xref: /openbmc/qemu/hw/i386/pc.c (revision a6896ebc)
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.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 "sysemu/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 "sysemu/sysemu.h"
43 #include "sysemu/xen.h"
44 #include "sysemu/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_1[] = {};
83 const size_t pc_compat_9_1_len = G_N_ELEMENTS(pc_compat_9_1);
84 
85 GlobalProperty pc_compat_9_0[] = {
86     { TYPE_X86_CPU, "x-amd-topoext-features-only", "false" },
87     { TYPE_X86_CPU, "x-l1-cache-per-thread", "false" },
88     { TYPE_X86_CPU, "guest-phys-bits", "0" },
89     { "sev-guest", "legacy-vm-type", "on" },
90     { TYPE_X86_CPU, "legacy-multi-node", "on" },
91 };
92 const size_t pc_compat_9_0_len = G_N_ELEMENTS(pc_compat_9_0);
93 
94 GlobalProperty pc_compat_8_2[] = {};
95 const size_t pc_compat_8_2_len = G_N_ELEMENTS(pc_compat_8_2);
96 
97 GlobalProperty pc_compat_8_1[] = {};
98 const size_t pc_compat_8_1_len = G_N_ELEMENTS(pc_compat_8_1);
99 
100 GlobalProperty pc_compat_8_0[] = {
101     { "virtio-mem", "unplugged-inaccessible", "auto" },
102 };
103 const size_t pc_compat_8_0_len = G_N_ELEMENTS(pc_compat_8_0);
104 
105 GlobalProperty pc_compat_7_2[] = {
106     { "ICH9-LPC", "noreboot", "true" },
107 };
108 const size_t pc_compat_7_2_len = G_N_ELEMENTS(pc_compat_7_2);
109 
110 GlobalProperty pc_compat_7_1[] = {};
111 const size_t pc_compat_7_1_len = G_N_ELEMENTS(pc_compat_7_1);
112 
113 GlobalProperty pc_compat_7_0[] = {};
114 const size_t pc_compat_7_0_len = G_N_ELEMENTS(pc_compat_7_0);
115 
116 GlobalProperty pc_compat_6_2[] = {
117     { "virtio-mem", "unplugged-inaccessible", "off" },
118 };
119 const size_t pc_compat_6_2_len = G_N_ELEMENTS(pc_compat_6_2);
120 
121 GlobalProperty pc_compat_6_1[] = {
122     { TYPE_X86_CPU, "hv-version-id-build", "0x1bbc" },
123     { TYPE_X86_CPU, "hv-version-id-major", "0x0006" },
124     { TYPE_X86_CPU, "hv-version-id-minor", "0x0001" },
125     { "ICH9-LPC", "x-keep-pci-slot-hpc", "false" },
126 };
127 const size_t pc_compat_6_1_len = G_N_ELEMENTS(pc_compat_6_1);
128 
129 GlobalProperty pc_compat_6_0[] = {
130     { "qemu64" "-" TYPE_X86_CPU, "family", "6" },
131     { "qemu64" "-" TYPE_X86_CPU, "model", "6" },
132     { "qemu64" "-" TYPE_X86_CPU, "stepping", "3" },
133     { TYPE_X86_CPU, "x-vendor-cpuid-only", "off" },
134     { "ICH9-LPC", ACPI_PM_PROP_ACPI_PCIHP_BRIDGE, "off" },
135     { "ICH9-LPC", "x-keep-pci-slot-hpc", "true" },
136 };
137 const size_t pc_compat_6_0_len = G_N_ELEMENTS(pc_compat_6_0);
138 
139 GlobalProperty pc_compat_5_2[] = {
140     { "ICH9-LPC", "x-smi-cpu-hotunplug", "off" },
141 };
142 const size_t pc_compat_5_2_len = G_N_ELEMENTS(pc_compat_5_2);
143 
144 GlobalProperty pc_compat_5_1[] = {
145     { "ICH9-LPC", "x-smi-cpu-hotplug", "off" },
146     { TYPE_X86_CPU, "kvm-msi-ext-dest-id", "off" },
147 };
148 const size_t pc_compat_5_1_len = G_N_ELEMENTS(pc_compat_5_1);
149 
150 GlobalProperty pc_compat_5_0[] = {
151 };
152 const size_t pc_compat_5_0_len = G_N_ELEMENTS(pc_compat_5_0);
153 
154 GlobalProperty pc_compat_4_2[] = {
155     { "mch", "smbase-smram", "off" },
156 };
157 const size_t pc_compat_4_2_len = G_N_ELEMENTS(pc_compat_4_2);
158 
159 GlobalProperty pc_compat_4_1[] = {};
160 const size_t pc_compat_4_1_len = G_N_ELEMENTS(pc_compat_4_1);
161 
162 GlobalProperty pc_compat_4_0[] = {};
163 const size_t pc_compat_4_0_len = G_N_ELEMENTS(pc_compat_4_0);
164 
165 GlobalProperty pc_compat_3_1[] = {
166     { "intel-iommu", "dma-drain", "off" },
167     { "Opteron_G3" "-" TYPE_X86_CPU, "rdtscp", "off" },
168     { "Opteron_G4" "-" TYPE_X86_CPU, "rdtscp", "off" },
169     { "Opteron_G4" "-" TYPE_X86_CPU, "npt", "off" },
170     { "Opteron_G4" "-" TYPE_X86_CPU, "nrip-save", "off" },
171     { "Opteron_G5" "-" TYPE_X86_CPU, "rdtscp", "off" },
172     { "Opteron_G5" "-" TYPE_X86_CPU, "npt", "off" },
173     { "Opteron_G5" "-" TYPE_X86_CPU, "nrip-save", "off" },
174     { "EPYC" "-" TYPE_X86_CPU, "npt", "off" },
175     { "EPYC" "-" TYPE_X86_CPU, "nrip-save", "off" },
176     { "EPYC-IBPB" "-" TYPE_X86_CPU, "npt", "off" },
177     { "EPYC-IBPB" "-" TYPE_X86_CPU, "nrip-save", "off" },
178     { "Skylake-Client" "-" TYPE_X86_CPU,      "mpx", "on" },
179     { "Skylake-Client-IBRS" "-" TYPE_X86_CPU, "mpx", "on" },
180     { "Skylake-Server" "-" TYPE_X86_CPU,      "mpx", "on" },
181     { "Skylake-Server-IBRS" "-" TYPE_X86_CPU, "mpx", "on" },
182     { "Cascadelake-Server" "-" TYPE_X86_CPU,  "mpx", "on" },
183     { "Icelake-Client" "-" TYPE_X86_CPU,      "mpx", "on" },
184     { "Icelake-Server" "-" TYPE_X86_CPU,      "mpx", "on" },
185     { "Cascadelake-Server" "-" TYPE_X86_CPU, "stepping", "5" },
186     { TYPE_X86_CPU, "x-intel-pt-auto-level", "off" },
187 };
188 const size_t pc_compat_3_1_len = G_N_ELEMENTS(pc_compat_3_1);
189 
190 GlobalProperty pc_compat_3_0[] = {
191     { TYPE_X86_CPU, "x-hv-synic-kvm-only", "on" },
192     { "Skylake-Server" "-" TYPE_X86_CPU, "pku", "off" },
193     { "Skylake-Server-IBRS" "-" TYPE_X86_CPU, "pku", "off" },
194 };
195 const size_t pc_compat_3_0_len = G_N_ELEMENTS(pc_compat_3_0);
196 
197 GlobalProperty pc_compat_2_12[] = {
198     { TYPE_X86_CPU, "legacy-cache", "on" },
199     { TYPE_X86_CPU, "topoext", "off" },
200     { "EPYC-" TYPE_X86_CPU, "xlevel", "0x8000000a" },
201     { "EPYC-IBPB-" TYPE_X86_CPU, "xlevel", "0x8000000a" },
202 };
203 const size_t pc_compat_2_12_len = G_N_ELEMENTS(pc_compat_2_12);
204 
205 GlobalProperty pc_compat_2_11[] = {
206     { TYPE_X86_CPU, "x-migrate-smi-count", "off" },
207     { "Skylake-Server" "-" TYPE_X86_CPU, "clflushopt", "off" },
208 };
209 const size_t pc_compat_2_11_len = G_N_ELEMENTS(pc_compat_2_11);
210 
211 GlobalProperty pc_compat_2_10[] = {
212     { TYPE_X86_CPU, "x-hv-max-vps", "0x40" },
213     { "i440FX-pcihost", "x-pci-hole64-fix", "off" },
214     { "q35-pcihost", "x-pci-hole64-fix", "off" },
215 };
216 const size_t pc_compat_2_10_len = G_N_ELEMENTS(pc_compat_2_10);
217 
218 GlobalProperty pc_compat_2_9[] = {
219     { "mch", "extended-tseg-mbytes", "0" },
220 };
221 const size_t pc_compat_2_9_len = G_N_ELEMENTS(pc_compat_2_9);
222 
223 GlobalProperty pc_compat_2_8[] = {
224     { TYPE_X86_CPU, "tcg-cpuid", "off" },
225     { "kvmclock", "x-mach-use-reliable-get-clock", "off" },
226     { "ICH9-LPC", "x-smi-broadcast", "off" },
227     { TYPE_X86_CPU, "vmware-cpuid-freq", "off" },
228     { "Haswell-" TYPE_X86_CPU, "stepping", "1" },
229 };
230 const size_t pc_compat_2_8_len = G_N_ELEMENTS(pc_compat_2_8);
231 
232 GlobalProperty pc_compat_2_7[] = {
233     { TYPE_X86_CPU, "l3-cache", "off" },
234     { TYPE_X86_CPU, "full-cpuid-auto-level", "off" },
235     { "Opteron_G3" "-" TYPE_X86_CPU, "family", "15" },
236     { "Opteron_G3" "-" TYPE_X86_CPU, "model", "6" },
237     { "Opteron_G3" "-" TYPE_X86_CPU, "stepping", "1" },
238     { "isa-pcspk", "migrate", "off" },
239 };
240 const size_t pc_compat_2_7_len = G_N_ELEMENTS(pc_compat_2_7);
241 
242 GlobalProperty pc_compat_2_6[] = {
243     { TYPE_X86_CPU, "cpuid-0xb", "off" },
244     { "vmxnet3", "romfile", "" },
245     { TYPE_X86_CPU, "fill-mtrr-mask", "off" },
246     { "apic-common", "legacy-instance-id", "on", }
247 };
248 const size_t pc_compat_2_6_len = G_N_ELEMENTS(pc_compat_2_6);
249 
250 GlobalProperty pc_compat_2_5[] = {};
251 const size_t pc_compat_2_5_len = G_N_ELEMENTS(pc_compat_2_5);
252 
253 GlobalProperty pc_compat_2_4[] = {
254     PC_CPU_MODEL_IDS("2.4.0")
255     { "Haswell-" TYPE_X86_CPU, "abm", "off" },
256     { "Haswell-noTSX-" TYPE_X86_CPU, "abm", "off" },
257     { "Broadwell-" TYPE_X86_CPU, "abm", "off" },
258     { "Broadwell-noTSX-" TYPE_X86_CPU, "abm", "off" },
259     { "host" "-" TYPE_X86_CPU, "host-cache-info", "on" },
260     { TYPE_X86_CPU, "check", "off" },
261     { "qemu64" "-" TYPE_X86_CPU, "sse4a", "on" },
262     { "qemu64" "-" TYPE_X86_CPU, "abm", "on" },
263     { "qemu64" "-" TYPE_X86_CPU, "popcnt", "on" },
264     { "qemu32" "-" TYPE_X86_CPU, "popcnt", "on" },
265     { "Opteron_G2" "-" TYPE_X86_CPU, "rdtscp", "on" },
266     { "Opteron_G3" "-" TYPE_X86_CPU, "rdtscp", "on" },
267     { "Opteron_G4" "-" TYPE_X86_CPU, "rdtscp", "on" },
268     { "Opteron_G5" "-" TYPE_X86_CPU, "rdtscp", "on", }
269 };
270 const size_t pc_compat_2_4_len = G_N_ELEMENTS(pc_compat_2_4);
271 
272 /*
273  * @PC_FW_DATA:
274  * Size of the chunk of memory at the top of RAM for the BIOS ACPI tables
275  * and other BIOS datastructures.
276  *
277  * BIOS ACPI tables: 128K. Other BIOS datastructures: less than 4K
278  * reported to be used at the moment, 32K should be enough for a while.
279  */
280 #define PC_FW_DATA (0x20000 + 0x8000)
281 
282 GSIState *pc_gsi_create(qemu_irq **irqs, bool pci_enabled)
283 {
284     GSIState *s;
285 
286     s = g_new0(GSIState, 1);
287     if (kvm_ioapic_in_kernel()) {
288         kvm_pc_setup_irq_routing(pci_enabled);
289     }
290     *irqs = qemu_allocate_irqs(gsi_handler, s, IOAPIC_NUM_PINS);
291 
292     return s;
293 }
294 
295 static void ioport80_write(void *opaque, hwaddr addr, uint64_t data,
296                            unsigned size)
297 {
298 }
299 
300 static uint64_t ioport80_read(void *opaque, hwaddr addr, unsigned size)
301 {
302     return 0xffffffffffffffffULL;
303 }
304 
305 /* MS-DOS compatibility mode FPU exception support */
306 static void ioportF0_write(void *opaque, hwaddr addr, uint64_t data,
307                            unsigned size)
308 {
309     if (tcg_enabled()) {
310         cpu_set_ignne();
311     }
312 }
313 
314 static uint64_t ioportF0_read(void *opaque, hwaddr addr, unsigned size)
315 {
316     return 0xffffffffffffffffULL;
317 }
318 
319 /* PC cmos mappings */
320 
321 #define REG_EQUIPMENT_BYTE          0x14
322 
323 static void cmos_init_hd(MC146818RtcState *s, int type_ofs, int info_ofs,
324                          int16_t cylinders, int8_t heads, int8_t sectors)
325 {
326     mc146818rtc_set_cmos_data(s, type_ofs, 47);
327     mc146818rtc_set_cmos_data(s, info_ofs, cylinders);
328     mc146818rtc_set_cmos_data(s, info_ofs + 1, cylinders >> 8);
329     mc146818rtc_set_cmos_data(s, info_ofs + 2, heads);
330     mc146818rtc_set_cmos_data(s, info_ofs + 3, 0xff);
331     mc146818rtc_set_cmos_data(s, info_ofs + 4, 0xff);
332     mc146818rtc_set_cmos_data(s, info_ofs + 5, 0xc0 | ((heads > 8) << 3));
333     mc146818rtc_set_cmos_data(s, info_ofs + 6, cylinders);
334     mc146818rtc_set_cmos_data(s, info_ofs + 7, cylinders >> 8);
335     mc146818rtc_set_cmos_data(s, info_ofs + 8, sectors);
336 }
337 
338 /* convert boot_device letter to something recognizable by the bios */
339 static int boot_device2nibble(char boot_device)
340 {
341     switch(boot_device) {
342     case 'a':
343     case 'b':
344         return 0x01; /* floppy boot */
345     case 'c':
346         return 0x02; /* hard drive boot */
347     case 'd':
348         return 0x03; /* CD-ROM boot */
349     case 'n':
350         return 0x04; /* Network boot */
351     }
352     return 0;
353 }
354 
355 static void set_boot_dev(PCMachineState *pcms, MC146818RtcState *s,
356                          const char *boot_device, Error **errp)
357 {
358 #define PC_MAX_BOOT_DEVICES 3
359     int nbds, bds[3] = { 0, };
360     int i;
361 
362     nbds = strlen(boot_device);
363     if (nbds > PC_MAX_BOOT_DEVICES) {
364         error_setg(errp, "Too many boot devices for PC");
365         return;
366     }
367     for (i = 0; i < nbds; i++) {
368         bds[i] = boot_device2nibble(boot_device[i]);
369         if (bds[i] == 0) {
370             error_setg(errp, "Invalid boot device for PC: '%c'",
371                        boot_device[i]);
372             return;
373         }
374     }
375     mc146818rtc_set_cmos_data(s, 0x3d, (bds[1] << 4) | bds[0]);
376     mc146818rtc_set_cmos_data(s, 0x38, (bds[2] << 4) | !pcms->fd_bootchk);
377 }
378 
379 static void pc_boot_set(void *opaque, const char *boot_device, Error **errp)
380 {
381     PCMachineState *pcms = opaque;
382     X86MachineState *x86ms = X86_MACHINE(pcms);
383 
384     set_boot_dev(pcms, MC146818_RTC(x86ms->rtc), boot_device, errp);
385 }
386 
387 static void pc_cmos_init_floppy(MC146818RtcState *rtc_state, ISADevice *floppy)
388 {
389     int val, nb;
390     FloppyDriveType fd_type[2] = { FLOPPY_DRIVE_TYPE_NONE,
391                                    FLOPPY_DRIVE_TYPE_NONE };
392 
393 #ifdef CONFIG_FDC_ISA
394     /* floppy type */
395     if (floppy) {
396         for (int i = 0; i < 2; i++) {
397             fd_type[i] = isa_fdc_get_drive_type(floppy, i);
398         }
399     }
400 #endif
401 
402     val = (cmos_get_fd_drive_type(fd_type[0]) << 4) |
403         cmos_get_fd_drive_type(fd_type[1]);
404     mc146818rtc_set_cmos_data(rtc_state, 0x10, val);
405 
406     val = mc146818rtc_get_cmos_data(rtc_state, REG_EQUIPMENT_BYTE);
407     nb = 0;
408     if (fd_type[0] != FLOPPY_DRIVE_TYPE_NONE) {
409         nb++;
410     }
411     if (fd_type[1] != FLOPPY_DRIVE_TYPE_NONE) {
412         nb++;
413     }
414     switch (nb) {
415     case 0:
416         break;
417     case 1:
418         val |= 0x01; /* 1 drive, ready for boot */
419         break;
420     case 2:
421         val |= 0x41; /* 2 drives, ready for boot */
422         break;
423     }
424     mc146818rtc_set_cmos_data(rtc_state, REG_EQUIPMENT_BYTE, val);
425 }
426 
427 typedef struct check_fdc_state {
428     ISADevice *floppy;
429     bool multiple;
430 } CheckFdcState;
431 
432 static int check_fdc(Object *obj, void *opaque)
433 {
434     CheckFdcState *state = opaque;
435     Object *fdc;
436     uint32_t iobase;
437     Error *local_err = NULL;
438 
439     fdc = object_dynamic_cast(obj, TYPE_ISA_FDC);
440     if (!fdc) {
441         return 0;
442     }
443 
444     iobase = object_property_get_uint(obj, "iobase", &local_err);
445     if (local_err || iobase != 0x3f0) {
446         error_free(local_err);
447         return 0;
448     }
449 
450     if (state->floppy) {
451         state->multiple = true;
452     } else {
453         state->floppy = ISA_DEVICE(obj);
454     }
455     return 0;
456 }
457 
458 static const char * const fdc_container_path[] = {
459     "/unattached", "/peripheral", "/peripheral-anon"
460 };
461 
462 /*
463  * Locate the FDC at IO address 0x3f0, in order to configure the CMOS registers
464  * and ACPI objects.
465  */
466 static ISADevice *pc_find_fdc0(void)
467 {
468     int i;
469     Object *container;
470     CheckFdcState state = { 0 };
471 
472     for (i = 0; i < ARRAY_SIZE(fdc_container_path); i++) {
473         container = container_get(qdev_get_machine(), fdc_container_path[i]);
474         object_child_foreach(container, check_fdc, &state);
475     }
476 
477     if (state.multiple) {
478         warn_report("multiple floppy disk controllers with "
479                     "iobase=0x3f0 have been found");
480         error_printf("the one being picked for CMOS setup might not reflect "
481                      "your intent");
482     }
483 
484     return state.floppy;
485 }
486 
487 static void pc_cmos_init_late(PCMachineState *pcms)
488 {
489     X86MachineState *x86ms = X86_MACHINE(pcms);
490     MC146818RtcState *s = MC146818_RTC(x86ms->rtc);
491     int16_t cylinders;
492     int8_t heads, sectors;
493     int val;
494     int i, trans;
495 
496     val = 0;
497     if (pcms->idebus[0] &&
498         ide_get_geometry(pcms->idebus[0], 0,
499                          &cylinders, &heads, &sectors) >= 0) {
500         cmos_init_hd(s, 0x19, 0x1b, cylinders, heads, sectors);
501         val |= 0xf0;
502     }
503     if (pcms->idebus[0] &&
504         ide_get_geometry(pcms->idebus[0], 1,
505                          &cylinders, &heads, &sectors) >= 0) {
506         cmos_init_hd(s, 0x1a, 0x24, cylinders, heads, sectors);
507         val |= 0x0f;
508     }
509     mc146818rtc_set_cmos_data(s, 0x12, val);
510 
511     val = 0;
512     for (i = 0; i < 4; i++) {
513         /* NOTE: ide_get_geometry() returns the physical
514            geometry.  It is always such that: 1 <= sects <= 63, 1
515            <= heads <= 16, 1 <= cylinders <= 16383. The BIOS
516            geometry can be different if a translation is done. */
517         BusState *idebus = pcms->idebus[i / 2];
518         if (idebus &&
519             ide_get_geometry(idebus, i % 2,
520                              &cylinders, &heads, &sectors) >= 0) {
521             trans = ide_get_bios_chs_trans(idebus, i % 2) - 1;
522             assert((trans & ~3) == 0);
523             val |= trans << (i * 2);
524         }
525     }
526     mc146818rtc_set_cmos_data(s, 0x39, val);
527 
528     pc_cmos_init_floppy(s, pc_find_fdc0());
529 
530     /* various important CMOS locations needed by PC/Bochs bios */
531 
532     /* memory size */
533     /* base memory (first MiB) */
534     val = MIN(x86ms->below_4g_mem_size / KiB, 640);
535     mc146818rtc_set_cmos_data(s, 0x15, val);
536     mc146818rtc_set_cmos_data(s, 0x16, val >> 8);
537     /* extended memory (next 64MiB) */
538     if (x86ms->below_4g_mem_size > 1 * MiB) {
539         val = (x86ms->below_4g_mem_size - 1 * MiB) / KiB;
540     } else {
541         val = 0;
542     }
543     if (val > 65535)
544         val = 65535;
545     mc146818rtc_set_cmos_data(s, 0x17, val);
546     mc146818rtc_set_cmos_data(s, 0x18, val >> 8);
547     mc146818rtc_set_cmos_data(s, 0x30, val);
548     mc146818rtc_set_cmos_data(s, 0x31, val >> 8);
549     /* memory between 16MiB and 4GiB */
550     if (x86ms->below_4g_mem_size > 16 * MiB) {
551         val = (x86ms->below_4g_mem_size - 16 * MiB) / (64 * KiB);
552     } else {
553         val = 0;
554     }
555     if (val > 65535)
556         val = 65535;
557     mc146818rtc_set_cmos_data(s, 0x34, val);
558     mc146818rtc_set_cmos_data(s, 0x35, val >> 8);
559     /* memory above 4GiB */
560     val = x86ms->above_4g_mem_size / 65536;
561     mc146818rtc_set_cmos_data(s, 0x5b, val);
562     mc146818rtc_set_cmos_data(s, 0x5c, val >> 8);
563     mc146818rtc_set_cmos_data(s, 0x5d, val >> 16);
564 
565     val = 0;
566     val |= 0x02; /* FPU is there */
567     val |= 0x04; /* PS/2 mouse installed */
568     mc146818rtc_set_cmos_data(s, REG_EQUIPMENT_BYTE, val);
569 }
570 
571 static void handle_a20_line_change(void *opaque, int irq, int level)
572 {
573     X86CPU *cpu = opaque;
574 
575     /* XXX: send to all CPUs ? */
576     /* XXX: add logic to handle multiple A20 line sources */
577     x86_cpu_set_a20(cpu, level);
578 }
579 
580 #define NE2000_NB_MAX 6
581 
582 static const int ne2000_io[NE2000_NB_MAX] = { 0x300, 0x320, 0x340, 0x360,
583                                               0x280, 0x380 };
584 static const int ne2000_irq[NE2000_NB_MAX] = { 9, 10, 11, 3, 4, 5 };
585 
586 static gboolean pc_init_ne2k_isa(ISABus *bus, NICInfo *nd, Error **errp)
587 {
588     static int nb_ne2k = 0;
589 
590     if (nb_ne2k == NE2000_NB_MAX) {
591         error_setg(errp,
592                    "maximum number of ISA NE2000 devices exceeded");
593         return false;
594     }
595     isa_ne2000_init(bus, ne2000_io[nb_ne2k],
596                     ne2000_irq[nb_ne2k], nd);
597     nb_ne2k++;
598     return true;
599 }
600 
601 void pc_acpi_smi_interrupt(void *opaque, int irq, int level)
602 {
603     X86CPU *cpu = opaque;
604 
605     if (level) {
606         cpu_interrupt(CPU(cpu), CPU_INTERRUPT_SMI);
607     }
608 }
609 
610 static
611 void pc_machine_done(Notifier *notifier, void *data)
612 {
613     PCMachineState *pcms = container_of(notifier,
614                                         PCMachineState, machine_done);
615     X86MachineState *x86ms = X86_MACHINE(pcms);
616 
617     cxl_hook_up_pxb_registers(pcms->pcibus, &pcms->cxl_devices_state,
618                               &error_fatal);
619 
620     if (pcms->cxl_devices_state.is_enabled) {
621         cxl_fmws_link_targets(&pcms->cxl_devices_state, &error_fatal);
622     }
623 
624     /* set the number of CPUs */
625     x86_rtc_set_cpus_count(x86ms->rtc, x86ms->boot_cpus);
626 
627     fw_cfg_add_extra_pci_roots(pcms->pcibus, x86ms->fw_cfg);
628 
629     acpi_setup();
630     if (x86ms->fw_cfg) {
631         fw_cfg_build_smbios(pcms, x86ms->fw_cfg, pcms->smbios_entry_point_type);
632         fw_cfg_add_e820(x86ms->fw_cfg);
633         fw_cfg_build_feature_control(MACHINE(pcms), x86ms->fw_cfg);
634         /* update FW_CFG_NB_CPUS to account for -device added CPUs */
635         fw_cfg_modify_i16(x86ms->fw_cfg, FW_CFG_NB_CPUS, x86ms->boot_cpus);
636     }
637 
638     pc_cmos_init_late(pcms);
639 }
640 
641 /* setup pci memory address space mapping into system address space */
642 void pc_pci_as_mapping_init(MemoryRegion *system_memory,
643                             MemoryRegion *pci_address_space)
644 {
645     /* Set to lower priority than RAM */
646     memory_region_add_subregion_overlap(system_memory, 0x0,
647                                         pci_address_space, -1);
648 }
649 
650 void xen_load_linux(PCMachineState *pcms)
651 {
652     int i;
653     FWCfgState *fw_cfg;
654     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
655     X86MachineState *x86ms = X86_MACHINE(pcms);
656 
657     assert(MACHINE(pcms)->kernel_filename != NULL);
658 
659     fw_cfg = fw_cfg_init_io_dma(FW_CFG_IO_BASE, FW_CFG_IO_BASE + 4,
660                                 &address_space_memory);
661     fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, x86ms->boot_cpus);
662     rom_set_fw(fw_cfg);
663 
664     x86_load_linux(x86ms, fw_cfg, PC_FW_DATA, pcmc->pvh_enabled);
665     for (i = 0; i < nb_option_roms; i++) {
666         assert(!strcmp(option_rom[i].name, "linuxboot.bin") ||
667                !strcmp(option_rom[i].name, "linuxboot_dma.bin") ||
668                !strcmp(option_rom[i].name, "pvh.bin") ||
669                !strcmp(option_rom[i].name, "multiboot.bin") ||
670                !strcmp(option_rom[i].name, "multiboot_dma.bin"));
671         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
672     }
673     x86ms->fw_cfg = fw_cfg;
674 }
675 
676 #define PC_ROM_MIN_VGA     0xc0000
677 #define PC_ROM_MIN_OPTION  0xc8000
678 #define PC_ROM_MAX         0xe0000
679 #define PC_ROM_ALIGN       0x800
680 #define PC_ROM_SIZE        (PC_ROM_MAX - PC_ROM_MIN_VGA)
681 
682 static hwaddr pc_above_4g_end(PCMachineState *pcms)
683 {
684     X86MachineState *x86ms = X86_MACHINE(pcms);
685 
686     if (pcms->sgx_epc.size != 0) {
687         return sgx_epc_above_4g_end(&pcms->sgx_epc);
688     }
689 
690     return x86ms->above_4g_mem_start + x86ms->above_4g_mem_size;
691 }
692 
693 static void pc_get_device_memory_range(PCMachineState *pcms,
694                                        hwaddr *base,
695                                        ram_addr_t *device_mem_size)
696 {
697     MachineState *machine = MACHINE(pcms);
698     ram_addr_t size;
699     hwaddr addr;
700 
701     size = machine->maxram_size - machine->ram_size;
702     addr = ROUND_UP(pc_above_4g_end(pcms), 1 * GiB);
703 
704     /* size device region assuming 1G page max alignment per slot */
705     size += (1 * GiB) * machine->ram_slots;
706 
707     *base = addr;
708     *device_mem_size = size;
709 }
710 
711 static uint64_t pc_get_cxl_range_start(PCMachineState *pcms)
712 {
713     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
714     MachineState *ms = MACHINE(pcms);
715     hwaddr cxl_base;
716     ram_addr_t size;
717 
718     if (pcmc->has_reserved_memory &&
719         (ms->ram_size < ms->maxram_size)) {
720         pc_get_device_memory_range(pcms, &cxl_base, &size);
721         cxl_base += size;
722     } else {
723         cxl_base = pc_above_4g_end(pcms);
724     }
725 
726     return cxl_base;
727 }
728 
729 static uint64_t pc_get_cxl_range_end(PCMachineState *pcms)
730 {
731     uint64_t start = pc_get_cxl_range_start(pcms) + MiB;
732 
733     if (pcms->cxl_devices_state.fixed_windows) {
734         GList *it;
735 
736         start = ROUND_UP(start, 256 * MiB);
737         for (it = pcms->cxl_devices_state.fixed_windows; it; it = it->next) {
738             CXLFixedWindow *fw = it->data;
739             start += fw->size;
740         }
741     }
742 
743     return start;
744 }
745 
746 static hwaddr pc_max_used_gpa(PCMachineState *pcms, uint64_t pci_hole64_size)
747 {
748     X86CPU *cpu = X86_CPU(first_cpu);
749     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
750     MachineState *ms = MACHINE(pcms);
751 
752     if (cpu->env.features[FEAT_8000_0001_EDX] & CPUID_EXT2_LM) {
753         /* 64-bit systems */
754         return pc_pci_hole64_start() + pci_hole64_size - 1;
755     }
756 
757     /* 32-bit systems */
758     if (pcmc->broken_32bit_mem_addr_check) {
759         /* old value for compatibility reasons */
760         return ((hwaddr)1 << cpu->phys_bits) - 1;
761     }
762 
763     /*
764      * 32-bit systems don't have hole64 but they might have a region for
765      * memory devices. Even if additional hotplugged memory devices might
766      * not be usable by most guest OSes, we need to still consider them for
767      * calculating the highest possible GPA so that we can properly report
768      * if someone configures them on a CPU that cannot possibly address them.
769      */
770     if (pcmc->has_reserved_memory &&
771         (ms->ram_size < ms->maxram_size)) {
772         hwaddr devmem_start;
773         ram_addr_t devmem_size;
774 
775         pc_get_device_memory_range(pcms, &devmem_start, &devmem_size);
776         devmem_start += devmem_size;
777         return devmem_start - 1;
778     }
779 
780     /* configuration without any memory hotplug */
781     return pc_above_4g_end(pcms) - 1;
782 }
783 
784 /*
785  * AMD systems with an IOMMU have an additional hole close to the
786  * 1Tb, which are special GPAs that cannot be DMA mapped. Depending
787  * on kernel version, VFIO may or may not let you DMA map those ranges.
788  * Starting Linux v5.4 we validate it, and can't create guests on AMD machines
789  * with certain memory sizes. It's also wrong to use those IOVA ranges
790  * in detriment of leading to IOMMU INVALID_DEVICE_REQUEST or worse.
791  * The ranges reserved for Hyper-Transport are:
792  *
793  * FD_0000_0000h - FF_FFFF_FFFFh
794  *
795  * The ranges represent the following:
796  *
797  * Base Address   Top Address  Use
798  *
799  * FD_0000_0000h FD_F7FF_FFFFh Reserved interrupt address space
800  * FD_F800_0000h FD_F8FF_FFFFh Interrupt/EOI IntCtl
801  * FD_F900_0000h FD_F90F_FFFFh Legacy PIC IACK
802  * FD_F910_0000h FD_F91F_FFFFh System Management
803  * FD_F920_0000h FD_FAFF_FFFFh Reserved Page Tables
804  * FD_FB00_0000h FD_FBFF_FFFFh Address Translation
805  * FD_FC00_0000h FD_FDFF_FFFFh I/O Space
806  * FD_FE00_0000h FD_FFFF_FFFFh Configuration
807  * FE_0000_0000h FE_1FFF_FFFFh Extended Configuration/Device Messages
808  * FE_2000_0000h FF_FFFF_FFFFh Reserved
809  *
810  * See AMD IOMMU spec, section 2.1.2 "IOMMU Logical Topology",
811  * Table 3: Special Address Controls (GPA) for more information.
812  */
813 #define AMD_HT_START         0xfd00000000UL
814 #define AMD_HT_END           0xffffffffffUL
815 #define AMD_ABOVE_1TB_START  (AMD_HT_END + 1)
816 #define AMD_HT_SIZE          (AMD_ABOVE_1TB_START - AMD_HT_START)
817 
818 void pc_memory_init(PCMachineState *pcms,
819                     MemoryRegion *system_memory,
820                     MemoryRegion *rom_memory,
821                     uint64_t pci_hole64_size)
822 {
823     int linux_boot, i;
824     MemoryRegion *option_rom_mr;
825     MemoryRegion *ram_below_4g, *ram_above_4g;
826     FWCfgState *fw_cfg;
827     MachineState *machine = MACHINE(pcms);
828     MachineClass *mc = MACHINE_GET_CLASS(machine);
829     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
830     X86MachineState *x86ms = X86_MACHINE(pcms);
831     hwaddr maxphysaddr, maxusedaddr;
832     hwaddr cxl_base, cxl_resv_end = 0;
833     X86CPU *cpu = X86_CPU(first_cpu);
834 
835     assert(machine->ram_size == x86ms->below_4g_mem_size +
836                                 x86ms->above_4g_mem_size);
837 
838     linux_boot = (machine->kernel_filename != NULL);
839 
840     /*
841      * The HyperTransport range close to the 1T boundary is unique to AMD
842      * hosts with IOMMUs enabled. Restrict the ram-above-4g relocation
843      * to above 1T to AMD vCPUs only. @enforce_amd_1tb_hole is only false in
844      * older machine types (<= 7.0) for compatibility purposes.
845      */
846     if (IS_AMD_CPU(&cpu->env) && pcmc->enforce_amd_1tb_hole) {
847         /* Bail out if max possible address does not cross HT range */
848         if (pc_max_used_gpa(pcms, pci_hole64_size) >= AMD_HT_START) {
849             x86ms->above_4g_mem_start = AMD_ABOVE_1TB_START;
850         }
851 
852         /*
853          * Advertise the HT region if address space covers the reserved
854          * region or if we relocate.
855          */
856         if (cpu->phys_bits >= 40) {
857             e820_add_entry(AMD_HT_START, AMD_HT_SIZE, E820_RESERVED);
858         }
859     }
860 
861     /*
862      * phys-bits is required to be appropriately configured
863      * to make sure max used GPA is reachable.
864      */
865     maxusedaddr = pc_max_used_gpa(pcms, pci_hole64_size);
866     maxphysaddr = ((hwaddr)1 << cpu->phys_bits) - 1;
867     if (maxphysaddr < maxusedaddr) {
868         error_report("Address space limit 0x%"PRIx64" < 0x%"PRIx64
869                      " phys-bits too low (%u)",
870                      maxphysaddr, maxusedaddr, cpu->phys_bits);
871         exit(EXIT_FAILURE);
872     }
873 
874     /*
875      * Split single memory region and use aliases to address portions of it,
876      * done for backwards compatibility with older qemus.
877      */
878     ram_below_4g = g_malloc(sizeof(*ram_below_4g));
879     memory_region_init_alias(ram_below_4g, NULL, "ram-below-4g", machine->ram,
880                              0, x86ms->below_4g_mem_size);
881     memory_region_add_subregion(system_memory, 0, ram_below_4g);
882     e820_add_entry(0, x86ms->below_4g_mem_size, E820_RAM);
883     if (x86ms->above_4g_mem_size > 0) {
884         ram_above_4g = g_malloc(sizeof(*ram_above_4g));
885         memory_region_init_alias(ram_above_4g, NULL, "ram-above-4g",
886                                  machine->ram,
887                                  x86ms->below_4g_mem_size,
888                                  x86ms->above_4g_mem_size);
889         memory_region_add_subregion(system_memory, x86ms->above_4g_mem_start,
890                                     ram_above_4g);
891         e820_add_entry(x86ms->above_4g_mem_start, x86ms->above_4g_mem_size,
892                        E820_RAM);
893     }
894 
895     if (pcms->sgx_epc.size != 0) {
896         e820_add_entry(pcms->sgx_epc.base, pcms->sgx_epc.size, E820_RESERVED);
897     }
898 
899     if (!pcmc->has_reserved_memory &&
900         (machine->ram_slots ||
901          (machine->maxram_size > machine->ram_size))) {
902 
903         error_report("\"-memory 'slots|maxmem'\" is not supported by: %s",
904                      mc->name);
905         exit(EXIT_FAILURE);
906     }
907 
908     /* initialize device memory address space */
909     if (pcmc->has_reserved_memory &&
910         (machine->ram_size < machine->maxram_size)) {
911         ram_addr_t device_mem_size;
912         hwaddr device_mem_base;
913 
914         if (machine->ram_slots > ACPI_MAX_RAM_SLOTS) {
915             error_report("unsupported amount of memory slots: %"PRIu64,
916                          machine->ram_slots);
917             exit(EXIT_FAILURE);
918         }
919 
920         if (QEMU_ALIGN_UP(machine->maxram_size,
921                           TARGET_PAGE_SIZE) != machine->maxram_size) {
922             error_report("maximum memory size must by aligned to multiple of "
923                          "%d bytes", TARGET_PAGE_SIZE);
924             exit(EXIT_FAILURE);
925         }
926 
927         pc_get_device_memory_range(pcms, &device_mem_base, &device_mem_size);
928 
929         if (device_mem_base + device_mem_size < device_mem_size) {
930             error_report("unsupported amount of maximum memory: " RAM_ADDR_FMT,
931                          machine->maxram_size);
932             exit(EXIT_FAILURE);
933         }
934         machine_memory_devices_init(machine, device_mem_base, device_mem_size);
935     }
936 
937     if (pcms->cxl_devices_state.is_enabled) {
938         MemoryRegion *mr = &pcms->cxl_devices_state.host_mr;
939         hwaddr cxl_size = MiB;
940 
941         cxl_base = pc_get_cxl_range_start(pcms);
942         memory_region_init(mr, OBJECT(machine), "cxl_host_reg", cxl_size);
943         memory_region_add_subregion(system_memory, cxl_base, mr);
944         cxl_resv_end = cxl_base + cxl_size;
945         if (pcms->cxl_devices_state.fixed_windows) {
946             hwaddr cxl_fmw_base;
947             GList *it;
948 
949             cxl_fmw_base = ROUND_UP(cxl_base + cxl_size, 256 * MiB);
950             for (it = pcms->cxl_devices_state.fixed_windows; it; it = it->next) {
951                 CXLFixedWindow *fw = it->data;
952 
953                 fw->base = cxl_fmw_base;
954                 memory_region_init_io(&fw->mr, OBJECT(machine), &cfmws_ops, fw,
955                                       "cxl-fixed-memory-region", fw->size);
956                 memory_region_add_subregion(system_memory, fw->base, &fw->mr);
957                 cxl_fmw_base += fw->size;
958                 cxl_resv_end = cxl_fmw_base;
959             }
960         }
961     }
962 
963     /* Initialize PC system firmware */
964     pc_system_firmware_init(pcms, rom_memory);
965 
966     option_rom_mr = g_malloc(sizeof(*option_rom_mr));
967     if (machine_require_guest_memfd(machine)) {
968         memory_region_init_ram_guest_memfd(option_rom_mr, NULL, "pc.rom",
969                                            PC_ROM_SIZE, &error_fatal);
970     } else {
971         memory_region_init_ram(option_rom_mr, NULL, "pc.rom", PC_ROM_SIZE,
972                                &error_fatal);
973         if (pcmc->pci_enabled) {
974             memory_region_set_readonly(option_rom_mr, true);
975         }
976     }
977     memory_region_add_subregion_overlap(rom_memory,
978                                         PC_ROM_MIN_VGA,
979                                         option_rom_mr,
980                                         1);
981 
982     fw_cfg = fw_cfg_arch_create(machine,
983                                 x86ms->boot_cpus, x86ms->apic_id_limit);
984 
985     rom_set_fw(fw_cfg);
986 
987     if (machine->device_memory) {
988         uint64_t *val = g_malloc(sizeof(*val));
989         uint64_t res_mem_end = machine->device_memory->base;
990 
991         if (!pcmc->broken_reserved_end) {
992             res_mem_end += memory_region_size(&machine->device_memory->mr);
993         }
994 
995         if (pcms->cxl_devices_state.is_enabled) {
996             res_mem_end = cxl_resv_end;
997         }
998         *val = cpu_to_le64(ROUND_UP(res_mem_end, 1 * GiB));
999         fw_cfg_add_file(fw_cfg, "etc/reserved-memory-end", val, sizeof(*val));
1000     }
1001 
1002     if (linux_boot) {
1003         x86_load_linux(x86ms, fw_cfg, PC_FW_DATA, pcmc->pvh_enabled);
1004     }
1005 
1006     for (i = 0; i < nb_option_roms; i++) {
1007         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1008     }
1009     x86ms->fw_cfg = fw_cfg;
1010 
1011     /* Init default IOAPIC address space */
1012     x86ms->ioapic_as = &address_space_memory;
1013 
1014     /* Init ACPI memory hotplug IO base address */
1015     pcms->memhp_io_base = ACPI_MEMORY_HOTPLUG_BASE;
1016 }
1017 
1018 /*
1019  * The 64bit pci hole starts after "above 4G RAM" and
1020  * potentially the space reserved for memory hotplug.
1021  */
1022 uint64_t pc_pci_hole64_start(void)
1023 {
1024     PCMachineState *pcms = PC_MACHINE(qdev_get_machine());
1025     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1026     MachineState *ms = MACHINE(pcms);
1027     uint64_t hole64_start = 0;
1028     ram_addr_t size = 0;
1029 
1030     if (pcms->cxl_devices_state.is_enabled) {
1031         hole64_start = pc_get_cxl_range_end(pcms);
1032     } else if (pcmc->has_reserved_memory && (ms->ram_size < ms->maxram_size)) {
1033         pc_get_device_memory_range(pcms, &hole64_start, &size);
1034         if (!pcmc->broken_reserved_end) {
1035             hole64_start += size;
1036         }
1037     } else {
1038         hole64_start = pc_above_4g_end(pcms);
1039     }
1040 
1041     return ROUND_UP(hole64_start, 1 * GiB);
1042 }
1043 
1044 DeviceState *pc_vga_init(ISABus *isa_bus, PCIBus *pci_bus)
1045 {
1046     DeviceState *dev = NULL;
1047 
1048     rom_set_order_override(FW_CFG_ORDER_OVERRIDE_VGA);
1049     if (pci_bus) {
1050         PCIDevice *pcidev = pci_vga_init(pci_bus);
1051         dev = pcidev ? &pcidev->qdev : NULL;
1052     } else if (isa_bus) {
1053         ISADevice *isadev = isa_vga_init(isa_bus);
1054         dev = isadev ? DEVICE(isadev) : NULL;
1055     }
1056     rom_reset_order_override();
1057     return dev;
1058 }
1059 
1060 static const MemoryRegionOps ioport80_io_ops = {
1061     .write = ioport80_write,
1062     .read = ioport80_read,
1063     .endianness = DEVICE_NATIVE_ENDIAN,
1064     .impl = {
1065         .min_access_size = 1,
1066         .max_access_size = 1,
1067     },
1068 };
1069 
1070 static const MemoryRegionOps ioportF0_io_ops = {
1071     .write = ioportF0_write,
1072     .read = ioportF0_read,
1073     .endianness = DEVICE_NATIVE_ENDIAN,
1074     .impl = {
1075         .min_access_size = 1,
1076         .max_access_size = 1,
1077     },
1078 };
1079 
1080 static void pc_superio_init(ISABus *isa_bus, bool create_fdctrl,
1081                             bool create_i8042, bool no_vmport, Error **errp)
1082 {
1083     int i;
1084     DriveInfo *fd[MAX_FD];
1085     qemu_irq *a20_line;
1086     ISADevice *i8042, *port92, *vmmouse;
1087 
1088     serial_hds_isa_init(isa_bus, 0, MAX_ISA_SERIAL_PORTS);
1089     parallel_hds_isa_init(isa_bus, MAX_PARALLEL_PORTS);
1090 
1091     for (i = 0; i < MAX_FD; i++) {
1092         fd[i] = drive_get(IF_FLOPPY, 0, i);
1093         create_fdctrl |= !!fd[i];
1094     }
1095     if (create_fdctrl) {
1096 #ifdef CONFIG_FDC_ISA
1097         ISADevice *fdc = isa_new(TYPE_ISA_FDC);
1098         if (fdc) {
1099             isa_realize_and_unref(fdc, isa_bus, &error_fatal);
1100             isa_fdc_init_drives(fdc, fd);
1101         }
1102 #endif
1103     }
1104 
1105     if (!create_i8042) {
1106         if (!no_vmport) {
1107             error_setg(errp,
1108                        "vmport requires the i8042 controller to be enabled");
1109         }
1110         return;
1111     }
1112 
1113     i8042 = isa_create_simple(isa_bus, TYPE_I8042);
1114     if (!no_vmport) {
1115         isa_create_simple(isa_bus, TYPE_VMPORT);
1116         vmmouse = isa_try_new("vmmouse");
1117     } else {
1118         vmmouse = NULL;
1119     }
1120     if (vmmouse) {
1121         object_property_set_link(OBJECT(vmmouse), TYPE_I8042, OBJECT(i8042),
1122                                  &error_abort);
1123         isa_realize_and_unref(vmmouse, isa_bus, &error_fatal);
1124     }
1125     port92 = isa_create_simple(isa_bus, TYPE_PORT92);
1126 
1127     a20_line = qemu_allocate_irqs(handle_a20_line_change, first_cpu, 2);
1128     qdev_connect_gpio_out_named(DEVICE(i8042),
1129                                 I8042_A20_LINE, 0, a20_line[0]);
1130     qdev_connect_gpio_out_named(DEVICE(port92),
1131                                 PORT92_A20_LINE, 0, a20_line[1]);
1132     g_free(a20_line);
1133 }
1134 
1135 void pc_basic_device_init(struct PCMachineState *pcms,
1136                           ISABus *isa_bus, qemu_irq *gsi,
1137                           ISADevice *rtc_state,
1138                           bool create_fdctrl,
1139                           uint32_t hpet_irqs)
1140 {
1141     int i;
1142     DeviceState *hpet = NULL;
1143     int pit_isa_irq = 0;
1144     qemu_irq pit_alt_irq = NULL;
1145     ISADevice *pit = NULL;
1146     MemoryRegion *ioport80_io = g_new(MemoryRegion, 1);
1147     MemoryRegion *ioportF0_io = g_new(MemoryRegion, 1);
1148     X86MachineState *x86ms = X86_MACHINE(pcms);
1149 
1150     memory_region_init_io(ioport80_io, NULL, &ioport80_io_ops, NULL, "ioport80", 1);
1151     memory_region_add_subregion(isa_bus->address_space_io, 0x80, ioport80_io);
1152 
1153     memory_region_init_io(ioportF0_io, NULL, &ioportF0_io_ops, NULL, "ioportF0", 1);
1154     memory_region_add_subregion(isa_bus->address_space_io, 0xf0, ioportF0_io);
1155 
1156     /*
1157      * Check if an HPET shall be created.
1158      */
1159     if (pcms->hpet_enabled) {
1160         qemu_irq rtc_irq;
1161 
1162         hpet = qdev_try_new(TYPE_HPET);
1163         if (!hpet) {
1164             error_report("couldn't create HPET device");
1165             exit(1);
1166         }
1167         /*
1168          * For pc-piix-*, hpet's intcap is always IRQ2. For pc-q35-*,
1169          * use IRQ16~23, IRQ8 and IRQ2.  If the user has already set
1170          * the property, use whatever mask they specified.
1171          */
1172         uint8_t compat = object_property_get_uint(OBJECT(hpet),
1173                 HPET_INTCAP, NULL);
1174         if (!compat) {
1175             qdev_prop_set_uint32(hpet, HPET_INTCAP, hpet_irqs);
1176         }
1177         sysbus_realize_and_unref(SYS_BUS_DEVICE(hpet), &error_fatal);
1178         sysbus_mmio_map(SYS_BUS_DEVICE(hpet), 0, HPET_BASE);
1179 
1180         for (i = 0; i < IOAPIC_NUM_PINS; i++) {
1181             sysbus_connect_irq(SYS_BUS_DEVICE(hpet), i, gsi[i]);
1182         }
1183         pit_isa_irq = -1;
1184         pit_alt_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_PIT_INT);
1185         rtc_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_RTC_INT);
1186 
1187         /* overwrite connection created by south bridge */
1188         qdev_connect_gpio_out(DEVICE(rtc_state), 0, rtc_irq);
1189     }
1190 
1191     object_property_add_alias(OBJECT(pcms), "rtc-time", OBJECT(rtc_state),
1192                               "date");
1193 
1194 #ifdef CONFIG_XEN_EMU
1195     if (xen_mode == XEN_EMULATE) {
1196         xen_overlay_create();
1197         xen_evtchn_create(IOAPIC_NUM_PINS, gsi);
1198         xen_gnttab_create();
1199         xen_xenstore_create();
1200         if (pcms->pcibus) {
1201             pci_create_simple(pcms->pcibus, -1, "xen-platform");
1202         }
1203         xen_bus_init();
1204     }
1205 #endif
1206 
1207     qemu_register_boot_set(pc_boot_set, pcms);
1208     set_boot_dev(pcms, MC146818_RTC(rtc_state),
1209                  MACHINE(pcms)->boot_config.order, &error_fatal);
1210 
1211     if (!xen_enabled() &&
1212         (x86ms->pit == ON_OFF_AUTO_AUTO || x86ms->pit == ON_OFF_AUTO_ON)) {
1213         if (kvm_pit_in_kernel()) {
1214             pit = kvm_pit_init(isa_bus, 0x40);
1215         } else {
1216             pit = i8254_pit_init(isa_bus, 0x40, pit_isa_irq, pit_alt_irq);
1217         }
1218         if (hpet) {
1219             /* connect PIT to output control line of the HPET */
1220             qdev_connect_gpio_out(hpet, 0, qdev_get_gpio_in(DEVICE(pit), 0));
1221         }
1222         object_property_set_link(OBJECT(pcms->pcspk), "pit",
1223                                  OBJECT(pit), &error_fatal);
1224         isa_realize_and_unref(pcms->pcspk, isa_bus, &error_fatal);
1225     }
1226 
1227     assert(pcms->vmport >= 0 && pcms->vmport < ON_OFF_AUTO__MAX);
1228     if (pcms->vmport == ON_OFF_AUTO_AUTO) {
1229         pcms->vmport = (xen_enabled() || !pcms->i8042_enabled)
1230             ? ON_OFF_AUTO_OFF : ON_OFF_AUTO_ON;
1231     }
1232 
1233     /* Super I/O */
1234     pc_superio_init(isa_bus, create_fdctrl, pcms->i8042_enabled,
1235                     pcms->vmport != ON_OFF_AUTO_ON, &error_fatal);
1236 }
1237 
1238 void pc_nic_init(PCMachineClass *pcmc, ISABus *isa_bus, PCIBus *pci_bus)
1239 {
1240     MachineClass *mc = MACHINE_CLASS(pcmc);
1241     bool default_is_ne2k = g_str_equal(mc->default_nic, TYPE_ISA_NE2000);
1242     NICInfo *nd;
1243 
1244     rom_set_order_override(FW_CFG_ORDER_OVERRIDE_NIC);
1245 
1246     while ((nd = qemu_find_nic_info(TYPE_ISA_NE2000, default_is_ne2k, NULL))) {
1247         pc_init_ne2k_isa(isa_bus, nd, &error_fatal);
1248     }
1249 
1250     /* Anything remaining should be a PCI NIC */
1251     pci_init_nic_devices(pci_bus, mc->default_nic);
1252 
1253     rom_reset_order_override();
1254 }
1255 
1256 void pc_i8259_create(ISABus *isa_bus, qemu_irq *i8259_irqs)
1257 {
1258     qemu_irq *i8259;
1259 
1260     if (kvm_pic_in_kernel()) {
1261         i8259 = kvm_i8259_init(isa_bus);
1262     } else if (xen_enabled()) {
1263         i8259 = xen_interrupt_controller_init();
1264     } else {
1265         i8259 = i8259_init(isa_bus, x86_allocate_cpu_irq());
1266     }
1267 
1268     for (size_t i = 0; i < ISA_NUM_IRQS; i++) {
1269         i8259_irqs[i] = i8259[i];
1270     }
1271 
1272     g_free(i8259);
1273 }
1274 
1275 static void pc_memory_pre_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
1276                                Error **errp)
1277 {
1278     const X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
1279     const MachineState *ms = MACHINE(hotplug_dev);
1280     const bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
1281     Error *local_err = NULL;
1282 
1283     /*
1284      * When "acpi=off" is used with the Q35 machine type, no ACPI is built,
1285      * but pcms->acpi_dev is still created. Check !acpi_enabled in
1286      * addition to cover this case.
1287      */
1288     if (!x86ms->acpi_dev || !x86_machine_is_acpi_enabled(x86ms)) {
1289         error_setg(errp,
1290                    "memory hotplug is not enabled: missing acpi device or acpi disabled");
1291         return;
1292     }
1293 
1294     if (is_nvdimm && !ms->nvdimms_state->is_enabled) {
1295         error_setg(errp, "nvdimm is not enabled: missing 'nvdimm' in '-M'");
1296         return;
1297     }
1298 
1299     hotplug_handler_pre_plug(x86ms->acpi_dev, dev, &local_err);
1300     if (local_err) {
1301         error_propagate(errp, local_err);
1302         return;
1303     }
1304 
1305     pc_dimm_pre_plug(PC_DIMM(dev), MACHINE(hotplug_dev), errp);
1306 }
1307 
1308 static void pc_memory_plug(HotplugHandler *hotplug_dev,
1309                            DeviceState *dev, Error **errp)
1310 {
1311     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1312     X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
1313     MachineState *ms = MACHINE(hotplug_dev);
1314     bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
1315 
1316     pc_dimm_plug(PC_DIMM(dev), MACHINE(pcms));
1317 
1318     if (is_nvdimm) {
1319         nvdimm_plug(ms->nvdimms_state);
1320     }
1321 
1322     hotplug_handler_plug(x86ms->acpi_dev, dev, &error_abort);
1323 }
1324 
1325 static void pc_memory_unplug_request(HotplugHandler *hotplug_dev,
1326                                      DeviceState *dev, Error **errp)
1327 {
1328     X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
1329 
1330     /*
1331      * When "acpi=off" is used with the Q35 machine type, no ACPI is built,
1332      * but pcms->acpi_dev is still created. Check !acpi_enabled in
1333      * addition to cover this case.
1334      */
1335     if (!x86ms->acpi_dev || !x86_machine_is_acpi_enabled(x86ms)) {
1336         error_setg(errp,
1337                    "memory hotplug is not enabled: missing acpi device or acpi disabled");
1338         return;
1339     }
1340 
1341     if (object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM)) {
1342         error_setg(errp, "nvdimm device hot unplug is not supported yet.");
1343         return;
1344     }
1345 
1346     hotplug_handler_unplug_request(x86ms->acpi_dev, dev,
1347                                    errp);
1348 }
1349 
1350 static void pc_memory_unplug(HotplugHandler *hotplug_dev,
1351                              DeviceState *dev, Error **errp)
1352 {
1353     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1354     X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
1355     Error *local_err = NULL;
1356 
1357     hotplug_handler_unplug(x86ms->acpi_dev, dev, &local_err);
1358     if (local_err) {
1359         goto out;
1360     }
1361 
1362     pc_dimm_unplug(PC_DIMM(dev), MACHINE(pcms));
1363     qdev_unrealize(dev);
1364  out:
1365     error_propagate(errp, local_err);
1366 }
1367 
1368 static void pc_hv_balloon_pre_plug(HotplugHandler *hotplug_dev,
1369                                    DeviceState *dev, Error **errp)
1370 {
1371     /* The vmbus handler has no hotplug handler; we should never end up here. */
1372     g_assert(!dev->hotplugged);
1373     memory_device_pre_plug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev), errp);
1374 }
1375 
1376 static void pc_hv_balloon_plug(HotplugHandler *hotplug_dev,
1377                                DeviceState *dev, Error **errp)
1378 {
1379     memory_device_plug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev));
1380 }
1381 
1382 static void pc_machine_device_pre_plug_cb(HotplugHandler *hotplug_dev,
1383                                           DeviceState *dev, Error **errp)
1384 {
1385     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1386         pc_memory_pre_plug(hotplug_dev, dev, errp);
1387     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1388         x86_cpu_pre_plug(hotplug_dev, dev, errp);
1389     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI)) {
1390         virtio_md_pci_pre_plug(VIRTIO_MD_PCI(dev), MACHINE(hotplug_dev), errp);
1391     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
1392         /* Declare the APIC range as the reserved MSI region */
1393         char *resv_prop_str = g_strdup_printf("0xfee00000:0xfeefffff:%d",
1394                                               VIRTIO_IOMMU_RESV_MEM_T_MSI);
1395         QList *reserved_regions = qlist_new();
1396 
1397         qlist_append_str(reserved_regions, resv_prop_str);
1398         qdev_prop_set_array(dev, "reserved-regions", reserved_regions);
1399 
1400         g_free(resv_prop_str);
1401     }
1402 
1403     if (object_dynamic_cast(OBJECT(dev), TYPE_X86_IOMMU_DEVICE) ||
1404         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
1405         PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1406 
1407         if (pcms->iommu) {
1408             error_setg(errp, "QEMU does not support multiple vIOMMUs "
1409                        "for x86 yet.");
1410             return;
1411         }
1412         pcms->iommu = dev;
1413     } else if (object_dynamic_cast(OBJECT(dev), TYPE_HV_BALLOON)) {
1414         pc_hv_balloon_pre_plug(hotplug_dev, dev, errp);
1415     }
1416 }
1417 
1418 static void pc_machine_device_plug_cb(HotplugHandler *hotplug_dev,
1419                                       DeviceState *dev, Error **errp)
1420 {
1421     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1422         pc_memory_plug(hotplug_dev, dev, errp);
1423     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1424         x86_cpu_plug(hotplug_dev, dev, errp);
1425     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI)) {
1426         virtio_md_pci_plug(VIRTIO_MD_PCI(dev), MACHINE(hotplug_dev), errp);
1427     } else if (object_dynamic_cast(OBJECT(dev), TYPE_HV_BALLOON)) {
1428         pc_hv_balloon_plug(hotplug_dev, dev, errp);
1429     }
1430 }
1431 
1432 static void pc_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
1433                                                 DeviceState *dev, Error **errp)
1434 {
1435     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1436         pc_memory_unplug_request(hotplug_dev, dev, errp);
1437     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1438         x86_cpu_unplug_request_cb(hotplug_dev, dev, errp);
1439     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI)) {
1440         virtio_md_pci_unplug_request(VIRTIO_MD_PCI(dev), MACHINE(hotplug_dev),
1441                                      errp);
1442     } else {
1443         error_setg(errp, "acpi: device unplug request for not supported device"
1444                    " type: %s", object_get_typename(OBJECT(dev)));
1445     }
1446 }
1447 
1448 static void pc_machine_device_unplug_cb(HotplugHandler *hotplug_dev,
1449                                         DeviceState *dev, Error **errp)
1450 {
1451     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1452         pc_memory_unplug(hotplug_dev, dev, errp);
1453     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1454         x86_cpu_unplug_cb(hotplug_dev, dev, errp);
1455     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI)) {
1456         virtio_md_pci_unplug(VIRTIO_MD_PCI(dev), MACHINE(hotplug_dev), errp);
1457     } else {
1458         error_setg(errp, "acpi: device unplug for not supported device"
1459                    " type: %s", object_get_typename(OBJECT(dev)));
1460     }
1461 }
1462 
1463 static HotplugHandler *pc_get_hotplug_handler(MachineState *machine,
1464                                              DeviceState *dev)
1465 {
1466     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) ||
1467         object_dynamic_cast(OBJECT(dev), TYPE_CPU) ||
1468         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI) ||
1469         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI) ||
1470         object_dynamic_cast(OBJECT(dev), TYPE_HV_BALLOON) ||
1471         object_dynamic_cast(OBJECT(dev), TYPE_X86_IOMMU_DEVICE)) {
1472         return HOTPLUG_HANDLER(machine);
1473     }
1474 
1475     return NULL;
1476 }
1477 
1478 static void pc_machine_get_vmport(Object *obj, Visitor *v, const char *name,
1479                                   void *opaque, Error **errp)
1480 {
1481     PCMachineState *pcms = PC_MACHINE(obj);
1482     OnOffAuto vmport = pcms->vmport;
1483 
1484     visit_type_OnOffAuto(v, name, &vmport, errp);
1485 }
1486 
1487 static void pc_machine_set_vmport(Object *obj, Visitor *v, const char *name,
1488                                   void *opaque, Error **errp)
1489 {
1490     PCMachineState *pcms = PC_MACHINE(obj);
1491 
1492     visit_type_OnOffAuto(v, name, &pcms->vmport, errp);
1493 }
1494 
1495 static bool pc_machine_get_fd_bootchk(Object *obj, Error **errp)
1496 {
1497     PCMachineState *pcms = PC_MACHINE(obj);
1498 
1499     return pcms->fd_bootchk;
1500 }
1501 
1502 static void pc_machine_set_fd_bootchk(Object *obj, bool value, Error **errp)
1503 {
1504     PCMachineState *pcms = PC_MACHINE(obj);
1505 
1506     pcms->fd_bootchk = value;
1507 }
1508 
1509 static bool pc_machine_get_smbus(Object *obj, Error **errp)
1510 {
1511     PCMachineState *pcms = PC_MACHINE(obj);
1512 
1513     return pcms->smbus_enabled;
1514 }
1515 
1516 static void pc_machine_set_smbus(Object *obj, bool value, Error **errp)
1517 {
1518     PCMachineState *pcms = PC_MACHINE(obj);
1519 
1520     pcms->smbus_enabled = value;
1521 }
1522 
1523 static bool pc_machine_get_sata(Object *obj, Error **errp)
1524 {
1525     PCMachineState *pcms = PC_MACHINE(obj);
1526 
1527     return pcms->sata_enabled;
1528 }
1529 
1530 static void pc_machine_set_sata(Object *obj, bool value, Error **errp)
1531 {
1532     PCMachineState *pcms = PC_MACHINE(obj);
1533 
1534     pcms->sata_enabled = value;
1535 }
1536 
1537 static bool pc_machine_get_hpet(Object *obj, Error **errp)
1538 {
1539     PCMachineState *pcms = PC_MACHINE(obj);
1540 
1541     return pcms->hpet_enabled;
1542 }
1543 
1544 static void pc_machine_set_hpet(Object *obj, bool value, Error **errp)
1545 {
1546     PCMachineState *pcms = PC_MACHINE(obj);
1547 
1548     pcms->hpet_enabled = value;
1549 }
1550 
1551 static bool pc_machine_get_i8042(Object *obj, Error **errp)
1552 {
1553     PCMachineState *pcms = PC_MACHINE(obj);
1554 
1555     return pcms->i8042_enabled;
1556 }
1557 
1558 static void pc_machine_set_i8042(Object *obj, bool value, Error **errp)
1559 {
1560     PCMachineState *pcms = PC_MACHINE(obj);
1561 
1562     pcms->i8042_enabled = value;
1563 }
1564 
1565 static bool pc_machine_get_default_bus_bypass_iommu(Object *obj, Error **errp)
1566 {
1567     PCMachineState *pcms = PC_MACHINE(obj);
1568 
1569     return pcms->default_bus_bypass_iommu;
1570 }
1571 
1572 static void pc_machine_set_default_bus_bypass_iommu(Object *obj, bool value,
1573                                                     Error **errp)
1574 {
1575     PCMachineState *pcms = PC_MACHINE(obj);
1576 
1577     pcms->default_bus_bypass_iommu = value;
1578 }
1579 
1580 static void pc_machine_get_smbios_ep(Object *obj, Visitor *v, const char *name,
1581                                      void *opaque, Error **errp)
1582 {
1583     PCMachineState *pcms = PC_MACHINE(obj);
1584     SmbiosEntryPointType smbios_entry_point_type = pcms->smbios_entry_point_type;
1585 
1586     visit_type_SmbiosEntryPointType(v, name, &smbios_entry_point_type, errp);
1587 }
1588 
1589 static void pc_machine_set_smbios_ep(Object *obj, Visitor *v, const char *name,
1590                                      void *opaque, Error **errp)
1591 {
1592     PCMachineState *pcms = PC_MACHINE(obj);
1593 
1594     visit_type_SmbiosEntryPointType(v, name, &pcms->smbios_entry_point_type, errp);
1595 }
1596 
1597 static void pc_machine_get_max_ram_below_4g(Object *obj, Visitor *v,
1598                                             const char *name, void *opaque,
1599                                             Error **errp)
1600 {
1601     PCMachineState *pcms = PC_MACHINE(obj);
1602     uint64_t value = pcms->max_ram_below_4g;
1603 
1604     visit_type_size(v, name, &value, errp);
1605 }
1606 
1607 static void pc_machine_set_max_ram_below_4g(Object *obj, Visitor *v,
1608                                             const char *name, void *opaque,
1609                                             Error **errp)
1610 {
1611     PCMachineState *pcms = PC_MACHINE(obj);
1612     uint64_t value;
1613 
1614     if (!visit_type_size(v, name, &value, errp)) {
1615         return;
1616     }
1617     if (value > 4 * GiB) {
1618         error_setg(errp,
1619                    "Machine option 'max-ram-below-4g=%"PRIu64
1620                    "' expects size less than or equal to 4G", value);
1621         return;
1622     }
1623 
1624     if (value < 1 * MiB) {
1625         warn_report("Only %" PRIu64 " bytes of RAM below the 4GiB boundary,"
1626                     "BIOS may not work with less than 1MiB", value);
1627     }
1628 
1629     pcms->max_ram_below_4g = value;
1630 }
1631 
1632 static void pc_machine_get_max_fw_size(Object *obj, Visitor *v,
1633                                        const char *name, void *opaque,
1634                                        Error **errp)
1635 {
1636     PCMachineState *pcms = PC_MACHINE(obj);
1637     uint64_t value = pcms->max_fw_size;
1638 
1639     visit_type_size(v, name, &value, errp);
1640 }
1641 
1642 static void pc_machine_set_max_fw_size(Object *obj, Visitor *v,
1643                                        const char *name, void *opaque,
1644                                        Error **errp)
1645 {
1646     PCMachineState *pcms = PC_MACHINE(obj);
1647     uint64_t value;
1648 
1649     if (!visit_type_size(v, name, &value, errp)) {
1650         return;
1651     }
1652 
1653     /*
1654      * We don't have a theoretically justifiable exact lower bound on the base
1655      * address of any flash mapping. In practice, the IO-APIC MMIO range is
1656      * [0xFEE00000..0xFEE01000] -- see IO_APIC_DEFAULT_ADDRESS --, leaving free
1657      * only 18MiB-4KiB below 4GiB. For now, restrict the cumulative mapping to
1658      * 16MiB in size.
1659      */
1660     if (value > 16 * MiB) {
1661         error_setg(errp,
1662                    "User specified max allowed firmware size %" PRIu64 " is "
1663                    "greater than 16MiB. If combined firmware size exceeds "
1664                    "16MiB the system may not boot, or experience intermittent"
1665                    "stability issues.",
1666                    value);
1667         return;
1668     }
1669 
1670     pcms->max_fw_size = value;
1671 }
1672 
1673 
1674 static void pc_machine_initfn(Object *obj)
1675 {
1676     PCMachineState *pcms = PC_MACHINE(obj);
1677     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1678 
1679 #ifdef CONFIG_VMPORT
1680     pcms->vmport = ON_OFF_AUTO_AUTO;
1681 #else
1682     pcms->vmport = ON_OFF_AUTO_OFF;
1683 #endif /* CONFIG_VMPORT */
1684     pcms->max_ram_below_4g = 0; /* use default */
1685     pcms->smbios_entry_point_type = pcmc->default_smbios_ep_type;
1686     pcms->south_bridge = pcmc->default_south_bridge;
1687 
1688     /* acpi build is enabled by default if machine supports it */
1689     pcms->acpi_build_enabled = pcmc->has_acpi_build;
1690     pcms->smbus_enabled = true;
1691     pcms->sata_enabled = true;
1692     pcms->i8042_enabled = true;
1693     pcms->max_fw_size = 8 * MiB;
1694 #ifdef CONFIG_HPET
1695     pcms->hpet_enabled = true;
1696 #endif
1697     pcms->fd_bootchk = true;
1698     pcms->default_bus_bypass_iommu = false;
1699 
1700     pc_system_flash_create(pcms);
1701     pcms->pcspk = isa_new(TYPE_PC_SPEAKER);
1702     object_property_add_alias(OBJECT(pcms), "pcspk-audiodev",
1703                               OBJECT(pcms->pcspk), "audiodev");
1704     if (pcmc->pci_enabled) {
1705         cxl_machine_init(obj, &pcms->cxl_devices_state);
1706     }
1707 
1708     pcms->machine_done.notify = pc_machine_done;
1709     qemu_add_machine_init_done_notifier(&pcms->machine_done);
1710 }
1711 
1712 static void pc_machine_reset(MachineState *machine, ShutdownCause reason)
1713 {
1714     CPUState *cs;
1715     X86CPU *cpu;
1716 
1717     qemu_devices_reset(reason);
1718 
1719     /* Reset APIC after devices have been reset to cancel
1720      * any changes that qemu_devices_reset() might have done.
1721      */
1722     CPU_FOREACH(cs) {
1723         cpu = X86_CPU(cs);
1724 
1725         x86_cpu_after_reset(cpu);
1726     }
1727 }
1728 
1729 static void pc_machine_wakeup(MachineState *machine)
1730 {
1731     cpu_synchronize_all_states();
1732     pc_machine_reset(machine, SHUTDOWN_CAUSE_NONE);
1733     cpu_synchronize_all_post_reset();
1734 }
1735 
1736 static bool pc_hotplug_allowed(MachineState *ms, DeviceState *dev, Error **errp)
1737 {
1738     X86IOMMUState *iommu = x86_iommu_get_default();
1739     IntelIOMMUState *intel_iommu;
1740 
1741     if (iommu &&
1742         object_dynamic_cast((Object *)iommu, TYPE_INTEL_IOMMU_DEVICE) &&
1743         object_dynamic_cast((Object *)dev, "vfio-pci")) {
1744         intel_iommu = INTEL_IOMMU_DEVICE(iommu);
1745         if (!intel_iommu->caching_mode) {
1746             error_setg(errp, "Device assignment is not allowed without "
1747                        "enabling caching-mode=on for Intel IOMMU.");
1748             return false;
1749         }
1750     }
1751 
1752     return true;
1753 }
1754 
1755 static void pc_machine_class_init(ObjectClass *oc, void *data)
1756 {
1757     MachineClass *mc = MACHINE_CLASS(oc);
1758     X86MachineClass *x86mc = X86_MACHINE_CLASS(oc);
1759     PCMachineClass *pcmc = PC_MACHINE_CLASS(oc);
1760     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
1761 
1762     pcmc->pci_enabled = true;
1763     pcmc->has_acpi_build = true;
1764     pcmc->smbios_defaults = true;
1765     pcmc->gigabyte_align = true;
1766     pcmc->has_reserved_memory = true;
1767     pcmc->enforce_amd_1tb_hole = true;
1768     pcmc->isa_bios_alias = true;
1769     pcmc->pvh_enabled = true;
1770     pcmc->kvmclock_create_always = true;
1771     x86mc->apic_xrupt_override = true;
1772     assert(!mc->get_hotplug_handler);
1773     mc->get_hotplug_handler = pc_get_hotplug_handler;
1774     mc->hotplug_allowed = pc_hotplug_allowed;
1775     mc->auto_enable_numa_with_memhp = true;
1776     mc->auto_enable_numa_with_memdev = true;
1777     mc->has_hotpluggable_cpus = true;
1778     mc->default_boot_order = "cad";
1779     mc->block_default_type = IF_IDE;
1780     mc->max_cpus = 255;
1781     mc->reset = pc_machine_reset;
1782     mc->wakeup = pc_machine_wakeup;
1783     hc->pre_plug = pc_machine_device_pre_plug_cb;
1784     hc->plug = pc_machine_device_plug_cb;
1785     hc->unplug_request = pc_machine_device_unplug_request_cb;
1786     hc->unplug = pc_machine_device_unplug_cb;
1787     mc->default_cpu_type = TARGET_DEFAULT_CPU_TYPE;
1788     mc->nvdimm_supported = true;
1789     mc->smp_props.dies_supported = true;
1790     mc->smp_props.modules_supported = true;
1791     mc->default_ram_id = "pc.ram";
1792     pcmc->default_smbios_ep_type = SMBIOS_ENTRY_POINT_TYPE_AUTO;
1793 
1794     object_class_property_add(oc, PC_MACHINE_MAX_RAM_BELOW_4G, "size",
1795         pc_machine_get_max_ram_below_4g, pc_machine_set_max_ram_below_4g,
1796         NULL, NULL);
1797     object_class_property_set_description(oc, PC_MACHINE_MAX_RAM_BELOW_4G,
1798         "Maximum ram below the 4G boundary (32bit boundary)");
1799 
1800     object_class_property_add(oc, PC_MACHINE_VMPORT, "OnOffAuto",
1801         pc_machine_get_vmport, pc_machine_set_vmport,
1802         NULL, NULL);
1803     object_class_property_set_description(oc, PC_MACHINE_VMPORT,
1804         "Enable vmport (pc & q35)");
1805 
1806     object_class_property_add_bool(oc, PC_MACHINE_SMBUS,
1807         pc_machine_get_smbus, pc_machine_set_smbus);
1808     object_class_property_set_description(oc, PC_MACHINE_SMBUS,
1809         "Enable/disable system management bus");
1810 
1811     object_class_property_add_bool(oc, PC_MACHINE_SATA,
1812         pc_machine_get_sata, pc_machine_set_sata);
1813     object_class_property_set_description(oc, PC_MACHINE_SATA,
1814         "Enable/disable Serial ATA bus");
1815 
1816     object_class_property_add_bool(oc, "hpet",
1817         pc_machine_get_hpet, pc_machine_set_hpet);
1818     object_class_property_set_description(oc, "hpet",
1819         "Enable/disable high precision event timer emulation");
1820 
1821     object_class_property_add_bool(oc, PC_MACHINE_I8042,
1822         pc_machine_get_i8042, pc_machine_set_i8042);
1823 
1824     object_class_property_add_bool(oc, "default-bus-bypass-iommu",
1825         pc_machine_get_default_bus_bypass_iommu,
1826         pc_machine_set_default_bus_bypass_iommu);
1827 
1828     object_class_property_add(oc, PC_MACHINE_MAX_FW_SIZE, "size",
1829         pc_machine_get_max_fw_size, pc_machine_set_max_fw_size,
1830         NULL, NULL);
1831     object_class_property_set_description(oc, PC_MACHINE_MAX_FW_SIZE,
1832         "Maximum combined firmware size");
1833 
1834     object_class_property_add(oc, PC_MACHINE_SMBIOS_EP, "str",
1835         pc_machine_get_smbios_ep, pc_machine_set_smbios_ep,
1836         NULL, NULL);
1837     object_class_property_set_description(oc, PC_MACHINE_SMBIOS_EP,
1838         "SMBIOS Entry Point type [32, 64]");
1839 
1840     object_class_property_add_bool(oc, "fd-bootchk",
1841         pc_machine_get_fd_bootchk,
1842         pc_machine_set_fd_bootchk);
1843 }
1844 
1845 static const TypeInfo pc_machine_info = {
1846     .name = TYPE_PC_MACHINE,
1847     .parent = TYPE_X86_MACHINE,
1848     .abstract = true,
1849     .instance_size = sizeof(PCMachineState),
1850     .instance_init = pc_machine_initfn,
1851     .class_size = sizeof(PCMachineClass),
1852     .class_init = pc_machine_class_init,
1853     .interfaces = (InterfaceInfo[]) {
1854          { TYPE_HOTPLUG_HANDLER },
1855          { }
1856     },
1857 };
1858 
1859 static void pc_machine_register_types(void)
1860 {
1861     type_register_static(&pc_machine_info);
1862 }
1863 
1864 type_init(pc_machine_register_types)
1865