xref: /qemu/hw/i386/pc.c (revision 29b62a10)
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/x86.h"
28 #include "hw/i386/pc.h"
29 #include "hw/char/serial.h"
30 #include "hw/char/parallel.h"
31 #include "hw/i386/apic.h"
32 #include "hw/i386/topology.h"
33 #include "hw/i386/fw_cfg.h"
34 #include "hw/i386/vmport.h"
35 #include "sysemu/cpus.h"
36 #include "hw/block/fdc.h"
37 #include "hw/ide.h"
38 #include "hw/pci/pci.h"
39 #include "hw/pci/pci_bus.h"
40 #include "hw/pci-bridge/pci_expander_bridge.h"
41 #include "hw/nvram/fw_cfg.h"
42 #include "hw/timer/hpet.h"
43 #include "hw/firmware/smbios.h"
44 #include "hw/loader.h"
45 #include "elf.h"
46 #include "migration/vmstate.h"
47 #include "multiboot.h"
48 #include "hw/rtc/mc146818rtc.h"
49 #include "hw/intc/i8259.h"
50 #include "hw/timer/i8254.h"
51 #include "hw/input/i8042.h"
52 #include "hw/irq.h"
53 #include "hw/audio/pcspk.h"
54 #include "hw/pci/msi.h"
55 #include "hw/sysbus.h"
56 #include "sysemu/sysemu.h"
57 #include "sysemu/tcg.h"
58 #include "sysemu/numa.h"
59 #include "sysemu/kvm.h"
60 #include "sysemu/xen.h"
61 #include "sysemu/reset.h"
62 #include "sysemu/runstate.h"
63 #include "kvm/kvm_i386.h"
64 #include "hw/xen/xen.h"
65 #include "hw/xen/start_info.h"
66 #include "ui/qemu-spice.h"
67 #include "exec/memory.h"
68 #include "qemu/bitmap.h"
69 #include "qemu/config-file.h"
70 #include "qemu/error-report.h"
71 #include "qemu/option.h"
72 #include "qemu/cutils.h"
73 #include "hw/acpi/acpi.h"
74 #include "hw/acpi/cpu_hotplug.h"
75 #include "acpi-build.h"
76 #include "hw/mem/pc-dimm.h"
77 #include "hw/mem/nvdimm.h"
78 #include "hw/cxl/cxl.h"
79 #include "hw/cxl/cxl_host.h"
80 #include "qapi/error.h"
81 #include "qapi/qapi-visit-common.h"
82 #include "qapi/qapi-visit-machine.h"
83 #include "qapi/visitor.h"
84 #include "hw/core/cpu.h"
85 #include "hw/usb.h"
86 #include "hw/i386/intel_iommu.h"
87 #include "hw/net/ne2000-isa.h"
88 #include "standard-headers/asm-x86/bootparam.h"
89 #include "hw/virtio/virtio-iommu.h"
90 #include "hw/virtio/virtio-pmem-pci.h"
91 #include "hw/virtio/virtio-mem-pci.h"
92 #include "hw/mem/memory-device.h"
93 #include "sysemu/replay.h"
94 #include "target/i386/cpu.h"
95 #include "qapi/qmp/qerror.h"
96 #include "e820_memory_layout.h"
97 #include "fw_cfg.h"
98 #include "trace.h"
99 #include CONFIG_DEVICES
100 
101 /*
102  * Helper for setting model-id for CPU models that changed model-id
103  * depending on QEMU versions up to QEMU 2.4.
104  */
105 #define PC_CPU_MODEL_IDS(v) \
106     { "qemu32-" TYPE_X86_CPU, "model-id", "QEMU Virtual CPU version " v, },\
107     { "qemu64-" TYPE_X86_CPU, "model-id", "QEMU Virtual CPU version " v, },\
108     { "athlon-" TYPE_X86_CPU, "model-id", "QEMU Virtual CPU version " v, },
109 
110 GlobalProperty pc_compat_7_2[] = {
111     { "ICH9-LPC", "noreboot", "true" },
112 };
113 const size_t pc_compat_7_2_len = G_N_ELEMENTS(pc_compat_7_2);
114 
115 GlobalProperty pc_compat_7_1[] = {};
116 const size_t pc_compat_7_1_len = G_N_ELEMENTS(pc_compat_7_1);
117 
118 GlobalProperty pc_compat_7_0[] = {};
119 const size_t pc_compat_7_0_len = G_N_ELEMENTS(pc_compat_7_0);
120 
121 GlobalProperty pc_compat_6_2[] = {
122     { "virtio-mem", "unplugged-inaccessible", "off" },
123 };
124 const size_t pc_compat_6_2_len = G_N_ELEMENTS(pc_compat_6_2);
125 
126 GlobalProperty pc_compat_6_1[] = {
127     { TYPE_X86_CPU, "hv-version-id-build", "0x1bbc" },
128     { TYPE_X86_CPU, "hv-version-id-major", "0x0006" },
129     { TYPE_X86_CPU, "hv-version-id-minor", "0x0001" },
130     { "ICH9-LPC", "x-keep-pci-slot-hpc", "false" },
131 };
132 const size_t pc_compat_6_1_len = G_N_ELEMENTS(pc_compat_6_1);
133 
134 GlobalProperty pc_compat_6_0[] = {
135     { "qemu64" "-" TYPE_X86_CPU, "family", "6" },
136     { "qemu64" "-" TYPE_X86_CPU, "model", "6" },
137     { "qemu64" "-" TYPE_X86_CPU, "stepping", "3" },
138     { TYPE_X86_CPU, "x-vendor-cpuid-only", "off" },
139     { "ICH9-LPC", ACPI_PM_PROP_ACPI_PCIHP_BRIDGE, "off" },
140     { "ICH9-LPC", "x-keep-pci-slot-hpc", "true" },
141 };
142 const size_t pc_compat_6_0_len = G_N_ELEMENTS(pc_compat_6_0);
143 
144 GlobalProperty pc_compat_5_2[] = {
145     { "ICH9-LPC", "x-smi-cpu-hotunplug", "off" },
146 };
147 const size_t pc_compat_5_2_len = G_N_ELEMENTS(pc_compat_5_2);
148 
149 GlobalProperty pc_compat_5_1[] = {
150     { "ICH9-LPC", "x-smi-cpu-hotplug", "off" },
151     { TYPE_X86_CPU, "kvm-msi-ext-dest-id", "off" },
152 };
153 const size_t pc_compat_5_1_len = G_N_ELEMENTS(pc_compat_5_1);
154 
155 GlobalProperty pc_compat_5_0[] = {
156 };
157 const size_t pc_compat_5_0_len = G_N_ELEMENTS(pc_compat_5_0);
158 
159 GlobalProperty pc_compat_4_2[] = {
160     { "mch", "smbase-smram", "off" },
161 };
162 const size_t pc_compat_4_2_len = G_N_ELEMENTS(pc_compat_4_2);
163 
164 GlobalProperty pc_compat_4_1[] = {};
165 const size_t pc_compat_4_1_len = G_N_ELEMENTS(pc_compat_4_1);
166 
167 GlobalProperty pc_compat_4_0[] = {};
168 const size_t pc_compat_4_0_len = G_N_ELEMENTS(pc_compat_4_0);
169 
170 GlobalProperty pc_compat_3_1[] = {
171     { "intel-iommu", "dma-drain", "off" },
172     { "Opteron_G3" "-" TYPE_X86_CPU, "rdtscp", "off" },
173     { "Opteron_G4" "-" TYPE_X86_CPU, "rdtscp", "off" },
174     { "Opteron_G4" "-" TYPE_X86_CPU, "npt", "off" },
175     { "Opteron_G4" "-" TYPE_X86_CPU, "nrip-save", "off" },
176     { "Opteron_G5" "-" TYPE_X86_CPU, "rdtscp", "off" },
177     { "Opteron_G5" "-" TYPE_X86_CPU, "npt", "off" },
178     { "Opteron_G5" "-" TYPE_X86_CPU, "nrip-save", "off" },
179     { "EPYC" "-" TYPE_X86_CPU, "npt", "off" },
180     { "EPYC" "-" TYPE_X86_CPU, "nrip-save", "off" },
181     { "EPYC-IBPB" "-" TYPE_X86_CPU, "npt", "off" },
182     { "EPYC-IBPB" "-" TYPE_X86_CPU, "nrip-save", "off" },
183     { "Skylake-Client" "-" TYPE_X86_CPU,      "mpx", "on" },
184     { "Skylake-Client-IBRS" "-" TYPE_X86_CPU, "mpx", "on" },
185     { "Skylake-Server" "-" TYPE_X86_CPU,      "mpx", "on" },
186     { "Skylake-Server-IBRS" "-" TYPE_X86_CPU, "mpx", "on" },
187     { "Cascadelake-Server" "-" TYPE_X86_CPU,  "mpx", "on" },
188     { "Icelake-Client" "-" TYPE_X86_CPU,      "mpx", "on" },
189     { "Icelake-Server" "-" TYPE_X86_CPU,      "mpx", "on" },
190     { "Cascadelake-Server" "-" TYPE_X86_CPU, "stepping", "5" },
191     { TYPE_X86_CPU, "x-intel-pt-auto-level", "off" },
192 };
193 const size_t pc_compat_3_1_len = G_N_ELEMENTS(pc_compat_3_1);
194 
195 GlobalProperty pc_compat_3_0[] = {
196     { TYPE_X86_CPU, "x-hv-synic-kvm-only", "on" },
197     { "Skylake-Server" "-" TYPE_X86_CPU, "pku", "off" },
198     { "Skylake-Server-IBRS" "-" TYPE_X86_CPU, "pku", "off" },
199 };
200 const size_t pc_compat_3_0_len = G_N_ELEMENTS(pc_compat_3_0);
201 
202 GlobalProperty pc_compat_2_12[] = {
203     { TYPE_X86_CPU, "legacy-cache", "on" },
204     { TYPE_X86_CPU, "topoext", "off" },
205     { "EPYC-" TYPE_X86_CPU, "xlevel", "0x8000000a" },
206     { "EPYC-IBPB-" TYPE_X86_CPU, "xlevel", "0x8000000a" },
207 };
208 const size_t pc_compat_2_12_len = G_N_ELEMENTS(pc_compat_2_12);
209 
210 GlobalProperty pc_compat_2_11[] = {
211     { TYPE_X86_CPU, "x-migrate-smi-count", "off" },
212     { "Skylake-Server" "-" TYPE_X86_CPU, "clflushopt", "off" },
213 };
214 const size_t pc_compat_2_11_len = G_N_ELEMENTS(pc_compat_2_11);
215 
216 GlobalProperty pc_compat_2_10[] = {
217     { TYPE_X86_CPU, "x-hv-max-vps", "0x40" },
218     { "i440FX-pcihost", "x-pci-hole64-fix", "off" },
219     { "q35-pcihost", "x-pci-hole64-fix", "off" },
220 };
221 const size_t pc_compat_2_10_len = G_N_ELEMENTS(pc_compat_2_10);
222 
223 GlobalProperty pc_compat_2_9[] = {
224     { "mch", "extended-tseg-mbytes", "0" },
225 };
226 const size_t pc_compat_2_9_len = G_N_ELEMENTS(pc_compat_2_9);
227 
228 GlobalProperty pc_compat_2_8[] = {
229     { TYPE_X86_CPU, "tcg-cpuid", "off" },
230     { "kvmclock", "x-mach-use-reliable-get-clock", "off" },
231     { "ICH9-LPC", "x-smi-broadcast", "off" },
232     { TYPE_X86_CPU, "vmware-cpuid-freq", "off" },
233     { "Haswell-" TYPE_X86_CPU, "stepping", "1" },
234 };
235 const size_t pc_compat_2_8_len = G_N_ELEMENTS(pc_compat_2_8);
236 
237 GlobalProperty pc_compat_2_7[] = {
238     { TYPE_X86_CPU, "l3-cache", "off" },
239     { TYPE_X86_CPU, "full-cpuid-auto-level", "off" },
240     { "Opteron_G3" "-" TYPE_X86_CPU, "family", "15" },
241     { "Opteron_G3" "-" TYPE_X86_CPU, "model", "6" },
242     { "Opteron_G3" "-" TYPE_X86_CPU, "stepping", "1" },
243     { "isa-pcspk", "migrate", "off" },
244 };
245 const size_t pc_compat_2_7_len = G_N_ELEMENTS(pc_compat_2_7);
246 
247 GlobalProperty pc_compat_2_6[] = {
248     { TYPE_X86_CPU, "cpuid-0xb", "off" },
249     { "vmxnet3", "romfile", "" },
250     { TYPE_X86_CPU, "fill-mtrr-mask", "off" },
251     { "apic-common", "legacy-instance-id", "on", }
252 };
253 const size_t pc_compat_2_6_len = G_N_ELEMENTS(pc_compat_2_6);
254 
255 GlobalProperty pc_compat_2_5[] = {};
256 const size_t pc_compat_2_5_len = G_N_ELEMENTS(pc_compat_2_5);
257 
258 GlobalProperty pc_compat_2_4[] = {
259     PC_CPU_MODEL_IDS("2.4.0")
260     { "Haswell-" TYPE_X86_CPU, "abm", "off" },
261     { "Haswell-noTSX-" TYPE_X86_CPU, "abm", "off" },
262     { "Broadwell-" TYPE_X86_CPU, "abm", "off" },
263     { "Broadwell-noTSX-" TYPE_X86_CPU, "abm", "off" },
264     { "host" "-" TYPE_X86_CPU, "host-cache-info", "on" },
265     { TYPE_X86_CPU, "check", "off" },
266     { "qemu64" "-" TYPE_X86_CPU, "sse4a", "on" },
267     { "qemu64" "-" TYPE_X86_CPU, "abm", "on" },
268     { "qemu64" "-" TYPE_X86_CPU, "popcnt", "on" },
269     { "qemu32" "-" TYPE_X86_CPU, "popcnt", "on" },
270     { "Opteron_G2" "-" TYPE_X86_CPU, "rdtscp", "on" },
271     { "Opteron_G3" "-" TYPE_X86_CPU, "rdtscp", "on" },
272     { "Opteron_G4" "-" TYPE_X86_CPU, "rdtscp", "on" },
273     { "Opteron_G5" "-" TYPE_X86_CPU, "rdtscp", "on", }
274 };
275 const size_t pc_compat_2_4_len = G_N_ELEMENTS(pc_compat_2_4);
276 
277 GlobalProperty pc_compat_2_3[] = {
278     PC_CPU_MODEL_IDS("2.3.0")
279     { TYPE_X86_CPU, "arat", "off" },
280     { "qemu64" "-" TYPE_X86_CPU, "min-level", "4" },
281     { "kvm64" "-" TYPE_X86_CPU, "min-level", "5" },
282     { "pentium3" "-" TYPE_X86_CPU, "min-level", "2" },
283     { "n270" "-" TYPE_X86_CPU, "min-level", "5" },
284     { "Conroe" "-" TYPE_X86_CPU, "min-level", "4" },
285     { "Penryn" "-" TYPE_X86_CPU, "min-level", "4" },
286     { "Nehalem" "-" TYPE_X86_CPU, "min-level", "4" },
287     { "n270" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
288     { "Penryn" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
289     { "Conroe" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
290     { "Nehalem" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
291     { "Westmere" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
292     { "SandyBridge" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
293     { "IvyBridge" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
294     { "Haswell" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
295     { "Haswell-noTSX" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
296     { "Broadwell" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
297     { "Broadwell-noTSX" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
298     { TYPE_X86_CPU, "kvm-no-smi-migration", "on" },
299 };
300 const size_t pc_compat_2_3_len = G_N_ELEMENTS(pc_compat_2_3);
301 
302 GlobalProperty pc_compat_2_2[] = {
303     PC_CPU_MODEL_IDS("2.2.0")
304     { "kvm64" "-" TYPE_X86_CPU, "vme", "off" },
305     { "kvm32" "-" TYPE_X86_CPU, "vme", "off" },
306     { "Conroe" "-" TYPE_X86_CPU, "vme", "off" },
307     { "Penryn" "-" TYPE_X86_CPU, "vme", "off" },
308     { "Nehalem" "-" TYPE_X86_CPU, "vme", "off" },
309     { "Westmere" "-" TYPE_X86_CPU, "vme", "off" },
310     { "SandyBridge" "-" TYPE_X86_CPU, "vme", "off" },
311     { "Haswell" "-" TYPE_X86_CPU, "vme", "off" },
312     { "Broadwell" "-" TYPE_X86_CPU, "vme", "off" },
313     { "Opteron_G1" "-" TYPE_X86_CPU, "vme", "off" },
314     { "Opteron_G2" "-" TYPE_X86_CPU, "vme", "off" },
315     { "Opteron_G3" "-" TYPE_X86_CPU, "vme", "off" },
316     { "Opteron_G4" "-" TYPE_X86_CPU, "vme", "off" },
317     { "Opteron_G5" "-" TYPE_X86_CPU, "vme", "off" },
318     { "Haswell" "-" TYPE_X86_CPU, "f16c", "off" },
319     { "Haswell" "-" TYPE_X86_CPU, "rdrand", "off" },
320     { "Broadwell" "-" TYPE_X86_CPU, "f16c", "off" },
321     { "Broadwell" "-" TYPE_X86_CPU, "rdrand", "off" },
322 };
323 const size_t pc_compat_2_2_len = G_N_ELEMENTS(pc_compat_2_2);
324 
325 GlobalProperty pc_compat_2_1[] = {
326     PC_CPU_MODEL_IDS("2.1.0")
327     { "coreduo" "-" TYPE_X86_CPU, "vmx", "on" },
328     { "core2duo" "-" TYPE_X86_CPU, "vmx", "on" },
329 };
330 const size_t pc_compat_2_1_len = G_N_ELEMENTS(pc_compat_2_1);
331 
332 GlobalProperty pc_compat_2_0[] = {
333     PC_CPU_MODEL_IDS("2.0.0")
334     { "virtio-scsi-pci", "any_layout", "off" },
335     { "PIIX4_PM", "memory-hotplug-support", "off" },
336     { "apic", "version", "0x11" },
337     { "nec-usb-xhci", "superspeed-ports-first", "off" },
338     { "nec-usb-xhci", "force-pcie-endcap", "on" },
339     { "pci-serial", "prog_if", "0" },
340     { "pci-serial-2x", "prog_if", "0" },
341     { "pci-serial-4x", "prog_if", "0" },
342     { "virtio-net-pci", "guest_announce", "off" },
343     { "ICH9-LPC", "memory-hotplug-support", "off" },
344 };
345 const size_t pc_compat_2_0_len = G_N_ELEMENTS(pc_compat_2_0);
346 
347 GlobalProperty pc_compat_1_7[] = {
348     PC_CPU_MODEL_IDS("1.7.0")
349     { TYPE_USB_DEVICE, "msos-desc", "no" },
350     { "PIIX4_PM", ACPI_PM_PROP_ACPI_PCIHP_BRIDGE, "off" },
351     { "hpet", HPET_INTCAP, "4" },
352 };
353 const size_t pc_compat_1_7_len = G_N_ELEMENTS(pc_compat_1_7);
354 
355 GlobalProperty pc_compat_1_6[] = {
356     PC_CPU_MODEL_IDS("1.6.0")
357     { "e1000", "mitigation", "off" },
358     { "qemu64-" TYPE_X86_CPU, "model", "2" },
359     { "qemu32-" TYPE_X86_CPU, "model", "3" },
360     { "i440FX-pcihost", "short_root_bus", "1" },
361     { "q35-pcihost", "short_root_bus", "1" },
362 };
363 const size_t pc_compat_1_6_len = G_N_ELEMENTS(pc_compat_1_6);
364 
365 GlobalProperty pc_compat_1_5[] = {
366     PC_CPU_MODEL_IDS("1.5.0")
367     { "Conroe-" TYPE_X86_CPU, "model", "2" },
368     { "Conroe-" TYPE_X86_CPU, "min-level", "2" },
369     { "Penryn-" TYPE_X86_CPU, "model", "2" },
370     { "Penryn-" TYPE_X86_CPU, "min-level", "2" },
371     { "Nehalem-" TYPE_X86_CPU, "model", "2" },
372     { "Nehalem-" TYPE_X86_CPU, "min-level", "2" },
373     { "virtio-net-pci", "any_layout", "off" },
374     { TYPE_X86_CPU, "pmu", "on" },
375     { "i440FX-pcihost", "short_root_bus", "0" },
376     { "q35-pcihost", "short_root_bus", "0" },
377 };
378 const size_t pc_compat_1_5_len = G_N_ELEMENTS(pc_compat_1_5);
379 
380 GlobalProperty pc_compat_1_4[] = {
381     PC_CPU_MODEL_IDS("1.4.0")
382     { "scsi-hd", "discard_granularity", "0" },
383     { "scsi-cd", "discard_granularity", "0" },
384     { "ide-hd", "discard_granularity", "0" },
385     { "ide-cd", "discard_granularity", "0" },
386     { "virtio-blk-pci", "discard_granularity", "0" },
387     /* DEV_NVECTORS_UNSPECIFIED as a uint32_t string: */
388     { "virtio-serial-pci", "vectors", "0xFFFFFFFF" },
389     { "virtio-net-pci", "ctrl_guest_offloads", "off" },
390     { "e1000", "romfile", "pxe-e1000.rom" },
391     { "ne2k_pci", "romfile", "pxe-ne2k_pci.rom" },
392     { "pcnet", "romfile", "pxe-pcnet.rom" },
393     { "rtl8139", "romfile", "pxe-rtl8139.rom" },
394     { "virtio-net-pci", "romfile", "pxe-virtio.rom" },
395     { "486-" TYPE_X86_CPU, "model", "0" },
396     { "n270" "-" TYPE_X86_CPU, "movbe", "off" },
397     { "Westmere" "-" TYPE_X86_CPU, "pclmulqdq", "off" },
398 };
399 const size_t pc_compat_1_4_len = G_N_ELEMENTS(pc_compat_1_4);
400 
401 GSIState *pc_gsi_create(qemu_irq **irqs, bool pci_enabled)
402 {
403     GSIState *s;
404 
405     s = g_new0(GSIState, 1);
406     if (kvm_ioapic_in_kernel()) {
407         kvm_pc_setup_irq_routing(pci_enabled);
408     }
409     *irqs = qemu_allocate_irqs(gsi_handler, s, GSI_NUM_PINS);
410 
411     return s;
412 }
413 
414 static void ioport80_write(void *opaque, hwaddr addr, uint64_t data,
415                            unsigned size)
416 {
417 }
418 
419 static uint64_t ioport80_read(void *opaque, hwaddr addr, unsigned size)
420 {
421     return 0xffffffffffffffffULL;
422 }
423 
424 /* MSDOS compatibility mode FPU exception support */
425 static void ioportF0_write(void *opaque, hwaddr addr, uint64_t data,
426                            unsigned size)
427 {
428     if (tcg_enabled()) {
429         cpu_set_ignne();
430     }
431 }
432 
433 static uint64_t ioportF0_read(void *opaque, hwaddr addr, unsigned size)
434 {
435     return 0xffffffffffffffffULL;
436 }
437 
438 /* PC cmos mappings */
439 
440 #define REG_EQUIPMENT_BYTE          0x14
441 
442 static void cmos_init_hd(ISADevice *s, int type_ofs, int info_ofs,
443                          int16_t cylinders, int8_t heads, int8_t sectors)
444 {
445     rtc_set_memory(s, type_ofs, 47);
446     rtc_set_memory(s, info_ofs, cylinders);
447     rtc_set_memory(s, info_ofs + 1, cylinders >> 8);
448     rtc_set_memory(s, info_ofs + 2, heads);
449     rtc_set_memory(s, info_ofs + 3, 0xff);
450     rtc_set_memory(s, info_ofs + 4, 0xff);
451     rtc_set_memory(s, info_ofs + 5, 0xc0 | ((heads > 8) << 3));
452     rtc_set_memory(s, info_ofs + 6, cylinders);
453     rtc_set_memory(s, info_ofs + 7, cylinders >> 8);
454     rtc_set_memory(s, info_ofs + 8, sectors);
455 }
456 
457 /* convert boot_device letter to something recognizable by the bios */
458 static int boot_device2nibble(char boot_device)
459 {
460     switch(boot_device) {
461     case 'a':
462     case 'b':
463         return 0x01; /* floppy boot */
464     case 'c':
465         return 0x02; /* hard drive boot */
466     case 'd':
467         return 0x03; /* CD-ROM boot */
468     case 'n':
469         return 0x04; /* Network boot */
470     }
471     return 0;
472 }
473 
474 static void set_boot_dev(ISADevice *s, const char *boot_device, Error **errp)
475 {
476 #define PC_MAX_BOOT_DEVICES 3
477     int nbds, bds[3] = { 0, };
478     int i;
479 
480     nbds = strlen(boot_device);
481     if (nbds > PC_MAX_BOOT_DEVICES) {
482         error_setg(errp, "Too many boot devices for PC");
483         return;
484     }
485     for (i = 0; i < nbds; i++) {
486         bds[i] = boot_device2nibble(boot_device[i]);
487         if (bds[i] == 0) {
488             error_setg(errp, "Invalid boot device for PC: '%c'",
489                        boot_device[i]);
490             return;
491         }
492     }
493     rtc_set_memory(s, 0x3d, (bds[1] << 4) | bds[0]);
494     rtc_set_memory(s, 0x38, (bds[2] << 4) | (fd_bootchk ? 0x0 : 0x1));
495 }
496 
497 static void pc_boot_set(void *opaque, const char *boot_device, Error **errp)
498 {
499     set_boot_dev(opaque, boot_device, errp);
500 }
501 
502 static void pc_cmos_init_floppy(ISADevice *rtc_state, ISADevice *floppy)
503 {
504     int val, nb, i;
505     FloppyDriveType fd_type[2] = { FLOPPY_DRIVE_TYPE_NONE,
506                                    FLOPPY_DRIVE_TYPE_NONE };
507 
508     /* floppy type */
509     if (floppy) {
510         for (i = 0; i < 2; i++) {
511             fd_type[i] = isa_fdc_get_drive_type(floppy, i);
512         }
513     }
514     val = (cmos_get_fd_drive_type(fd_type[0]) << 4) |
515         cmos_get_fd_drive_type(fd_type[1]);
516     rtc_set_memory(rtc_state, 0x10, val);
517 
518     val = rtc_get_memory(rtc_state, REG_EQUIPMENT_BYTE);
519     nb = 0;
520     if (fd_type[0] != FLOPPY_DRIVE_TYPE_NONE) {
521         nb++;
522     }
523     if (fd_type[1] != FLOPPY_DRIVE_TYPE_NONE) {
524         nb++;
525     }
526     switch (nb) {
527     case 0:
528         break;
529     case 1:
530         val |= 0x01; /* 1 drive, ready for boot */
531         break;
532     case 2:
533         val |= 0x41; /* 2 drives, ready for boot */
534         break;
535     }
536     rtc_set_memory(rtc_state, REG_EQUIPMENT_BYTE, val);
537 }
538 
539 typedef struct pc_cmos_init_late_arg {
540     ISADevice *rtc_state;
541     BusState *idebus[2];
542 } pc_cmos_init_late_arg;
543 
544 typedef struct check_fdc_state {
545     ISADevice *floppy;
546     bool multiple;
547 } CheckFdcState;
548 
549 static int check_fdc(Object *obj, void *opaque)
550 {
551     CheckFdcState *state = opaque;
552     Object *fdc;
553     uint32_t iobase;
554     Error *local_err = NULL;
555 
556     fdc = object_dynamic_cast(obj, TYPE_ISA_FDC);
557     if (!fdc) {
558         return 0;
559     }
560 
561     iobase = object_property_get_uint(obj, "iobase", &local_err);
562     if (local_err || iobase != 0x3f0) {
563         error_free(local_err);
564         return 0;
565     }
566 
567     if (state->floppy) {
568         state->multiple = true;
569     } else {
570         state->floppy = ISA_DEVICE(obj);
571     }
572     return 0;
573 }
574 
575 static const char * const fdc_container_path[] = {
576     "/unattached", "/peripheral", "/peripheral-anon"
577 };
578 
579 /*
580  * Locate the FDC at IO address 0x3f0, in order to configure the CMOS registers
581  * and ACPI objects.
582  */
583 static ISADevice *pc_find_fdc0(void)
584 {
585     int i;
586     Object *container;
587     CheckFdcState state = { 0 };
588 
589     for (i = 0; i < ARRAY_SIZE(fdc_container_path); i++) {
590         container = container_get(qdev_get_machine(), fdc_container_path[i]);
591         object_child_foreach(container, check_fdc, &state);
592     }
593 
594     if (state.multiple) {
595         warn_report("multiple floppy disk controllers with "
596                     "iobase=0x3f0 have been found");
597         error_printf("the one being picked for CMOS setup might not reflect "
598                      "your intent");
599     }
600 
601     return state.floppy;
602 }
603 
604 static void pc_cmos_init_late(void *opaque)
605 {
606     pc_cmos_init_late_arg *arg = opaque;
607     ISADevice *s = arg->rtc_state;
608     int16_t cylinders;
609     int8_t heads, sectors;
610     int val;
611     int i, trans;
612 
613     val = 0;
614     if (arg->idebus[0] && ide_get_geometry(arg->idebus[0], 0,
615                                            &cylinders, &heads, &sectors) >= 0) {
616         cmos_init_hd(s, 0x19, 0x1b, cylinders, heads, sectors);
617         val |= 0xf0;
618     }
619     if (arg->idebus[0] && ide_get_geometry(arg->idebus[0], 1,
620                                            &cylinders, &heads, &sectors) >= 0) {
621         cmos_init_hd(s, 0x1a, 0x24, cylinders, heads, sectors);
622         val |= 0x0f;
623     }
624     rtc_set_memory(s, 0x12, val);
625 
626     val = 0;
627     for (i = 0; i < 4; i++) {
628         /* NOTE: ide_get_geometry() returns the physical
629            geometry.  It is always such that: 1 <= sects <= 63, 1
630            <= heads <= 16, 1 <= cylinders <= 16383. The BIOS
631            geometry can be different if a translation is done. */
632         if (arg->idebus[i / 2] &&
633             ide_get_geometry(arg->idebus[i / 2], i % 2,
634                              &cylinders, &heads, &sectors) >= 0) {
635             trans = ide_get_bios_chs_trans(arg->idebus[i / 2], i % 2) - 1;
636             assert((trans & ~3) == 0);
637             val |= trans << (i * 2);
638         }
639     }
640     rtc_set_memory(s, 0x39, val);
641 
642     pc_cmos_init_floppy(s, pc_find_fdc0());
643 
644     qemu_unregister_reset(pc_cmos_init_late, opaque);
645 }
646 
647 void pc_cmos_init(PCMachineState *pcms,
648                   BusState *idebus0, BusState *idebus1,
649                   ISADevice *s)
650 {
651     int val;
652     static pc_cmos_init_late_arg arg;
653     X86MachineState *x86ms = X86_MACHINE(pcms);
654 
655     /* various important CMOS locations needed by PC/Bochs bios */
656 
657     /* memory size */
658     /* base memory (first MiB) */
659     val = MIN(x86ms->below_4g_mem_size / KiB, 640);
660     rtc_set_memory(s, 0x15, val);
661     rtc_set_memory(s, 0x16, val >> 8);
662     /* extended memory (next 64MiB) */
663     if (x86ms->below_4g_mem_size > 1 * MiB) {
664         val = (x86ms->below_4g_mem_size - 1 * MiB) / KiB;
665     } else {
666         val = 0;
667     }
668     if (val > 65535)
669         val = 65535;
670     rtc_set_memory(s, 0x17, val);
671     rtc_set_memory(s, 0x18, val >> 8);
672     rtc_set_memory(s, 0x30, val);
673     rtc_set_memory(s, 0x31, val >> 8);
674     /* memory between 16MiB and 4GiB */
675     if (x86ms->below_4g_mem_size > 16 * MiB) {
676         val = (x86ms->below_4g_mem_size - 16 * MiB) / (64 * KiB);
677     } else {
678         val = 0;
679     }
680     if (val > 65535)
681         val = 65535;
682     rtc_set_memory(s, 0x34, val);
683     rtc_set_memory(s, 0x35, val >> 8);
684     /* memory above 4GiB */
685     val = x86ms->above_4g_mem_size / 65536;
686     rtc_set_memory(s, 0x5b, val);
687     rtc_set_memory(s, 0x5c, val >> 8);
688     rtc_set_memory(s, 0x5d, val >> 16);
689 
690     object_property_add_link(OBJECT(pcms), "rtc_state",
691                              TYPE_ISA_DEVICE,
692                              (Object **)&x86ms->rtc,
693                              object_property_allow_set_link,
694                              OBJ_PROP_LINK_STRONG);
695     object_property_set_link(OBJECT(pcms), "rtc_state", OBJECT(s),
696                              &error_abort);
697 
698     set_boot_dev(s, MACHINE(pcms)->boot_config.order, &error_fatal);
699 
700     val = 0;
701     val |= 0x02; /* FPU is there */
702     val |= 0x04; /* PS/2 mouse installed */
703     rtc_set_memory(s, REG_EQUIPMENT_BYTE, val);
704 
705     /* hard drives and FDC */
706     arg.rtc_state = s;
707     arg.idebus[0] = idebus0;
708     arg.idebus[1] = idebus1;
709     qemu_register_reset(pc_cmos_init_late, &arg);
710 }
711 
712 static void handle_a20_line_change(void *opaque, int irq, int level)
713 {
714     X86CPU *cpu = opaque;
715 
716     /* XXX: send to all CPUs ? */
717     /* XXX: add logic to handle multiple A20 line sources */
718     x86_cpu_set_a20(cpu, level);
719 }
720 
721 #define NE2000_NB_MAX 6
722 
723 static const int ne2000_io[NE2000_NB_MAX] = { 0x300, 0x320, 0x340, 0x360,
724                                               0x280, 0x380 };
725 static const int ne2000_irq[NE2000_NB_MAX] = { 9, 10, 11, 3, 4, 5 };
726 
727 static void pc_init_ne2k_isa(ISABus *bus, NICInfo *nd)
728 {
729     static int nb_ne2k = 0;
730 
731     if (nb_ne2k == NE2000_NB_MAX)
732         return;
733     isa_ne2000_init(bus, ne2000_io[nb_ne2k],
734                     ne2000_irq[nb_ne2k], nd);
735     nb_ne2k++;
736 }
737 
738 void pc_acpi_smi_interrupt(void *opaque, int irq, int level)
739 {
740     X86CPU *cpu = opaque;
741 
742     if (level) {
743         cpu_interrupt(CPU(cpu), CPU_INTERRUPT_SMI);
744     }
745 }
746 
747 static
748 void pc_machine_done(Notifier *notifier, void *data)
749 {
750     PCMachineState *pcms = container_of(notifier,
751                                         PCMachineState, machine_done);
752     X86MachineState *x86ms = X86_MACHINE(pcms);
753 
754     cxl_hook_up_pxb_registers(pcms->bus, &pcms->cxl_devices_state,
755                               &error_fatal);
756 
757     if (pcms->cxl_devices_state.is_enabled) {
758         cxl_fmws_link_targets(&pcms->cxl_devices_state, &error_fatal);
759     }
760 
761     /* set the number of CPUs */
762     x86_rtc_set_cpus_count(x86ms->rtc, x86ms->boot_cpus);
763 
764     fw_cfg_add_extra_pci_roots(pcms->bus, x86ms->fw_cfg);
765 
766     acpi_setup();
767     if (x86ms->fw_cfg) {
768         fw_cfg_build_smbios(MACHINE(pcms), x86ms->fw_cfg);
769         fw_cfg_build_feature_control(MACHINE(pcms), x86ms->fw_cfg);
770         /* update FW_CFG_NB_CPUS to account for -device added CPUs */
771         fw_cfg_modify_i16(x86ms->fw_cfg, FW_CFG_NB_CPUS, x86ms->boot_cpus);
772     }
773 }
774 
775 void pc_guest_info_init(PCMachineState *pcms)
776 {
777     X86MachineState *x86ms = X86_MACHINE(pcms);
778 
779     x86ms->apic_xrupt_override = true;
780     pcms->machine_done.notify = pc_machine_done;
781     qemu_add_machine_init_done_notifier(&pcms->machine_done);
782 }
783 
784 /* setup pci memory address space mapping into system address space */
785 void pc_pci_as_mapping_init(MemoryRegion *system_memory,
786                             MemoryRegion *pci_address_space)
787 {
788     /* Set to lower priority than RAM */
789     memory_region_add_subregion_overlap(system_memory, 0x0,
790                                         pci_address_space, -1);
791 }
792 
793 void xen_load_linux(PCMachineState *pcms)
794 {
795     int i;
796     FWCfgState *fw_cfg;
797     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
798     X86MachineState *x86ms = X86_MACHINE(pcms);
799 
800     assert(MACHINE(pcms)->kernel_filename != NULL);
801 
802     fw_cfg = fw_cfg_init_io(FW_CFG_IO_BASE);
803     fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, x86ms->boot_cpus);
804     rom_set_fw(fw_cfg);
805 
806     x86_load_linux(x86ms, fw_cfg, pcmc->acpi_data_size,
807                    pcmc->pvh_enabled, pcmc->legacy_no_rng_seed);
808     for (i = 0; i < nb_option_roms; i++) {
809         assert(!strcmp(option_rom[i].name, "linuxboot.bin") ||
810                !strcmp(option_rom[i].name, "linuxboot_dma.bin") ||
811                !strcmp(option_rom[i].name, "pvh.bin") ||
812                !strcmp(option_rom[i].name, "multiboot.bin") ||
813                !strcmp(option_rom[i].name, "multiboot_dma.bin"));
814         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
815     }
816     x86ms->fw_cfg = fw_cfg;
817 }
818 
819 #define PC_ROM_MIN_VGA     0xc0000
820 #define PC_ROM_MIN_OPTION  0xc8000
821 #define PC_ROM_MAX         0xe0000
822 #define PC_ROM_ALIGN       0x800
823 #define PC_ROM_SIZE        (PC_ROM_MAX - PC_ROM_MIN_VGA)
824 
825 static hwaddr pc_above_4g_end(PCMachineState *pcms)
826 {
827     X86MachineState *x86ms = X86_MACHINE(pcms);
828 
829     if (pcms->sgx_epc.size != 0) {
830         return sgx_epc_above_4g_end(&pcms->sgx_epc);
831     }
832 
833     return x86ms->above_4g_mem_start + x86ms->above_4g_mem_size;
834 }
835 
836 static void pc_get_device_memory_range(PCMachineState *pcms,
837                                        hwaddr *base,
838                                        ram_addr_t *device_mem_size)
839 {
840     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
841     MachineState *machine = MACHINE(pcms);
842     ram_addr_t size;
843     hwaddr addr;
844 
845     size = machine->maxram_size - machine->ram_size;
846     addr = ROUND_UP(pc_above_4g_end(pcms), 1 * GiB);
847 
848     if (pcmc->enforce_aligned_dimm) {
849         /* size device region assuming 1G page max alignment per slot */
850         size += (1 * GiB) * machine->ram_slots;
851     }
852 
853     *base = addr;
854     *device_mem_size = size;
855 }
856 
857 static uint64_t pc_get_cxl_range_start(PCMachineState *pcms)
858 {
859     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
860     hwaddr cxl_base;
861     ram_addr_t size;
862 
863     if (pcmc->has_reserved_memory) {
864         pc_get_device_memory_range(pcms, &cxl_base, &size);
865         cxl_base += size;
866     } else {
867         cxl_base = pc_above_4g_end(pcms);
868     }
869 
870     return cxl_base;
871 }
872 
873 static uint64_t pc_get_cxl_range_end(PCMachineState *pcms)
874 {
875     uint64_t start = pc_get_cxl_range_start(pcms) + MiB;
876 
877     if (pcms->cxl_devices_state.fixed_windows) {
878         GList *it;
879 
880         start = ROUND_UP(start, 256 * MiB);
881         for (it = pcms->cxl_devices_state.fixed_windows; it; it = it->next) {
882             CXLFixedWindow *fw = it->data;
883             start += fw->size;
884         }
885     }
886 
887     return start;
888 }
889 
890 static hwaddr pc_max_used_gpa(PCMachineState *pcms, uint64_t pci_hole64_size)
891 {
892     X86CPU *cpu = X86_CPU(first_cpu);
893 
894     /* 32-bit systems don't have hole64 thus return max CPU address */
895     if (cpu->phys_bits <= 32) {
896         return ((hwaddr)1 << cpu->phys_bits) - 1;
897     }
898 
899     return pc_pci_hole64_start() + pci_hole64_size - 1;
900 }
901 
902 /*
903  * AMD systems with an IOMMU have an additional hole close to the
904  * 1Tb, which are special GPAs that cannot be DMA mapped. Depending
905  * on kernel version, VFIO may or may not let you DMA map those ranges.
906  * Starting Linux v5.4 we validate it, and can't create guests on AMD machines
907  * with certain memory sizes. It's also wrong to use those IOVA ranges
908  * in detriment of leading to IOMMU INVALID_DEVICE_REQUEST or worse.
909  * The ranges reserved for Hyper-Transport are:
910  *
911  * FD_0000_0000h - FF_FFFF_FFFFh
912  *
913  * The ranges represent the following:
914  *
915  * Base Address   Top Address  Use
916  *
917  * FD_0000_0000h FD_F7FF_FFFFh Reserved interrupt address space
918  * FD_F800_0000h FD_F8FF_FFFFh Interrupt/EOI IntCtl
919  * FD_F900_0000h FD_F90F_FFFFh Legacy PIC IACK
920  * FD_F910_0000h FD_F91F_FFFFh System Management
921  * FD_F920_0000h FD_FAFF_FFFFh Reserved Page Tables
922  * FD_FB00_0000h FD_FBFF_FFFFh Address Translation
923  * FD_FC00_0000h FD_FDFF_FFFFh I/O Space
924  * FD_FE00_0000h FD_FFFF_FFFFh Configuration
925  * FE_0000_0000h FE_1FFF_FFFFh Extended Configuration/Device Messages
926  * FE_2000_0000h FF_FFFF_FFFFh Reserved
927  *
928  * See AMD IOMMU spec, section 2.1.2 "IOMMU Logical Topology",
929  * Table 3: Special Address Controls (GPA) for more information.
930  */
931 #define AMD_HT_START         0xfd00000000UL
932 #define AMD_HT_END           0xffffffffffUL
933 #define AMD_ABOVE_1TB_START  (AMD_HT_END + 1)
934 #define AMD_HT_SIZE          (AMD_ABOVE_1TB_START - AMD_HT_START)
935 
936 void pc_memory_init(PCMachineState *pcms,
937                     MemoryRegion *system_memory,
938                     MemoryRegion *rom_memory,
939                     MemoryRegion **ram_memory,
940                     uint64_t pci_hole64_size)
941 {
942     int linux_boot, i;
943     MemoryRegion *option_rom_mr;
944     MemoryRegion *ram_below_4g, *ram_above_4g;
945     FWCfgState *fw_cfg;
946     MachineState *machine = MACHINE(pcms);
947     MachineClass *mc = MACHINE_GET_CLASS(machine);
948     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
949     X86MachineState *x86ms = X86_MACHINE(pcms);
950     hwaddr maxphysaddr, maxusedaddr;
951     hwaddr cxl_base, cxl_resv_end = 0;
952     X86CPU *cpu = X86_CPU(first_cpu);
953 
954     assert(machine->ram_size == x86ms->below_4g_mem_size +
955                                 x86ms->above_4g_mem_size);
956 
957     linux_boot = (machine->kernel_filename != NULL);
958 
959     /*
960      * The HyperTransport range close to the 1T boundary is unique to AMD
961      * hosts with IOMMUs enabled. Restrict the ram-above-4g relocation
962      * to above 1T to AMD vCPUs only. @enforce_amd_1tb_hole is only false in
963      * older machine types (<= 7.0) for compatibility purposes.
964      */
965     if (IS_AMD_CPU(&cpu->env) && pcmc->enforce_amd_1tb_hole) {
966         /* Bail out if max possible address does not cross HT range */
967         if (pc_max_used_gpa(pcms, pci_hole64_size) >= AMD_HT_START) {
968             x86ms->above_4g_mem_start = AMD_ABOVE_1TB_START;
969         }
970 
971         /*
972          * Advertise the HT region if address space covers the reserved
973          * region or if we relocate.
974          */
975         if (cpu->phys_bits >= 40) {
976             e820_add_entry(AMD_HT_START, AMD_HT_SIZE, E820_RESERVED);
977         }
978     }
979 
980     /*
981      * phys-bits is required to be appropriately configured
982      * to make sure max used GPA is reachable.
983      */
984     maxusedaddr = pc_max_used_gpa(pcms, pci_hole64_size);
985     maxphysaddr = ((hwaddr)1 << cpu->phys_bits) - 1;
986     if (maxphysaddr < maxusedaddr) {
987         error_report("Address space limit 0x%"PRIx64" < 0x%"PRIx64
988                      " phys-bits too low (%u)",
989                      maxphysaddr, maxusedaddr, cpu->phys_bits);
990         exit(EXIT_FAILURE);
991     }
992 
993     /*
994      * Split single memory region and use aliases to address portions of it,
995      * done for backwards compatibility with older qemus.
996      */
997     *ram_memory = machine->ram;
998     ram_below_4g = g_malloc(sizeof(*ram_below_4g));
999     memory_region_init_alias(ram_below_4g, NULL, "ram-below-4g", machine->ram,
1000                              0, x86ms->below_4g_mem_size);
1001     memory_region_add_subregion(system_memory, 0, ram_below_4g);
1002     e820_add_entry(0, x86ms->below_4g_mem_size, E820_RAM);
1003     if (x86ms->above_4g_mem_size > 0) {
1004         ram_above_4g = g_malloc(sizeof(*ram_above_4g));
1005         memory_region_init_alias(ram_above_4g, NULL, "ram-above-4g",
1006                                  machine->ram,
1007                                  x86ms->below_4g_mem_size,
1008                                  x86ms->above_4g_mem_size);
1009         memory_region_add_subregion(system_memory, x86ms->above_4g_mem_start,
1010                                     ram_above_4g);
1011         e820_add_entry(x86ms->above_4g_mem_start, x86ms->above_4g_mem_size,
1012                        E820_RAM);
1013     }
1014 
1015     if (pcms->sgx_epc.size != 0) {
1016         e820_add_entry(pcms->sgx_epc.base, pcms->sgx_epc.size, E820_RESERVED);
1017     }
1018 
1019     if (!pcmc->has_reserved_memory &&
1020         (machine->ram_slots ||
1021          (machine->maxram_size > machine->ram_size))) {
1022 
1023         error_report("\"-memory 'slots|maxmem'\" is not supported by: %s",
1024                      mc->name);
1025         exit(EXIT_FAILURE);
1026     }
1027 
1028     /* always allocate the device memory information */
1029     machine->device_memory = g_malloc0(sizeof(*machine->device_memory));
1030 
1031     /* initialize device memory address space */
1032     if (pcmc->has_reserved_memory &&
1033         (machine->ram_size < machine->maxram_size)) {
1034         ram_addr_t device_mem_size;
1035 
1036         if (machine->ram_slots > ACPI_MAX_RAM_SLOTS) {
1037             error_report("unsupported amount of memory slots: %"PRIu64,
1038                          machine->ram_slots);
1039             exit(EXIT_FAILURE);
1040         }
1041 
1042         if (QEMU_ALIGN_UP(machine->maxram_size,
1043                           TARGET_PAGE_SIZE) != machine->maxram_size) {
1044             error_report("maximum memory size must by aligned to multiple of "
1045                          "%d bytes", TARGET_PAGE_SIZE);
1046             exit(EXIT_FAILURE);
1047         }
1048 
1049         pc_get_device_memory_range(pcms, &machine->device_memory->base, &device_mem_size);
1050 
1051         if ((machine->device_memory->base + device_mem_size) <
1052             device_mem_size) {
1053             error_report("unsupported amount of maximum memory: " RAM_ADDR_FMT,
1054                          machine->maxram_size);
1055             exit(EXIT_FAILURE);
1056         }
1057 
1058         memory_region_init(&machine->device_memory->mr, OBJECT(pcms),
1059                            "device-memory", device_mem_size);
1060         memory_region_add_subregion(system_memory, machine->device_memory->base,
1061                                     &machine->device_memory->mr);
1062     }
1063 
1064     if (pcms->cxl_devices_state.is_enabled) {
1065         MemoryRegion *mr = &pcms->cxl_devices_state.host_mr;
1066         hwaddr cxl_size = MiB;
1067 
1068         cxl_base = pc_get_cxl_range_start(pcms);
1069         memory_region_init(mr, OBJECT(machine), "cxl_host_reg", cxl_size);
1070         memory_region_add_subregion(system_memory, cxl_base, mr);
1071         cxl_resv_end = cxl_base + cxl_size;
1072         if (pcms->cxl_devices_state.fixed_windows) {
1073             hwaddr cxl_fmw_base;
1074             GList *it;
1075 
1076             cxl_fmw_base = ROUND_UP(cxl_base + cxl_size, 256 * MiB);
1077             for (it = pcms->cxl_devices_state.fixed_windows; it; it = it->next) {
1078                 CXLFixedWindow *fw = it->data;
1079 
1080                 fw->base = cxl_fmw_base;
1081                 memory_region_init_io(&fw->mr, OBJECT(machine), &cfmws_ops, fw,
1082                                       "cxl-fixed-memory-region", fw->size);
1083                 memory_region_add_subregion(system_memory, fw->base, &fw->mr);
1084                 cxl_fmw_base += fw->size;
1085                 cxl_resv_end = cxl_fmw_base;
1086             }
1087         }
1088     }
1089 
1090     /* Initialize PC system firmware */
1091     pc_system_firmware_init(pcms, rom_memory);
1092 
1093     option_rom_mr = g_malloc(sizeof(*option_rom_mr));
1094     memory_region_init_ram(option_rom_mr, NULL, "pc.rom", PC_ROM_SIZE,
1095                            &error_fatal);
1096     if (pcmc->pci_enabled) {
1097         memory_region_set_readonly(option_rom_mr, true);
1098     }
1099     memory_region_add_subregion_overlap(rom_memory,
1100                                         PC_ROM_MIN_VGA,
1101                                         option_rom_mr,
1102                                         1);
1103 
1104     fw_cfg = fw_cfg_arch_create(machine,
1105                                 x86ms->boot_cpus, x86ms->apic_id_limit);
1106 
1107     rom_set_fw(fw_cfg);
1108 
1109     if (pcmc->has_reserved_memory && machine->device_memory->base) {
1110         uint64_t *val = g_malloc(sizeof(*val));
1111         PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1112         uint64_t res_mem_end = machine->device_memory->base;
1113 
1114         if (!pcmc->broken_reserved_end) {
1115             res_mem_end += memory_region_size(&machine->device_memory->mr);
1116         }
1117 
1118         if (pcms->cxl_devices_state.is_enabled) {
1119             res_mem_end = cxl_resv_end;
1120         }
1121         *val = cpu_to_le64(ROUND_UP(res_mem_end, 1 * GiB));
1122         fw_cfg_add_file(fw_cfg, "etc/reserved-memory-end", val, sizeof(*val));
1123     }
1124 
1125     if (linux_boot) {
1126         x86_load_linux(x86ms, fw_cfg, pcmc->acpi_data_size,
1127                        pcmc->pvh_enabled, pcmc->legacy_no_rng_seed);
1128     }
1129 
1130     for (i = 0; i < nb_option_roms; i++) {
1131         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1132     }
1133     x86ms->fw_cfg = fw_cfg;
1134 
1135     /* Init default IOAPIC address space */
1136     x86ms->ioapic_as = &address_space_memory;
1137 
1138     /* Init ACPI memory hotplug IO base address */
1139     pcms->memhp_io_base = ACPI_MEMORY_HOTPLUG_BASE;
1140 }
1141 
1142 /*
1143  * The 64bit pci hole starts after "above 4G RAM" and
1144  * potentially the space reserved for memory hotplug.
1145  */
1146 uint64_t pc_pci_hole64_start(void)
1147 {
1148     PCMachineState *pcms = PC_MACHINE(qdev_get_machine());
1149     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1150     MachineState *ms = MACHINE(pcms);
1151     uint64_t hole64_start = 0;
1152     ram_addr_t size = 0;
1153 
1154     if (pcms->cxl_devices_state.is_enabled) {
1155         hole64_start = pc_get_cxl_range_end(pcms);
1156     } else if (pcmc->has_reserved_memory && (ms->ram_size < ms->maxram_size)) {
1157         pc_get_device_memory_range(pcms, &hole64_start, &size);
1158         if (!pcmc->broken_reserved_end) {
1159             hole64_start += size;
1160         }
1161     } else {
1162         hole64_start = pc_above_4g_end(pcms);
1163     }
1164 
1165     return ROUND_UP(hole64_start, 1 * GiB);
1166 }
1167 
1168 DeviceState *pc_vga_init(ISABus *isa_bus, PCIBus *pci_bus)
1169 {
1170     DeviceState *dev = NULL;
1171 
1172     rom_set_order_override(FW_CFG_ORDER_OVERRIDE_VGA);
1173     if (pci_bus) {
1174         PCIDevice *pcidev = pci_vga_init(pci_bus);
1175         dev = pcidev ? &pcidev->qdev : NULL;
1176     } else if (isa_bus) {
1177         ISADevice *isadev = isa_vga_init(isa_bus);
1178         dev = isadev ? DEVICE(isadev) : NULL;
1179     }
1180     rom_reset_order_override();
1181     return dev;
1182 }
1183 
1184 static const MemoryRegionOps ioport80_io_ops = {
1185     .write = ioport80_write,
1186     .read = ioport80_read,
1187     .endianness = DEVICE_NATIVE_ENDIAN,
1188     .impl = {
1189         .min_access_size = 1,
1190         .max_access_size = 1,
1191     },
1192 };
1193 
1194 static const MemoryRegionOps ioportF0_io_ops = {
1195     .write = ioportF0_write,
1196     .read = ioportF0_read,
1197     .endianness = DEVICE_NATIVE_ENDIAN,
1198     .impl = {
1199         .min_access_size = 1,
1200         .max_access_size = 1,
1201     },
1202 };
1203 
1204 static void pc_superio_init(ISABus *isa_bus, bool create_fdctrl,
1205                             bool create_i8042, bool no_vmport)
1206 {
1207     int i;
1208     DriveInfo *fd[MAX_FD];
1209     qemu_irq *a20_line;
1210     ISADevice *fdc, *i8042, *port92, *vmmouse;
1211 
1212     serial_hds_isa_init(isa_bus, 0, MAX_ISA_SERIAL_PORTS);
1213     parallel_hds_isa_init(isa_bus, MAX_PARALLEL_PORTS);
1214 
1215     for (i = 0; i < MAX_FD; i++) {
1216         fd[i] = drive_get(IF_FLOPPY, 0, i);
1217         create_fdctrl |= !!fd[i];
1218     }
1219     if (create_fdctrl) {
1220         fdc = isa_new(TYPE_ISA_FDC);
1221         if (fdc) {
1222             isa_realize_and_unref(fdc, isa_bus, &error_fatal);
1223             isa_fdc_init_drives(fdc, fd);
1224         }
1225     }
1226 
1227     if (!create_i8042) {
1228         return;
1229     }
1230 
1231     i8042 = isa_create_simple(isa_bus, TYPE_I8042);
1232     if (!no_vmport) {
1233         isa_create_simple(isa_bus, TYPE_VMPORT);
1234         vmmouse = isa_try_new("vmmouse");
1235     } else {
1236         vmmouse = NULL;
1237     }
1238     if (vmmouse) {
1239         object_property_set_link(OBJECT(vmmouse), TYPE_I8042, OBJECT(i8042),
1240                                  &error_abort);
1241         isa_realize_and_unref(vmmouse, isa_bus, &error_fatal);
1242     }
1243     port92 = isa_create_simple(isa_bus, TYPE_PORT92);
1244 
1245     a20_line = qemu_allocate_irqs(handle_a20_line_change, first_cpu, 2);
1246     i8042_setup_a20_line(i8042, a20_line[0]);
1247     qdev_connect_gpio_out_named(DEVICE(port92),
1248                                 PORT92_A20_LINE, 0, a20_line[1]);
1249     g_free(a20_line);
1250 }
1251 
1252 void pc_basic_device_init(struct PCMachineState *pcms,
1253                           ISABus *isa_bus, qemu_irq *gsi,
1254                           ISADevice **rtc_state,
1255                           bool create_fdctrl,
1256                           uint32_t hpet_irqs)
1257 {
1258     int i;
1259     DeviceState *hpet = NULL;
1260     int pit_isa_irq = 0;
1261     qemu_irq pit_alt_irq = NULL;
1262     qemu_irq rtc_irq = NULL;
1263     ISADevice *pit = NULL;
1264     MemoryRegion *ioport80_io = g_new(MemoryRegion, 1);
1265     MemoryRegion *ioportF0_io = g_new(MemoryRegion, 1);
1266     X86MachineState *x86ms = X86_MACHINE(pcms);
1267 
1268     memory_region_init_io(ioport80_io, NULL, &ioport80_io_ops, NULL, "ioport80", 1);
1269     memory_region_add_subregion(isa_bus->address_space_io, 0x80, ioport80_io);
1270 
1271     memory_region_init_io(ioportF0_io, NULL, &ioportF0_io_ops, NULL, "ioportF0", 1);
1272     memory_region_add_subregion(isa_bus->address_space_io, 0xf0, ioportF0_io);
1273 
1274     /*
1275      * Check if an HPET shall be created.
1276      *
1277      * Without KVM_CAP_PIT_STATE2, we cannot switch off the in-kernel PIT
1278      * when the HPET wants to take over. Thus we have to disable the latter.
1279      */
1280     if (pcms->hpet_enabled && (!kvm_irqchip_in_kernel() ||
1281                                kvm_has_pit_state2())) {
1282         hpet = qdev_try_new(TYPE_HPET);
1283         if (!hpet) {
1284             error_report("couldn't create HPET device");
1285             exit(1);
1286         }
1287         /*
1288          * For pc-piix-*, hpet's intcap is always IRQ2. For pc-q35-1.7 and
1289          * earlier, use IRQ2 for compat. Otherwise, use IRQ16~23, IRQ8 and
1290          * IRQ2.
1291          */
1292         uint8_t compat = object_property_get_uint(OBJECT(hpet),
1293                 HPET_INTCAP, NULL);
1294         if (!compat) {
1295             qdev_prop_set_uint32(hpet, HPET_INTCAP, hpet_irqs);
1296         }
1297         sysbus_realize_and_unref(SYS_BUS_DEVICE(hpet), &error_fatal);
1298         sysbus_mmio_map(SYS_BUS_DEVICE(hpet), 0, HPET_BASE);
1299 
1300         for (i = 0; i < GSI_NUM_PINS; i++) {
1301             sysbus_connect_irq(SYS_BUS_DEVICE(hpet), i, gsi[i]);
1302         }
1303         pit_isa_irq = -1;
1304         pit_alt_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_PIT_INT);
1305         rtc_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_RTC_INT);
1306     }
1307     *rtc_state = mc146818_rtc_init(isa_bus, 2000, rtc_irq);
1308 
1309     qemu_register_boot_set(pc_boot_set, *rtc_state);
1310 
1311     if (!xen_enabled() &&
1312         (x86ms->pit == ON_OFF_AUTO_AUTO || x86ms->pit == ON_OFF_AUTO_ON)) {
1313         if (kvm_pit_in_kernel()) {
1314             pit = kvm_pit_init(isa_bus, 0x40);
1315         } else {
1316             pit = i8254_pit_init(isa_bus, 0x40, pit_isa_irq, pit_alt_irq);
1317         }
1318         if (hpet) {
1319             /* connect PIT to output control line of the HPET */
1320             qdev_connect_gpio_out(hpet, 0, qdev_get_gpio_in(DEVICE(pit), 0));
1321         }
1322         pcspk_init(pcms->pcspk, isa_bus, pit);
1323     }
1324 
1325     /* Super I/O */
1326     pc_superio_init(isa_bus, create_fdctrl, pcms->i8042_enabled,
1327                     pcms->vmport != ON_OFF_AUTO_ON);
1328 }
1329 
1330 void pc_nic_init(PCMachineClass *pcmc, ISABus *isa_bus, PCIBus *pci_bus)
1331 {
1332     int i;
1333 
1334     rom_set_order_override(FW_CFG_ORDER_OVERRIDE_NIC);
1335     for (i = 0; i < nb_nics; i++) {
1336         NICInfo *nd = &nd_table[i];
1337         const char *model = nd->model ? nd->model : pcmc->default_nic_model;
1338 
1339         if (g_str_equal(model, "ne2k_isa")) {
1340             pc_init_ne2k_isa(isa_bus, nd);
1341         } else {
1342             pci_nic_init_nofail(nd, pci_bus, model, NULL);
1343         }
1344     }
1345     rom_reset_order_override();
1346 }
1347 
1348 void pc_i8259_create(ISABus *isa_bus, qemu_irq *i8259_irqs)
1349 {
1350     qemu_irq *i8259;
1351 
1352     if (kvm_pic_in_kernel()) {
1353         i8259 = kvm_i8259_init(isa_bus);
1354     } else if (xen_enabled()) {
1355         i8259 = xen_interrupt_controller_init();
1356     } else {
1357         i8259 = i8259_init(isa_bus, x86_allocate_cpu_irq());
1358     }
1359 
1360     for (size_t i = 0; i < ISA_NUM_IRQS; i++) {
1361         i8259_irqs[i] = i8259[i];
1362     }
1363 
1364     g_free(i8259);
1365 }
1366 
1367 static void pc_memory_pre_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
1368                                Error **errp)
1369 {
1370     const PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1371     const X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
1372     const PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1373     const MachineState *ms = MACHINE(hotplug_dev);
1374     const bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
1375     const uint64_t legacy_align = TARGET_PAGE_SIZE;
1376     Error *local_err = NULL;
1377 
1378     /*
1379      * When -no-acpi is used with Q35 machine type, no ACPI is built,
1380      * but pcms->acpi_dev is still created. Check !acpi_enabled in
1381      * addition to cover this case.
1382      */
1383     if (!x86ms->acpi_dev || !x86_machine_is_acpi_enabled(x86ms)) {
1384         error_setg(errp,
1385                    "memory hotplug is not enabled: missing acpi device or acpi disabled");
1386         return;
1387     }
1388 
1389     if (is_nvdimm && !ms->nvdimms_state->is_enabled) {
1390         error_setg(errp, "nvdimm is not enabled: missing 'nvdimm' in '-M'");
1391         return;
1392     }
1393 
1394     hotplug_handler_pre_plug(x86ms->acpi_dev, dev, &local_err);
1395     if (local_err) {
1396         error_propagate(errp, local_err);
1397         return;
1398     }
1399 
1400     pc_dimm_pre_plug(PC_DIMM(dev), MACHINE(hotplug_dev),
1401                      pcmc->enforce_aligned_dimm ? NULL : &legacy_align, errp);
1402 }
1403 
1404 static void pc_memory_plug(HotplugHandler *hotplug_dev,
1405                            DeviceState *dev, Error **errp)
1406 {
1407     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1408     X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
1409     MachineState *ms = MACHINE(hotplug_dev);
1410     bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
1411 
1412     pc_dimm_plug(PC_DIMM(dev), MACHINE(pcms));
1413 
1414     if (is_nvdimm) {
1415         nvdimm_plug(ms->nvdimms_state);
1416     }
1417 
1418     hotplug_handler_plug(x86ms->acpi_dev, dev, &error_abort);
1419 }
1420 
1421 static void pc_memory_unplug_request(HotplugHandler *hotplug_dev,
1422                                      DeviceState *dev, Error **errp)
1423 {
1424     X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
1425 
1426     /*
1427      * When -no-acpi is used with Q35 machine type, no ACPI is built,
1428      * but pcms->acpi_dev is still created. Check !acpi_enabled in
1429      * addition to cover this case.
1430      */
1431     if (!x86ms->acpi_dev || !x86_machine_is_acpi_enabled(x86ms)) {
1432         error_setg(errp,
1433                    "memory hotplug is not enabled: missing acpi device or acpi disabled");
1434         return;
1435     }
1436 
1437     if (object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM)) {
1438         error_setg(errp, "nvdimm device hot unplug is not supported yet.");
1439         return;
1440     }
1441 
1442     hotplug_handler_unplug_request(x86ms->acpi_dev, dev,
1443                                    errp);
1444 }
1445 
1446 static void pc_memory_unplug(HotplugHandler *hotplug_dev,
1447                              DeviceState *dev, Error **errp)
1448 {
1449     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1450     X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
1451     Error *local_err = NULL;
1452 
1453     hotplug_handler_unplug(x86ms->acpi_dev, dev, &local_err);
1454     if (local_err) {
1455         goto out;
1456     }
1457 
1458     pc_dimm_unplug(PC_DIMM(dev), MACHINE(pcms));
1459     qdev_unrealize(dev);
1460  out:
1461     error_propagate(errp, local_err);
1462 }
1463 
1464 static void pc_virtio_md_pci_pre_plug(HotplugHandler *hotplug_dev,
1465                                       DeviceState *dev, Error **errp)
1466 {
1467     HotplugHandler *hotplug_dev2 = qdev_get_bus_hotplug_handler(dev);
1468     Error *local_err = NULL;
1469 
1470     if (!hotplug_dev2 && dev->hotplugged) {
1471         /*
1472          * Without a bus hotplug handler, we cannot control the plug/unplug
1473          * order. We should never reach this point when hotplugging on x86,
1474          * however, better add a safety net.
1475          */
1476         error_setg(errp, "hotplug of virtio based memory devices not supported"
1477                    " on this bus.");
1478         return;
1479     }
1480     /*
1481      * First, see if we can plug this memory device at all. If that
1482      * succeeds, branch of to the actual hotplug handler.
1483      */
1484     memory_device_pre_plug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev), NULL,
1485                            &local_err);
1486     if (!local_err && hotplug_dev2) {
1487         hotplug_handler_pre_plug(hotplug_dev2, dev, &local_err);
1488     }
1489     error_propagate(errp, local_err);
1490 }
1491 
1492 static void pc_virtio_md_pci_plug(HotplugHandler *hotplug_dev,
1493                                   DeviceState *dev, Error **errp)
1494 {
1495     HotplugHandler *hotplug_dev2 = qdev_get_bus_hotplug_handler(dev);
1496     Error *local_err = NULL;
1497 
1498     /*
1499      * Plug the memory device first and then branch off to the actual
1500      * hotplug handler. If that one fails, we can easily undo the memory
1501      * device bits.
1502      */
1503     memory_device_plug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev));
1504     if (hotplug_dev2) {
1505         hotplug_handler_plug(hotplug_dev2, dev, &local_err);
1506         if (local_err) {
1507             memory_device_unplug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev));
1508         }
1509     }
1510     error_propagate(errp, local_err);
1511 }
1512 
1513 static void pc_virtio_md_pci_unplug_request(HotplugHandler *hotplug_dev,
1514                                             DeviceState *dev, Error **errp)
1515 {
1516     /* We don't support hot unplug of virtio based memory devices */
1517     error_setg(errp, "virtio based memory devices cannot be unplugged.");
1518 }
1519 
1520 static void pc_virtio_md_pci_unplug(HotplugHandler *hotplug_dev,
1521                                     DeviceState *dev, Error **errp)
1522 {
1523     /* We don't support hot unplug of virtio based memory devices */
1524 }
1525 
1526 static void pc_machine_device_pre_plug_cb(HotplugHandler *hotplug_dev,
1527                                           DeviceState *dev, Error **errp)
1528 {
1529     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1530         pc_memory_pre_plug(hotplug_dev, dev, errp);
1531     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1532         x86_cpu_pre_plug(hotplug_dev, dev, errp);
1533     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI) ||
1534                object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI)) {
1535         pc_virtio_md_pci_pre_plug(hotplug_dev, dev, errp);
1536     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
1537         /* Declare the APIC range as the reserved MSI region */
1538         char *resv_prop_str = g_strdup_printf("0xfee00000:0xfeefffff:%d",
1539                                               VIRTIO_IOMMU_RESV_MEM_T_MSI);
1540 
1541         object_property_set_uint(OBJECT(dev), "len-reserved-regions", 1, errp);
1542         object_property_set_str(OBJECT(dev), "reserved-regions[0]",
1543                                 resv_prop_str, errp);
1544         g_free(resv_prop_str);
1545     }
1546 
1547     if (object_dynamic_cast(OBJECT(dev), TYPE_X86_IOMMU_DEVICE) ||
1548         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
1549         PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1550 
1551         if (pcms->iommu) {
1552             error_setg(errp, "QEMU does not support multiple vIOMMUs "
1553                        "for x86 yet.");
1554             return;
1555         }
1556         pcms->iommu = dev;
1557     }
1558 }
1559 
1560 static void pc_machine_device_plug_cb(HotplugHandler *hotplug_dev,
1561                                       DeviceState *dev, Error **errp)
1562 {
1563     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1564         pc_memory_plug(hotplug_dev, dev, errp);
1565     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1566         x86_cpu_plug(hotplug_dev, dev, errp);
1567     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI) ||
1568                object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI)) {
1569         pc_virtio_md_pci_plug(hotplug_dev, dev, errp);
1570     }
1571 }
1572 
1573 static void pc_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
1574                                                 DeviceState *dev, Error **errp)
1575 {
1576     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1577         pc_memory_unplug_request(hotplug_dev, dev, errp);
1578     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1579         x86_cpu_unplug_request_cb(hotplug_dev, dev, errp);
1580     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI) ||
1581                object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI)) {
1582         pc_virtio_md_pci_unplug_request(hotplug_dev, dev, errp);
1583     } else {
1584         error_setg(errp, "acpi: device unplug request for not supported device"
1585                    " type: %s", object_get_typename(OBJECT(dev)));
1586     }
1587 }
1588 
1589 static void pc_machine_device_unplug_cb(HotplugHandler *hotplug_dev,
1590                                         DeviceState *dev, Error **errp)
1591 {
1592     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1593         pc_memory_unplug(hotplug_dev, dev, errp);
1594     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1595         x86_cpu_unplug_cb(hotplug_dev, dev, errp);
1596     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI) ||
1597                object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI)) {
1598         pc_virtio_md_pci_unplug(hotplug_dev, dev, errp);
1599     } else {
1600         error_setg(errp, "acpi: device unplug for not supported device"
1601                    " type: %s", object_get_typename(OBJECT(dev)));
1602     }
1603 }
1604 
1605 static HotplugHandler *pc_get_hotplug_handler(MachineState *machine,
1606                                              DeviceState *dev)
1607 {
1608     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) ||
1609         object_dynamic_cast(OBJECT(dev), TYPE_CPU) ||
1610         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI) ||
1611         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI) ||
1612         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI) ||
1613         object_dynamic_cast(OBJECT(dev), TYPE_X86_IOMMU_DEVICE)) {
1614         return HOTPLUG_HANDLER(machine);
1615     }
1616 
1617     return NULL;
1618 }
1619 
1620 static void
1621 pc_machine_get_device_memory_region_size(Object *obj, Visitor *v,
1622                                          const char *name, void *opaque,
1623                                          Error **errp)
1624 {
1625     MachineState *ms = MACHINE(obj);
1626     int64_t value = 0;
1627 
1628     if (ms->device_memory) {
1629         value = memory_region_size(&ms->device_memory->mr);
1630     }
1631 
1632     visit_type_int(v, name, &value, errp);
1633 }
1634 
1635 static void pc_machine_get_vmport(Object *obj, Visitor *v, const char *name,
1636                                   void *opaque, Error **errp)
1637 {
1638     PCMachineState *pcms = PC_MACHINE(obj);
1639     OnOffAuto vmport = pcms->vmport;
1640 
1641     visit_type_OnOffAuto(v, name, &vmport, errp);
1642 }
1643 
1644 static void pc_machine_set_vmport(Object *obj, Visitor *v, const char *name,
1645                                   void *opaque, Error **errp)
1646 {
1647     PCMachineState *pcms = PC_MACHINE(obj);
1648 
1649     visit_type_OnOffAuto(v, name, &pcms->vmport, errp);
1650 }
1651 
1652 static bool pc_machine_get_smbus(Object *obj, Error **errp)
1653 {
1654     PCMachineState *pcms = PC_MACHINE(obj);
1655 
1656     return pcms->smbus_enabled;
1657 }
1658 
1659 static void pc_machine_set_smbus(Object *obj, bool value, Error **errp)
1660 {
1661     PCMachineState *pcms = PC_MACHINE(obj);
1662 
1663     pcms->smbus_enabled = value;
1664 }
1665 
1666 static bool pc_machine_get_sata(Object *obj, Error **errp)
1667 {
1668     PCMachineState *pcms = PC_MACHINE(obj);
1669 
1670     return pcms->sata_enabled;
1671 }
1672 
1673 static void pc_machine_set_sata(Object *obj, bool value, Error **errp)
1674 {
1675     PCMachineState *pcms = PC_MACHINE(obj);
1676 
1677     pcms->sata_enabled = value;
1678 }
1679 
1680 static bool pc_machine_get_hpet(Object *obj, Error **errp)
1681 {
1682     PCMachineState *pcms = PC_MACHINE(obj);
1683 
1684     return pcms->hpet_enabled;
1685 }
1686 
1687 static void pc_machine_set_hpet(Object *obj, bool value, Error **errp)
1688 {
1689     PCMachineState *pcms = PC_MACHINE(obj);
1690 
1691     pcms->hpet_enabled = value;
1692 }
1693 
1694 static bool pc_machine_get_i8042(Object *obj, Error **errp)
1695 {
1696     PCMachineState *pcms = PC_MACHINE(obj);
1697 
1698     return pcms->i8042_enabled;
1699 }
1700 
1701 static void pc_machine_set_i8042(Object *obj, bool value, Error **errp)
1702 {
1703     PCMachineState *pcms = PC_MACHINE(obj);
1704 
1705     pcms->i8042_enabled = value;
1706 }
1707 
1708 static bool pc_machine_get_default_bus_bypass_iommu(Object *obj, Error **errp)
1709 {
1710     PCMachineState *pcms = PC_MACHINE(obj);
1711 
1712     return pcms->default_bus_bypass_iommu;
1713 }
1714 
1715 static void pc_machine_set_default_bus_bypass_iommu(Object *obj, bool value,
1716                                                     Error **errp)
1717 {
1718     PCMachineState *pcms = PC_MACHINE(obj);
1719 
1720     pcms->default_bus_bypass_iommu = value;
1721 }
1722 
1723 static void pc_machine_get_smbios_ep(Object *obj, Visitor *v, const char *name,
1724                                      void *opaque, Error **errp)
1725 {
1726     PCMachineState *pcms = PC_MACHINE(obj);
1727     SmbiosEntryPointType smbios_entry_point_type = pcms->smbios_entry_point_type;
1728 
1729     visit_type_SmbiosEntryPointType(v, name, &smbios_entry_point_type, errp);
1730 }
1731 
1732 static void pc_machine_set_smbios_ep(Object *obj, Visitor *v, const char *name,
1733                                      void *opaque, Error **errp)
1734 {
1735     PCMachineState *pcms = PC_MACHINE(obj);
1736 
1737     visit_type_SmbiosEntryPointType(v, name, &pcms->smbios_entry_point_type, errp);
1738 }
1739 
1740 static void pc_machine_get_max_ram_below_4g(Object *obj, Visitor *v,
1741                                             const char *name, void *opaque,
1742                                             Error **errp)
1743 {
1744     PCMachineState *pcms = PC_MACHINE(obj);
1745     uint64_t value = pcms->max_ram_below_4g;
1746 
1747     visit_type_size(v, name, &value, errp);
1748 }
1749 
1750 static void pc_machine_set_max_ram_below_4g(Object *obj, Visitor *v,
1751                                             const char *name, void *opaque,
1752                                             Error **errp)
1753 {
1754     PCMachineState *pcms = PC_MACHINE(obj);
1755     uint64_t value;
1756 
1757     if (!visit_type_size(v, name, &value, errp)) {
1758         return;
1759     }
1760     if (value > 4 * GiB) {
1761         error_setg(errp,
1762                    "Machine option 'max-ram-below-4g=%"PRIu64
1763                    "' expects size less than or equal to 4G", value);
1764         return;
1765     }
1766 
1767     if (value < 1 * MiB) {
1768         warn_report("Only %" PRIu64 " bytes of RAM below the 4GiB boundary,"
1769                     "BIOS may not work with less than 1MiB", value);
1770     }
1771 
1772     pcms->max_ram_below_4g = value;
1773 }
1774 
1775 static void pc_machine_get_max_fw_size(Object *obj, Visitor *v,
1776                                        const char *name, void *opaque,
1777                                        Error **errp)
1778 {
1779     PCMachineState *pcms = PC_MACHINE(obj);
1780     uint64_t value = pcms->max_fw_size;
1781 
1782     visit_type_size(v, name, &value, errp);
1783 }
1784 
1785 static void pc_machine_set_max_fw_size(Object *obj, Visitor *v,
1786                                        const char *name, void *opaque,
1787                                        Error **errp)
1788 {
1789     PCMachineState *pcms = PC_MACHINE(obj);
1790     uint64_t value;
1791 
1792     if (!visit_type_size(v, name, &value, errp)) {
1793         return;
1794     }
1795 
1796     /*
1797     * We don't have a theoretically justifiable exact lower bound on the base
1798     * address of any flash mapping. In practice, the IO-APIC MMIO range is
1799     * [0xFEE00000..0xFEE01000] -- see IO_APIC_DEFAULT_ADDRESS --, leaving free
1800     * only 18MB-4KB below 4G. For now, restrict the cumulative mapping to 8MB in
1801     * size.
1802     */
1803     if (value > 16 * MiB) {
1804         error_setg(errp,
1805                    "User specified max allowed firmware size %" PRIu64 " is "
1806                    "greater than 16MiB. If combined firwmare size exceeds "
1807                    "16MiB the system may not boot, or experience intermittent"
1808                    "stability issues.",
1809                    value);
1810         return;
1811     }
1812 
1813     pcms->max_fw_size = value;
1814 }
1815 
1816 
1817 static void pc_machine_initfn(Object *obj)
1818 {
1819     PCMachineState *pcms = PC_MACHINE(obj);
1820 
1821 #ifdef CONFIG_VMPORT
1822     pcms->vmport = ON_OFF_AUTO_AUTO;
1823 #else
1824     pcms->vmport = ON_OFF_AUTO_OFF;
1825 #endif /* CONFIG_VMPORT */
1826     pcms->max_ram_below_4g = 0; /* use default */
1827     pcms->smbios_entry_point_type = SMBIOS_ENTRY_POINT_TYPE_32;
1828 
1829     /* acpi build is enabled by default if machine supports it */
1830     pcms->acpi_build_enabled = PC_MACHINE_GET_CLASS(pcms)->has_acpi_build;
1831     pcms->smbus_enabled = true;
1832     pcms->sata_enabled = true;
1833     pcms->i8042_enabled = true;
1834     pcms->max_fw_size = 8 * MiB;
1835 #ifdef CONFIG_HPET
1836     pcms->hpet_enabled = true;
1837 #endif
1838     pcms->default_bus_bypass_iommu = false;
1839 
1840     pc_system_flash_create(pcms);
1841     pcms->pcspk = isa_new(TYPE_PC_SPEAKER);
1842     object_property_add_alias(OBJECT(pcms), "pcspk-audiodev",
1843                               OBJECT(pcms->pcspk), "audiodev");
1844     cxl_machine_init(obj, &pcms->cxl_devices_state);
1845 }
1846 
1847 static void pc_machine_reset(MachineState *machine, ShutdownCause reason)
1848 {
1849     CPUState *cs;
1850     X86CPU *cpu;
1851 
1852     qemu_devices_reset(reason);
1853 
1854     /* Reset APIC after devices have been reset to cancel
1855      * any changes that qemu_devices_reset() might have done.
1856      */
1857     CPU_FOREACH(cs) {
1858         cpu = X86_CPU(cs);
1859 
1860         x86_cpu_after_reset(cpu);
1861     }
1862 }
1863 
1864 static void pc_machine_wakeup(MachineState *machine)
1865 {
1866     cpu_synchronize_all_states();
1867     pc_machine_reset(machine, SHUTDOWN_CAUSE_NONE);
1868     cpu_synchronize_all_post_reset();
1869 }
1870 
1871 static bool pc_hotplug_allowed(MachineState *ms, DeviceState *dev, Error **errp)
1872 {
1873     X86IOMMUState *iommu = x86_iommu_get_default();
1874     IntelIOMMUState *intel_iommu;
1875 
1876     if (iommu &&
1877         object_dynamic_cast((Object *)iommu, TYPE_INTEL_IOMMU_DEVICE) &&
1878         object_dynamic_cast((Object *)dev, "vfio-pci")) {
1879         intel_iommu = INTEL_IOMMU_DEVICE(iommu);
1880         if (!intel_iommu->caching_mode) {
1881             error_setg(errp, "Device assignment is not allowed without "
1882                        "enabling caching-mode=on for Intel IOMMU.");
1883             return false;
1884         }
1885     }
1886 
1887     return true;
1888 }
1889 
1890 static void pc_machine_class_init(ObjectClass *oc, void *data)
1891 {
1892     MachineClass *mc = MACHINE_CLASS(oc);
1893     PCMachineClass *pcmc = PC_MACHINE_CLASS(oc);
1894     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
1895 
1896     pcmc->pci_enabled = true;
1897     pcmc->has_acpi_build = true;
1898     pcmc->rsdp_in_ram = true;
1899     pcmc->smbios_defaults = true;
1900     pcmc->smbios_uuid_encoded = true;
1901     pcmc->gigabyte_align = true;
1902     pcmc->has_reserved_memory = true;
1903     pcmc->kvmclock_enabled = true;
1904     pcmc->enforce_aligned_dimm = true;
1905     pcmc->enforce_amd_1tb_hole = true;
1906     /* BIOS ACPI tables: 128K. Other BIOS datastructures: less than 4K reported
1907      * to be used at the moment, 32K should be enough for a while.  */
1908     pcmc->acpi_data_size = 0x20000 + 0x8000;
1909     pcmc->pvh_enabled = true;
1910     pcmc->kvmclock_create_always = true;
1911     assert(!mc->get_hotplug_handler);
1912     mc->get_hotplug_handler = pc_get_hotplug_handler;
1913     mc->hotplug_allowed = pc_hotplug_allowed;
1914     mc->cpu_index_to_instance_props = x86_cpu_index_to_props;
1915     mc->get_default_cpu_node_id = x86_get_default_cpu_node_id;
1916     mc->possible_cpu_arch_ids = x86_possible_cpu_arch_ids;
1917     mc->auto_enable_numa_with_memhp = true;
1918     mc->auto_enable_numa_with_memdev = true;
1919     mc->has_hotpluggable_cpus = true;
1920     mc->default_boot_order = "cad";
1921     mc->block_default_type = IF_IDE;
1922     mc->max_cpus = 255;
1923     mc->reset = pc_machine_reset;
1924     mc->wakeup = pc_machine_wakeup;
1925     hc->pre_plug = pc_machine_device_pre_plug_cb;
1926     hc->plug = pc_machine_device_plug_cb;
1927     hc->unplug_request = pc_machine_device_unplug_request_cb;
1928     hc->unplug = pc_machine_device_unplug_cb;
1929     mc->default_cpu_type = TARGET_DEFAULT_CPU_TYPE;
1930     mc->nvdimm_supported = true;
1931     mc->smp_props.dies_supported = true;
1932     mc->default_ram_id = "pc.ram";
1933 
1934     object_class_property_add(oc, PC_MACHINE_MAX_RAM_BELOW_4G, "size",
1935         pc_machine_get_max_ram_below_4g, pc_machine_set_max_ram_below_4g,
1936         NULL, NULL);
1937     object_class_property_set_description(oc, PC_MACHINE_MAX_RAM_BELOW_4G,
1938         "Maximum ram below the 4G boundary (32bit boundary)");
1939 
1940     object_class_property_add(oc, PC_MACHINE_DEVMEM_REGION_SIZE, "int",
1941         pc_machine_get_device_memory_region_size, NULL,
1942         NULL, NULL);
1943 
1944     object_class_property_add(oc, PC_MACHINE_VMPORT, "OnOffAuto",
1945         pc_machine_get_vmport, pc_machine_set_vmport,
1946         NULL, NULL);
1947     object_class_property_set_description(oc, PC_MACHINE_VMPORT,
1948         "Enable vmport (pc & q35)");
1949 
1950     object_class_property_add_bool(oc, PC_MACHINE_SMBUS,
1951         pc_machine_get_smbus, pc_machine_set_smbus);
1952     object_class_property_set_description(oc, PC_MACHINE_SMBUS,
1953         "Enable/disable system management bus");
1954 
1955     object_class_property_add_bool(oc, PC_MACHINE_SATA,
1956         pc_machine_get_sata, pc_machine_set_sata);
1957     object_class_property_set_description(oc, PC_MACHINE_SATA,
1958         "Enable/disable Serial ATA bus");
1959 
1960     object_class_property_add_bool(oc, "hpet",
1961         pc_machine_get_hpet, pc_machine_set_hpet);
1962     object_class_property_set_description(oc, "hpet",
1963         "Enable/disable high precision event timer emulation");
1964 
1965     object_class_property_add_bool(oc, PC_MACHINE_I8042,
1966         pc_machine_get_i8042, pc_machine_set_i8042);
1967 
1968     object_class_property_add_bool(oc, "default-bus-bypass-iommu",
1969         pc_machine_get_default_bus_bypass_iommu,
1970         pc_machine_set_default_bus_bypass_iommu);
1971 
1972     object_class_property_add(oc, PC_MACHINE_MAX_FW_SIZE, "size",
1973         pc_machine_get_max_fw_size, pc_machine_set_max_fw_size,
1974         NULL, NULL);
1975     object_class_property_set_description(oc, PC_MACHINE_MAX_FW_SIZE,
1976         "Maximum combined firmware size");
1977 
1978     object_class_property_add(oc, PC_MACHINE_SMBIOS_EP, "str",
1979         pc_machine_get_smbios_ep, pc_machine_set_smbios_ep,
1980         NULL, NULL);
1981     object_class_property_set_description(oc, PC_MACHINE_SMBIOS_EP,
1982         "SMBIOS Entry Point type [32, 64]");
1983 }
1984 
1985 static const TypeInfo pc_machine_info = {
1986     .name = TYPE_PC_MACHINE,
1987     .parent = TYPE_X86_MACHINE,
1988     .abstract = true,
1989     .instance_size = sizeof(PCMachineState),
1990     .instance_init = pc_machine_initfn,
1991     .class_size = sizeof(PCMachineClass),
1992     .class_init = pc_machine_class_init,
1993     .interfaces = (InterfaceInfo[]) {
1994          { TYPE_HOTPLUG_HANDLER },
1995          { }
1996     },
1997 };
1998 
1999 static void pc_machine_register_types(void)
2000 {
2001     type_register_static(&pc_machine_info);
2002 }
2003 
2004 type_init(pc_machine_register_types)
2005