_MAX] = { 9, 10, 11, 3, 4, 5 };
592 
593 static gboolean pc_init_ne2k_isa(ISABus *bus, NICInfo *nd, Error **errp)
594 {
595     static int nb_ne2k = 0;
596 
597     if (nb_ne2k == NE2000_NB_MAX) {
598         error_setg(errp,
599                    "maximum number of ISA NE2000 devices exceeded");
600         return false;
601     }
602     isa_ne2000_init(bus, ne2000_io[nb_ne2k],
603                     ne2000_irq[nb_ne2k], nd);
604     nb_ne2k++;
605     return true;
606 }
607 
608 void pc_acpi_smi_interrupt(void *opaque, int irq, int level)
609 {
610     X86CPU *cpu = opaque;
611 
612     if (level) {
613         cpu_interrupt(CPU(cpu), CPU_INTERRUPT_SMI);
614     }
615 }
616 
617 static
618 void pc_machine_done(Notifier *notifier, void *data)
619 {
620     PCMachineState *pcms = container_of(notifier,
621                                         PCMachineState, machine_done);
622     X86MachineState *x86ms = X86_MACHINE(pcms);
623 
624     cxl_hook_up_pxb_registers(pcms->pcibus, &pcms->cxl_devices_state,
625                               &error_fatal);
626 
627     if (pcms->cxl_devices_state.is_enabled) {
628         cxl_fmws_link_targets(&pcms->cxl_devices_state, &error_fatal);
629     }
630 
631     /* set the number of CPUs */
632     x86_rtc_set_cpus_count(x86ms->rtc, x86ms->boot_cpus);
633 
634     pci_bus_add_fw_cfg_extra_pci_roots(x86ms->fw_cfg, pcms->pcibus,
635                                        &error_abort);
636 
637     acpi_setup();
638     if (x86ms->fw_cfg) {
639         fw_cfg_build_smbios(pcms, x86ms->fw_cfg, pcms->smbios_entry_point_type);
640         fw_cfg_add_e820(x86ms->fw_cfg);
641         fw_cfg_build_feature_control(MACHINE(pcms), x86ms->fw_cfg);
642         /* update FW_CFG_NB_CPUS to account for -device added CPUs */
643         fw_cfg_modify_i16(x86ms->fw_cfg, FW_CFG_NB_CPUS, x86ms->boot_cpus);
644     }
645 
646     pc_cmos_init_late(pcms);
647 }
648 
649 /* setup pci memory address space mapping into system address space */
650 void pc_pci_as_mapping_init(MemoryRegion *system_memory,
651                             MemoryRegion *pci_address_space)
652 {
653     /* Set to lower priority than RAM */
654     memory_region_add_subregion_overlap(system_memory, 0x0,
655                                         pci_address_space, -1);
656 }
657 
658 void xen_load_linux(PCMachineState *pcms)
659 {
660     int i;
661     FWCfgState *fw_cfg;
662     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
663     X86MachineState *x86ms = X86_MACHINE(pcms);
664 
665     assert(MACHINE(pcms)->kernel_filename != NULL);
666 
667     fw_cfg = fw_cfg_init_io_dma(FW_CFG_IO_BASE, FW_CFG_IO_BASE + 4,
668                                 &address_space_memory);
669     fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, x86ms->boot_cpus);
670     rom_set_fw(fw_cfg);
671 
672     x86_load_linux(x86ms, fw_cfg, PC_FW_DATA, pcmc->pvh_enabled);
673     for (i = 0; i < nb_option_roms; i++) {
674         assert(!strcmp(option_rom[i].name, "linuxboot.bin") ||
675                !strcmp(option_rom[i].name, "linuxboot_dma.bin") ||
676                !strcmp(option_rom[i].name, "pvh.bin") ||
677                !strcmp(option_rom[i].name, "multiboot.bin") ||
678                !strcmp(option_rom[i].name, "multiboot_dma.bin"));
679         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
680     }
681     x86ms->fw_cfg = fw_cfg;
682 }
683 
684 #define PC_ROM_MIN_VGA     0xc0000
685 #define PC_ROM_MIN_OPTION  0xc8000
686 #define PC_ROM_MAX         0xe0000
687 #define PC_ROM_ALIGN       0x800
688 #define PC_ROM_SIZE        (PC_ROM_MAX - PC_ROM_MIN_VGA)
689 
690 static hwaddr pc_above_4g_end(PCMachineState *pcms)
691 {
692     X86MachineState *x86ms = X86_MACHINE(pcms);
693 
694     if (pcms->sgx_epc.size != 0) {
695         return sgx_epc_above_4g_end(&pcms->sgx_epc);
696     }
697 
698     return x86ms->above_4g_mem_start + x86ms->above_4g_mem_size;
699 }
700 
701 static void pc_get_device_memory_range(PCMachineState *pcms,
702                                        hwaddr *base,
703                                        ram_addr_t *device_mem_size)
704 {
705     MachineState *machine = MACHINE(pcms);
706     ram_addr_t size;
707     hwaddr addr;
708 
709     size = machine->maxram_size - machine->ram_size;
710     addr = ROUND_UP(pc_above_4g_end(pcms), 1 * GiB);
711 
712     /* size device region assuming 1G page max alignment per slot */
713     size += (1 * GiB) * machine->ram_slots;
714 
715     *base = addr;
716     *device_mem_size = size;
717 }
718 
719 static uint64_t pc_get_cxl_range_start(PCMachineState *pcms)
720 {
721     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
722     MachineState *ms = MACHINE(pcms);
723     hwaddr cxl_base;
724     ram_addr_t size;
725 
726     if (pcmc->has_reserved_memory &&
727         (ms->ram_size < ms->maxram_size)) {
728         pc_get_device_memory_range(pcms, &cxl_base, &size);
729         cxl_base += size;
730     } else {
731         cxl_base = pc_above_4g_end(pcms);
732     }
733 
734     return cxl_base;
735 }
736 
737 static uint64_t pc_get_cxl_range_end(PCMachineState *pcms)
738 {
739     uint64_t start = pc_get_cxl_range_start(pcms) + MiB;
740 
741     if (pcms->cxl_devices_state.fixed_windows) {
742         GList *it;
743 
744         start = ROUND_UP(start, 256 * MiB);
745         for (it = pcms->cxl_devices_state.fixed_windows; it; it = it->next) {
746             CXLFixedWindow *fw = it->data;
747             start += fw->size;
748         }
749     }
750 
751     return start;
752 }
753 
754 static hwaddr pc_max_used_gpa(PCMachineState *pcms, uint64_t pci_hole64_size)
755 {
756     X86CPU *cpu = X86_CPU(first_cpu);
757     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
758     MachineState *ms = MACHINE(pcms);
759 
760     if (cpu->env.features[FEAT_8000_0001_EDX] & CPUID_EXT2_LM) {
761         /* 64-bit systems */
762         return pc_pci_hole64_start() + pci_hole64_size - 1;
763     }
764 
765     /* 32-bit systems */
766     if (pcmc->broken_32bit_mem_addr_check) {
767         /* old value for compatibility reasons */
768         return ((hwaddr)1 << cpu->phys_bits) - 1;
769     }
770 
771     /*
772      * 32-bit systems don't have hole64 but they might have a region for
773      * memory devices. Even if additional hotplugged memory devices might
774      * not be usable by most guest OSes, we need to still consider them for
775      * calculating the highest possible GPA so that we can properly report
776      * if someone configures them on a CPU that cannot possibly address them.
777      */
778     if (pcmc->has_reserved_memory &&
779         (ms->ram_size < ms->maxram_size)) {
780         hwaddr devmem_start;
781         ram_addr_t devmem_size;
782 
783         pc_get_device_memory_range(pcms, &devmem_start, &devmem_size);
784         devmem_start += devmem_size;
785         return devmem_start - 1;
786     }
787 
788     /* configuration without any memory hotplug */
789     return pc_above_4g_end(pcms) - 1;
790 }
791 
792 /*
793  * AMD systems with an IOMMU have an additional hole close to the
794  * 1Tb, which are special GPAs that cannot be DMA mapped. Depending
795  * on kernel version, VFIO may or may not let you DMA map those ranges.
796  * Starting Linux v5.4 we validate it, and can't create guests on AMD machines
797  * with certain memory sizes. It's also wrong to use those IOVA ranges
798  * in detriment of leading to IOMMU INVALID_DEVICE_REQUEST or worse.
799  * The ranges reserved for Hyper-Transport are:
800  *
801  * FD_0000_0000h - FF_FFFF_FFFFh
802  *
803  * The ranges represent the following:
804  *
805  * Base Address   Top Address  Use
806  *
807  * FD_0000_0000h FD_F7FF_FFFFh Reserved interrupt address space
808  * FD_F800_0000h FD_F8FF_FFFFh Interrupt/EOI IntCtl
809  * FD_F900_0000h FD_F90F_FFFFh Legacy PIC IACK
810  * FD_F910_0000h FD_F91F_FFFFh System Management
811  * FD_F920_0000h FD_FAFF_FFFFh Reserved Page Tables
812  * FD_FB00_0000h FD_FBFF_FFFFh Address Translation
813  * FD_FC00_0000h FD_FDFF_FFFFh I/O Space
814  * FD_FE00_0000h FD_FFFF_FFFFh Configuration
815  * FE_0000_0000h FE_1FFF_FFFFh Extended Configuration/Device Messages
816  * FE_2000_0000h FF_FFFF_FFFFh Reserved
817  *
818  * See AMD IOMMU spec, section 2.1.2 "IOMMU Logical Topology",
819  * Table 3: Special Address Controls (GPA) for more information.
820  */
821 #define AMD_HT_START         0xfd00000000UL
822 #define AMD_HT_END           0xffffffffffUL
823 #define AMD_ABOVE_1TB_START  (AMD_HT_END + 1)
824 #define AMD_HT_SIZE          (AMD_ABOVE_1TB_START - AMD_HT_START)
825 
826 void pc_memory_init(PCMachineState *pcms,
827                     MemoryRegion *system_memory,
828                     MemoryRegion *rom_memory,
829                     uint64_t pci_hole64_size)
830 {
831     int linux_boot, i;
832     MemoryRegion *option_rom_mr;
833     MemoryRegion *ram_below_4g, *ram_above_4g;
834     FWCfgState *fw_cfg;
835     MachineState *machine = MACHINE(pcms);
836     MachineClass *mc = MACHINE_GET_CLASS(machine);
837     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
838     X86MachineState *x86ms = X86_MACHINE(pcms);
839     hwaddr maxphysaddr, maxusedaddr;
840     hwaddr cxl_base, cxl_resv_end = 0;
841     X86CPU *cpu = X86_CPU(first_cpu);
842 
843     assert(machine->ram_size == x86ms->below_4g_mem_size +
844                                 x86ms->above_4g_mem_size);
845 
846     linux_boot = (machine->kernel_filename != NULL);
847 
848     /*
849      * The HyperTransport range close to the 1T boundary is unique to AMD
850      * hosts with IOMMUs enabled. Restrict the ram-above-4g relocation
851      * to above 1T to AMD vCPUs only. @enforce_amd_1tb_hole is only false in
852      * older machine types (<= 7.0) for compatibility purposes.
853      */
854     if (IS_AMD_CPU(&cpu->env) && pcmc->enforce_amd_1tb_hole) {
855         /* Bail out if max possible address does not cross HT range */
856         if (pc_max_used_gpa(pcms, pci_hole64_size) >= AMD_HT_START) {
857             x86ms->above_4g_mem_start = AMD_ABOVE_1TB_START;
858         }
859 
860         /*
861          * Advertise the HT region if address space covers the reserved
862          * region or if we relocate.
863          */
864         if (cpu->phys_bits >= 40) {
865             e820_add_entry(AMD_HT_START, AMD_HT_SIZE, E820_RESERVED);
866         }
867     }
868 
869     /*
870      * phys-bits is required to be appropriately configured
871      * to make sure max used GPA is reachable.
872      */
873     maxusedaddr = pc_max_used_gpa(pcms, pci_hole64_size);
874     maxphysaddr = ((hwaddr)1 << cpu->phys_bits) - 1;
875     if (maxphysaddr < maxusedaddr) {
876         error_report("Address space limit 0x%"PRIx64" < 0x%"PRIx64
877                      " phys-bits too low (%u)",
878                      maxphysaddr, maxusedaddr, cpu->phys_bits);
879         exit(EXIT_FAILURE);
880     }
881 
882     /*
883      * Split single memory region and use aliases to address portions of it,
884      * done for backwards compatibility with older qemus.
885      */
886     ram_below_4g = g_malloc(sizeof(*ram_below_4g));
887     memory_region_init_alias(ram_below_4g, NULL, "ram-below-4g", machine->ram,
888                              0, x86ms->below_4g_mem_size);
889     memory_region_add_subregion(system_memory, 0, ram_below_4g);
890     e820_add_entry(0, x86ms->below_4g_mem_size, E820_RAM);
891     if (x86ms->above_4g_mem_size > 0) {
892         ram_above_4g = g_malloc(sizeof(*ram_above_4g));
893         memory_region_init_alias(ram_above_4g, NULL, "ram-above-4g",
894                                  machine->ram,
895                                  x86ms->below_4g_mem_size,
896                                  x86ms->above_4g_mem_size);
897         memory_region_add_subregion(system_memory, x86ms->above_4g_mem_start,
898                                     ram_above_4g);
899         e820_add_entry(x86ms->above_4g_mem_start, x86ms->above_4g_mem_size,
900                        E820_RAM);
901     }
902 
903     if (pcms->sgx_epc.size != 0) {
904         e820_add_entry(pcms->sgx_epc.base, pcms->sgx_epc.size, E820_RESERVED);
905     }
906 
907     if (!pcmc->has_reserved_memory &&
908         (machine->ram_slots ||
909          (machine->maxram_size > machine->ram_size))) {
910 
911         error_report("\"-memory 'slots|maxmem'\" is not supported by: %s",
912                      mc->name);
913         exit(EXIT_FAILURE);
914     }
915 
916     /* initialize device memory address space */
917     if (pcmc->has_reserved_memory &&
918         (machine->ram_size < machine->maxram_size)) {
919         ram_addr_t device_mem_size;
920         hwaddr device_mem_base;
921 
922         if (machine->ram_slots > ACPI_MAX_RAM_SLOTS) {
923             error_report("unsupported amount of memory slots: %"PRIu64,
924                          machine->ram_slots);
925             exit(EXIT_FAILURE);
926         }
927 
928         if (QEMU_ALIGN_UP(machine->maxram_size,
929                           TARGET_PAGE_SIZE) != machine->maxram_size) {
930             error_report("maximum memory size must by aligned to multiple of "
931                          "%d bytes", TARGET_PAGE_SIZE);
932             exit(EXIT_FAILURE);
933         }
934 
935         pc_get_device_memory_range(pcms, &device_mem_base, &device_mem_size);
936 
937         if (device_mem_base + device_mem_size < device_mem_size) {
938             error_report("unsupported amount of maximum memory: " RAM_ADDR_FMT,
939                          machine->maxram_size);
940             exit(EXIT_FAILURE);
941         }
942         machine_memory_devices_init(machine, device_mem_base, device_mem_size);
943     }
944 
945     if (pcms->cxl_devices_state.is_enabled) {
946         MemoryRegion *mr = &pcms->cxl_devices_state.host_mr;
947         hwaddr cxl_size = MiB;
948 
949         cxl_base = pc_get_cxl_range_start(pcms);
950         memory_region_init(mr, OBJECT(machine), "cxl_host_reg", cxl_size);
951         memory_region_add_subregion(system_memory, cxl_base, mr);
952         cxl_resv_end = cxl_base + cxl_size;
953         if (pcms->cxl_devices_state.fixed_windows) {
954             hwaddr cxl_fmw_base;
955             GList *it;
956 
957             cxl_fmw_base = ROUND_UP(cxl_base + cxl_size, 256 * MiB);
958             for (it = pcms->cxl_devices_state.fixed_windows; it; it = it->next) {
959                 CXLFixedWindow *fw = it->data;
960 
961                 fw->base = cxl_fmw_base;
962                 memory_region_init_io(&fw->mr, OBJECT(machine), &cfmws_ops, fw,
963                                       "cxl-fixed-memory-region", fw->size);
964                 memory_region_add_subregion(system_memory, fw->base, &fw->mr);
965                 cxl_fmw_base += fw->size;
966                 cxl_resv_end = cxl_fmw_base;
967             }
968         }
969     }
970 
971     /* Initialize PC system firmware */
972     pc_system_firmware_init(pcms, rom_memory);
973 
974     option_rom_mr = g_malloc(sizeof(*option_rom_mr));
975     if (machine_require_guest_memfd(machine)) {
976         memory_region_init_ram_guest_memfd(option_rom_mr, NULL, "pc.rom",
977                                            PC_ROM_SIZE, &error_fatal);
978     } else {
979         memory_region_init_ram(option_rom_mr, NULL, "pc.rom", PC_ROM_SIZE,
980                                &error_fatal);
981         if (pcmc->pci_enabled) {
982             memory_region_set_readonly(option_rom_mr, true);
983         }
984     }
985     memory_region_add_subregion_overlap(rom_memory,
986                                         PC_ROM_MIN_VGA,
987                                         option_rom_mr,
988                                         1);
989 
990     fw_cfg = fw_cfg_arch_create(machine,
991                                 x86ms->boot_cpus, x86ms->apic_id_limit);
992 
993     rom_set_fw(fw_cfg);
994 
995     if (machine->device_memory) {
996         uint64_t *val = g_malloc(sizeof(*val));
997         uint64_t res_mem_end = machine->device_memory->base;
998 
999         if (!pcmc->broken_reserved_end) {
1000             res_mem_end += memory_region_size(&machine->device_memory->mr);
1001         }
1002 
1003         if (pcms->cxl_devices_state.is_enabled) {
1004             res_mem_end = cxl_resv_end;
1005         }
1006         *val = cpu_to_le64(ROUND_UP(res_mem_end, 1 * GiB));
1007         fw_cfg_add_file(fw_cfg, "etc/reserved-memory-end", val, sizeof(*val));
1008     }
1009 
1010     if (linux_boot) {
1011         x86_load_linux(x86ms, fw_cfg, PC_FW_DATA, pcmc->pvh_enabled);
1012     }
1013 
1014     for (i = 0; i < nb_option_roms; i++) {
1015         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1016     }
1017     x86ms->fw_cfg = fw_cfg;
1018 
1019     /* Init default IOAPIC address space */
1020     x86ms->ioapic_as = &address_space_memory;
1021 
1022     /* Init ACPI memory hotplug IO base address */
1023     pcms->memhp_io_base = ACPI_MEMORY_HOTPLUG_BASE;
1024 }
1025 
1026 /*
1027  * The 64bit pci hole starts after "above 4G RAM" and
1028  * potentially the space reserved for memory hotplug.
1029  */
1030 uint64_t pc_pci_hole64_start(void)
1031 {
1032     PCMachineState *pcms = PC_MACHINE(qdev_get_machine());
1033     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1034     MachineState *ms = MACHINE(pcms);
1035     uint64_t hole64_start = 0;
1036     ram_addr_t size = 0;
1037 
1038     if (pcms->cxl_devices_state.is_enabled) {
1039         hole64_start = pc_get_cxl_range_end(pcms);
1040     } else if (pcmc->has_reserved_memory && (ms->ram_size < ms->maxram_size)) {
1041         pc_get_device_memory_range(pcms, &hole64_start, &size);
1042         if (!pcmc->broken_reserved_end) {
1043             hole64_start += size;
1044         }
1045     } else {
1046         hole64_start = pc_above_4g_end(pcms);
1047     }
1048 
1049     return ROUND_UP(hole64_start, 1 * GiB);
1050 }
1051 
1052 DeviceState *pc_vga_init(ISABus *isa_bus, PCIBus *pci_bus)
1053 {
1054     DeviceState *dev = NULL;
1055 
1056     rom_set_order_override(FW_CFG_ORDER_OVERRIDE_VGA);
1057     if (pci_bus) {
1058         PCIDevice *pcidev = pci_vga_init(pci_bus);
1059         dev = pcidev ? &pcidev->qdev : NULL;
1060     } else if (isa_bus) {
1061         ISADevice *isadev = isa_vga_init(isa_bus);
1062         dev = isadev ? DEVICE(isadev) : NULL;
1063     }
1064     rom_reset_order_override();
1065     return dev;
1066 }
1067 
1068 static const MemoryRegionOps ioport80_io_ops = {
1069     .write = ioport80_write,
1070     .read = ioport80_read,
1071     .endianness = DEVICE_LITTLE_ENDIAN,
1072     .impl = {
1073         .min_access_size = 1,
1074         .max_access_size = 1,
1075     },
1076 };
1077 
1078 static const MemoryRegionOps ioportF0_io_ops = {
1079     .write = ioportF0_write,
1080     .read = ioportF0_read,
1081     .endianness = DEVICE_LITTLE_ENDIAN,
1082     .impl = {
1083         .min_access_size = 1,
1084         .max_access_size = 1,
1085     },
1086 };
1087 
1088 static void pc_superio_init(ISABus *isa_bus, bool create_fdctrl,
1089                             bool create_i8042, bool no_vmport, Error **errp)
1090 {
1091     int i;
1092     DriveInfo *fd[MAX_FD];
1093     qemu_irq *a20_line;
1094     ISADevice *i8042, *port92, *vmmouse;
1095 
1096     serial_hds_isa_init(isa_bus, 0, MAX_ISA_SERIAL_PORTS);
1097     parallel_hds_isa_init(isa_bus, MAX_PARALLEL_PORTS);
1098 
1099     for (i = 0; i < MAX_FD; i++) {
1100         fd[i] = drive_get(IF_FLOPPY, 0, i);
1101         create_fdctrl |= !!fd[i];
1102     }
1103     if (create_fdctrl) {
1104 #ifdef CONFIG_FDC_ISA
1105         ISADevice *fdc = isa_new(TYPE_ISA_FDC);
1106         if (fdc) {
1107             isa_realize_and_unref(fdc, isa_bus, &error_fatal);
1108             isa_fdc_init_drives(fdc, fd);
1109         }
1110 #endif
1111     }
1112 
1113     if (!create_i8042) {
1114         if (!no_vmport) {
1115             error_setg(errp,
1116                        "vmport requires the i8042 controller to be enabled");
1117         }
1118         return;
1119     }
1120 
1121     i8042 = isa_create_simple(isa_bus, TYPE_I8042);
1122     if (!no_vmport) {
1123         isa_create_simple(isa_bus, TYPE_VMPORT);
1124         vmmouse = isa_try_new("vmmouse");
1125     } else {
1126         vmmouse = NULL;
1127     }
1128     if (vmmouse) {
1129         object_property_set_link(OBJECT(vmmouse), TYPE_I8042, OBJECT(i8042),
1130                                  &error_abort);
1131         isa_realize_and_unref(vmmouse, isa_bus, &error_fatal);
1132     }
1133     port92 = isa_create_simple(isa_bus, TYPE_PORT92);
1134 
1135     a20_line = qemu_allocate_irqs(handle_a20_line_change, first_cpu, 2);
1136     qdev_connect_gpio_out_named(DEVICE(i8042),
1137                                 I8042_A20_LINE, 0, a20_line[0]);
1138     qdev_connect_gpio_out_named(DEVICE(port92),
1139                                 PORT92_A20_LINE, 0, a20_line[1]);
1140     g_free(a20_line);
1141 }
1142 
1143 void pc_basic_device_init(struct PCMachineState *pcms,
1144                           ISABus *isa_bus, qemu_irq *gsi,
1145                           ISADevice *rtc_state,
1146                           bool create_fdctrl,
1147                           uint32_t hpet_irqs)
1148 {
1149     int i;
1150     DeviceState *hpet = NULL;
1151     int pit_isa_irq = 0;
1152     qemu_irq pit_alt_irq = NULL;
1153     ISADevice *pit = NULL;
1154     MemoryRegion *ioport80_io = g_new(MemoryRegion, 1);
1155     MemoryRegion *ioportF0_io = g_new(MemoryRegion, 1);
1156     X86MachineState *x86ms = X86_MACHINE(pcms);
1157 
1158     memory_region_init_io(ioport80_io, NULL, &ioport80_io_ops, NULL, "ioport80", 1);
1159     memory_region_add_subregion(isa_bus->address_space_io, 0x80, ioport80_io);
1160 
1161     memory_region_init_io(ioportF0_io, NULL, &ioportF0_io_ops, NULL, "ioportF0", 1);
1162     memory_region_add_subregion(isa_bus->address_space_io, 0xf0, ioportF0_io);
1163 
1164     /*
1165      * Check if an HPET shall be created.
1166      */
1167     if (pcms->hpet_enabled) {
1168         qemu_irq rtc_irq;
1169 
1170         hpet = qdev_try_new(TYPE_HPET);
1171         if (!hpet) {
1172             error_report("couldn't create HPET device");
1173             exit(1);
1174         }
1175         /*
1176          * For pc-piix-*, hpet's intcap is always IRQ2. For pc-q35-*,
1177          * use IRQ16~23, IRQ8 and IRQ2.  If the user has already set
1178          * the property, use whatever mask they specified.
1179          */
1180         uint8_t compat = object_property_get_uint(OBJECT(hpet),
1181                 HPET_INTCAP, NULL);
1182         if (!compat) {
1183             qdev_prop_set_uint32(hpet, HPET_INTCAP, hpet_irqs);
1184         }
1185         sysbus_realize_and_unref(SYS_BUS_DEVICE(hpet), &error_fatal);
1186         sysbus_mmio_map(SYS_BUS_DEVICE(hpet), 0, HPET_BASE);
1187 
1188         for (i = 0; i < IOAPIC_NUM_PINS; i++) {
1189             sysbus_connect_irq(SYS_BUS_DEVICE(hpet), i, gsi[i]);
1190         }
1191         pit_isa_irq = -1;
1192         pit_alt_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_PIT_INT);
1193         rtc_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_RTC_INT);
1194 
1195         /* overwrite connection created by south bridge */
1196         qdev_connect_gpio_out(DEVICE(rtc_state), 0, rtc_irq);
1197     }
1198 
1199     object_property_add_alias(OBJECT(pcms), "rtc-time", OBJECT(rtc_state),
1200                               "date");
1201 
1202 #ifdef CONFIG_XEN_EMU
1203     if (xen_mode == XEN_EMULATE) {
1204         xen_overlay_create();
1205         xen_evtchn_create(IOAPIC_NUM_PINS, gsi);
1206         xen_gnttab_create();
1207         xen_xenstore_create();
1208         if (pcms->pcibus) {
1209             pci_create_simple(pcms->pcibus, -1, "xen-platform");
1210         }
1211         xen_bus_init();
1212     }
1213 #endif
1214 
1215     qemu_register_boot_set(pc_boot_set, pcms);
1216     set_boot_dev(pcms, MC146818_RTC(rtc_state),
1217                  MACHINE(pcms)->boot_config.order, &error_fatal);
1218 
1219     if (!xen_enabled() &&
1220         (x86ms->pit == ON_OFF_AUTO_AUTO || x86ms->pit == ON_OFF_AUTO_ON)) {
1221         if (kvm_pit_in_kernel()) {
1222             pit = kvm_pit_init(isa_bus, 0x40);
1223         } else {
1224             pit = i8254_pit_init(isa_bus, 0x40, pit_isa_irq, pit_alt_irq);
1225         }
1226         if (hpet) {
1227             /* connect PIT to output control line of the HPET */
1228             qdev_connect_gpio_out(hpet, 0, qdev_get_gpio_in(DEVICE(pit), 0));
1229         }
1230         object_property_set_link(OBJECT(pcms->pcspk), "pit",
1231                                  OBJECT(pit), &error_fatal);
1232         isa_realize_and_unref(pcms->pcspk, isa_bus, &error_fatal);
1233     }
1234 
1235     if (pcms->vmport == ON_OFF_AUTO_AUTO) {
1236         pcms->vmport = (xen_enabled() || !pcms->i8042_enabled)
1237             ? ON_OFF_AUTO_OFF : ON_OFF_AUTO_ON;
1238     }
1239 
1240     /* Super I/O */
1241     pc_superio_init(isa_bus, create_fdctrl, pcms->i8042_enabled,
1242                     pcms->vmport != ON_OFF_AUTO_ON, &error_fatal);
1243 }
1244 
1245 void pc_nic_init(PCMachineClass *pcmc, ISABus *isa_bus, PCIBus *pci_bus)
1246 {
1247     MachineClass *mc = MACHINE_CLASS(pcmc);
1248     bool default_is_ne2k = g_str_equal(mc->default_nic, TYPE_ISA_NE2000);
1249     NICInfo *nd;
1250 
1251     rom_set_order_override(FW_CFG_ORDER_OVERRIDE_NIC);
1252 
1253     while ((nd = qemu_find_nic_info(TYPE_ISA_NE2000, default_is_ne2k, NULL))) {
1254         pc_init_ne2k_isa(isa_bus, nd, &error_fatal);
1255     }
1256 
1257     /* Anything remaining should be a PCI NIC */
1258     if (pci_bus) {
1259         pci_init_nic_devices(pci_bus, mc->default_nic);
1260     }
1261 
1262     rom_reset_order_override();
1263 }
1264 
1265 void pc_i8259_create(ISABus *isa_bus, qemu_irq *i8259_irqs)
1266 {
1267     qemu_irq *i8259;
1268 
1269     if (kvm_pic_in_kernel()) {
1270         i8259 = kvm_i8259_init(isa_bus);
1271     } else if (xen_enabled()) {
1272         i8259 = xen_interrupt_controller_init();
1273     } else {
1274         i8259 = i8259_init(isa_bus, x86_allocate_cpu_irq());
1275     }
1276 
1277     for (size_t i = 0; i < ISA_NUM_IRQS; i++) {
1278         i8259_irqs[i] = i8259[i];
1279     }
1280 
1281     g_free(i8259);
1282 }
1283 
1284 static void pc_memory_pre_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
1285                                Error **errp)
1286 {
1287     const X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
1288     const MachineState *ms = MACHINE(hotplug_dev);
1289     const bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
1290     Error *local_err = NULL;
1291 
1292     /*
1293      * When "acpi=off" is used with the Q35 machine type, no ACPI is built,
1294      * but pcms->acpi_dev is still created. Check !acpi_enabled in
1295      * addition to cover this case.
1296      */
1297     if (!x86ms->acpi_dev || !x86_machine_is_acpi_enabled(x86ms)) {
1298         error_setg(errp,
1299                    "memory hotplug is not enabled: missing acpi device or acpi disabled");
1300         return;
1301     }
1302 
1303     if (is_nvdimm && !ms->nvdimms_state->is_enabled) {
1304         error_setg(errp, "nvdimm is not enabled: missing 'nvdimm' in '-M'");
1305         return;
1306     }
1307 
1308     hotplug_handler_pre_plug(x86ms->acpi_dev, dev, &local_err);
1309     if (local_err) {
1310         error_propagate(errp, local_err);
1311         return;
1312     }
1313 
1314     pc_dimm_pre_plug(PC_DIMM(dev), MACHINE(hotplug_dev), errp);
1315 }
1316 
1317 static void pc_memory_plug(HotplugHandler *hotplug_dev,
1318                            DeviceState *dev, Error **errp)
1319 {
1320     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1321     X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
1322     MachineState *ms = MACHINE(hotplug_dev);
1323     bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
1324 
1325     pc_dimm_plug(PC_DIMM(dev), MACHINE(pcms));
1326 
1327     if (is_nvdimm) {
1328         nvdimm_plug(ms->nvdimms_state);
1329     }
1330 
1331     hotplug_handler_plug(x86ms->acpi_dev, dev, &error_abort);
1332 }
1333 
1334 static void pc_memory_unplug_request(HotplugHandler *hotplug_dev,
1335                                      DeviceState *dev, Error **errp)
1336 {
1337     X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
1338 
1339     /*
1340      * When "acpi=off" is used with the Q35 machine type, no ACPI is built,
1341      * but pcms->acpi_dev is still created. Check !acpi_enabled in
1342      * addition to cover this case.
1343      */
1344     if (!x86ms->acpi_dev || !x86_machine_is_acpi_enabled(x86ms)) {
1345         error_setg(errp,
1346                    "memory hotplug is not enabled: missing acpi device or acpi disabled");
1347         return;
1348     }
1349 
1350     if (object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM)) {
1351         error_setg(errp, "nvdimm device hot unplug is not supported yet.");
1352         return;
1353     }
1354 
1355     hotplug_handler_unplug_request(x86ms->acpi_dev, dev,
1356                                    errp);
1357 }
1358 
1359 static void pc_memory_unplug(HotplugHandler *hotplug_dev,
1360                              DeviceState *dev, Error **errp)
1361 {
1362     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1363     X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
1364     Error *local_err = NULL;
1365 
1366     hotplug_handler_unplug(x86ms->acpi_dev, dev, &local_err);
1367     if (local_err) {
1368         goto out;
1369     }
1370 
1371     pc_dimm_unplug(PC_DIMM(dev), MACHINE(pcms));
1372     qdev_unrealize(dev);
1373  out:
1374     error_propagate(errp, local_err);
1375 }
1376 
1377 static void pc_hv_balloon_pre_plug(HotplugHandler *hotplug_dev,
1378                                    DeviceState *dev, Error **errp)
1379 {
1380     /* The vmbus handler has no hotplug handler; we should never end up here. */
1381     g_assert(!dev->hotplugged);
1382     memory_device_pre_plug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev), errp);
1383 }
1384 
1385 static void pc_hv_balloon_plug(HotplugHandler *hotplug_dev,
1386                                DeviceState *dev, Error **errp)
1387 {
1388     memory_device_plug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev));
1389 }
1390 
1391 static void pc_machine_device_pre_plug_cb(HotplugHandler *hotplug_dev,
1392                                           DeviceState *dev, Error **errp)
1393 {
1394     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1395         pc_memory_pre_plug(hotplug_dev, dev, errp);
1396     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1397         x86_cpu_pre_plug(hotplug_dev, dev, errp);
1398     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI)) {
1399         virtio_md_pci_pre_plug(VIRTIO_MD_PCI(dev), MACHINE(hotplug_dev), errp);
1400     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
1401         /* Declare the APIC range as the reserved MSI region */
1402         char *resv_prop_str = g_strdup_printf("0xfee00000:0xfeefffff:%d",
1403                                               VIRTIO_IOMMU_RESV_MEM_T_MSI);
1404         QList *reserved_regions = qlist_new();
1405 
1406         qlist_append_str(reserved_regions, resv_prop_str);
1407         qdev_prop_set_array(dev, "reserved-regions", reserved_regions);
1408 
1409         g_free(resv_prop_str);
1410     }
1411 
1412     if (object_dynamic_cast(OBJECT(dev), TYPE_X86_IOMMU_DEVICE) ||
1413         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
1414         PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1415 
1416         if (pcms->iommu) {
1417             error_setg(errp, "QEMU does not support multiple vIOMMUs "
1418                        "for x86 yet.");
1419             return;
1420         }
1421         pcms->iommu = dev;
1422     } else if (object_dynamic_cast(OBJECT(dev), TYPE_HV_BALLOON)) {
1423         pc_hv_balloon_pre_plug(hotplug_dev, dev, errp);
1424     }
1425 }
1426 
1427 static void pc_machine_device_plug_cb(HotplugHandler *hotplug_dev,
1428                                       DeviceState *dev, Error **errp)
1429 {
1430     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1431         pc_memory_plug(hotplug_dev, dev, errp);
1432     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1433         x86_cpu_plug(hotplug_dev, dev, errp);
1434     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI)) {
1435         virtio_md_pci_plug(VIRTIO_MD_PCI(dev), MACHINE(hotplug_dev), errp);
1436     } else if (object_dynamic_cast(OBJECT(dev), TYPE_HV_BALLOON)) {
1437         pc_hv_balloon_plug(hotplug_dev, dev, errp);
1438     }
1439 }
1440 
1441 static void pc_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
1442                                                 DeviceState *dev, Error **errp)
1443 {
1444     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1445         pc_memory_unplug_request(hotplug_dev, dev, errp);
1446     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1447         x86_cpu_unplug_request_cb(hotplug_dev, dev, errp);
1448     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI)) {
1449         virtio_md_pci_unplug_request(VIRTIO_MD_PCI(dev), MACHINE(hotplug_dev),
1450                                      errp);
1451     } else {
1452         error_setg(errp, "acpi: device unplug request for not supported device"
1453                    " type: %s", object_get_typename(OBJECT(dev)));
1454     }
1455 }
1456 
1457 static void pc_machine_device_unplug_cb(HotplugHandler *hotplug_dev,
1458                                         DeviceState *dev, Error **errp)
1459 {
1460     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1461         pc_memory_unplug(hotplug_dev, dev, errp);
1462     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1463         x86_cpu_unplug_cb(hotplug_dev, dev, errp);
1464     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI)) {
1465         virtio_md_pci_unplug(VIRTIO_MD_PCI(dev), MACHINE(hotplug_dev), errp);
1466     } else {
1467         error_setg(errp, "acpi: device unplug for not supported device"
1468                    " type: %s", object_get_typename(OBJECT(dev)));
1469     }
1470 }
1471 
1472 static HotplugHandler *pc_get_hotplug_handler(MachineState *machine,
1473                                              DeviceState *dev)
1474 {
1475     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) ||
1476         object_dynamic_cast(OBJECT(dev), TYPE_CPU) ||
1477         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI) ||
1478         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI) ||
1479         object_dynamic_cast(OBJECT(dev), TYPE_HV_BALLOON) ||
1480         object_dynamic_cast(OBJECT(dev), TYPE_X86_IOMMU_DEVICE)) {
1481         return HOTPLUG_HANDLER(machine);
1482     }
1483 
1484     return NULL;
1485 }
1486 
1487 static void pc_machine_get_vmport(Object *obj, Visitor *v, const char *name,
1488                                   void *opaque, Error **errp)
1489 {
1490     PCMachineState *pcms = PC_MACHINE(obj);
1491     OnOffAuto vmport = pcms->vmport;
1492 
1493     visit_type_OnOffAuto(v, name, &vmport, errp);
1494 }
1495 
1496 static void pc_machine_set_vmport(Object *obj, Visitor *v, const char *name,
1497                                   void *opaque, Error **errp)
1498 {
1499     PCMachineState *pcms = PC_MACHINE(obj);
1500 
1501     visit_type_OnOffAuto(v, name, &pcms->vmport, errp);
1502 }
1503 
1504 static bool pc_machine_get_fd_bootchk(Object *obj, Error **errp)
1505 {
1506     PCMachineState *pcms = PC_MACHINE(obj);
1507 
1508     return pcms->fd_bootchk;
1509 }
1510 
1511 static void pc_machine_set_fd_bootchk(Object *obj, bool value, Error **errp)
1512 {
1513     PCMachineState *pcms = PC_MACHINE(obj);
1514 
1515     pcms->fd_bootchk = value;
1516 }
1517 
1518 static bool pc_machine_get_smbus(Object *obj, Error **errp)
1519 {
1520     PCMachineState *pcms = PC_MACHINE(obj);
1521 
1522     return pcms->smbus_enabled;
1523 }
1524 
1525 static void pc_machine_set_smbus(Object *obj, bool value, Error **errp)
1526 {
1527     PCMachineState *pcms = PC_MACHINE(obj);
1528 
1529     pcms->smbus_enabled = value;
1530 }
1531 
1532 static bool pc_machine_get_sata(Object *obj, Error **errp)
1533 {
1534     PCMachineState *pcms = PC_MACHINE(obj);
1535 
1536     return pcms->sata_enabled;
1537 }
1538 
1539 static void pc_machine_set_sata(Object *obj, bool value, Error **errp)
1540 {
1541     PCMachineState *pcms = PC_MACHINE(obj);
1542 
1543     pcms->sata_enabled = value;
1544 }
1545 
1546 static bool pc_machine_get_hpet(Object *obj, Error **errp)
1547 {
1548     PCMachineState *pcms = PC_MACHINE(obj);
1549 
1550     return pcms->hpet_enabled;
1551 }
1552 
1553 static void pc_machine_set_hpet(Object *obj, bool value, Error **errp)
1554 {
1555     PCMachineState *pcms = PC_MACHINE(obj);
1556 
1557     pcms->hpet_enabled = value;
1558 }
1559 
1560 static bool pc_machine_get_i8042(Object *obj, Error **errp)
1561 {
1562     PCMachineState *pcms = PC_MACHINE(obj);
1563 
1564     return pcms->i8042_enabled;
1565 }
1566 
1567 static void pc_machine_set_i8042(Object *obj, bool value, Error **errp)
1568 {
1569     PCMachineState *pcms = PC_MACHINE(obj);
1570 
1571     pcms->i8042_enabled = value;
1572 }
1573 
1574 static bool pc_machine_get_default_bus_bypass_iommu(Object *obj, Error **errp)
1575 {
1576     PCMachineState *pcms = PC_MACHINE(obj);
1577 
1578     return pcms->default_bus_bypass_iommu;
1579 }
1580 
1581 static void pc_machine_set_default_bus_bypass_iommu(Object *obj, bool value,
1582                                                     Error **errp)
1583 {
1584     PCMachineState *pcms = PC_MACHINE(obj);
1585 
1586     pcms->default_bus_bypass_iommu = value;
1587 }
1588 
1589 static void pc_machine_get_smbios_ep(Object *obj, Visitor *v, const char *name,
1590                                      void *opaque, Error **errp)
1591 {
1592     PCMachineState *pcms = PC_MACHINE(obj);
1593     SmbiosEntryPointType smbios_entry_point_type = pcms->smbios_entry_point_type;
1594 
1595     visit_type_SmbiosEntryPointType(v, name, &smbios_entry_point_type, errp);
1596 }
1597 
1598 static void pc_machine_set_smbios_ep(Object *obj, Visitor *v, const char *name,
1599                                      void *opaque, Error **errp)
1600 {
1601     PCMachineState *pcms = PC_MACHINE(obj);
1602 
1603     visit_type_SmbiosEntryPointType(v, name, &pcms->smbios_entry_point_type, errp);
1604 }
1605 
1606 static void pc_machine_get_max_ram_below_4g(Object *obj, Visitor *v,
1607                                             const char *name, void *opaque,
1608                                             Error **errp)
1609 {
1610     PCMachineState *pcms = PC_MACHINE(obj);
1611     uint64_t value = pcms->max_ram_below_4g;
1612 
1613     visit_type_size(v, name, &value, errp);
1614 }
1615 
1616 static void pc_machine_set_max_ram_below_4g(Object *obj, Visitor *v,
1617                                             const char *name, void *opaque,
1618                                             Error **errp)
1619 {
1620     PCMachineState *pcms = PC_MACHINE(obj);
1621     uint64_t value;
1622 
1623     if (!visit_type_size(v, name, &value, errp)) {
1624         return;
1625     }
1626     if (value > 4 * GiB) {
1627         error_setg(errp,
1628                    "Machine option 'max-ram-below-4g=%"PRIu64
1629                    "' expects size less than or equal to 4G", value);
1630         return;
1631     }
1632 
1633     if (value < 1 * MiB) {
1634         warn_report("Only %" PRIu64 " bytes of RAM below the 4GiB boundary,"
1635                     "BIOS may not work with less than 1MiB", value);
1636     }
1637 
1638     pcms->max_ram_below_4g = value;
1639 }
1640 
1641 static void pc_machine_get_max_fw_size(Object *obj, Visitor *v,
1642                                        const char *name, void *opaque,
1643                                        Error **errp)
1644 {
1645     PCMachineState *pcms = PC_MACHINE(obj);
1646     uint64_t value = pcms->max_fw_size;
1647 
1648     visit_type_size(v, name, &value, errp);
1649 }
1650 
1651 static void pc_machine_set_max_fw_size(Object *obj, Visitor *v,
1652                                        const char *name, void *opaque,
1653                                        Error **errp)
1654 {
1655     PCMachineState *pcms = PC_MACHINE(obj);
1656     uint64_t value;
1657 
1658     if (!visit_type_size(v, name, &value, errp)) {
1659         return;
1660     }
1661 
1662     /*
1663      * We don't have a theoretically justifiable exact lower bound on the base
1664      * address of any flash mapping. In practice, the IO-APIC MMIO range is
1665      * [0xFEE00000..0xFEE01000] -- see IO_APIC_DEFAULT_ADDRESS --, leaving free
1666      * only 18MiB-4KiB below 4GiB. For now, restrict the cumulative mapping to
1667      * 16MiB in size.
1668      */
1669     if (value > 16 * MiB) {
1670         error_setg(errp,
1671                    "User specified max allowed firmware size %" PRIu64 " is "
1672                    "greater than 16MiB. If combined firmware size exceeds "
1673                    "16MiB the system may not boot, or experience intermittent"
1674                    "stability issues.",
1675                    value);
1676         return;
1677     }
1678 
1679     pcms->max_fw_size = value;
1680 }
1681 
1682 
1683 static void pc_machine_initfn(Object *obj)
1684 {
1685     PCMachineState *pcms = PC_MACHINE(obj);
1686     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1687 
1688 #ifdef CONFIG_VMPORT
1689     pcms->vmport = ON_OFF_AUTO_AUTO;
1690 #else
1691     pcms->vmport = ON_OFF_AUTO_OFF;
1692 #endif /* CONFIG_VMPORT */
1693     pcms->max_ram_below_4g = 0; /* use default */
1694     pcms->smbios_entry_point_type = pcmc->default_smbios_ep_type;
1695     pcms->south_bridge = pcmc->default_south_bridge;
1696 
1697     /* acpi build is enabled by default if machine supports it */
1698     pcms->acpi_build_enabled = pcmc->has_acpi_build;
1699     pcms->smbus_enabled = true;
1700     pcms->sata_enabled = true;
1701     pcms->i8042_enabled = true;
1702     pcms->max_fw_size = 8 * MiB;
1703 #ifdef CONFIG_HPET
1704     pcms->hpet_enabled = true;
1705 #endif
1706     pcms->fd_bootchk = true;
1707     pcms->default_bus_bypass_iommu = false;
1708 
1709     pc_system_flash_create(pcms);
1710     pcms->pcspk = isa_new(TYPE_PC_SPEAKER);
1711     object_property_add_alias(OBJECT(pcms), "pcspk-audiodev",
1712                               OBJECT(pcms->pcspk), "audiodev");
1713     if (pcmc->pci_enabled) {
1714         cxl_machine_init(obj, &pcms->cxl_devices_state);
1715     }
1716 
1717     pcms->machine_done.notify = pc_machine_done;
1718     qemu_add_machine_init_done_notifier(&pcms->machine_done);
1719 }
1720 
1721 static void pc_machine_reset(MachineState *machine, ResetType type)
1722 {
1723     CPUState *cs;
1724     X86CPU *cpu;
1725 
1726     qemu_devices_reset(type);
1727 
1728     /* Reset APIC after devices have been reset to cancel
1729      * any changes that qemu_devices_reset() might have done.
1730      */
1731     CPU_FOREACH(cs) {
1732         cpu = X86_CPU(cs);
1733 
1734         x86_cpu_after_reset(cpu);
1735     }
1736 }
1737 
1738 static void pc_machine_wakeup(MachineState *machine)
1739 {
1740     cpu_synchronize_all_states();
1741     pc_machine_reset(machine, RESET_TYPE_WAKEUP);
1742     cpu_synchronize_all_post_reset();
1743 }
1744 
1745 static bool pc_hotplug_allowed(MachineState *ms, DeviceState *dev, Error **errp)
1746 {
1747     X86IOMMUState *iommu = x86_iommu_get_default();
1748     IntelIOMMUState *intel_iommu;
1749 
1750     if (iommu &&
1751         object_dynamic_cast((Object *)iommu, TYPE_INTEL_IOMMU_DEVICE) &&
1752         object_dynamic_cast((Object *)dev, "vfio-pci")) {
1753         intel_iommu = INTEL_IOMMU_DEVICE(iommu);
1754         if (!intel_iommu->caching_mode) {
1755             error_setg(errp, "Device assignment is not allowed without "
1756                        "enabling caching-mode=on for Intel IOMMU.");
1757             return false;
1758         }
1759     }
1760 
1761     return true;
1762 }
1763 
1764 static void pc_machine_class_init(ObjectClass *oc, void *data)
1765 {
1766     MachineClass *mc = MACHINE_CLASS(oc);
1767     X86MachineClass *x86mc = X86_MACHINE_CLASS(oc);
1768     PCMachineClass *pcmc = PC_MACHINE_CLASS(oc);
1769     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
1770 
1771     pcmc->pci_enabled = true;
1772     pcmc->has_acpi_build = true;
1773     pcmc->smbios_defaults = true;
1774     pcmc->gigabyte_align = true;
1775     pcmc->has_reserved_memory = true;
1776     pcmc->enforce_amd_1tb_hole = true;
1777     pcmc->isa_bios_alias = true;
1778     pcmc->pvh_enabled = true;
1779     pcmc->kvmclock_create_always = true;
1780     x86mc->apic_xrupt_override = true;
1781     assert(!mc->get_hotplug_handler);
1782     mc->get_hotplug_handler = pc_get_hotplug_handler;
1783     mc->hotplug_allowed = pc_hotplug_allowed;
1784     mc->auto_enable_numa_with_memhp = true;
1785     mc->auto_enable_numa_with_memdev = true;
1786     mc->has_hotpluggable_cpus = true;
1787     mc->default_boot_order = "cad";
1788     mc->block_default_type = IF_IDE;
1789     mc->max_cpus = 255;
1790     mc->reset = pc_machine_reset;
1791     mc->wakeup = pc_machine_wakeup;
1792     hc->pre_plug = pc_machine_device_pre_plug_cb;
1793     hc->plug = pc_machine_device_plug_cb;
1794     hc->unplug_request = pc_machine_device_unplug_request_cb;
1795     hc->unplug = pc_machine_device_unplug_cb;
1796     mc->default_cpu_type = TARGET_DEFAULT_CPU_TYPE;
1797     mc->nvdimm_supported = true;
1798     mc->smp_props.dies_supported = true;
1799     mc->smp_props.modules_supported = true;
1800     mc->default_ram_id = "pc.ram";
1801     pcmc->default_smbios_ep_type = SMBIOS_ENTRY_POINT_TYPE_AUTO;
1802 
1803     object_class_property_add(oc, PC_MACHINE_MAX_RAM_BELOW_4G, "size",
1804         pc_machine_get_max_ram_below_4g, pc_machine_set_max_ram_below_4g,
1805         NULL, NULL);
1806     object_class_property_set_description(oc, PC_MACHINE_MAX_RAM_BELOW_4G,
1807         "Maximum ram below the 4G boundary (32bit boundary)");
1808 
1809     object_class_property_add(oc, PC_MACHINE_VMPORT, "OnOffAuto",
1810         pc_machine_get_vmport, pc_machine_set_vmport,
1811         NULL, NULL);
1812     object_class_property_set_description(oc, PC_MACHINE_VMPORT,
1813         "Enable vmport (pc & q35)");
1814 
1815     object_class_property_add_bool(oc, PC_MACHINE_SMBUS,
1816         pc_machine_get_smbus, pc_machine_set_smbus);
1817     object_class_property_set_description(oc, PC_MACHINE_SMBUS,
1818         "Enable/disable system management bus");
1819 
1820     object_class_property_add_bool(oc, PC_MACHINE_SATA,
1821         pc_machine_get_sata, pc_machine_set_sata);
1822     object_class_property_set_description(oc, PC_MACHINE_SATA,
1823         "Enable/disable Serial ATA bus");
1824 
1825     object_class_property_add_bool(oc, "hpet",
1826         pc_machine_get_hpet, pc_machine_set_hpet);
1827     object_class_property_set_description(oc, "hpet",
1828         "Enable/disable high precision event timer emulation");
1829 
1830     object_class_property_add_bool(oc, PC_MACHINE_I8042,
1831         pc_machine_get_i8042, pc_machine_set_i8042);
1832     object_class_property_set_description(oc, PC_MACHINE_I8042,
1833         "Enable/disable Intel 8042 PS/2 controller emulation");
1834 
1835     object_class_property_add_bool(oc, "default-bus-bypass-iommu",
1836         pc_machine_get_default_bus_bypass_iommu,
1837         pc_machine_set_default_bus_bypass_iommu);
1838 
1839     object_class_property_add(oc, PC_MACHINE_MAX_FW_SIZE, "size",
1840         pc_machine_get_max_fw_size, pc_machine_set_max_fw_size,
1841         NULL, NULL);
1842     object_class_property_set_description(oc, PC_MACHINE_MAX_FW_SIZE,
1843         "Maximum combined firmware size");
1844 
1845     object_class_property_add(oc, PC_MACHINE_SMBIOS_EP, "str",
1846         pc_machine_get_smbios_ep, pc_machine_set_smbios_ep,
1847         NULL, NULL);
1848     object_class_property_set_description(oc, PC_MACHINE_SMBIOS_EP,
1849         "SMBIOS Entry Point type [32, 64]");
1850 
1851     object_class_property_add_bool(oc, "fd-bootchk",
1852         pc_machine_get_fd_bootchk,
1853         pc_machine_set_fd_bootchk);
1854 }
1855 
1856 static const TypeInfo pc_machine_info = {
1857     .name = TYPE_PC_MACHINE,
1858     .parent = TYPE_X86_MACHINE,
1859     .abstract = true,
1860     .instance_size = sizeof(PCMachineState),
1861     .instance_init = pc_machine_initfn,
1862     .class_size = sizeof(PCMachineClass),
1863     .class_init = pc_machine_class_init,
1864     .interfaces = (InterfaceInfo[]) {
1865          { TYPE_HOTPLUG_HANDLER },
1866          { }
1867     },
1868 };
1869 
1870 static void pc_machine_register_types(void)
1871 {
1872     type_register_static(&pc_machine_info);
1873 }
1874 
1875 type_init(pc_machine_register_types)
1876