xref: /openbmc/qemu/hw/i386/pc.c (revision ef7e76a2)
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 "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     { "ICH9-LPC", "x-smi-swsmi-timer", "off" },
84     { "ICH9-LPC", "x-smi-periodic-timer", "off" },
85     { TYPE_INTEL_IOMMU_DEVICE, "stale-tm", "on" },
86 };
87 const size_t pc_compat_9_1_len = G_N_ELEMENTS(pc_compat_9_1);
88 
89 GlobalProperty pc_compat_9_0[] = {
90     { TYPE_X86_CPU, "x-amd-topoext-features-only", "false" },
91     { TYPE_X86_CPU, "x-l1-cache-per-thread", "false" },
92     { TYPE_X86_CPU, "guest-phys-bits", "0" },
93     { "sev-guest", "legacy-vm-type", "on" },
94     { TYPE_X86_CPU, "legacy-multi-node", "on" },
95 };
96 const size_t pc_compat_9_0_len = G_N_ELEMENTS(pc_compat_9_0);
97 
98 GlobalProperty pc_compat_8_2[] = {};
99 const size_t pc_compat_8_2_len = G_N_ELEMENTS(pc_compat_8_2);
100 
101 GlobalProperty pc_compat_8_1[] = {};
102 const size_t pc_compat_8_1_len = G_N_ELEMENTS(pc_compat_8_1);
103 
104 GlobalProperty pc_compat_8_0[] = {
105     { "virtio-mem", "unplugged-inaccessible", "auto" },
106 };
107 const size_t pc_compat_8_0_len = G_N_ELEMENTS(pc_compat_8_0);
108 
109 GlobalProperty pc_compat_7_2[] = {
110     { "ICH9-LPC", "noreboot", "true" },
111 };
112 const size_t pc_compat_7_2_len = G_N_ELEMENTS(pc_compat_7_2);
113 
114 GlobalProperty pc_compat_7_1[] = {};
115 const size_t pc_compat_7_1_len = G_N_ELEMENTS(pc_compat_7_1);
116 
117 GlobalProperty pc_compat_7_0[] = {};
118 const size_t pc_compat_7_0_len = G_N_ELEMENTS(pc_compat_7_0);
119 
120 GlobalProperty pc_compat_6_2[] = {
121     { "virtio-mem", "unplugged-inaccessible", "off" },
122 };
123 const size_t pc_compat_6_2_len = G_N_ELEMENTS(pc_compat_6_2);
124 
125 GlobalProperty pc_compat_6_1[] = {
126     { TYPE_X86_CPU, "hv-version-id-build", "0x1bbc" },
127     { TYPE_X86_CPU, "hv-version-id-major", "0x0006" },
128     { TYPE_X86_CPU, "hv-version-id-minor", "0x0001" },
129     { "ICH9-LPC", "x-keep-pci-slot-hpc", "false" },
130 };
131 const size_t pc_compat_6_1_len = G_N_ELEMENTS(pc_compat_6_1);
132 
133 GlobalProperty pc_compat_6_0[] = {
134     { "qemu64" "-" TYPE_X86_CPU, "family", "6" },
135     { "qemu64" "-" TYPE_X86_CPU, "model", "6" },
136     { "qemu64" "-" TYPE_X86_CPU, "stepping", "3" },
137     { TYPE_X86_CPU, "x-vendor-cpuid-only", "off" },
138     { "ICH9-LPC", ACPI_PM_PROP_ACPI_PCIHP_BRIDGE, "off" },
139     { "ICH9-LPC", "x-keep-pci-slot-hpc", "true" },
140 };
141 const size_t pc_compat_6_0_len = G_N_ELEMENTS(pc_compat_6_0);
142 
143 GlobalProperty pc_compat_5_2[] = {
144     { "ICH9-LPC", "x-smi-cpu-hotunplug", "off" },
145 };
146 const size_t pc_compat_5_2_len = G_N_ELEMENTS(pc_compat_5_2);
147 
148 GlobalProperty pc_compat_5_1[] = {
149     { "ICH9-LPC", "x-smi-cpu-hotplug", "off" },
150     { TYPE_X86_CPU, "kvm-msi-ext-dest-id", "off" },
151 };
152 const size_t pc_compat_5_1_len = G_N_ELEMENTS(pc_compat_5_1);
153 
154 GlobalProperty pc_compat_5_0[] = {
155 };
156 const size_t pc_compat_5_0_len = G_N_ELEMENTS(pc_compat_5_0);
157 
158 GlobalProperty pc_compat_4_2[] = {
159     { "mch", "smbase-smram", "off" },
160 };
161 const size_t pc_compat_4_2_len = G_N_ELEMENTS(pc_compat_4_2);
162 
163 GlobalProperty pc_compat_4_1[] = {};
164 const size_t pc_compat_4_1_len = G_N_ELEMENTS(pc_compat_4_1);
165 
166 GlobalProperty pc_compat_4_0[] = {};
167 const size_t pc_compat_4_0_len = G_N_ELEMENTS(pc_compat_4_0);
168 
169 GlobalProperty pc_compat_3_1[] = {
170     { "intel-iommu", "dma-drain", "off" },
171     { "Opteron_G3" "-" TYPE_X86_CPU, "rdtscp", "off" },
172     { "Opteron_G4" "-" TYPE_X86_CPU, "rdtscp", "off" },
173     { "Opteron_G4" "-" TYPE_X86_CPU, "npt", "off" },
174     { "Opteron_G4" "-" TYPE_X86_CPU, "nrip-save", "off" },
175     { "Opteron_G5" "-" TYPE_X86_CPU, "rdtscp", "off" },
176     { "Opteron_G5" "-" TYPE_X86_CPU, "npt", "off" },
177     { "Opteron_G5" "-" TYPE_X86_CPU, "nrip-save", "off" },
178     { "EPYC" "-" TYPE_X86_CPU, "npt", "off" },
179     { "EPYC" "-" TYPE_X86_CPU, "nrip-save", "off" },
180     { "EPYC-IBPB" "-" TYPE_X86_CPU, "npt", "off" },
181     { "EPYC-IBPB" "-" TYPE_X86_CPU, "nrip-save", "off" },
182     { "Skylake-Client" "-" TYPE_X86_CPU,      "mpx", "on" },
183     { "Skylake-Client-IBRS" "-" TYPE_X86_CPU, "mpx", "on" },
184     { "Skylake-Server" "-" TYPE_X86_CPU,      "mpx", "on" },
185     { "Skylake-Server-IBRS" "-" TYPE_X86_CPU, "mpx", "on" },
186     { "Cascadelake-Server" "-" TYPE_X86_CPU,  "mpx", "on" },
187     { "Icelake-Client" "-" TYPE_X86_CPU,      "mpx", "on" },
188     { "Icelake-Server" "-" TYPE_X86_CPU,      "mpx", "on" },
189     { "Cascadelake-Server" "-" TYPE_X86_CPU, "stepping", "5" },
190     { TYPE_X86_CPU, "x-intel-pt-auto-level", "off" },
191 };
192 const size_t pc_compat_3_1_len = G_N_ELEMENTS(pc_compat_3_1);
193 
194 GlobalProperty pc_compat_3_0[] = {
195     { TYPE_X86_CPU, "x-hv-synic-kvm-only", "on" },
196     { "Skylake-Server" "-" TYPE_X86_CPU, "pku", "off" },
197     { "Skylake-Server-IBRS" "-" TYPE_X86_CPU, "pku", "off" },
198 };
199 const size_t pc_compat_3_0_len = G_N_ELEMENTS(pc_compat_3_0);
200 
201 GlobalProperty pc_compat_2_12[] = {
202     { TYPE_X86_CPU, "legacy-cache", "on" },
203     { TYPE_X86_CPU, "topoext", "off" },
204     { "EPYC-" TYPE_X86_CPU, "xlevel", "0x8000000a" },
205     { "EPYC-IBPB-" TYPE_X86_CPU, "xlevel", "0x8000000a" },
206 };
207 const size_t pc_compat_2_12_len = G_N_ELEMENTS(pc_compat_2_12);
208 
209 GlobalProperty pc_compat_2_11[] = {
210     { TYPE_X86_CPU, "x-migrate-smi-count", "off" },
211     { "Skylake-Server" "-" TYPE_X86_CPU, "clflushopt", "off" },
212 };
213 const size_t pc_compat_2_11_len = G_N_ELEMENTS(pc_compat_2_11);
214 
215 GlobalProperty pc_compat_2_10[] = {
216     { TYPE_X86_CPU, "x-hv-max-vps", "0x40" },
217     { "i440FX-pcihost", "x-pci-hole64-fix", "off" },
218     { "q35-pcihost", "x-pci-hole64-fix", "off" },
219 };
220 const size_t pc_compat_2_10_len = G_N_ELEMENTS(pc_compat_2_10);
221 
222 GlobalProperty pc_compat_2_9[] = {
223     { "mch", "extended-tseg-mbytes", "0" },
224 };
225 const size_t pc_compat_2_9_len = G_N_ELEMENTS(pc_compat_2_9);
226 
227 GlobalProperty pc_compat_2_8[] = {
228     { TYPE_X86_CPU, "tcg-cpuid", "off" },
229     { "kvmclock", "x-mach-use-reliable-get-clock", "off" },
230     { "ICH9-LPC", "x-smi-broadcast", "off" },
231     { TYPE_X86_CPU, "vmware-cpuid-freq", "off" },
232     { "Haswell-" TYPE_X86_CPU, "stepping", "1" },
233 };
234 const size_t pc_compat_2_8_len = G_N_ELEMENTS(pc_compat_2_8);
235 
236 GlobalProperty pc_compat_2_7[] = {
237     { TYPE_X86_CPU, "l3-cache", "off" },
238     { TYPE_X86_CPU, "full-cpuid-auto-level", "off" },
239     { "Opteron_G3" "-" TYPE_X86_CPU, "family", "15" },
240     { "Opteron_G3" "-" TYPE_X86_CPU, "model", "6" },
241     { "Opteron_G3" "-" TYPE_X86_CPU, "stepping", "1" },
242     { "isa-pcspk", "migrate", "off" },
243 };
244 const size_t pc_compat_2_7_len = G_N_ELEMENTS(pc_compat_2_7);
245 
246 GlobalProperty pc_compat_2_6[] = {
247     { TYPE_X86_CPU, "cpuid-0xb", "off" },
248     { "vmxnet3", "romfile", "" },
249     { TYPE_X86_CPU, "fill-mtrr-mask", "off" },
250     { "apic-common", "legacy-instance-id", "on", }
251 };
252 const size_t pc_compat_2_6_len = G_N_ELEMENTS(pc_compat_2_6);
253 
254 GlobalProperty pc_compat_2_5[] = {};
255 const size_t pc_compat_2_5_len = G_N_ELEMENTS(pc_compat_2_5);
256 
257 GlobalProperty pc_compat_2_4[] = {
258     PC_CPU_MODEL_IDS("2.4.0")
259     { "Haswell-" TYPE_X86_CPU, "abm", "off" },
260     { "Haswell-noTSX-" TYPE_X86_CPU, "abm", "off" },
261     { "Broadwell-" TYPE_X86_CPU, "abm", "off" },
262     { "Broadwell-noTSX-" TYPE_X86_CPU, "abm", "off" },
263     { "host" "-" TYPE_X86_CPU, "host-cache-info", "on" },
264     { TYPE_X86_CPU, "check", "off" },
265     { "qemu64" "-" TYPE_X86_CPU, "sse4a", "on" },
266     { "qemu64" "-" TYPE_X86_CPU, "abm", "on" },
267     { "qemu64" "-" TYPE_X86_CPU, "popcnt", "on" },
268     { "qemu32" "-" TYPE_X86_CPU, "popcnt", "on" },
269     { "Opteron_G2" "-" TYPE_X86_CPU, "rdtscp", "on" },
270     { "Opteron_G3" "-" TYPE_X86_CPU, "rdtscp", "on" },
271     { "Opteron_G4" "-" TYPE_X86_CPU, "rdtscp", "on" },
272     { "Opteron_G5" "-" TYPE_X86_CPU, "rdtscp", "on", }
273 };
274 const size_t pc_compat_2_4_len = G_N_ELEMENTS(pc_compat_2_4);
275 
276 /*
277  * @PC_FW_DATA:
278  * Size of the chunk of memory at the top of RAM for the BIOS ACPI tables
279  * and other BIOS datastructures.
280  *
281  * BIOS ACPI tables: 128K. Other BIOS datastructures: less than 4K
282  * reported to be used at the moment, 32K should be enough for a while.
283  */
284 #define PC_FW_DATA (0x20000 + 0x8000)
285 
286 GSIState *pc_gsi_create(qemu_irq **irqs, bool pci_enabled)
287 {
288     GSIState *s;
289 
290     s = g_new0(GSIState, 1);
291     if (kvm_ioapic_in_kernel()) {
292         kvm_pc_setup_irq_routing(pci_enabled);
293     }
294     *irqs = qemu_allocate_irqs(gsi_handler, s, IOAPIC_NUM_PINS);
295 
296     return s;
297 }
298 
299 static void ioport80_write(void *opaque, hwaddr addr, uint64_t data,
300                            unsigned size)
301 {
302 }
303 
304 static uint64_t ioport80_read(void *opaque, hwaddr addr, unsigned size)
305 {
306     return 0xffffffffffffffffULL;
307 }
308 
309 /* MS-DOS compatibility mode FPU exception support */
310 static void ioportF0_write(void *opaque, hwaddr addr, uint64_t data,
311                            unsigned size)
312 {
313     if (tcg_enabled()) {
314         cpu_set_ignne();
315     }
316 }
317 
318 static uint64_t ioportF0_read(void *opaque, hwaddr addr, unsigned size)
319 {
320     return 0xffffffffffffffffULL;
321 }
322 
323 /* PC cmos mappings */
324 
325 #define REG_EQUIPMENT_BYTE          0x14
326 
327 static void cmos_init_hd(MC146818RtcState *s, int type_ofs, int info_ofs,
328                          int16_t cylinders, int8_t heads, int8_t sectors)
329 {
330     mc146818rtc_set_cmos_data(s, type_ofs, 47);
331     mc146818rtc_set_cmos_data(s, info_ofs, cylinders);
332     mc146818rtc_set_cmos_data(s, info_ofs + 1, cylinders >> 8);
333     mc146818rtc_set_cmos_data(s, info_ofs + 2, heads);
334     mc146818rtc_set_cmos_data(s, info_ofs + 3, 0xff);
335     mc146818rtc_set_cmos_data(s, info_ofs + 4, 0xff);
336     mc146818rtc_set_cmos_data(s, info_ofs + 5, 0xc0 | ((heads > 8) << 3));
337     mc146818rtc_set_cmos_data(s, info_ofs + 6, cylinders);
338     mc146818rtc_set_cmos_data(s, info_ofs + 7, cylinders >> 8);
339     mc146818rtc_set_cmos_data(s, info_ofs + 8, sectors);
340 }
341 
342 /* convert boot_device letter to something recognizable by the bios */
343 static int boot_device2nibble(char boot_device)
344 {
345     switch(boot_device) {
346     case 'a':
347     case 'b':
348         return 0x01; /* floppy boot */
349     case 'c':
350         return 0x02; /* hard drive boot */
351     case 'd':
352         return 0x03; /* CD-ROM boot */
353     case 'n':
354         return 0x04; /* Network boot */
355     }
356     return 0;
357 }
358 
359 static void set_boot_dev(PCMachineState *pcms, MC146818RtcState *s,
360                          const char *boot_device, Error **errp)
361 {
362 #define PC_MAX_BOOT_DEVICES 3
363     int nbds, bds[3] = { 0, };
364     int i;
365 
366     nbds = strlen(boot_device);
367     if (nbds > PC_MAX_BOOT_DEVICES) {
368         error_setg(errp, "Too many boot devices for PC");
369         return;
370     }
371     for (i = 0; i < nbds; i++) {
372         bds[i] = boot_device2nibble(boot_device[i]);
373         if (bds[i] == 0) {
374             error_setg(errp, "Invalid boot device for PC: '%c'",
375                        boot_device[i]);
376             return;
377         }
378     }
379     mc146818rtc_set_cmos_data(s, 0x3d, (bds[1] << 4) | bds[0]);
380     mc146818rtc_set_cmos_data(s, 0x38, (bds[2] << 4) | !pcms->fd_bootchk);
381 }
382 
383 static void pc_boot_set(void *opaque, const char *boot_device, Error **errp)
384 {
385     PCMachineState *pcms = opaque;
386     X86MachineState *x86ms = X86_MACHINE(pcms);
387 
388     set_boot_dev(pcms, MC146818_RTC(x86ms->rtc), boot_device, errp);
389 }
390 
391 static void pc_cmos_init_floppy(MC146818RtcState *rtc_state, ISADevice *floppy)
392 {
393     int val, nb;
394     FloppyDriveType fd_type[2] = { FLOPPY_DRIVE_TYPE_NONE,
395                                    FLOPPY_DRIVE_TYPE_NONE };
396 
397 #ifdef CONFIG_FDC_ISA
398     /* floppy type */
399     if (floppy) {
400         for (int i = 0; i < 2; i++) {
401             fd_type[i] = isa_fdc_get_drive_type(floppy, i);
402         }
403     }
404 #endif
405 
406     val = (cmos_get_fd_drive_type(fd_type[0]) << 4) |
407         cmos_get_fd_drive_type(fd_type[1]);
408     mc146818rtc_set_cmos_data(rtc_state, 0x10, val);
409 
410     val = mc146818rtc_get_cmos_data(rtc_state, REG_EQUIPMENT_BYTE);
411     nb = 0;
412     if (fd_type[0] != FLOPPY_DRIVE_TYPE_NONE) {
413         nb++;
414     }
415     if (fd_type[1] != FLOPPY_DRIVE_TYPE_NONE) {
416         nb++;
417     }
418     switch (nb) {
419     case 0:
420         break;
421     case 1:
422         val |= 0x01; /* 1 drive, ready for boot */
423         break;
424     case 2:
425         val |= 0x41; /* 2 drives, ready for boot */
426         break;
427     }
428     mc146818rtc_set_cmos_data(rtc_state, REG_EQUIPMENT_BYTE, val);
429 }
430 
431 typedef struct check_fdc_state {
432     ISADevice *floppy;
433     bool multiple;
434 } CheckFdcState;
435 
436 static int check_fdc(Object *obj, void *opaque)
437 {
438     CheckFdcState *state = opaque;
439     Object *fdc;
440     uint32_t iobase;
441     Error *local_err = NULL;
442 
443     fdc = object_dynamic_cast(obj, TYPE_ISA_FDC);
444     if (!fdc) {
445         return 0;
446     }
447 
448     iobase = object_property_get_uint(obj, "iobase", &local_err);
449     if (local_err || iobase != 0x3f0) {
450         error_free(local_err);
451         return 0;
452     }
453 
454     if (state->floppy) {
455         state->multiple = true;
456     } else {
457         state->floppy = ISA_DEVICE(obj);
458     }
459     return 0;
460 }
461 
462 static const char * const fdc_container_path[] = {
463     "/unattached", "/peripheral", "/peripheral-anon"
464 };
465 
466 /*
467  * Locate the FDC at IO address 0x3f0, in order to configure the CMOS registers
468  * and ACPI objects.
469  */
470 static ISADevice *pc_find_fdc0(void)
471 {
472     int i;
473     Object *container;
474     CheckFdcState state = { 0 };
475 
476     for (i = 0; i < ARRAY_SIZE(fdc_container_path); i++) {
477         container = container_get(qdev_get_machine(), fdc_container_path[i]);
478         object_child_foreach(container, check_fdc, &state);
479     }
480 
481     if (state.multiple) {
482         warn_report("multiple floppy disk controllers with "
483                     "iobase=0x3f0 have been found");
484         error_printf("the one being picked for CMOS setup might not reflect "
485                      "your intent");
486     }
487 
488     return state.floppy;
489 }
490 
491 static void pc_cmos_init_late(PCMachineState *pcms)
492 {
493     X86MachineState *x86ms = X86_MACHINE(pcms);
494     MC146818RtcState *s = MC146818_RTC(x86ms->rtc);
495     int16_t cylinders;
496     int8_t heads, sectors;
497     int val;
498     int i, trans;
499 
500     val = 0;
501     if (pcms->idebus[0] &&
502         ide_get_geometry(pcms->idebus[0], 0,
503                          &cylinders, &heads, &sectors) >= 0) {
504         cmos_init_hd(s, 0x19, 0x1b, cylinders, heads, sectors);
505         val |= 0xf0;
506     }
507     if (pcms->idebus[0] &&
508         ide_get_geometry(pcms->idebus[0], 1,
509                          &cylinders, &heads, &sectors) >= 0) {
510         cmos_init_hd(s, 0x1a, 0x24, cylinders, heads, sectors);
511         val |= 0x0f;
512     }
513     mc146818rtc_set_cmos_data(s, 0x12, val);
514 
515     val = 0;
516     for (i = 0; i < 4; i++) {
517         /* NOTE: ide_get_geometry() returns the physical
518            geometry.  It is always such that: 1 <= sects <= 63, 1
519            <= heads <= 16, 1 <= cylinders <= 16383. The BIOS
520            geometry can be different if a translation is done. */
521         BusState *idebus = pcms->idebus[i / 2];
522         if (idebus &&
523             ide_get_geometry(idebus, i % 2,
524                              &cylinders, &heads, &sectors) >= 0) {
525             trans = ide_get_bios_chs_trans(idebus, i % 2) - 1;
526             assert((trans & ~3) == 0);
527             val |= trans << (i * 2);
528         }
529     }
530     mc146818rtc_set_cmos_data(s, 0x39, val);
531 
532     pc_cmos_init_floppy(s, pc_find_fdc0());
533 
534     /* various important CMOS locations needed by PC/Bochs bios */
535 
536     /* memory size */
537     /* base memory (first MiB) */
538     val = MIN(x86ms->below_4g_mem_size / KiB, 640);
539     mc146818rtc_set_cmos_data(s, 0x15, val);
540     mc146818rtc_set_cmos_data(s, 0x16, val >> 8);
541     /* extended memory (next 64MiB) */
542     if (x86ms->below_4g_mem_size > 1 * MiB) {
543         val = (x86ms->below_4g_mem_size - 1 * MiB) / KiB;
544     } else {
545         val = 0;
546     }
547     if (val > 65535)
548         val = 65535;
549     mc146818rtc_set_cmos_data(s, 0x17, val);
550     mc146818rtc_set_cmos_data(s, 0x18, val >> 8);
551     mc146818rtc_set_cmos_data(s, 0x30, val);
552     mc146818rtc_set_cmos_data(s, 0x31, val >> 8);
553     /* memory between 16MiB and 4GiB */
554     if (x86ms->below_4g_mem_size > 16 * MiB) {
555         val = (x86ms->below_4g_mem_size - 16 * MiB) / (64 * KiB);
556     } else {
557         val = 0;
558     }
559     if (val > 65535)
560         val = 65535;
561     mc146818rtc_set_cmos_data(s, 0x34, val);
562     mc146818rtc_set_cmos_data(s, 0x35, val >> 8);
563     /* memory above 4GiB */
564     val = x86ms->above_4g_mem_size / 65536;
565     mc146818rtc_set_cmos_data(s, 0x5b, val);
566     mc146818rtc_set_cmos_data(s, 0x5c, val >> 8);
567     mc146818rtc_set_cmos_data(s, 0x5d, val >> 16);
568 
569     val = 0;
570     val |= 0x02; /* FPU is there */
571     val |= 0x04; /* PS/2 mouse installed */
572     mc146818rtc_set_cmos_data(s, REG_EQUIPMENT_BYTE, val);
573 }
574 
575 static void handle_a20_line_change(void *opaque, int irq, int level)
576 {
577     X86CPU *cpu = opaque;
578 
579     /* XXX: send to all CPUs ? */
580     /* XXX: add logic to handle multiple A20 line sources */
581     x86_cpu_set_a20(cpu, level);
582 }
583 
584 #define NE2000_NB_MAX 6
585 
586 static const int ne2000_io[NE2000_NB_MAX] = { 0x300, 0x320, 0x340, 0x360,
587                                               0x280, 0x380 };
588 static const int ne2000_irq[NE2000_NB_MAX] = { 9, 10, 11, 3, 4, 5 };
589 
590 static gboolean pc_init_ne2k_isa(ISABus *bus, NICInfo *nd, Error **errp)
591 {
592     static int nb_ne2k = 0;
593 
594     if (nb_ne2k == NE2000_NB_MAX) {
595         error_setg(errp,
596                    "maximum number of ISA NE2000 devices exceeded");
597         return false;
598     }
599     isa_ne2000_init(bus, ne2000_io[nb_ne2k],
600                     ne2000_irq[nb_ne2k], nd);
601     nb_ne2k++;
602     return true;
603 }
604 
605 void pc_acpi_smi_interrupt(void *opaque, int irq, int level)
606 {
607     X86CPU *cpu = opaque;
608 
609     if (level) {
610         cpu_interrupt(CPU(cpu), CPU_INTERRUPT_SMI);
611     }
612 }
613 
614 static
615 void pc_machine_done(Notifier *notifier, void *data)
616 {
617     PCMachineState *pcms = container_of(notifier,
618                                         PCMachineState, machine_done);
619     X86MachineState *x86ms = X86_MACHINE(pcms);
620 
621     cxl_hook_up_pxb_registers(pcms->pcibus, &pcms->cxl_devices_state,
622                               &error_fatal);
623 
624     if (pcms->cxl_devices_state.is_enabled) {
625         cxl_fmws_link_targets(&pcms->cxl_devices_state, &error_fatal);
626     }
627 
628     /* set the number of CPUs */
629     x86_rtc_set_cpus_count(x86ms->rtc, x86ms->boot_cpus);
630 
631     fw_cfg_add_extra_pci_roots(pcms->pcibus, x86ms->fw_cfg);
632 
633     acpi_setup();
634     if (x86ms->fw_cfg) {
635         fw_cfg_build_smbios(pcms, x86ms->fw_cfg, pcms->smbios_entry_point_type);
636         fw_cfg_add_e820(x86ms->fw_cfg);
637         fw_cfg_build_feature_control(MACHINE(pcms), x86ms->fw_cfg);
638         /* update FW_CFG_NB_CPUS to account for -device added CPUs */
639         fw_cfg_modify_i16(x86ms->fw_cfg, FW_CFG_NB_CPUS, x86ms->boot_cpus);
640     }
641 
642     pc_cmos_init_late(pcms);
643 }
644 
645 /* setup pci memory address space mapping into system address space */
646 void pc_pci_as_mapping_init(MemoryRegion *system_memory,
647                             MemoryRegion *pci_address_space)
648 {
649     /* Set to lower priority than RAM */
650     memory_region_add_subregion_overlap(system_memory, 0x0,
651                                         pci_address_space, -1);
652 }
653 
654 void xen_load_linux(PCMachineState *pcms)
655 {
656     int i;
657     FWCfgState *fw_cfg;
658     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
659     X86MachineState *x86ms = X86_MACHINE(pcms);
660 
661     assert(MACHINE(pcms)->kernel_filename != NULL);
662 
663     fw_cfg = fw_cfg_init_io_dma(FW_CFG_IO_BASE, FW_CFG_IO_BASE + 4,
664                                 &address_space_memory);
665     fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, x86ms->boot_cpus);
666     rom_set_fw(fw_cfg);
667 
668     x86_load_linux(x86ms, fw_cfg, PC_FW_DATA, pcmc->pvh_enabled);
669     for (i = 0; i < nb_option_roms; i++) {
670         assert(!strcmp(option_rom[i].name, "linuxboot.bin") ||
671                !strcmp(option_rom[i].name, "linuxboot_dma.bin") ||
672                !strcmp(option_rom[i].name, "pvh.bin") ||
673                !strcmp(option_rom[i].name, "multiboot.bin") ||
674                !strcmp(option_rom[i].name, "multiboot_dma.bin"));
675         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
676     }
677     x86ms->fw_cfg = fw_cfg;
678 }
679 
680 #define PC_ROM_MIN_VGA     0xc0000
681 #define PC_ROM_MIN_OPTION  0xc8000
682 #define PC_ROM_MAX         0xe0000
683 #define PC_ROM_ALIGN       0x800
684 #define PC_ROM_SIZE        (PC_ROM_MAX - PC_ROM_MIN_VGA)
685 
686 static hwaddr pc_above_4g_end(PCMachineState *pcms)
687 {
688     X86MachineState *x86ms = X86_MACHINE(pcms);
689 
690     if (pcms->sgx_epc.size != 0) {
691         return sgx_epc_above_4g_end(&pcms->sgx_epc);
692     }
693 
694     return x86ms->above_4g_mem_start + x86ms->above_4g_mem_size;
695 }
696 
697 static void pc_get_device_memory_range(PCMachineState *pcms,
698                                        hwaddr *base,
699                                        ram_addr_t *device_mem_size)
700 {
701     MachineState *machine = MACHINE(pcms);
702     ram_addr_t size;
703     hwaddr addr;
704 
705     size = machine->maxram_size - machine->ram_size;
706     addr = ROUND_UP(pc_above_4g_end(pcms), 1 * GiB);
707 
708     /* size device region assuming 1G page max alignment per slot */
709     size += (1 * GiB) * machine->ram_slots;
710 
711     *base = addr;
712     *device_mem_size = size;
713 }
714 
715 static uint64_t pc_get_cxl_range_start(PCMachineState *pcms)
716 {
717     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
718     MachineState *ms = MACHINE(pcms);
719     hwaddr cxl_base;
720     ram_addr_t size;
721 
722     if (pcmc->has_reserved_memory &&
723         (ms->ram_size < ms->maxram_size)) {
724         pc_get_device_memory_range(pcms, &cxl_base, &size);
725         cxl_base += size;
726     } else {
727         cxl_base = pc_above_4g_end(pcms);
728     }
729 
730     return cxl_base;
731 }
732 
733 static uint64_t pc_get_cxl_range_end(PCMachineState *pcms)
734 {
735     uint64_t start = pc_get_cxl_range_start(pcms) + MiB;
736 
737     if (pcms->cxl_devices_state.fixed_windows) {
738         GList *it;
739 
740         start = ROUND_UP(start, 256 * MiB);
741         for (it = pcms->cxl_devices_state.fixed_windows; it; it = it->next) {
742             CXLFixedWindow *fw = it->data;
743             start += fw->size;
744         }
745     }
746 
747     return start;
748 }
749 
750 static hwaddr pc_max_used_gpa(PCMachineState *pcms, uint64_t pci_hole64_size)
751 {
752     X86CPU *cpu = X86_CPU(first_cpu);
753     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
754     MachineState *ms = MACHINE(pcms);
755 
756     if (cpu->env.features[FEAT_8000_0001_EDX] & CPUID_EXT2_LM) {
757         /* 64-bit systems */
758         return pc_pci_hole64_start() + pci_hole64_size - 1;
759     }
760 
761     /* 32-bit systems */
762     if (pcmc->broken_32bit_mem_addr_check) {
763         /* old value for compatibility reasons */
764         return ((hwaddr)1 << cpu->phys_bits) - 1;
765     }
766 
767     /*
768      * 32-bit systems don't have hole64 but they might have a region for
769      * memory devices. Even if additional hotplugged memory devices might
770      * not be usable by most guest OSes, we need to still consider them for
771      * calculating the highest possible GPA so that we can properly report
772      * if someone configures them on a CPU that cannot possibly address them.
773      */
774     if (pcmc->has_reserved_memory &&
775         (ms->ram_size < ms->maxram_size)) {
776         hwaddr devmem_start;
777         ram_addr_t devmem_size;
778 
779         pc_get_device_memory_range(pcms, &devmem_start, &devmem_size);
780         devmem_start += devmem_size;
781         return devmem_start - 1;
782     }
783 
784     /* configuration without any memory hotplug */
785     return pc_above_4g_end(pcms) - 1;
786 }
787 
788 /*
789  * AMD systems with an IOMMU have an additional hole close to the
790  * 1Tb, which are special GPAs that cannot be DMA mapped. Depending
791  * on kernel version, VFIO may or may not let you DMA map those ranges.
792  * Starting Linux v5.4 we validate it, and can't create guests on AMD machines
793  * with certain memory sizes. It's also wrong to use those IOVA ranges
794  * in detriment of leading to IOMMU INVALID_DEVICE_REQUEST or worse.
795  * The ranges reserved for Hyper-Transport are:
796  *
797  * FD_0000_0000h - FF_FFFF_FFFFh
798  *
799  * The ranges represent the following:
800  *
801  * Base Address   Top Address  Use
802  *
803  * FD_0000_0000h FD_F7FF_FFFFh Reserved interrupt address space
804  * FD_F800_0000h FD_F8FF_FFFFh Interrupt/EOI IntCtl
805  * FD_F900_0000h FD_F90F_FFFFh Legacy PIC IACK
806  * FD_F910_0000h FD_F91F_FFFFh System Management
807  * FD_F920_0000h FD_FAFF_FFFFh Reserved Page Tables
808  * FD_FB00_0000h FD_FBFF_FFFFh Address Translation
809  * FD_FC00_0000h FD_FDFF_FFFFh I/O Space
810  * FD_FE00_0000h FD_FFFF_FFFFh Configuration
811  * FE_0000_0000h FE_1FFF_FFFFh Extended Configuration/Device Messages
812  * FE_2000_0000h FF_FFFF_FFFFh Reserved
813  *
814  * See AMD IOMMU spec, section 2.1.2 "IOMMU Logical Topology",
815  * Table 3: Special Address Controls (GPA) for more information.
816  */
817 #define AMD_HT_START         0xfd00000000UL
818 #define AMD_HT_END           0xffffffffffUL
819 #define AMD_ABOVE_1TB_START  (AMD_HT_END + 1)
820 #define AMD_HT_SIZE          (AMD_ABOVE_1TB_START - AMD_HT_START)
821 
822 void pc_memory_init(PCMachineState *pcms,
823                     MemoryRegion *system_memory,
824                     MemoryRegion *rom_memory,
825                     uint64_t pci_hole64_size)
826 {
827     int linux_boot, i;
828     MemoryRegion *option_rom_mr;
829     MemoryRegion *ram_below_4g, *ram_above_4g;
830     FWCfgState *fw_cfg;
831     MachineState *machine = MACHINE(pcms);
832     MachineClass *mc = MACHINE_GET_CLASS(machine);
833     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
834     X86MachineState *x86ms = X86_MACHINE(pcms);
835     hwaddr maxphysaddr, maxusedaddr;
836     hwaddr cxl_base, cxl_resv_end = 0;
837     X86CPU *cpu = X86_CPU(first_cpu);
838 
839     assert(machine->ram_size == x86ms->below_4g_mem_size +
840                                 x86ms->above_4g_mem_size);
841 
842     linux_boot = (machine->kernel_filename != NULL);
843 
844     /*
845      * The HyperTransport range close to the 1T boundary is unique to AMD
846      * hosts with IOMMUs enabled. Restrict the ram-above-4g relocation
847      * to above 1T to AMD vCPUs only. @enforce_amd_1tb_hole is only false in
848      * older machine types (<= 7.0) for compatibility purposes.
849      */
850     if (IS_AMD_CPU(&cpu->env) && pcmc->enforce_amd_1tb_hole) {
851         /* Bail out if max possible address does not cross HT range */
852         if (pc_max_used_gpa(pcms, pci_hole64_size) >= AMD_HT_START) {
853             x86ms->above_4g_mem_start = AMD_ABOVE_1TB_START;
854         }
855 
856         /*
857          * Advertise the HT region if address space covers the reserved
858          * region or if we relocate.
859          */
860         if (cpu->phys_bits >= 40) {
861             e820_add_entry(AMD_HT_START, AMD_HT_SIZE, E820_RESERVED);
862         }
863     }
864 
865     /*
866      * phys-bits is required to be appropriately configured
867      * to make sure max used GPA is reachable.
868      */
869     maxusedaddr = pc_max_used_gpa(pcms, pci_hole64_size);
870     maxphysaddr = ((hwaddr)1 << cpu->phys_bits) - 1;
871     if (maxphysaddr < maxusedaddr) {
872         error_report("Address space limit 0x%"PRIx64" < 0x%"PRIx64
873                      " phys-bits too low (%u)",
874                      maxphysaddr, maxusedaddr, cpu->phys_bits);
875         exit(EXIT_FAILURE);
876     }
877 
878     /*
879      * Split single memory region and use aliases to address portions of it,
880      * done for backwards compatibility with older qemus.
881      */
882     ram_below_4g = g_malloc(sizeof(*ram_below_4g));
883     memory_region_init_alias(ram_below_4g, NULL, "ram-below-4g", machine->ram,
884                              0, x86ms->below_4g_mem_size);
885     memory_region_add_subregion(system_memory, 0, ram_below_4g);
886     e820_add_entry(0, x86ms->below_4g_mem_size, E820_RAM);
887     if (x86ms->above_4g_mem_size > 0) {
888         ram_above_4g = g_malloc(sizeof(*ram_above_4g));
889         memory_region_init_alias(ram_above_4g, NULL, "ram-above-4g",
890                                  machine->ram,
891                                  x86ms->below_4g_mem_size,
892                                  x86ms->above_4g_mem_size);
893         memory_region_add_subregion(system_memory, x86ms->above_4g_mem_start,
894                                     ram_above_4g);
895         e820_add_entry(x86ms->above_4g_mem_start, x86ms->above_4g_mem_size,
896                        E820_RAM);
897     }
898 
899     if (pcms->sgx_epc.size != 0) {
900         e820_add_entry(pcms->sgx_epc.base, pcms->sgx_epc.size, E820_RESERVED);
901     }
902 
903     if (!pcmc->has_reserved_memory &&
904         (machine->ram_slots ||
905          (machine->maxram_size > machine->ram_size))) {
906 
907         error_report("\"-memory 'slots|maxmem'\" is not supported by: %s",
908                      mc->name);
909         exit(EXIT_FAILURE);
910     }
911 
912     /* initialize device memory address space */
913     if (pcmc->has_reserved_memory &&
914         (machine->ram_size < machine->maxram_size)) {
915         ram_addr_t device_mem_size;
916         hwaddr device_mem_base;
917 
918         if (machine->ram_slots > ACPI_MAX_RAM_SLOTS) {
919             error_report("unsupported amount of memory slots: %"PRIu64,
920                          machine->ram_slots);
921             exit(EXIT_FAILURE);
922         }
923 
924         if (QEMU_ALIGN_UP(machine->maxram_size,
925                           TARGET_PAGE_SIZE) != machine->maxram_size) {
926             error_report("maximum memory size must by aligned to multiple of "
927                          "%d bytes", TARGET_PAGE_SIZE);
928             exit(EXIT_FAILURE);
929         }
930 
931         pc_get_device_memory_range(pcms, &device_mem_base, &device_mem_size);
932 
933         if (device_mem_base + device_mem_size < device_mem_size) {
934             error_report("unsupported amount of maximum memory: " RAM_ADDR_FMT,
935                          machine->maxram_size);
936             exit(EXIT_FAILURE);
937         }
938         machine_memory_devices_init(machine, device_mem_base, device_mem_size);
939     }
940 
941     if (pcms->cxl_devices_state.is_enabled) {
942         MemoryRegion *mr = &pcms->cxl_devices_state.host_mr;
943         hwaddr cxl_size = MiB;
944 
945         cxl_base = pc_get_cxl_range_start(pcms);
946         memory_region_init(mr, OBJECT(machine), "cxl_host_reg", cxl_size);
947         memory_region_add_subregion(system_memory, cxl_base, mr);
948         cxl_resv_end = cxl_base + cxl_size;
949         if (pcms->cxl_devices_state.fixed_windows) {
950             hwaddr cxl_fmw_base;
951             GList *it;
952 
953             cxl_fmw_base = ROUND_UP(cxl_base + cxl_size, 256 * MiB);
954             for (it = pcms->cxl_devices_state.fixed_windows; it; it = it->next) {
955                 CXLFixedWindow *fw = it->data;
956 
957                 fw->base = cxl_fmw_base;
958                 memory_region_init_io(&fw->mr, OBJECT(machine), &cfmws_ops, fw,
959                                       "cxl-fixed-memory-region", fw->size);
960                 memory_region_add_subregion(system_memory, fw->base, &fw->mr);
961                 cxl_fmw_base += fw->size;
962                 cxl_resv_end = cxl_fmw_base;
963             }
964         }
965     }
966 
967     /* Initialize PC system firmware */
968     pc_system_firmware_init(pcms, rom_memory);
969 
970     option_rom_mr = g_malloc(sizeof(*option_rom_mr));
971     if (machine_require_guest_memfd(machine)) {
972         memory_region_init_ram_guest_memfd(option_rom_mr, NULL, "pc.rom",
973                                            PC_ROM_SIZE, &error_fatal);
974     } else {
975         memory_region_init_ram(option_rom_mr, NULL, "pc.rom", PC_ROM_SIZE,
976                                &error_fatal);
977         if (pcmc->pci_enabled) {
978             memory_region_set_readonly(option_rom_mr, true);
979         }
980     }
981     memory_region_add_subregion_overlap(rom_memory,
982                                         PC_ROM_MIN_VGA,
983                                         option_rom_mr,
984                                         1);
985 
986     fw_cfg = fw_cfg_arch_create(machine,
987                                 x86ms->boot_cpus, x86ms->apic_id_limit);
988 
989     rom_set_fw(fw_cfg);
990 
991     if (machine->device_memory) {
992         uint64_t *val = g_malloc(sizeof(*val));
993         uint64_t res_mem_end = machine->device_memory->base;
994 
995         if (!pcmc->broken_reserved_end) {
996             res_mem_end += memory_region_size(&machine->device_memory->mr);
997         }
998 
999         if (pcms->cxl_devices_state.is_enabled) {
1000             res_mem_end = cxl_resv_end;
1001         }
1002         *val = cpu_to_le64(ROUND_UP(res_mem_end, 1 * GiB));
1003         fw_cfg_add_file(fw_cfg, "etc/reserved-memory-end", val, sizeof(*val));
1004     }
1005 
1006     if (linux_boot) {
1007         x86_load_linux(x86ms, fw_cfg, PC_FW_DATA, pcmc->pvh_enabled);
1008     }
1009 
1010     for (i = 0; i < nb_option_roms; i++) {
1011         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1012     }
1013     x86ms->fw_cfg = fw_cfg;
1014 
1015     /* Init default IOAPIC address space */
1016     x86ms->ioapic_as = &address_space_memory;
1017 
1018     /* Init ACPI memory hotplug IO base address */
1019     pcms->memhp_io_base = ACPI_MEMORY_HOTPLUG_BASE;
1020 }
1021 
1022 /*
1023  * The 64bit pci hole starts after "above 4G RAM" and
1024  * potentially the space reserved for memory hotplug.
1025  */
1026 uint64_t pc_pci_hole64_start(void)
1027 {
1028     PCMachineState *pcms = PC_MACHINE(qdev_get_machine());
1029     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1030     MachineState *ms = MACHINE(pcms);
1031     uint64_t hole64_start = 0;
1032     ram_addr_t size = 0;
1033 
1034     if (pcms->cxl_devices_state.is_enabled) {
1035         hole64_start = pc_get_cxl_range_end(pcms);
1036     } else if (pcmc->has_reserved_memory && (ms->ram_size < ms->maxram_size)) {
1037         pc_get_device_memory_range(pcms, &hole64_start, &size);
1038         if (!pcmc->broken_reserved_end) {
1039             hole64_start += size;
1040         }
1041     } else {
1042         hole64_start = pc_above_4g_end(pcms);
1043     }
1044 
1045     return ROUND_UP(hole64_start, 1 * GiB);
1046 }
1047 
1048 DeviceState *pc_vga_init(ISABus *isa_bus, PCIBus *pci_bus)
1049 {
1050     DeviceState *dev = NULL;
1051 
1052     rom_set_order_override(FW_CFG_ORDER_OVERRIDE_VGA);
1053     if (pci_bus) {
1054         PCIDevice *pcidev = pci_vga_init(pci_bus);
1055         dev = pcidev ? &pcidev->qdev : NULL;
1056     } else if (isa_bus) {
1057         ISADevice *isadev = isa_vga_init(isa_bus);
1058         dev = isadev ? DEVICE(isadev) : NULL;
1059     }
1060     rom_reset_order_override();
1061     return dev;
1062 }
1063 
1064 static const MemoryRegionOps ioport80_io_ops = {
1065     .write = ioport80_write,
1066     .read = ioport80_read,
1067     .endianness = DEVICE_NATIVE_ENDIAN,
1068     .impl = {
1069         .min_access_size = 1,
1070         .max_access_size = 1,
1071     },
1072 };
1073 
1074 static const MemoryRegionOps ioportF0_io_ops = {
1075     .write = ioportF0_write,
1076     .read = ioportF0_read,
1077     .endianness = DEVICE_NATIVE_ENDIAN,
1078     .impl = {
1079         .min_access_size = 1,
1080         .max_access_size = 1,
1081     },
1082 };
1083 
1084 static void pc_superio_init(ISABus *isa_bus, bool create_fdctrl,
1085                             bool create_i8042, bool no_vmport, Error **errp)
1086 {
1087     int i;
1088     DriveInfo *fd[MAX_FD];
1089     qemu_irq *a20_line;
1090     ISADevice *i8042, *port92, *vmmouse;
1091 
1092     serial_hds_isa_init(isa_bus, 0, MAX_ISA_SERIAL_PORTS);
1093     parallel_hds_isa_init(isa_bus, MAX_PARALLEL_PORTS);
1094 
1095     for (i = 0; i < MAX_FD; i++) {
1096         fd[i] = drive_get(IF_FLOPPY, 0, i);
1097         create_fdctrl |= !!fd[i];
1098     }
1099     if (create_fdctrl) {
1100 #ifdef CONFIG_FDC_ISA
1101         ISADevice *fdc = isa_new(TYPE_ISA_FDC);
1102         if (fdc) {
1103             isa_realize_and_unref(fdc, isa_bus, &error_fatal);
1104             isa_fdc_init_drives(fdc, fd);
1105         }
1106 #endif
1107     }
1108 
1109     if (!create_i8042) {
1110         if (!no_vmport) {
1111             error_setg(errp,
1112                        "vmport requires the i8042 controller to be enabled");
1113         }
1114         return;
1115     }
1116 
1117     i8042 = isa_create_simple(isa_bus, TYPE_I8042);
1118     if (!no_vmport) {
1119         isa_create_simple(isa_bus, TYPE_VMPORT);
1120         vmmouse = isa_try_new("vmmouse");
1121     } else {
1122         vmmouse = NULL;
1123     }
1124     if (vmmouse) {
1125         object_property_set_link(OBJECT(vmmouse), TYPE_I8042, OBJECT(i8042),
1126                                  &error_abort);
1127         isa_realize_and_unref(vmmouse, isa_bus, &error_fatal);
1128     }
1129     port92 = isa_create_simple(isa_bus, TYPE_PORT92);
1130 
1131     a20_line = qemu_allocate_irqs(handle_a20_line_change, first_cpu, 2);
1132     qdev_connect_gpio_out_named(DEVICE(i8042),
1133                                 I8042_A20_LINE, 0, a20_line[0]);
1134     qdev_connect_gpio_out_named(DEVICE(port92),
1135                                 PORT92_A20_LINE, 0, a20_line[1]);
1136     g_free(a20_line);
1137 }
1138 
1139 void pc_basic_device_init(struct PCMachineState *pcms,
1140                           ISABus *isa_bus, qemu_irq *gsi,
1141                           ISADevice *rtc_state,
1142                           bool create_fdctrl,
1143                           uint32_t hpet_irqs)
1144 {
1145     int i;
1146     DeviceState *hpet = NULL;
1147     int pit_isa_irq = 0;
1148     qemu_irq pit_alt_irq = NULL;
1149     ISADevice *pit = NULL;
1150     MemoryRegion *ioport80_io = g_new(MemoryRegion, 1);
1151     MemoryRegion *ioportF0_io = g_new(MemoryRegion, 1);
1152     X86MachineState *x86ms = X86_MACHINE(pcms);
1153 
1154     memory_region_init_io(ioport80_io, NULL, &ioport80_io_ops, NULL, "ioport80", 1);
1155     memory_region_add_subregion(isa_bus->address_space_io, 0x80, ioport80_io);
1156 
1157     memory_region_init_io(ioportF0_io, NULL, &ioportF0_io_ops, NULL, "ioportF0", 1);
1158     memory_region_add_subregion(isa_bus->address_space_io, 0xf0, ioportF0_io);
1159 
1160     /*
1161      * Check if an HPET shall be created.
1162      */
1163     if (pcms->hpet_enabled) {
1164         qemu_irq rtc_irq;
1165 
1166         hpet = qdev_try_new(TYPE_HPET);
1167         if (!hpet) {
1168             error_report("couldn't create HPET device");
1169             exit(1);
1170         }
1171         /*
1172          * For pc-piix-*, hpet's intcap is always IRQ2. For pc-q35-*,
1173          * use IRQ16~23, IRQ8 and IRQ2.  If the user has already set
1174          * the property, use whatever mask they specified.
1175          */
1176         uint8_t compat = object_property_get_uint(OBJECT(hpet),
1177                 HPET_INTCAP, NULL);
1178         if (!compat) {
1179             qdev_prop_set_uint32(hpet, HPET_INTCAP, hpet_irqs);
1180         }
1181         sysbus_realize_and_unref(SYS_BUS_DEVICE(hpet), &error_fatal);
1182         sysbus_mmio_map(SYS_BUS_DEVICE(hpet), 0, HPET_BASE);
1183 
1184         for (i = 0; i < IOAPIC_NUM_PINS; i++) {
1185             sysbus_connect_irq(SYS_BUS_DEVICE(hpet), i, gsi[i]);
1186         }
1187         pit_isa_irq = -1;
1188         pit_alt_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_PIT_INT);
1189         rtc_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_RTC_INT);
1190 
1191         /* overwrite connection created by south bridge */
1192         qdev_connect_gpio_out(DEVICE(rtc_state), 0, rtc_irq);
1193     }
1194 
1195     object_property_add_alias(OBJECT(pcms), "rtc-time", OBJECT(rtc_state),
1196                               "date");
1197 
1198 #ifdef CONFIG_XEN_EMU
1199     if (xen_mode == XEN_EMULATE) {
1200         xen_overlay_create();
1201         xen_evtchn_create(IOAPIC_NUM_PINS, gsi);
1202         xen_gnttab_create();
1203         xen_xenstore_create();
1204         if (pcms->pcibus) {
1205             pci_create_simple(pcms->pcibus, -1, "xen-platform");
1206         }
1207         xen_bus_init();
1208     }
1209 #endif
1210 
1211     qemu_register_boot_set(pc_boot_set, pcms);
1212     set_boot_dev(pcms, MC146818_RTC(rtc_state),
1213                  MACHINE(pcms)->boot_config.order, &error_fatal);
1214 
1215     if (!xen_enabled() &&
1216         (x86ms->pit == ON_OFF_AUTO_AUTO || x86ms->pit == ON_OFF_AUTO_ON)) {
1217         if (kvm_pit_in_kernel()) {
1218             pit = kvm_pit_init(isa_bus, 0x40);
1219         } else {
1220             pit = i8254_pit_init(isa_bus, 0x40, pit_isa_irq, pit_alt_irq);
1221         }
1222         if (hpet) {
1223             /* connect PIT to output control line of the HPET */
1224             qdev_connect_gpio_out(hpet, 0, qdev_get_gpio_in(DEVICE(pit), 0));
1225         }
1226         object_property_set_link(OBJECT(pcms->pcspk), "pit",
1227                                  OBJECT(pit), &error_fatal);
1228         isa_realize_and_unref(pcms->pcspk, isa_bus, &error_fatal);
1229     }
1230 
1231     assert(pcms->vmport >= 0 && pcms->vmport < ON_OFF_AUTO__MAX);
1232     if (pcms->vmport == ON_OFF_AUTO_AUTO) {
1233         pcms->vmport = (xen_enabled() || !pcms->i8042_enabled)
1234             ? ON_OFF_AUTO_OFF : ON_OFF_AUTO_ON;
1235     }
1236 
1237     /* Super I/O */
1238     pc_superio_init(isa_bus, create_fdctrl, pcms->i8042_enabled,
1239                     pcms->vmport != ON_OFF_AUTO_ON, &error_fatal);
1240 }
1241 
1242 void pc_nic_init(PCMachineClass *pcmc, ISABus *isa_bus, PCIBus *pci_bus)
1243 {
1244     MachineClass *mc = MACHINE_CLASS(pcmc);
1245     bool default_is_ne2k = g_str_equal(mc->default_nic, TYPE_ISA_NE2000);
1246     NICInfo *nd;
1247 
1248     rom_set_order_override(FW_CFG_ORDER_OVERRIDE_NIC);
1249 
1250     while ((nd = qemu_find_nic_info(TYPE_ISA_NE2000, default_is_ne2k, NULL))) {
1251         pc_init_ne2k_isa(isa_bus, nd, &error_fatal);
1252     }
1253 
1254     /* Anything remaining should be a PCI NIC */
1255     if (pci_bus) {
1256         pci_init_nic_devices(pci_bus, mc->default_nic);
1257     }
1258 
1259     rom_reset_order_override();
1260 }
1261 
1262 void pc_i8259_create(ISABus *isa_bus, qemu_irq *i8259_irqs)
1263 {
1264     qemu_irq *i8259;
1265 
1266     if (kvm_pic_in_kernel()) {
1267         i8259 = kvm_i8259_init(isa_bus);
1268     } else if (xen_enabled()) {
1269         i8259 = xen_interrupt_controller_init();
1270     } else {
1271         i8259 = i8259_init(isa_bus, x86_allocate_cpu_irq());
1272     }
1273 
1274     for (size_t i = 0; i < ISA_NUM_IRQS; i++) {
1275         i8259_irqs[i] = i8259[i];
1276     }
1277 
1278     g_free(i8259);
1279 }
1280 
1281 static void pc_memory_pre_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
1282                                Error **errp)
1283 {
1284     const X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
1285     const MachineState *ms = MACHINE(hotplug_dev);
1286     const bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
1287     Error *local_err = NULL;
1288 
1289     /*
1290      * When "acpi=off" is used with the Q35 machine type, no ACPI is built,
1291      * but pcms->acpi_dev is still created. Check !acpi_enabled in
1292      * addition to cover this case.
1293      */
1294     if (!x86ms->acpi_dev || !x86_machine_is_acpi_enabled(x86ms)) {
1295         error_setg(errp,
1296                    "memory hotplug is not enabled: missing acpi device or acpi disabled");
1297         return;
1298     }
1299 
1300     if (is_nvdimm && !ms->nvdimms_state->is_enabled) {
1301         error_setg(errp, "nvdimm is not enabled: missing 'nvdimm' in '-M'");
1302         return;
1303     }
1304 
1305     hotplug_handler_pre_plug(x86ms->acpi_dev, dev, &local_err);
1306     if (local_err) {
1307         error_propagate(errp, local_err);
1308         return;
1309     }
1310 
1311     pc_dimm_pre_plug(PC_DIMM(dev), MACHINE(hotplug_dev), errp);
1312 }
1313 
1314 static void pc_memory_plug(HotplugHandler *hotplug_dev,
1315                            DeviceState *dev, Error **errp)
1316 {
1317     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1318     X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
1319     MachineState *ms = MACHINE(hotplug_dev);
1320     bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
1321 
1322     pc_dimm_plug(PC_DIMM(dev), MACHINE(pcms));
1323 
1324     if (is_nvdimm) {
1325         nvdimm_plug(ms->nvdimms_state);
1326     }
1327 
1328     hotplug_handler_plug(x86ms->acpi_dev, dev, &error_abort);
1329 }
1330 
1331 static void pc_memory_unplug_request(HotplugHandler *hotplug_dev,
1332                                      DeviceState *dev, Error **errp)
1333 {
1334     X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
1335 
1336     /*
1337      * When "acpi=off" is used with the Q35 machine type, no ACPI is built,
1338      * but pcms->acpi_dev is still created. Check !acpi_enabled in
1339      * addition to cover this case.
1340      */
1341     if (!x86ms->acpi_dev || !x86_machine_is_acpi_enabled(x86ms)) {
1342         error_setg(errp,
1343                    "memory hotplug is not enabled: missing acpi device or acpi disabled");
1344         return;
1345     }
1346 
1347     if (object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM)) {
1348         error_setg(errp, "nvdimm device hot unplug is not supported yet.");
1349         return;
1350     }
1351 
1352     hotplug_handler_unplug_request(x86ms->acpi_dev, dev,
1353                                    errp);
1354 }
1355 
1356 static void pc_memory_unplug(HotplugHandler *hotplug_dev,
1357                              DeviceState *dev, Error **errp)
1358 {
1359     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1360     X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
1361     Error *local_err = NULL;
1362 
1363     hotplug_handler_unplug(x86ms->acpi_dev, dev, &local_err);
1364     if (local_err) {
1365         goto out;
1366     }
1367 
1368     pc_dimm_unplug(PC_DIMM(dev), MACHINE(pcms));
1369     qdev_unrealize(dev);
1370  out:
1371     error_propagate(errp, local_err);
1372 }
1373 
1374 static void pc_hv_balloon_pre_plug(HotplugHandler *hotplug_dev,
1375                                    DeviceState *dev, Error **errp)
1376 {
1377     /* The vmbus handler has no hotplug handler; we should never end up here. */
1378     g_assert(!dev->hotplugged);
1379     memory_device_pre_plug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev), errp);
1380 }
1381 
1382 static void pc_hv_balloon_plug(HotplugHandler *hotplug_dev,
1383                                DeviceState *dev, Error **errp)
1384 {
1385     memory_device_plug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev));
1386 }
1387 
1388 static void pc_machine_device_pre_plug_cb(HotplugHandler *hotplug_dev,
1389                                           DeviceState *dev, Error **errp)
1390 {
1391     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1392         pc_memory_pre_plug(hotplug_dev, dev, errp);
1393     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1394         x86_cpu_pre_plug(hotplug_dev, dev, errp);
1395     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI)) {
1396         virtio_md_pci_pre_plug(VIRTIO_MD_PCI(dev), MACHINE(hotplug_dev), errp);
1397     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
1398         /* Declare the APIC range as the reserved MSI region */
1399         char *resv_prop_str = g_strdup_printf("0xfee00000:0xfeefffff:%d",
1400                                               VIRTIO_IOMMU_RESV_MEM_T_MSI);
1401         QList *reserved_regions = qlist_new();
1402 
1403         qlist_append_str(reserved_regions, resv_prop_str);
1404         qdev_prop_set_array(dev, "reserved-regions", reserved_regions);
1405 
1406         g_free(resv_prop_str);
1407     }
1408 
1409     if (object_dynamic_cast(OBJECT(dev), TYPE_X86_IOMMU_DEVICE) ||
1410         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
1411         PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1412 
1413         if (pcms->iommu) {
1414             error_setg(errp, "QEMU does not support multiple vIOMMUs "
1415                        "for x86 yet.");
1416             return;
1417         }
1418         pcms->iommu = dev;
1419     } else if (object_dynamic_cast(OBJECT(dev), TYPE_HV_BALLOON)) {
1420         pc_hv_balloon_pre_plug(hotplug_dev, dev, errp);
1421     }
1422 }
1423 
1424 static void pc_machine_device_plug_cb(HotplugHandler *hotplug_dev,
1425                                       DeviceState *dev, Error **errp)
1426 {
1427     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1428         pc_memory_plug(hotplug_dev, dev, errp);
1429     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1430         x86_cpu_plug(hotplug_dev, dev, errp);
1431     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI)) {
1432         virtio_md_pci_plug(VIRTIO_MD_PCI(dev), MACHINE(hotplug_dev), errp);
1433     } else if (object_dynamic_cast(OBJECT(dev), TYPE_HV_BALLOON)) {
1434         pc_hv_balloon_plug(hotplug_dev, dev, errp);
1435     }
1436 }
1437 
1438 static void pc_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
1439                                                 DeviceState *dev, Error **errp)
1440 {
1441     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1442         pc_memory_unplug_request(hotplug_dev, dev, errp);
1443     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1444         x86_cpu_unplug_request_cb(hotplug_dev, dev, errp);
1445     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI)) {
1446         virtio_md_pci_unplug_request(VIRTIO_MD_PCI(dev), MACHINE(hotplug_dev),
1447                                      errp);
1448     } else {
1449         error_setg(errp, "acpi: device unplug request for not supported device"
1450                    " type: %s", object_get_typename(OBJECT(dev)));
1451     }
1452 }
1453 
1454 static void pc_machine_device_unplug_cb(HotplugHandler *hotplug_dev,
1455                                         DeviceState *dev, Error **errp)
1456 {
1457     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1458         pc_memory_unplug(hotplug_dev, dev, errp);
1459     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1460         x86_cpu_unplug_cb(hotplug_dev, dev, errp);
1461     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI)) {
1462         virtio_md_pci_unplug(VIRTIO_MD_PCI(dev), MACHINE(hotplug_dev), errp);
1463     } else {
1464         error_setg(errp, "acpi: device unplug for not supported device"
1465                    " type: %s", object_get_typename(OBJECT(dev)));
1466     }
1467 }
1468 
1469 static HotplugHandler *pc_get_hotplug_handler(MachineState *machine,
1470                                              DeviceState *dev)
1471 {
1472     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) ||
1473         object_dynamic_cast(OBJECT(dev), TYPE_CPU) ||
1474         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI) ||
1475         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI) ||
1476         object_dynamic_cast(OBJECT(dev), TYPE_HV_BALLOON) ||
1477         object_dynamic_cast(OBJECT(dev), TYPE_X86_IOMMU_DEVICE)) {
1478         return HOTPLUG_HANDLER(machine);
1479     }
1480 
1481     return NULL;
1482 }
1483 
1484 static void pc_machine_get_vmport(Object *obj, Visitor *v, const char *name,
1485                                   void *opaque, Error **errp)
1486 {
1487     PCMachineState *pcms = PC_MACHINE(obj);
1488     OnOffAuto vmport = pcms->vmport;
1489 
1490     visit_type_OnOffAuto(v, name, &vmport, errp);
1491 }
1492 
1493 static void pc_machine_set_vmport(Object *obj, Visitor *v, const char *name,
1494                                   void *opaque, Error **errp)
1495 {
1496     PCMachineState *pcms = PC_MACHINE(obj);
1497 
1498     visit_type_OnOffAuto(v, name, &pcms->vmport, errp);
1499 }
1500 
1501 static bool pc_machine_get_fd_bootchk(Object *obj, Error **errp)
1502 {
1503     PCMachineState *pcms = PC_MACHINE(obj);
1504 
1505     return pcms->fd_bootchk;
1506 }
1507 
1508 static void pc_machine_set_fd_bootchk(Object *obj, bool value, Error **errp)
1509 {
1510     PCMachineState *pcms = PC_MACHINE(obj);
1511 
1512     pcms->fd_bootchk = value;
1513 }
1514 
1515 static bool pc_machine_get_smbus(Object *obj, Error **errp)
1516 {
1517     PCMachineState *pcms = PC_MACHINE(obj);
1518 
1519     return pcms->smbus_enabled;
1520 }
1521 
1522 static void pc_machine_set_smbus(Object *obj, bool value, Error **errp)
1523 {
1524     PCMachineState *pcms = PC_MACHINE(obj);
1525 
1526     pcms->smbus_enabled = value;
1527 }
1528 
1529 static bool pc_machine_get_sata(Object *obj, Error **errp)
1530 {
1531     PCMachineState *pcms = PC_MACHINE(obj);
1532 
1533     return pcms->sata_enabled;
1534 }
1535 
1536 static void pc_machine_set_sata(Object *obj, bool value, Error **errp)
1537 {
1538     PCMachineState *pcms = PC_MACHINE(obj);
1539 
1540     pcms->sata_enabled = value;
1541 }
1542 
1543 static bool pc_machine_get_hpet(Object *obj, Error **errp)
1544 {
1545     PCMachineState *pcms = PC_MACHINE(obj);
1546 
1547     return pcms->hpet_enabled;
1548 }
1549 
1550 static void pc_machine_set_hpet(Object *obj, bool value, Error **errp)
1551 {
1552     PCMachineState *pcms = PC_MACHINE(obj);
1553 
1554     pcms->hpet_enabled = value;
1555 }
1556 
1557 static bool pc_machine_get_i8042(Object *obj, Error **errp)
1558 {
1559     PCMachineState *pcms = PC_MACHINE(obj);
1560 
1561     return pcms->i8042_enabled;
1562 }
1563 
1564 static void pc_machine_set_i8042(Object *obj, bool value, Error **errp)
1565 {
1566     PCMachineState *pcms = PC_MACHINE(obj);
1567 
1568     pcms->i8042_enabled = value;
1569 }
1570 
1571 static bool pc_machine_get_default_bus_bypass_iommu(Object *obj, Error **errp)
1572 {
1573     PCMachineState *pcms = PC_MACHINE(obj);
1574 
1575     return pcms->default_bus_bypass_iommu;
1576 }
1577 
1578 static void pc_machine_set_default_bus_bypass_iommu(Object *obj, bool value,
1579                                                     Error **errp)
1580 {
1581     PCMachineState *pcms = PC_MACHINE(obj);
1582 
1583     pcms->default_bus_bypass_iommu = value;
1584 }
1585 
1586 static void pc_machine_get_smbios_ep(Object *obj, Visitor *v, const char *name,
1587                                      void *opaque, Error **errp)
1588 {
1589     PCMachineState *pcms = PC_MACHINE(obj);
1590     SmbiosEntryPointType smbios_entry_point_type = pcms->smbios_entry_point_type;
1591 
1592     visit_type_SmbiosEntryPointType(v, name, &smbios_entry_point_type, errp);
1593 }
1594 
1595 static void pc_machine_set_smbios_ep(Object *obj, Visitor *v, const char *name,
1596                                      void *opaque, Error **errp)
1597 {
1598     PCMachineState *pcms = PC_MACHINE(obj);
1599 
1600     visit_type_SmbiosEntryPointType(v, name, &pcms->smbios_entry_point_type, errp);
1601 }
1602 
1603 static void pc_machine_get_max_ram_below_4g(Object *obj, Visitor *v,
1604                                             const char *name, void *opaque,
1605                                             Error **errp)
1606 {
1607     PCMachineState *pcms = PC_MACHINE(obj);
1608     uint64_t value = pcms->max_ram_below_4g;
1609 
1610     visit_type_size(v, name, &value, errp);
1611 }
1612 
1613 static void pc_machine_set_max_ram_below_4g(Object *obj, Visitor *v,
1614                                             const char *name, void *opaque,
1615                                             Error **errp)
1616 {
1617     PCMachineState *pcms = PC_MACHINE(obj);
1618     uint64_t value;
1619 
1620     if (!visit_type_size(v, name, &value, errp)) {
1621         return;
1622     }
1623     if (value > 4 * GiB) {
1624         error_setg(errp,
1625                    "Machine option 'max-ram-below-4g=%"PRIu64
1626                    "' expects size less than or equal to 4G", value);
1627         return;
1628     }
1629 
1630     if (value < 1 * MiB) {
1631         warn_report("Only %" PRIu64 " bytes of RAM below the 4GiB boundary,"
1632                     "BIOS may not work with less than 1MiB", value);
1633     }
1634 
1635     pcms->max_ram_below_4g = value;
1636 }
1637 
1638 static void pc_machine_get_max_fw_size(Object *obj, Visitor *v,
1639                                        const char *name, void *opaque,
1640                                        Error **errp)
1641 {
1642     PCMachineState *pcms = PC_MACHINE(obj);
1643     uint64_t value = pcms->max_fw_size;
1644 
1645     visit_type_size(v, name, &value, errp);
1646 }
1647 
1648 static void pc_machine_set_max_fw_size(Object *obj, Visitor *v,
1649                                        const char *name, void *opaque,
1650                                        Error **errp)
1651 {
1652     PCMachineState *pcms = PC_MACHINE(obj);
1653     uint64_t value;
1654 
1655     if (!visit_type_size(v, name, &value, errp)) {
1656         return;
1657     }
1658 
1659     /*
1660      * We don't have a theoretically justifiable exact lower bound on the base
1661      * address of any flash mapping. In practice, the IO-APIC MMIO range is
1662      * [0xFEE00000..0xFEE01000] -- see IO_APIC_DEFAULT_ADDRESS --, leaving free
1663      * only 18MiB-4KiB below 4GiB. For now, restrict the cumulative mapping to
1664      * 16MiB in size.
1665      */
1666     if (value > 16 * MiB) {
1667         error_setg(errp,
1668                    "User specified max allowed firmware size %" PRIu64 " is "
1669                    "greater than 16MiB. If combined firmware size exceeds "
1670                    "16MiB the system may not boot, or experience intermittent"
1671                    "stability issues.",
1672                    value);
1673         return;
1674     }
1675 
1676     pcms->max_fw_size = value;
1677 }
1678 
1679 
1680 static void pc_machine_initfn(Object *obj)
1681 {
1682     PCMachineState *pcms = PC_MACHINE(obj);
1683     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1684 
1685 #ifdef CONFIG_VMPORT
1686     pcms->vmport = ON_OFF_AUTO_AUTO;
1687 #else
1688     pcms->vmport = ON_OFF_AUTO_OFF;
1689 #endif /* CONFIG_VMPORT */
1690     pcms->max_ram_below_4g = 0; /* use default */
1691     pcms->smbios_entry_point_type = pcmc->default_smbios_ep_type;
1692     pcms->south_bridge = pcmc->default_south_bridge;
1693 
1694     /* acpi build is enabled by default if machine supports it */
1695     pcms->acpi_build_enabled = pcmc->has_acpi_build;
1696     pcms->smbus_enabled = true;
1697     pcms->sata_enabled = true;
1698     pcms->i8042_enabled = true;
1699     pcms->max_fw_size = 8 * MiB;
1700 #ifdef CONFIG_HPET
1701     pcms->hpet_enabled = true;
1702 #endif
1703     pcms->fd_bootchk = true;
1704     pcms->default_bus_bypass_iommu = false;
1705 
1706     pc_system_flash_create(pcms);
1707     pcms->pcspk = isa_new(TYPE_PC_SPEAKER);
1708     object_property_add_alias(OBJECT(pcms), "pcspk-audiodev",
1709                               OBJECT(pcms->pcspk), "audiodev");
1710     if (pcmc->pci_enabled) {
1711         cxl_machine_init(obj, &pcms->cxl_devices_state);
1712     }
1713 
1714     pcms->machine_done.notify = pc_machine_done;
1715     qemu_add_machine_init_done_notifier(&pcms->machine_done);
1716 }
1717 
1718 static void pc_machine_reset(MachineState *machine, ResetType type)
1719 {
1720     CPUState *cs;
1721     X86CPU *cpu;
1722 
1723     qemu_devices_reset(type);
1724 
1725     /* Reset APIC after devices have been reset to cancel
1726      * any changes that qemu_devices_reset() might have done.
1727      */
1728     CPU_FOREACH(cs) {
1729         cpu = X86_CPU(cs);
1730 
1731         x86_cpu_after_reset(cpu);
1732     }
1733 }
1734 
1735 static void pc_machine_wakeup(MachineState *machine)
1736 {
1737     cpu_synchronize_all_states();
1738     pc_machine_reset(machine, RESET_TYPE_WAKEUP);
1739     cpu_synchronize_all_post_reset();
1740 }
1741 
1742 static bool pc_hotplug_allowed(MachineState *ms, DeviceState *dev, Error **errp)
1743 {
1744     X86IOMMUState *iommu = x86_iommu_get_default();
1745     IntelIOMMUState *intel_iommu;
1746 
1747     if (iommu &&
1748         object_dynamic_cast((Object *)iommu, TYPE_INTEL_IOMMU_DEVICE) &&
1749         object_dynamic_cast((Object *)dev, "vfio-pci")) {
1750         intel_iommu = INTEL_IOMMU_DEVICE(iommu);
1751         if (!intel_iommu->caching_mode) {
1752             error_setg(errp, "Device assignment is not allowed without "
1753                        "enabling caching-mode=on for Intel IOMMU.");
1754             return false;
1755         }
1756     }
1757 
1758     return true;
1759 }
1760 
1761 static void pc_machine_class_init(ObjectClass *oc, void *data)
1762 {
1763     MachineClass *mc = MACHINE_CLASS(oc);
1764     X86MachineClass *x86mc = X86_MACHINE_CLASS(oc);
1765     PCMachineClass *pcmc = PC_MACHINE_CLASS(oc);
1766     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
1767 
1768     pcmc->pci_enabled = true;
1769     pcmc->has_acpi_build = true;
1770     pcmc->smbios_defaults = true;
1771     pcmc->gigabyte_align = true;
1772     pcmc->has_reserved_memory = true;
1773     pcmc->enforce_amd_1tb_hole = true;
1774     pcmc->isa_bios_alias = true;
1775     pcmc->pvh_enabled = true;
1776     pcmc->kvmclock_create_always = true;
1777     x86mc->apic_xrupt_override = true;
1778     assert(!mc->get_hotplug_handler);
1779     mc->get_hotplug_handler = pc_get_hotplug_handler;
1780     mc->hotplug_allowed = pc_hotplug_allowed;
1781     mc->auto_enable_numa_with_memhp = true;
1782     mc->auto_enable_numa_with_memdev = true;
1783     mc->has_hotpluggable_cpus = true;
1784     mc->default_boot_order = "cad";
1785     mc->block_default_type = IF_IDE;
1786     mc->max_cpus = 255;
1787     mc->reset = pc_machine_reset;
1788     mc->wakeup = pc_machine_wakeup;
1789     hc->pre_plug = pc_machine_device_pre_plug_cb;
1790     hc->plug = pc_machine_device_plug_cb;
1791     hc->unplug_request = pc_machine_device_unplug_request_cb;
1792     hc->unplug = pc_machine_device_unplug_cb;
1793     mc->default_cpu_type = TARGET_DEFAULT_CPU_TYPE;
1794     mc->nvdimm_supported = true;
1795     mc->smp_props.dies_supported = true;
1796     mc->smp_props.modules_supported = true;
1797     mc->default_ram_id = "pc.ram";
1798     pcmc->default_smbios_ep_type = SMBIOS_ENTRY_POINT_TYPE_AUTO;
1799 
1800     object_class_property_add(oc, PC_MACHINE_MAX_RAM_BELOW_4G, "size",
1801         pc_machine_get_max_ram_below_4g, pc_machine_set_max_ram_below_4g,
1802         NULL, NULL);
1803     object_class_property_set_description(oc, PC_MACHINE_MAX_RAM_BELOW_4G,
1804         "Maximum ram below the 4G boundary (32bit boundary)");
1805 
1806     object_class_property_add(oc, PC_MACHINE_VMPORT, "OnOffAuto",
1807         pc_machine_get_vmport, pc_machine_set_vmport,
1808         NULL, NULL);
1809     object_class_property_set_description(oc, PC_MACHINE_VMPORT,
1810         "Enable vmport (pc & q35)");
1811 
1812     object_class_property_add_bool(oc, PC_MACHINE_SMBUS,
1813         pc_machine_get_smbus, pc_machine_set_smbus);
1814     object_class_property_set_description(oc, PC_MACHINE_SMBUS,
1815         "Enable/disable system management bus");
1816 
1817     object_class_property_add_bool(oc, PC_MACHINE_SATA,
1818         pc_machine_get_sata, pc_machine_set_sata);
1819     object_class_property_set_description(oc, PC_MACHINE_SATA,
1820         "Enable/disable Serial ATA bus");
1821 
1822     object_class_property_add_bool(oc, "hpet",
1823         pc_machine_get_hpet, pc_machine_set_hpet);
1824     object_class_property_set_description(oc, "hpet",
1825         "Enable/disable high precision event timer emulation");
1826 
1827     object_class_property_add_bool(oc, PC_MACHINE_I8042,
1828         pc_machine_get_i8042, pc_machine_set_i8042);
1829     object_class_property_set_description(oc, PC_MACHINE_I8042,
1830         "Enable/disable Intel 8042 PS/2 controller emulation");
1831 
1832     object_class_property_add_bool(oc, "default-bus-bypass-iommu",
1833         pc_machine_get_default_bus_bypass_iommu,
1834         pc_machine_set_default_bus_bypass_iommu);
1835 
1836     object_class_property_add(oc, PC_MACHINE_MAX_FW_SIZE, "size",
1837         pc_machine_get_max_fw_size, pc_machine_set_max_fw_size,
1838         NULL, NULL);
1839     object_class_property_set_description(oc, PC_MACHINE_MAX_FW_SIZE,
1840         "Maximum combined firmware size");
1841 
1842     object_class_property_add(oc, PC_MACHINE_SMBIOS_EP, "str",
1843         pc_machine_get_smbios_ep, pc_machine_set_smbios_ep,
1844         NULL, NULL);
1845     object_class_property_set_description(oc, PC_MACHINE_SMBIOS_EP,
1846         "SMBIOS Entry Point type [32, 64]");
1847 
1848     object_class_property_add_bool(oc, "fd-bootchk",
1849         pc_machine_get_fd_bootchk,
1850         pc_machine_set_fd_bootchk);
1851 }
1852 
1853 static const TypeInfo pc_machine_info = {
1854     .name = TYPE_PC_MACHINE,
1855     .parent = TYPE_X86_MACHINE,
1856     .abstract = true,
1857     .instance_size = sizeof(PCMachineState),
1858     .instance_init = pc_machine_initfn,
1859     .class_size = sizeof(PCMachineClass),
1860     .class_init = pc_machine_class_init,
1861     .interfaces = (InterfaceInfo[]) {
1862          { TYPE_HOTPLUG_HANDLER },
1863          { }
1864     },
1865 };
1866 
1867 static void pc_machine_register_types(void)
1868 {
1869     type_register_static(&pc_machine_info);
1870 }
1871 
1872 type_init(pc_machine_register_types)
1873