xref: /qemu/hw/i386/pc.c (revision e995d5cc)
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/topology.h"
32 #include "hw/i386/fw_cfg.h"
33 #include "hw/i386/vmport.h"
34 #include "sysemu/cpus.h"
35 #include "hw/block/fdc.h"
36 #include "hw/ide/internal.h"
37 #include "hw/ide/isa.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/intc/ioapic.h"
51 #include "hw/timer/i8254.h"
52 #include "hw/input/i8042.h"
53 #include "hw/irq.h"
54 #include "hw/audio/pcspk.h"
55 #include "hw/pci/msi.h"
56 #include "hw/sysbus.h"
57 #include "sysemu/sysemu.h"
58 #include "sysemu/tcg.h"
59 #include "sysemu/numa.h"
60 #include "sysemu/kvm.h"
61 #include "sysemu/xen.h"
62 #include "sysemu/reset.h"
63 #include "sysemu/runstate.h"
64 #include "kvm/kvm_i386.h"
65 #include "hw/xen/xen.h"
66 #include "hw/xen/start_info.h"
67 #include "ui/qemu-spice.h"
68 #include "exec/memory.h"
69 #include "qemu/bitmap.h"
70 #include "qemu/config-file.h"
71 #include "qemu/error-report.h"
72 #include "qemu/option.h"
73 #include "qemu/cutils.h"
74 #include "hw/acpi/acpi.h"
75 #include "hw/acpi/cpu_hotplug.h"
76 #include "acpi-build.h"
77 #include "hw/mem/pc-dimm.h"
78 #include "hw/mem/nvdimm.h"
79 #include "hw/cxl/cxl.h"
80 #include "hw/cxl/cxl_host.h"
81 #include "qapi/error.h"
82 #include "qapi/qapi-visit-common.h"
83 #include "qapi/qapi-visit-machine.h"
84 #include "qapi/visitor.h"
85 #include "hw/core/cpu.h"
86 #include "hw/usb.h"
87 #include "hw/i386/intel_iommu.h"
88 #include "hw/net/ne2000-isa.h"
89 #include "standard-headers/asm-x86/bootparam.h"
90 #include "hw/virtio/virtio-iommu.h"
91 #include "hw/virtio/virtio-pmem-pci.h"
92 #include "hw/virtio/virtio-mem-pci.h"
93 #include "hw/i386/kvm/xen_overlay.h"
94 #include "hw/i386/kvm/xen_evtchn.h"
95 #include "hw/i386/kvm/xen_gnttab.h"
96 #include "hw/i386/kvm/xen_xenstore.h"
97 #include "hw/mem/memory-device.h"
98 #include "sysemu/replay.h"
99 #include "target/i386/cpu.h"
100 #include "e820_memory_layout.h"
101 #include "fw_cfg.h"
102 #include "trace.h"
103 #include CONFIG_DEVICES
104 
105 /*
106  * Helper for setting model-id for CPU models that changed model-id
107  * depending on QEMU versions up to QEMU 2.4.
108  */
109 #define PC_CPU_MODEL_IDS(v) \
110     { "qemu32-" TYPE_X86_CPU, "model-id", "QEMU Virtual CPU version " v, },\
111     { "qemu64-" TYPE_X86_CPU, "model-id", "QEMU Virtual CPU version " v, },\
112     { "athlon-" TYPE_X86_CPU, "model-id", "QEMU Virtual CPU version " v, },
113 
114 GlobalProperty pc_compat_7_2[] = {
115     { "ICH9-LPC", "noreboot", "true" },
116 };
117 const size_t pc_compat_7_2_len = G_N_ELEMENTS(pc_compat_7_2);
118 
119 GlobalProperty pc_compat_7_1[] = {};
120 const size_t pc_compat_7_1_len = G_N_ELEMENTS(pc_compat_7_1);
121 
122 GlobalProperty pc_compat_7_0[] = {};
123 const size_t pc_compat_7_0_len = G_N_ELEMENTS(pc_compat_7_0);
124 
125 GlobalProperty pc_compat_6_2[] = {
126     { "virtio-mem", "unplugged-inaccessible", "off" },
127 };
128 const size_t pc_compat_6_2_len = G_N_ELEMENTS(pc_compat_6_2);
129 
130 GlobalProperty pc_compat_6_1[] = {
131     { TYPE_X86_CPU, "hv-version-id-build", "0x1bbc" },
132     { TYPE_X86_CPU, "hv-version-id-major", "0x0006" },
133     { TYPE_X86_CPU, "hv-version-id-minor", "0x0001" },
134     { "ICH9-LPC", "x-keep-pci-slot-hpc", "false" },
135 };
136 const size_t pc_compat_6_1_len = G_N_ELEMENTS(pc_compat_6_1);
137 
138 GlobalProperty pc_compat_6_0[] = {
139     { "qemu64" "-" TYPE_X86_CPU, "family", "6" },
140     { "qemu64" "-" TYPE_X86_CPU, "model", "6" },
141     { "qemu64" "-" TYPE_X86_CPU, "stepping", "3" },
142     { TYPE_X86_CPU, "x-vendor-cpuid-only", "off" },
143     { "ICH9-LPC", ACPI_PM_PROP_ACPI_PCIHP_BRIDGE, "off" },
144     { "ICH9-LPC", "x-keep-pci-slot-hpc", "true" },
145 };
146 const size_t pc_compat_6_0_len = G_N_ELEMENTS(pc_compat_6_0);
147 
148 GlobalProperty pc_compat_5_2[] = {
149     { "ICH9-LPC", "x-smi-cpu-hotunplug", "off" },
150 };
151 const size_t pc_compat_5_2_len = G_N_ELEMENTS(pc_compat_5_2);
152 
153 GlobalProperty pc_compat_5_1[] = {
154     { "ICH9-LPC", "x-smi-cpu-hotplug", "off" },
155     { TYPE_X86_CPU, "kvm-msi-ext-dest-id", "off" },
156 };
157 const size_t pc_compat_5_1_len = G_N_ELEMENTS(pc_compat_5_1);
158 
159 GlobalProperty pc_compat_5_0[] = {
160 };
161 const size_t pc_compat_5_0_len = G_N_ELEMENTS(pc_compat_5_0);
162 
163 GlobalProperty pc_compat_4_2[] = {
164     { "mch", "smbase-smram", "off" },
165 };
166 const size_t pc_compat_4_2_len = G_N_ELEMENTS(pc_compat_4_2);
167 
168 GlobalProperty pc_compat_4_1[] = {};
169 const size_t pc_compat_4_1_len = G_N_ELEMENTS(pc_compat_4_1);
170 
171 GlobalProperty pc_compat_4_0[] = {};
172 const size_t pc_compat_4_0_len = G_N_ELEMENTS(pc_compat_4_0);
173 
174 GlobalProperty pc_compat_3_1[] = {
175     { "intel-iommu", "dma-drain", "off" },
176     { "Opteron_G3" "-" TYPE_X86_CPU, "rdtscp", "off" },
177     { "Opteron_G4" "-" TYPE_X86_CPU, "rdtscp", "off" },
178     { "Opteron_G4" "-" TYPE_X86_CPU, "npt", "off" },
179     { "Opteron_G4" "-" TYPE_X86_CPU, "nrip-save", "off" },
180     { "Opteron_G5" "-" TYPE_X86_CPU, "rdtscp", "off" },
181     { "Opteron_G5" "-" TYPE_X86_CPU, "npt", "off" },
182     { "Opteron_G5" "-" TYPE_X86_CPU, "nrip-save", "off" },
183     { "EPYC" "-" TYPE_X86_CPU, "npt", "off" },
184     { "EPYC" "-" TYPE_X86_CPU, "nrip-save", "off" },
185     { "EPYC-IBPB" "-" TYPE_X86_CPU, "npt", "off" },
186     { "EPYC-IBPB" "-" TYPE_X86_CPU, "nrip-save", "off" },
187     { "Skylake-Client" "-" TYPE_X86_CPU,      "mpx", "on" },
188     { "Skylake-Client-IBRS" "-" TYPE_X86_CPU, "mpx", "on" },
189     { "Skylake-Server" "-" TYPE_X86_CPU,      "mpx", "on" },
190     { "Skylake-Server-IBRS" "-" TYPE_X86_CPU, "mpx", "on" },
191     { "Cascadelake-Server" "-" TYPE_X86_CPU,  "mpx", "on" },
192     { "Icelake-Client" "-" TYPE_X86_CPU,      "mpx", "on" },
193     { "Icelake-Server" "-" TYPE_X86_CPU,      "mpx", "on" },
194     { "Cascadelake-Server" "-" TYPE_X86_CPU, "stepping", "5" },
195     { TYPE_X86_CPU, "x-intel-pt-auto-level", "off" },
196 };
197 const size_t pc_compat_3_1_len = G_N_ELEMENTS(pc_compat_3_1);
198 
199 GlobalProperty pc_compat_3_0[] = {
200     { TYPE_X86_CPU, "x-hv-synic-kvm-only", "on" },
201     { "Skylake-Server" "-" TYPE_X86_CPU, "pku", "off" },
202     { "Skylake-Server-IBRS" "-" TYPE_X86_CPU, "pku", "off" },
203 };
204 const size_t pc_compat_3_0_len = G_N_ELEMENTS(pc_compat_3_0);
205 
206 GlobalProperty pc_compat_2_12[] = {
207     { TYPE_X86_CPU, "legacy-cache", "on" },
208     { TYPE_X86_CPU, "topoext", "off" },
209     { "EPYC-" TYPE_X86_CPU, "xlevel", "0x8000000a" },
210     { "EPYC-IBPB-" TYPE_X86_CPU, "xlevel", "0x8000000a" },
211 };
212 const size_t pc_compat_2_12_len = G_N_ELEMENTS(pc_compat_2_12);
213 
214 GlobalProperty pc_compat_2_11[] = {
215     { TYPE_X86_CPU, "x-migrate-smi-count", "off" },
216     { "Skylake-Server" "-" TYPE_X86_CPU, "clflushopt", "off" },
217 };
218 const size_t pc_compat_2_11_len = G_N_ELEMENTS(pc_compat_2_11);
219 
220 GlobalProperty pc_compat_2_10[] = {
221     { TYPE_X86_CPU, "x-hv-max-vps", "0x40" },
222     { "i440FX-pcihost", "x-pci-hole64-fix", "off" },
223     { "q35-pcihost", "x-pci-hole64-fix", "off" },
224 };
225 const size_t pc_compat_2_10_len = G_N_ELEMENTS(pc_compat_2_10);
226 
227 GlobalProperty pc_compat_2_9[] = {
228     { "mch", "extended-tseg-mbytes", "0" },
229 };
230 const size_t pc_compat_2_9_len = G_N_ELEMENTS(pc_compat_2_9);
231 
232 GlobalProperty pc_compat_2_8[] = {
233     { TYPE_X86_CPU, "tcg-cpuid", "off" },
234     { "kvmclock", "x-mach-use-reliable-get-clock", "off" },
235     { "ICH9-LPC", "x-smi-broadcast", "off" },
236     { TYPE_X86_CPU, "vmware-cpuid-freq", "off" },
237     { "Haswell-" TYPE_X86_CPU, "stepping", "1" },
238 };
239 const size_t pc_compat_2_8_len = G_N_ELEMENTS(pc_compat_2_8);
240 
241 GlobalProperty pc_compat_2_7[] = {
242     { TYPE_X86_CPU, "l3-cache", "off" },
243     { TYPE_X86_CPU, "full-cpuid-auto-level", "off" },
244     { "Opteron_G3" "-" TYPE_X86_CPU, "family", "15" },
245     { "Opteron_G3" "-" TYPE_X86_CPU, "model", "6" },
246     { "Opteron_G3" "-" TYPE_X86_CPU, "stepping", "1" },
247     { "isa-pcspk", "migrate", "off" },
248 };
249 const size_t pc_compat_2_7_len = G_N_ELEMENTS(pc_compat_2_7);
250 
251 GlobalProperty pc_compat_2_6[] = {
252     { TYPE_X86_CPU, "cpuid-0xb", "off" },
253     { "vmxnet3", "romfile", "" },
254     { TYPE_X86_CPU, "fill-mtrr-mask", "off" },
255     { "apic-common", "legacy-instance-id", "on", }
256 };
257 const size_t pc_compat_2_6_len = G_N_ELEMENTS(pc_compat_2_6);
258 
259 GlobalProperty pc_compat_2_5[] = {};
260 const size_t pc_compat_2_5_len = G_N_ELEMENTS(pc_compat_2_5);
261 
262 GlobalProperty pc_compat_2_4[] = {
263     PC_CPU_MODEL_IDS("2.4.0")
264     { "Haswell-" TYPE_X86_CPU, "abm", "off" },
265     { "Haswell-noTSX-" TYPE_X86_CPU, "abm", "off" },
266     { "Broadwell-" TYPE_X86_CPU, "abm", "off" },
267     { "Broadwell-noTSX-" TYPE_X86_CPU, "abm", "off" },
268     { "host" "-" TYPE_X86_CPU, "host-cache-info", "on" },
269     { TYPE_X86_CPU, "check", "off" },
270     { "qemu64" "-" TYPE_X86_CPU, "sse4a", "on" },
271     { "qemu64" "-" TYPE_X86_CPU, "abm", "on" },
272     { "qemu64" "-" TYPE_X86_CPU, "popcnt", "on" },
273     { "qemu32" "-" TYPE_X86_CPU, "popcnt", "on" },
274     { "Opteron_G2" "-" TYPE_X86_CPU, "rdtscp", "on" },
275     { "Opteron_G3" "-" TYPE_X86_CPU, "rdtscp", "on" },
276     { "Opteron_G4" "-" TYPE_X86_CPU, "rdtscp", "on" },
277     { "Opteron_G5" "-" TYPE_X86_CPU, "rdtscp", "on", }
278 };
279 const size_t pc_compat_2_4_len = G_N_ELEMENTS(pc_compat_2_4);
280 
281 GlobalProperty pc_compat_2_3[] = {
282     PC_CPU_MODEL_IDS("2.3.0")
283     { TYPE_X86_CPU, "arat", "off" },
284     { "qemu64" "-" TYPE_X86_CPU, "min-level", "4" },
285     { "kvm64" "-" TYPE_X86_CPU, "min-level", "5" },
286     { "pentium3" "-" TYPE_X86_CPU, "min-level", "2" },
287     { "n270" "-" TYPE_X86_CPU, "min-level", "5" },
288     { "Conroe" "-" TYPE_X86_CPU, "min-level", "4" },
289     { "Penryn" "-" TYPE_X86_CPU, "min-level", "4" },
290     { "Nehalem" "-" TYPE_X86_CPU, "min-level", "4" },
291     { "n270" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
292     { "Penryn" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
293     { "Conroe" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
294     { "Nehalem" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
295     { "Westmere" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
296     { "SandyBridge" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
297     { "IvyBridge" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
298     { "Haswell" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
299     { "Haswell-noTSX" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
300     { "Broadwell" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
301     { "Broadwell-noTSX" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
302     { TYPE_X86_CPU, "kvm-no-smi-migration", "on" },
303 };
304 const size_t pc_compat_2_3_len = G_N_ELEMENTS(pc_compat_2_3);
305 
306 GlobalProperty pc_compat_2_2[] = {
307     PC_CPU_MODEL_IDS("2.2.0")
308     { "kvm64" "-" TYPE_X86_CPU, "vme", "off" },
309     { "kvm32" "-" TYPE_X86_CPU, "vme", "off" },
310     { "Conroe" "-" TYPE_X86_CPU, "vme", "off" },
311     { "Penryn" "-" TYPE_X86_CPU, "vme", "off" },
312     { "Nehalem" "-" TYPE_X86_CPU, "vme", "off" },
313     { "Westmere" "-" TYPE_X86_CPU, "vme", "off" },
314     { "SandyBridge" "-" TYPE_X86_CPU, "vme", "off" },
315     { "Haswell" "-" TYPE_X86_CPU, "vme", "off" },
316     { "Broadwell" "-" TYPE_X86_CPU, "vme", "off" },
317     { "Opteron_G1" "-" TYPE_X86_CPU, "vme", "off" },
318     { "Opteron_G2" "-" TYPE_X86_CPU, "vme", "off" },
319     { "Opteron_G3" "-" TYPE_X86_CPU, "vme", "off" },
320     { "Opteron_G4" "-" TYPE_X86_CPU, "vme", "off" },
321     { "Opteron_G5" "-" TYPE_X86_CPU, "vme", "off" },
322     { "Haswell" "-" TYPE_X86_CPU, "f16c", "off" },
323     { "Haswell" "-" TYPE_X86_CPU, "rdrand", "off" },
324     { "Broadwell" "-" TYPE_X86_CPU, "f16c", "off" },
325     { "Broadwell" "-" TYPE_X86_CPU, "rdrand", "off" },
326 };
327 const size_t pc_compat_2_2_len = G_N_ELEMENTS(pc_compat_2_2);
328 
329 GlobalProperty pc_compat_2_1[] = {
330     PC_CPU_MODEL_IDS("2.1.0")
331     { "coreduo" "-" TYPE_X86_CPU, "vmx", "on" },
332     { "core2duo" "-" TYPE_X86_CPU, "vmx", "on" },
333 };
334 const size_t pc_compat_2_1_len = G_N_ELEMENTS(pc_compat_2_1);
335 
336 GlobalProperty pc_compat_2_0[] = {
337     PC_CPU_MODEL_IDS("2.0.0")
338     { "virtio-scsi-pci", "any_layout", "off" },
339     { "PIIX4_PM", "memory-hotplug-support", "off" },
340     { "apic", "version", "0x11" },
341     { "nec-usb-xhci", "superspeed-ports-first", "off" },
342     { "nec-usb-xhci", "force-pcie-endcap", "on" },
343     { "pci-serial", "prog_if", "0" },
344     { "pci-serial-2x", "prog_if", "0" },
345     { "pci-serial-4x", "prog_if", "0" },
346     { "virtio-net-pci", "guest_announce", "off" },
347     { "ICH9-LPC", "memory-hotplug-support", "off" },
348 };
349 const size_t pc_compat_2_0_len = G_N_ELEMENTS(pc_compat_2_0);
350 
351 GlobalProperty pc_compat_1_7[] = {
352     PC_CPU_MODEL_IDS("1.7.0")
353     { TYPE_USB_DEVICE, "msos-desc", "no" },
354     { "PIIX4_PM", ACPI_PM_PROP_ACPI_PCIHP_BRIDGE, "off" },
355     { "hpet", HPET_INTCAP, "4" },
356 };
357 const size_t pc_compat_1_7_len = G_N_ELEMENTS(pc_compat_1_7);
358 
359 GlobalProperty pc_compat_1_6[] = {
360     PC_CPU_MODEL_IDS("1.6.0")
361     { "e1000", "mitigation", "off" },
362     { "qemu64-" TYPE_X86_CPU, "model", "2" },
363     { "qemu32-" TYPE_X86_CPU, "model", "3" },
364     { "i440FX-pcihost", "short_root_bus", "1" },
365     { "q35-pcihost", "short_root_bus", "1" },
366 };
367 const size_t pc_compat_1_6_len = G_N_ELEMENTS(pc_compat_1_6);
368 
369 GlobalProperty pc_compat_1_5[] = {
370     PC_CPU_MODEL_IDS("1.5.0")
371     { "Conroe-" TYPE_X86_CPU, "model", "2" },
372     { "Conroe-" TYPE_X86_CPU, "min-level", "2" },
373     { "Penryn-" TYPE_X86_CPU, "model", "2" },
374     { "Penryn-" TYPE_X86_CPU, "min-level", "2" },
375     { "Nehalem-" TYPE_X86_CPU, "model", "2" },
376     { "Nehalem-" TYPE_X86_CPU, "min-level", "2" },
377     { "virtio-net-pci", "any_layout", "off" },
378     { TYPE_X86_CPU, "pmu", "on" },
379     { "i440FX-pcihost", "short_root_bus", "0" },
380     { "q35-pcihost", "short_root_bus", "0" },
381 };
382 const size_t pc_compat_1_5_len = G_N_ELEMENTS(pc_compat_1_5);
383 
384 GlobalProperty pc_compat_1_4[] = {
385     PC_CPU_MODEL_IDS("1.4.0")
386     { "scsi-hd", "discard_granularity", "0" },
387     { "scsi-cd", "discard_granularity", "0" },
388     { "ide-hd", "discard_granularity", "0" },
389     { "ide-cd", "discard_granularity", "0" },
390     { "virtio-blk-pci", "discard_granularity", "0" },
391     /* DEV_NVECTORS_UNSPECIFIED as a uint32_t string: */
392     { "virtio-serial-pci", "vectors", "0xFFFFFFFF" },
393     { "virtio-net-pci", "ctrl_guest_offloads", "off" },
394     { "e1000", "romfile", "pxe-e1000.rom" },
395     { "ne2k_pci", "romfile", "pxe-ne2k_pci.rom" },
396     { "pcnet", "romfile", "pxe-pcnet.rom" },
397     { "rtl8139", "romfile", "pxe-rtl8139.rom" },
398     { "virtio-net-pci", "romfile", "pxe-virtio.rom" },
399     { "486-" TYPE_X86_CPU, "model", "0" },
400     { "n270" "-" TYPE_X86_CPU, "movbe", "off" },
401     { "Westmere" "-" TYPE_X86_CPU, "pclmulqdq", "off" },
402 };
403 const size_t pc_compat_1_4_len = G_N_ELEMENTS(pc_compat_1_4);
404 
405 GSIState *pc_gsi_create(qemu_irq **irqs, bool pci_enabled)
406 {
407     GSIState *s;
408 
409     s = g_new0(GSIState, 1);
410     if (kvm_ioapic_in_kernel()) {
411         kvm_pc_setup_irq_routing(pci_enabled);
412     }
413     *irqs = qemu_allocate_irqs(gsi_handler, s, IOAPIC_NUM_PINS);
414 
415     return s;
416 }
417 
418 static void ioport80_write(void *opaque, hwaddr addr, uint64_t data,
419                            unsigned size)
420 {
421 }
422 
423 static uint64_t ioport80_read(void *opaque, hwaddr addr, unsigned size)
424 {
425     return 0xffffffffffffffffULL;
426 }
427 
428 /* MSDOS compatibility mode FPU exception support */
429 static void ioportF0_write(void *opaque, hwaddr addr, uint64_t data,
430                            unsigned size)
431 {
432     if (tcg_enabled()) {
433         cpu_set_ignne();
434     }
435 }
436 
437 static uint64_t ioportF0_read(void *opaque, hwaddr addr, unsigned size)
438 {
439     return 0xffffffffffffffffULL;
440 }
441 
442 /* PC cmos mappings */
443 
444 #define REG_EQUIPMENT_BYTE          0x14
445 
446 static void cmos_init_hd(MC146818RtcState *s, int type_ofs, int info_ofs,
447                          int16_t cylinders, int8_t heads, int8_t sectors)
448 {
449     mc146818rtc_set_cmos_data(s, type_ofs, 47);
450     mc146818rtc_set_cmos_data(s, info_ofs, cylinders);
451     mc146818rtc_set_cmos_data(s, info_ofs + 1, cylinders >> 8);
452     mc146818rtc_set_cmos_data(s, info_ofs + 2, heads);
453     mc146818rtc_set_cmos_data(s, info_ofs + 3, 0xff);
454     mc146818rtc_set_cmos_data(s, info_ofs + 4, 0xff);
455     mc146818rtc_set_cmos_data(s, info_ofs + 5, 0xc0 | ((heads > 8) << 3));
456     mc146818rtc_set_cmos_data(s, info_ofs + 6, cylinders);
457     mc146818rtc_set_cmos_data(s, info_ofs + 7, cylinders >> 8);
458     mc146818rtc_set_cmos_data(s, info_ofs + 8, sectors);
459 }
460 
461 /* convert boot_device letter to something recognizable by the bios */
462 static int boot_device2nibble(char boot_device)
463 {
464     switch(boot_device) {
465     case 'a':
466     case 'b':
467         return 0x01; /* floppy boot */
468     case 'c':
469         return 0x02; /* hard drive boot */
470     case 'd':
471         return 0x03; /* CD-ROM boot */
472     case 'n':
473         return 0x04; /* Network boot */
474     }
475     return 0;
476 }
477 
478 static void set_boot_dev(MC146818RtcState *s, const char *boot_device,
479                          Error **errp)
480 {
481 #define PC_MAX_BOOT_DEVICES 3
482     int nbds, bds[3] = { 0, };
483     int i;
484 
485     nbds = strlen(boot_device);
486     if (nbds > PC_MAX_BOOT_DEVICES) {
487         error_setg(errp, "Too many boot devices for PC");
488         return;
489     }
490     for (i = 0; i < nbds; i++) {
491         bds[i] = boot_device2nibble(boot_device[i]);
492         if (bds[i] == 0) {
493             error_setg(errp, "Invalid boot device for PC: '%c'",
494                        boot_device[i]);
495             return;
496         }
497     }
498     mc146818rtc_set_cmos_data(s, 0x3d, (bds[1] << 4) | bds[0]);
499     mc146818rtc_set_cmos_data(s, 0x38, (bds[2] << 4) | (fd_bootchk ? 0x0 : 0x1));
500 }
501 
502 static void pc_boot_set(void *opaque, const char *boot_device, Error **errp)
503 {
504     set_boot_dev(opaque, boot_device, errp);
505 }
506 
507 static void pc_cmos_init_floppy(MC146818RtcState *rtc_state, ISADevice *floppy)
508 {
509     int val, nb, i;
510     FloppyDriveType fd_type[2] = { FLOPPY_DRIVE_TYPE_NONE,
511                                    FLOPPY_DRIVE_TYPE_NONE };
512 
513     /* floppy type */
514     if (floppy) {
515         for (i = 0; i < 2; i++) {
516             fd_type[i] = isa_fdc_get_drive_type(floppy, i);
517         }
518     }
519     val = (cmos_get_fd_drive_type(fd_type[0]) << 4) |
520         cmos_get_fd_drive_type(fd_type[1]);
521     mc146818rtc_set_cmos_data(rtc_state, 0x10, val);
522 
523     val = mc146818rtc_get_cmos_data(rtc_state, REG_EQUIPMENT_BYTE);
524     nb = 0;
525     if (fd_type[0] != FLOPPY_DRIVE_TYPE_NONE) {
526         nb++;
527     }
528     if (fd_type[1] != FLOPPY_DRIVE_TYPE_NONE) {
529         nb++;
530     }
531     switch (nb) {
532     case 0:
533         break;
534     case 1:
535         val |= 0x01; /* 1 drive, ready for boot */
536         break;
537     case 2:
538         val |= 0x41; /* 2 drives, ready for boot */
539         break;
540     }
541     mc146818rtc_set_cmos_data(rtc_state, REG_EQUIPMENT_BYTE, val);
542 }
543 
544 typedef struct pc_cmos_init_late_arg {
545     MC146818RtcState *rtc_state;
546     BusState *idebus[2];
547 } pc_cmos_init_late_arg;
548 
549 typedef struct check_fdc_state {
550     ISADevice *floppy;
551     bool multiple;
552 } CheckFdcState;
553 
554 static int check_fdc(Object *obj, void *opaque)
555 {
556     CheckFdcState *state = opaque;
557     Object *fdc;
558     uint32_t iobase;
559     Error *local_err = NULL;
560 
561     fdc = object_dynamic_cast(obj, TYPE_ISA_FDC);
562     if (!fdc) {
563         return 0;
564     }
565 
566     iobase = object_property_get_uint(obj, "iobase", &local_err);
567     if (local_err || iobase != 0x3f0) {
568         error_free(local_err);
569         return 0;
570     }
571 
572     if (state->floppy) {
573         state->multiple = true;
574     } else {
575         state->floppy = ISA_DEVICE(obj);
576     }
577     return 0;
578 }
579 
580 static const char * const fdc_container_path[] = {
581     "/unattached", "/peripheral", "/peripheral-anon"
582 };
583 
584 /*
585  * Locate the FDC at IO address 0x3f0, in order to configure the CMOS registers
586  * and ACPI objects.
587  */
588 static ISADevice *pc_find_fdc0(void)
589 {
590     int i;
591     Object *container;
592     CheckFdcState state = { 0 };
593 
594     for (i = 0; i < ARRAY_SIZE(fdc_container_path); i++) {
595         container = container_get(qdev_get_machine(), fdc_container_path[i]);
596         object_child_foreach(container, check_fdc, &state);
597     }
598 
599     if (state.multiple) {
600         warn_report("multiple floppy disk controllers with "
601                     "iobase=0x3f0 have been found");
602         error_printf("the one being picked for CMOS setup might not reflect "
603                      "your intent");
604     }
605 
606     return state.floppy;
607 }
608 
609 static void pc_cmos_init_late(void *opaque)
610 {
611     pc_cmos_init_late_arg *arg = opaque;
612     MC146818RtcState *s = arg->rtc_state;
613     int16_t cylinders;
614     int8_t heads, sectors;
615     int val;
616     int i, trans;
617 
618     val = 0;
619     if (arg->idebus[0] && ide_get_geometry(arg->idebus[0], 0,
620                                            &cylinders, &heads, &sectors) >= 0) {
621         cmos_init_hd(s, 0x19, 0x1b, cylinders, heads, sectors);
622         val |= 0xf0;
623     }
624     if (arg->idebus[0] && ide_get_geometry(arg->idebus[0], 1,
625                                            &cylinders, &heads, &sectors) >= 0) {
626         cmos_init_hd(s, 0x1a, 0x24, cylinders, heads, sectors);
627         val |= 0x0f;
628     }
629     mc146818rtc_set_cmos_data(s, 0x12, val);
630 
631     val = 0;
632     for (i = 0; i < 4; i++) {
633         /* NOTE: ide_get_geometry() returns the physical
634            geometry.  It is always such that: 1 <= sects <= 63, 1
635            <= heads <= 16, 1 <= cylinders <= 16383. The BIOS
636            geometry can be different if a translation is done. */
637         if (arg->idebus[i / 2] &&
638             ide_get_geometry(arg->idebus[i / 2], i % 2,
639                              &cylinders, &heads, &sectors) >= 0) {
640             trans = ide_get_bios_chs_trans(arg->idebus[i / 2], i % 2) - 1;
641             assert((trans & ~3) == 0);
642             val |= trans << (i * 2);
643         }
644     }
645     mc146818rtc_set_cmos_data(s, 0x39, val);
646 
647     pc_cmos_init_floppy(s, pc_find_fdc0());
648 
649     qemu_unregister_reset(pc_cmos_init_late, opaque);
650 }
651 
652 void pc_cmos_init(PCMachineState *pcms,
653                   BusState *idebus0, BusState *idebus1,
654                   ISADevice *rtc)
655 {
656     int val;
657     static pc_cmos_init_late_arg arg;
658     X86MachineState *x86ms = X86_MACHINE(pcms);
659     MC146818RtcState *s = MC146818_RTC(rtc);
660 
661     /* various important CMOS locations needed by PC/Bochs bios */
662 
663     /* memory size */
664     /* base memory (first MiB) */
665     val = MIN(x86ms->below_4g_mem_size / KiB, 640);
666     mc146818rtc_set_cmos_data(s, 0x15, val);
667     mc146818rtc_set_cmos_data(s, 0x16, val >> 8);
668     /* extended memory (next 64MiB) */
669     if (x86ms->below_4g_mem_size > 1 * MiB) {
670         val = (x86ms->below_4g_mem_size - 1 * MiB) / KiB;
671     } else {
672         val = 0;
673     }
674     if (val > 65535)
675         val = 65535;
676     mc146818rtc_set_cmos_data(s, 0x17, val);
677     mc146818rtc_set_cmos_data(s, 0x18, val >> 8);
678     mc146818rtc_set_cmos_data(s, 0x30, val);
679     mc146818rtc_set_cmos_data(s, 0x31, val >> 8);
680     /* memory between 16MiB and 4GiB */
681     if (x86ms->below_4g_mem_size > 16 * MiB) {
682         val = (x86ms->below_4g_mem_size - 16 * MiB) / (64 * KiB);
683     } else {
684         val = 0;
685     }
686     if (val > 65535)
687         val = 65535;
688     mc146818rtc_set_cmos_data(s, 0x34, val);
689     mc146818rtc_set_cmos_data(s, 0x35, val >> 8);
690     /* memory above 4GiB */
691     val = x86ms->above_4g_mem_size / 65536;
692     mc146818rtc_set_cmos_data(s, 0x5b, val);
693     mc146818rtc_set_cmos_data(s, 0x5c, val >> 8);
694     mc146818rtc_set_cmos_data(s, 0x5d, val >> 16);
695 
696     object_property_add_link(OBJECT(pcms), "rtc_state",
697                              TYPE_ISA_DEVICE,
698                              (Object **)&x86ms->rtc,
699                              object_property_allow_set_link,
700                              OBJ_PROP_LINK_STRONG);
701     object_property_set_link(OBJECT(pcms), "rtc_state", OBJECT(s),
702                              &error_abort);
703 
704     set_boot_dev(s, MACHINE(pcms)->boot_config.order, &error_fatal);
705 
706     val = 0;
707     val |= 0x02; /* FPU is there */
708     val |= 0x04; /* PS/2 mouse installed */
709     mc146818rtc_set_cmos_data(s, REG_EQUIPMENT_BYTE, val);
710 
711     /* hard drives and FDC */
712     arg.rtc_state = s;
713     arg.idebus[0] = idebus0;
714     arg.idebus[1] = idebus1;
715     qemu_register_reset(pc_cmos_init_late, &arg);
716 }
717 
718 static void handle_a20_line_change(void *opaque, int irq, int level)
719 {
720     X86CPU *cpu = opaque;
721 
722     /* XXX: send to all CPUs ? */
723     /* XXX: add logic to handle multiple A20 line sources */
724     x86_cpu_set_a20(cpu, level);
725 }
726 
727 #define NE2000_NB_MAX 6
728 
729 static const int ne2000_io[NE2000_NB_MAX] = { 0x300, 0x320, 0x340, 0x360,
730                                               0x280, 0x380 };
731 static const int ne2000_irq[NE2000_NB_MAX] = { 9, 10, 11, 3, 4, 5 };
732 
733 static void pc_init_ne2k_isa(ISABus *bus, NICInfo *nd)
734 {
735     static int nb_ne2k = 0;
736 
737     if (nb_ne2k == NE2000_NB_MAX)
738         return;
739     isa_ne2000_init(bus, ne2000_io[nb_ne2k],
740                     ne2000_irq[nb_ne2k], nd);
741     nb_ne2k++;
742 }
743 
744 void pc_acpi_smi_interrupt(void *opaque, int irq, int level)
745 {
746     X86CPU *cpu = opaque;
747 
748     if (level) {
749         cpu_interrupt(CPU(cpu), CPU_INTERRUPT_SMI);
750     }
751 }
752 
753 static
754 void pc_machine_done(Notifier *notifier, void *data)
755 {
756     PCMachineState *pcms = container_of(notifier,
757                                         PCMachineState, machine_done);
758     X86MachineState *x86ms = X86_MACHINE(pcms);
759 
760     cxl_hook_up_pxb_registers(pcms->bus, &pcms->cxl_devices_state,
761                               &error_fatal);
762 
763     if (pcms->cxl_devices_state.is_enabled) {
764         cxl_fmws_link_targets(&pcms->cxl_devices_state, &error_fatal);
765     }
766 
767     /* set the number of CPUs */
768     x86_rtc_set_cpus_count(x86ms->rtc, x86ms->boot_cpus);
769 
770     fw_cfg_add_extra_pci_roots(pcms->bus, x86ms->fw_cfg);
771 
772     acpi_setup();
773     if (x86ms->fw_cfg) {
774         fw_cfg_build_smbios(MACHINE(pcms), x86ms->fw_cfg);
775         fw_cfg_build_feature_control(MACHINE(pcms), x86ms->fw_cfg);
776         /* update FW_CFG_NB_CPUS to account for -device added CPUs */
777         fw_cfg_modify_i16(x86ms->fw_cfg, FW_CFG_NB_CPUS, x86ms->boot_cpus);
778     }
779 }
780 
781 void pc_guest_info_init(PCMachineState *pcms)
782 {
783     X86MachineState *x86ms = X86_MACHINE(pcms);
784 
785     x86ms->apic_xrupt_override = true;
786     pcms->machine_done.notify = pc_machine_done;
787     qemu_add_machine_init_done_notifier(&pcms->machine_done);
788 }
789 
790 /* setup pci memory address space mapping into system address space */
791 void pc_pci_as_mapping_init(MemoryRegion *system_memory,
792                             MemoryRegion *pci_address_space)
793 {
794     /* Set to lower priority than RAM */
795     memory_region_add_subregion_overlap(system_memory, 0x0,
796                                         pci_address_space, -1);
797 }
798 
799 void xen_load_linux(PCMachineState *pcms)
800 {
801     int i;
802     FWCfgState *fw_cfg;
803     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
804     X86MachineState *x86ms = X86_MACHINE(pcms);
805 
806     assert(MACHINE(pcms)->kernel_filename != NULL);
807 
808     fw_cfg = fw_cfg_init_io(FW_CFG_IO_BASE);
809     fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, x86ms->boot_cpus);
810     rom_set_fw(fw_cfg);
811 
812     x86_load_linux(x86ms, fw_cfg, pcmc->acpi_data_size,
813                    pcmc->pvh_enabled);
814     for (i = 0; i < nb_option_roms; i++) {
815         assert(!strcmp(option_rom[i].name, "linuxboot.bin") ||
816                !strcmp(option_rom[i].name, "linuxboot_dma.bin") ||
817                !strcmp(option_rom[i].name, "pvh.bin") ||
818                !strcmp(option_rom[i].name, "multiboot.bin") ||
819                !strcmp(option_rom[i].name, "multiboot_dma.bin"));
820         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
821     }
822     x86ms->fw_cfg = fw_cfg;
823 }
824 
825 #define PC_ROM_MIN_VGA     0xc0000
826 #define PC_ROM_MIN_OPTION  0xc8000
827 #define PC_ROM_MAX         0xe0000
828 #define PC_ROM_ALIGN       0x800
829 #define PC_ROM_SIZE        (PC_ROM_MAX - PC_ROM_MIN_VGA)
830 
831 static hwaddr pc_above_4g_end(PCMachineState *pcms)
832 {
833     X86MachineState *x86ms = X86_MACHINE(pcms);
834 
835     if (pcms->sgx_epc.size != 0) {
836         return sgx_epc_above_4g_end(&pcms->sgx_epc);
837     }
838 
839     return x86ms->above_4g_mem_start + x86ms->above_4g_mem_size;
840 }
841 
842 static void pc_get_device_memory_range(PCMachineState *pcms,
843                                        hwaddr *base,
844                                        ram_addr_t *device_mem_size)
845 {
846     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
847     MachineState *machine = MACHINE(pcms);
848     ram_addr_t size;
849     hwaddr addr;
850 
851     size = machine->maxram_size - machine->ram_size;
852     addr = ROUND_UP(pc_above_4g_end(pcms), 1 * GiB);
853 
854     if (pcmc->enforce_aligned_dimm) {
855         /* size device region assuming 1G page max alignment per slot */
856         size += (1 * GiB) * machine->ram_slots;
857     }
858 
859     *base = addr;
860     *device_mem_size = size;
861 }
862 
863 static uint64_t pc_get_cxl_range_start(PCMachineState *pcms)
864 {
865     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
866     hwaddr cxl_base;
867     ram_addr_t size;
868 
869     if (pcmc->has_reserved_memory) {
870         pc_get_device_memory_range(pcms, &cxl_base, &size);
871         cxl_base += size;
872     } else {
873         cxl_base = pc_above_4g_end(pcms);
874     }
875 
876     return cxl_base;
877 }
878 
879 static uint64_t pc_get_cxl_range_end(PCMachineState *pcms)
880 {
881     uint64_t start = pc_get_cxl_range_start(pcms) + MiB;
882 
883     if (pcms->cxl_devices_state.fixed_windows) {
884         GList *it;
885 
886         start = ROUND_UP(start, 256 * MiB);
887         for (it = pcms->cxl_devices_state.fixed_windows; it; it = it->next) {
888             CXLFixedWindow *fw = it->data;
889             start += fw->size;
890         }
891     }
892 
893     return start;
894 }
895 
896 static hwaddr pc_max_used_gpa(PCMachineState *pcms, uint64_t pci_hole64_size)
897 {
898     X86CPU *cpu = X86_CPU(first_cpu);
899 
900     /* 32-bit systems don't have hole64 thus return max CPU address */
901     if (cpu->phys_bits <= 32) {
902         return ((hwaddr)1 << cpu->phys_bits) - 1;
903     }
904 
905     return pc_pci_hole64_start() + pci_hole64_size - 1;
906 }
907 
908 /*
909  * AMD systems with an IOMMU have an additional hole close to the
910  * 1Tb, which are special GPAs that cannot be DMA mapped. Depending
911  * on kernel version, VFIO may or may not let you DMA map those ranges.
912  * Starting Linux v5.4 we validate it, and can't create guests on AMD machines
913  * with certain memory sizes. It's also wrong to use those IOVA ranges
914  * in detriment of leading to IOMMU INVALID_DEVICE_REQUEST or worse.
915  * The ranges reserved for Hyper-Transport are:
916  *
917  * FD_0000_0000h - FF_FFFF_FFFFh
918  *
919  * The ranges represent the following:
920  *
921  * Base Address   Top Address  Use
922  *
923  * FD_0000_0000h FD_F7FF_FFFFh Reserved interrupt address space
924  * FD_F800_0000h FD_F8FF_FFFFh Interrupt/EOI IntCtl
925  * FD_F900_0000h FD_F90F_FFFFh Legacy PIC IACK
926  * FD_F910_0000h FD_F91F_FFFFh System Management
927  * FD_F920_0000h FD_FAFF_FFFFh Reserved Page Tables
928  * FD_FB00_0000h FD_FBFF_FFFFh Address Translation
929  * FD_FC00_0000h FD_FDFF_FFFFh I/O Space
930  * FD_FE00_0000h FD_FFFF_FFFFh Configuration
931  * FE_0000_0000h FE_1FFF_FFFFh Extended Configuration/Device Messages
932  * FE_2000_0000h FF_FFFF_FFFFh Reserved
933  *
934  * See AMD IOMMU spec, section 2.1.2 "IOMMU Logical Topology",
935  * Table 3: Special Address Controls (GPA) for more information.
936  */
937 #define AMD_HT_START         0xfd00000000UL
938 #define AMD_HT_END           0xffffffffffUL
939 #define AMD_ABOVE_1TB_START  (AMD_HT_END + 1)
940 #define AMD_HT_SIZE          (AMD_ABOVE_1TB_START - AMD_HT_START)
941 
942 void pc_memory_init(PCMachineState *pcms,
943                     MemoryRegion *system_memory,
944                     MemoryRegion *rom_memory,
945                     MemoryRegion **ram_memory,
946                     uint64_t pci_hole64_size)
947 {
948     int linux_boot, i;
949     MemoryRegion *option_rom_mr;
950     MemoryRegion *ram_below_4g, *ram_above_4g;
951     FWCfgState *fw_cfg;
952     MachineState *machine = MACHINE(pcms);
953     MachineClass *mc = MACHINE_GET_CLASS(machine);
954     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
955     X86MachineState *x86ms = X86_MACHINE(pcms);
956     hwaddr maxphysaddr, maxusedaddr;
957     hwaddr cxl_base, cxl_resv_end = 0;
958     X86CPU *cpu = X86_CPU(first_cpu);
959 
960     assert(machine->ram_size == x86ms->below_4g_mem_size +
961                                 x86ms->above_4g_mem_size);
962 
963     linux_boot = (machine->kernel_filename != NULL);
964 
965     /*
966      * The HyperTransport range close to the 1T boundary is unique to AMD
967      * hosts with IOMMUs enabled. Restrict the ram-above-4g relocation
968      * to above 1T to AMD vCPUs only. @enforce_amd_1tb_hole is only false in
969      * older machine types (<= 7.0) for compatibility purposes.
970      */
971     if (IS_AMD_CPU(&cpu->env) && pcmc->enforce_amd_1tb_hole) {
972         /* Bail out if max possible address does not cross HT range */
973         if (pc_max_used_gpa(pcms, pci_hole64_size) >= AMD_HT_START) {
974             x86ms->above_4g_mem_start = AMD_ABOVE_1TB_START;
975         }
976 
977         /*
978          * Advertise the HT region if address space covers the reserved
979          * region or if we relocate.
980          */
981         if (cpu->phys_bits >= 40) {
982             e820_add_entry(AMD_HT_START, AMD_HT_SIZE, E820_RESERVED);
983         }
984     }
985 
986     /*
987      * phys-bits is required to be appropriately configured
988      * to make sure max used GPA is reachable.
989      */
990     maxusedaddr = pc_max_used_gpa(pcms, pci_hole64_size);
991     maxphysaddr = ((hwaddr)1 << cpu->phys_bits) - 1;
992     if (maxphysaddr < maxusedaddr) {
993         error_report("Address space limit 0x%"PRIx64" < 0x%"PRIx64
994                      " phys-bits too low (%u)",
995                      maxphysaddr, maxusedaddr, cpu->phys_bits);
996         exit(EXIT_FAILURE);
997     }
998 
999     /*
1000      * Split single memory region and use aliases to address portions of it,
1001      * done for backwards compatibility with older qemus.
1002      */
1003     *ram_memory = machine->ram;
1004     ram_below_4g = g_malloc(sizeof(*ram_below_4g));
1005     memory_region_init_alias(ram_below_4g, NULL, "ram-below-4g", machine->ram,
1006                              0, x86ms->below_4g_mem_size);
1007     memory_region_add_subregion(system_memory, 0, ram_below_4g);
1008     e820_add_entry(0, x86ms->below_4g_mem_size, E820_RAM);
1009     if (x86ms->above_4g_mem_size > 0) {
1010         ram_above_4g = g_malloc(sizeof(*ram_above_4g));
1011         memory_region_init_alias(ram_above_4g, NULL, "ram-above-4g",
1012                                  machine->ram,
1013                                  x86ms->below_4g_mem_size,
1014                                  x86ms->above_4g_mem_size);
1015         memory_region_add_subregion(system_memory, x86ms->above_4g_mem_start,
1016                                     ram_above_4g);
1017         e820_add_entry(x86ms->above_4g_mem_start, x86ms->above_4g_mem_size,
1018                        E820_RAM);
1019     }
1020 
1021     if (pcms->sgx_epc.size != 0) {
1022         e820_add_entry(pcms->sgx_epc.base, pcms->sgx_epc.size, E820_RESERVED);
1023     }
1024 
1025     if (!pcmc->has_reserved_memory &&
1026         (machine->ram_slots ||
1027          (machine->maxram_size > machine->ram_size))) {
1028 
1029         error_report("\"-memory 'slots|maxmem'\" is not supported by: %s",
1030                      mc->name);
1031         exit(EXIT_FAILURE);
1032     }
1033 
1034     /* always allocate the device memory information */
1035     machine->device_memory = g_malloc0(sizeof(*machine->device_memory));
1036 
1037     /* initialize device memory address space */
1038     if (pcmc->has_reserved_memory &&
1039         (machine->ram_size < machine->maxram_size)) {
1040         ram_addr_t device_mem_size;
1041 
1042         if (machine->ram_slots > ACPI_MAX_RAM_SLOTS) {
1043             error_report("unsupported amount of memory slots: %"PRIu64,
1044                          machine->ram_slots);
1045             exit(EXIT_FAILURE);
1046         }
1047 
1048         if (QEMU_ALIGN_UP(machine->maxram_size,
1049                           TARGET_PAGE_SIZE) != machine->maxram_size) {
1050             error_report("maximum memory size must by aligned to multiple of "
1051                          "%d bytes", TARGET_PAGE_SIZE);
1052             exit(EXIT_FAILURE);
1053         }
1054 
1055         pc_get_device_memory_range(pcms, &machine->device_memory->base, &device_mem_size);
1056 
1057         if ((machine->device_memory->base + device_mem_size) <
1058             device_mem_size) {
1059             error_report("unsupported amount of maximum memory: " RAM_ADDR_FMT,
1060                          machine->maxram_size);
1061             exit(EXIT_FAILURE);
1062         }
1063 
1064         memory_region_init(&machine->device_memory->mr, OBJECT(pcms),
1065                            "device-memory", device_mem_size);
1066         memory_region_add_subregion(system_memory, machine->device_memory->base,
1067                                     &machine->device_memory->mr);
1068     }
1069 
1070     if (pcms->cxl_devices_state.is_enabled) {
1071         MemoryRegion *mr = &pcms->cxl_devices_state.host_mr;
1072         hwaddr cxl_size = MiB;
1073 
1074         cxl_base = pc_get_cxl_range_start(pcms);
1075         memory_region_init(mr, OBJECT(machine), "cxl_host_reg", cxl_size);
1076         memory_region_add_subregion(system_memory, cxl_base, mr);
1077         cxl_resv_end = cxl_base + cxl_size;
1078         if (pcms->cxl_devices_state.fixed_windows) {
1079             hwaddr cxl_fmw_base;
1080             GList *it;
1081 
1082             cxl_fmw_base = ROUND_UP(cxl_base + cxl_size, 256 * MiB);
1083             for (it = pcms->cxl_devices_state.fixed_windows; it; it = it->next) {
1084                 CXLFixedWindow *fw = it->data;
1085 
1086                 fw->base = cxl_fmw_base;
1087                 memory_region_init_io(&fw->mr, OBJECT(machine), &cfmws_ops, fw,
1088                                       "cxl-fixed-memory-region", fw->size);
1089                 memory_region_add_subregion(system_memory, fw->base, &fw->mr);
1090                 cxl_fmw_base += fw->size;
1091                 cxl_resv_end = cxl_fmw_base;
1092             }
1093         }
1094     }
1095 
1096     /* Initialize PC system firmware */
1097     pc_system_firmware_init(pcms, rom_memory);
1098 
1099     option_rom_mr = g_malloc(sizeof(*option_rom_mr));
1100     memory_region_init_ram(option_rom_mr, NULL, "pc.rom", PC_ROM_SIZE,
1101                            &error_fatal);
1102     if (pcmc->pci_enabled) {
1103         memory_region_set_readonly(option_rom_mr, true);
1104     }
1105     memory_region_add_subregion_overlap(rom_memory,
1106                                         PC_ROM_MIN_VGA,
1107                                         option_rom_mr,
1108                                         1);
1109 
1110     fw_cfg = fw_cfg_arch_create(machine,
1111                                 x86ms->boot_cpus, x86ms->apic_id_limit);
1112 
1113     rom_set_fw(fw_cfg);
1114 
1115     if (pcmc->has_reserved_memory && machine->device_memory->base) {
1116         uint64_t *val = g_malloc(sizeof(*val));
1117         PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1118         uint64_t res_mem_end = machine->device_memory->base;
1119 
1120         if (!pcmc->broken_reserved_end) {
1121             res_mem_end += memory_region_size(&machine->device_memory->mr);
1122         }
1123 
1124         if (pcms->cxl_devices_state.is_enabled) {
1125             res_mem_end = cxl_resv_end;
1126         }
1127         *val = cpu_to_le64(ROUND_UP(res_mem_end, 1 * GiB));
1128         fw_cfg_add_file(fw_cfg, "etc/reserved-memory-end", val, sizeof(*val));
1129     }
1130 
1131     if (linux_boot) {
1132         x86_load_linux(x86ms, fw_cfg, pcmc->acpi_data_size,
1133                        pcmc->pvh_enabled);
1134     }
1135 
1136     for (i = 0; i < nb_option_roms; i++) {
1137         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1138     }
1139     x86ms->fw_cfg = fw_cfg;
1140 
1141     /* Init default IOAPIC address space */
1142     x86ms->ioapic_as = &address_space_memory;
1143 
1144     /* Init ACPI memory hotplug IO base address */
1145     pcms->memhp_io_base = ACPI_MEMORY_HOTPLUG_BASE;
1146 }
1147 
1148 /*
1149  * The 64bit pci hole starts after "above 4G RAM" and
1150  * potentially the space reserved for memory hotplug.
1151  */
1152 uint64_t pc_pci_hole64_start(void)
1153 {
1154     PCMachineState *pcms = PC_MACHINE(qdev_get_machine());
1155     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1156     MachineState *ms = MACHINE(pcms);
1157     uint64_t hole64_start = 0;
1158     ram_addr_t size = 0;
1159 
1160     if (pcms->cxl_devices_state.is_enabled) {
1161         hole64_start = pc_get_cxl_range_end(pcms);
1162     } else if (pcmc->has_reserved_memory && (ms->ram_size < ms->maxram_size)) {
1163         pc_get_device_memory_range(pcms, &hole64_start, &size);
1164         if (!pcmc->broken_reserved_end) {
1165             hole64_start += size;
1166         }
1167     } else {
1168         hole64_start = pc_above_4g_end(pcms);
1169     }
1170 
1171     return ROUND_UP(hole64_start, 1 * GiB);
1172 }
1173 
1174 DeviceState *pc_vga_init(ISABus *isa_bus, PCIBus *pci_bus)
1175 {
1176     DeviceState *dev = NULL;
1177 
1178     rom_set_order_override(FW_CFG_ORDER_OVERRIDE_VGA);
1179     if (pci_bus) {
1180         PCIDevice *pcidev = pci_vga_init(pci_bus);
1181         dev = pcidev ? &pcidev->qdev : NULL;
1182     } else if (isa_bus) {
1183         ISADevice *isadev = isa_vga_init(isa_bus);
1184         dev = isadev ? DEVICE(isadev) : NULL;
1185     }
1186     rom_reset_order_override();
1187     return dev;
1188 }
1189 
1190 static const MemoryRegionOps ioport80_io_ops = {
1191     .write = ioport80_write,
1192     .read = ioport80_read,
1193     .endianness = DEVICE_NATIVE_ENDIAN,
1194     .impl = {
1195         .min_access_size = 1,
1196         .max_access_size = 1,
1197     },
1198 };
1199 
1200 static const MemoryRegionOps ioportF0_io_ops = {
1201     .write = ioportF0_write,
1202     .read = ioportF0_read,
1203     .endianness = DEVICE_NATIVE_ENDIAN,
1204     .impl = {
1205         .min_access_size = 1,
1206         .max_access_size = 1,
1207     },
1208 };
1209 
1210 static void pc_superio_init(ISABus *isa_bus, bool create_fdctrl,
1211                             bool create_i8042, bool no_vmport)
1212 {
1213     int i;
1214     DriveInfo *fd[MAX_FD];
1215     qemu_irq *a20_line;
1216     ISADevice *fdc, *i8042, *port92, *vmmouse;
1217 
1218     serial_hds_isa_init(isa_bus, 0, MAX_ISA_SERIAL_PORTS);
1219     parallel_hds_isa_init(isa_bus, MAX_PARALLEL_PORTS);
1220 
1221     for (i = 0; i < MAX_FD; i++) {
1222         fd[i] = drive_get(IF_FLOPPY, 0, i);
1223         create_fdctrl |= !!fd[i];
1224     }
1225     if (create_fdctrl) {
1226         fdc = isa_new(TYPE_ISA_FDC);
1227         if (fdc) {
1228             isa_realize_and_unref(fdc, isa_bus, &error_fatal);
1229             isa_fdc_init_drives(fdc, fd);
1230         }
1231     }
1232 
1233     if (!create_i8042) {
1234         return;
1235     }
1236 
1237     i8042 = isa_create_simple(isa_bus, TYPE_I8042);
1238     if (!no_vmport) {
1239         isa_create_simple(isa_bus, TYPE_VMPORT);
1240         vmmouse = isa_try_new("vmmouse");
1241     } else {
1242         vmmouse = NULL;
1243     }
1244     if (vmmouse) {
1245         object_property_set_link(OBJECT(vmmouse), TYPE_I8042, OBJECT(i8042),
1246                                  &error_abort);
1247         isa_realize_and_unref(vmmouse, isa_bus, &error_fatal);
1248     }
1249     port92 = isa_create_simple(isa_bus, TYPE_PORT92);
1250 
1251     a20_line = qemu_allocate_irqs(handle_a20_line_change, first_cpu, 2);
1252     i8042_setup_a20_line(i8042, a20_line[0]);
1253     qdev_connect_gpio_out_named(DEVICE(port92),
1254                                 PORT92_A20_LINE, 0, a20_line[1]);
1255     g_free(a20_line);
1256 }
1257 
1258 void pc_basic_device_init(struct PCMachineState *pcms,
1259                           ISABus *isa_bus, qemu_irq *gsi,
1260                           ISADevice **rtc_state,
1261                           bool create_fdctrl,
1262                           uint32_t hpet_irqs)
1263 {
1264     int i;
1265     DeviceState *hpet = NULL;
1266     int pit_isa_irq = 0;
1267     qemu_irq pit_alt_irq = NULL;
1268     qemu_irq rtc_irq = NULL;
1269     ISADevice *pit = NULL;
1270     MemoryRegion *ioport80_io = g_new(MemoryRegion, 1);
1271     MemoryRegion *ioportF0_io = g_new(MemoryRegion, 1);
1272     X86MachineState *x86ms = X86_MACHINE(pcms);
1273 
1274     memory_region_init_io(ioport80_io, NULL, &ioport80_io_ops, NULL, "ioport80", 1);
1275     memory_region_add_subregion(isa_bus->address_space_io, 0x80, ioport80_io);
1276 
1277     memory_region_init_io(ioportF0_io, NULL, &ioportF0_io_ops, NULL, "ioportF0", 1);
1278     memory_region_add_subregion(isa_bus->address_space_io, 0xf0, ioportF0_io);
1279 
1280     /*
1281      * Check if an HPET shall be created.
1282      *
1283      * Without KVM_CAP_PIT_STATE2, we cannot switch off the in-kernel PIT
1284      * when the HPET wants to take over. Thus we have to disable the latter.
1285      */
1286     if (pcms->hpet_enabled && (!kvm_irqchip_in_kernel() ||
1287                                kvm_has_pit_state2())) {
1288         hpet = qdev_try_new(TYPE_HPET);
1289         if (!hpet) {
1290             error_report("couldn't create HPET device");
1291             exit(1);
1292         }
1293         /*
1294          * For pc-piix-*, hpet's intcap is always IRQ2. For pc-q35-1.7 and
1295          * earlier, use IRQ2 for compat. Otherwise, use IRQ16~23, IRQ8 and
1296          * IRQ2.
1297          */
1298         uint8_t compat = object_property_get_uint(OBJECT(hpet),
1299                 HPET_INTCAP, NULL);
1300         if (!compat) {
1301             qdev_prop_set_uint32(hpet, HPET_INTCAP, hpet_irqs);
1302         }
1303         sysbus_realize_and_unref(SYS_BUS_DEVICE(hpet), &error_fatal);
1304         sysbus_mmio_map(SYS_BUS_DEVICE(hpet), 0, HPET_BASE);
1305 
1306         for (i = 0; i < IOAPIC_NUM_PINS; i++) {
1307             sysbus_connect_irq(SYS_BUS_DEVICE(hpet), i, gsi[i]);
1308         }
1309         pit_isa_irq = -1;
1310         pit_alt_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_PIT_INT);
1311         rtc_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_RTC_INT);
1312     }
1313     *rtc_state = ISA_DEVICE(mc146818_rtc_init(isa_bus, 2000, rtc_irq));
1314 
1315 #ifdef CONFIG_XEN_EMU
1316     if (xen_mode == XEN_EMULATE) {
1317         xen_evtchn_connect_gsis(gsi);
1318         if (pcms->bus) {
1319             pci_create_simple(pcms->bus, -1, "xen-platform");
1320         }
1321     }
1322 #endif
1323 
1324     qemu_register_boot_set(pc_boot_set, *rtc_state);
1325 
1326     if (!xen_enabled() &&
1327         (x86ms->pit == ON_OFF_AUTO_AUTO || x86ms->pit == ON_OFF_AUTO_ON)) {
1328         if (kvm_pit_in_kernel()) {
1329             pit = kvm_pit_init(isa_bus, 0x40);
1330         } else {
1331             pit = i8254_pit_init(isa_bus, 0x40, pit_isa_irq, pit_alt_irq);
1332         }
1333         if (hpet) {
1334             /* connect PIT to output control line of the HPET */
1335             qdev_connect_gpio_out(hpet, 0, qdev_get_gpio_in(DEVICE(pit), 0));
1336         }
1337         pcspk_init(pcms->pcspk, isa_bus, pit);
1338     }
1339 
1340     /* Super I/O */
1341     pc_superio_init(isa_bus, create_fdctrl, pcms->i8042_enabled,
1342                     pcms->vmport != ON_OFF_AUTO_ON);
1343 }
1344 
1345 void pc_nic_init(PCMachineClass *pcmc, ISABus *isa_bus, PCIBus *pci_bus)
1346 {
1347     int i;
1348 
1349     rom_set_order_override(FW_CFG_ORDER_OVERRIDE_NIC);
1350     for (i = 0; i < nb_nics; i++) {
1351         NICInfo *nd = &nd_table[i];
1352         const char *model = nd->model ? nd->model : pcmc->default_nic_model;
1353 
1354         if (g_str_equal(model, "ne2k_isa")) {
1355             pc_init_ne2k_isa(isa_bus, nd);
1356         } else {
1357             pci_nic_init_nofail(nd, pci_bus, model, NULL);
1358         }
1359     }
1360     rom_reset_order_override();
1361 }
1362 
1363 void pc_i8259_create(ISABus *isa_bus, qemu_irq *i8259_irqs)
1364 {
1365     qemu_irq *i8259;
1366 
1367     if (kvm_pic_in_kernel()) {
1368         i8259 = kvm_i8259_init(isa_bus);
1369     } else if (xen_enabled()) {
1370         i8259 = xen_interrupt_controller_init();
1371     } else {
1372         i8259 = i8259_init(isa_bus, x86_allocate_cpu_irq());
1373     }
1374 
1375     for (size_t i = 0; i < ISA_NUM_IRQS; i++) {
1376         i8259_irqs[i] = i8259[i];
1377     }
1378 
1379     g_free(i8259);
1380 }
1381 
1382 static void pc_memory_pre_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
1383                                Error **errp)
1384 {
1385     const PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1386     const X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
1387     const PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1388     const MachineState *ms = MACHINE(hotplug_dev);
1389     const bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
1390     const uint64_t legacy_align = TARGET_PAGE_SIZE;
1391     Error *local_err = NULL;
1392 
1393     /*
1394      * When -no-acpi is used with Q35 machine type, no ACPI is built,
1395      * but pcms->acpi_dev is still created. Check !acpi_enabled in
1396      * addition to cover this case.
1397      */
1398     if (!x86ms->acpi_dev || !x86_machine_is_acpi_enabled(x86ms)) {
1399         error_setg(errp,
1400                    "memory hotplug is not enabled: missing acpi device or acpi disabled");
1401         return;
1402     }
1403 
1404     if (is_nvdimm && !ms->nvdimms_state->is_enabled) {
1405         error_setg(errp, "nvdimm is not enabled: missing 'nvdimm' in '-M'");
1406         return;
1407     }
1408 
1409     hotplug_handler_pre_plug(x86ms->acpi_dev, dev, &local_err);
1410     if (local_err) {
1411         error_propagate(errp, local_err);
1412         return;
1413     }
1414 
1415     pc_dimm_pre_plug(PC_DIMM(dev), MACHINE(hotplug_dev),
1416                      pcmc->enforce_aligned_dimm ? NULL : &legacy_align, errp);
1417 }
1418 
1419 static void pc_memory_plug(HotplugHandler *hotplug_dev,
1420                            DeviceState *dev, Error **errp)
1421 {
1422     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1423     X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
1424     MachineState *ms = MACHINE(hotplug_dev);
1425     bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
1426 
1427     pc_dimm_plug(PC_DIMM(dev), MACHINE(pcms));
1428 
1429     if (is_nvdimm) {
1430         nvdimm_plug(ms->nvdimms_state);
1431     }
1432 
1433     hotplug_handler_plug(x86ms->acpi_dev, dev, &error_abort);
1434 }
1435 
1436 static void pc_memory_unplug_request(HotplugHandler *hotplug_dev,
1437                                      DeviceState *dev, Error **errp)
1438 {
1439     X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
1440 
1441     /*
1442      * When -no-acpi is used with Q35 machine type, no ACPI is built,
1443      * but pcms->acpi_dev is still created. Check !acpi_enabled in
1444      * addition to cover this case.
1445      */
1446     if (!x86ms->acpi_dev || !x86_machine_is_acpi_enabled(x86ms)) {
1447         error_setg(errp,
1448                    "memory hotplug is not enabled: missing acpi device or acpi disabled");
1449         return;
1450     }
1451 
1452     if (object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM)) {
1453         error_setg(errp, "nvdimm device hot unplug is not supported yet.");
1454         return;
1455     }
1456 
1457     hotplug_handler_unplug_request(x86ms->acpi_dev, dev,
1458                                    errp);
1459 }
1460 
1461 static void pc_memory_unplug(HotplugHandler *hotplug_dev,
1462                              DeviceState *dev, Error **errp)
1463 {
1464     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1465     X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
1466     Error *local_err = NULL;
1467 
1468     hotplug_handler_unplug(x86ms->acpi_dev, dev, &local_err);
1469     if (local_err) {
1470         goto out;
1471     }
1472 
1473     pc_dimm_unplug(PC_DIMM(dev), MACHINE(pcms));
1474     qdev_unrealize(dev);
1475  out:
1476     error_propagate(errp, local_err);
1477 }
1478 
1479 static void pc_virtio_md_pci_pre_plug(HotplugHandler *hotplug_dev,
1480                                       DeviceState *dev, Error **errp)
1481 {
1482     HotplugHandler *hotplug_dev2 = qdev_get_bus_hotplug_handler(dev);
1483     Error *local_err = NULL;
1484 
1485     if (!hotplug_dev2 && dev->hotplugged) {
1486         /*
1487          * Without a bus hotplug handler, we cannot control the plug/unplug
1488          * order. We should never reach this point when hotplugging on x86,
1489          * however, better add a safety net.
1490          */
1491         error_setg(errp, "hotplug of virtio based memory devices not supported"
1492                    " on this bus.");
1493         return;
1494     }
1495     /*
1496      * First, see if we can plug this memory device at all. If that
1497      * succeeds, branch of to the actual hotplug handler.
1498      */
1499     memory_device_pre_plug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev), NULL,
1500                            &local_err);
1501     if (!local_err && hotplug_dev2) {
1502         hotplug_handler_pre_plug(hotplug_dev2, dev, &local_err);
1503     }
1504     error_propagate(errp, local_err);
1505 }
1506 
1507 static void pc_virtio_md_pci_plug(HotplugHandler *hotplug_dev,
1508                                   DeviceState *dev, Error **errp)
1509 {
1510     HotplugHandler *hotplug_dev2 = qdev_get_bus_hotplug_handler(dev);
1511     Error *local_err = NULL;
1512 
1513     /*
1514      * Plug the memory device first and then branch off to the actual
1515      * hotplug handler. If that one fails, we can easily undo the memory
1516      * device bits.
1517      */
1518     memory_device_plug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev));
1519     if (hotplug_dev2) {
1520         hotplug_handler_plug(hotplug_dev2, dev, &local_err);
1521         if (local_err) {
1522             memory_device_unplug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev));
1523         }
1524     }
1525     error_propagate(errp, local_err);
1526 }
1527 
1528 static void pc_virtio_md_pci_unplug_request(HotplugHandler *hotplug_dev,
1529                                             DeviceState *dev, Error **errp)
1530 {
1531     /* We don't support hot unplug of virtio based memory devices */
1532     error_setg(errp, "virtio based memory devices cannot be unplugged.");
1533 }
1534 
1535 static void pc_virtio_md_pci_unplug(HotplugHandler *hotplug_dev,
1536                                     DeviceState *dev, Error **errp)
1537 {
1538     /* We don't support hot unplug of virtio based memory devices */
1539 }
1540 
1541 static void pc_machine_device_pre_plug_cb(HotplugHandler *hotplug_dev,
1542                                           DeviceState *dev, Error **errp)
1543 {
1544     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1545         pc_memory_pre_plug(hotplug_dev, dev, errp);
1546     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1547         x86_cpu_pre_plug(hotplug_dev, dev, errp);
1548     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI) ||
1549                object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI)) {
1550         pc_virtio_md_pci_pre_plug(hotplug_dev, dev, errp);
1551     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
1552         /* Declare the APIC range as the reserved MSI region */
1553         char *resv_prop_str = g_strdup_printf("0xfee00000:0xfeefffff:%d",
1554                                               VIRTIO_IOMMU_RESV_MEM_T_MSI);
1555 
1556         object_property_set_uint(OBJECT(dev), "len-reserved-regions", 1, errp);
1557         object_property_set_str(OBJECT(dev), "reserved-regions[0]",
1558                                 resv_prop_str, errp);
1559         g_free(resv_prop_str);
1560     }
1561 
1562     if (object_dynamic_cast(OBJECT(dev), TYPE_X86_IOMMU_DEVICE) ||
1563         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
1564         PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1565 
1566         if (pcms->iommu) {
1567             error_setg(errp, "QEMU does not support multiple vIOMMUs "
1568                        "for x86 yet.");
1569             return;
1570         }
1571         pcms->iommu = dev;
1572     }
1573 }
1574 
1575 static void pc_machine_device_plug_cb(HotplugHandler *hotplug_dev,
1576                                       DeviceState *dev, Error **errp)
1577 {
1578     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1579         pc_memory_plug(hotplug_dev, dev, errp);
1580     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1581         x86_cpu_plug(hotplug_dev, dev, errp);
1582     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI) ||
1583                object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI)) {
1584         pc_virtio_md_pci_plug(hotplug_dev, dev, errp);
1585     }
1586 }
1587 
1588 static void pc_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
1589                                                 DeviceState *dev, Error **errp)
1590 {
1591     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1592         pc_memory_unplug_request(hotplug_dev, dev, errp);
1593     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1594         x86_cpu_unplug_request_cb(hotplug_dev, dev, errp);
1595     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI) ||
1596                object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI)) {
1597         pc_virtio_md_pci_unplug_request(hotplug_dev, dev, errp);
1598     } else {
1599         error_setg(errp, "acpi: device unplug request for not supported device"
1600                    " type: %s", object_get_typename(OBJECT(dev)));
1601     }
1602 }
1603 
1604 static void pc_machine_device_unplug_cb(HotplugHandler *hotplug_dev,
1605                                         DeviceState *dev, Error **errp)
1606 {
1607     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1608         pc_memory_unplug(hotplug_dev, dev, errp);
1609     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1610         x86_cpu_unplug_cb(hotplug_dev, dev, errp);
1611     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI) ||
1612                object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI)) {
1613         pc_virtio_md_pci_unplug(hotplug_dev, dev, errp);
1614     } else {
1615         error_setg(errp, "acpi: device unplug for not supported device"
1616                    " type: %s", object_get_typename(OBJECT(dev)));
1617     }
1618 }
1619 
1620 static HotplugHandler *pc_get_hotplug_handler(MachineState *machine,
1621                                              DeviceState *dev)
1622 {
1623     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) ||
1624         object_dynamic_cast(OBJECT(dev), TYPE_CPU) ||
1625         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI) ||
1626         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI) ||
1627         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI) ||
1628         object_dynamic_cast(OBJECT(dev), TYPE_X86_IOMMU_DEVICE)) {
1629         return HOTPLUG_HANDLER(machine);
1630     }
1631 
1632     return NULL;
1633 }
1634 
1635 static void
1636 pc_machine_get_device_memory_region_size(Object *obj, Visitor *v,
1637                                          const char *name, void *opaque,
1638                                          Error **errp)
1639 {
1640     MachineState *ms = MACHINE(obj);
1641     int64_t value = 0;
1642 
1643     if (ms->device_memory) {
1644         value = memory_region_size(&ms->device_memory->mr);
1645     }
1646 
1647     visit_type_int(v, name, &value, errp);
1648 }
1649 
1650 static void pc_machine_get_vmport(Object *obj, Visitor *v, const char *name,
1651                                   void *opaque, Error **errp)
1652 {
1653     PCMachineState *pcms = PC_MACHINE(obj);
1654     OnOffAuto vmport = pcms->vmport;
1655 
1656     visit_type_OnOffAuto(v, name, &vmport, errp);
1657 }
1658 
1659 static void pc_machine_set_vmport(Object *obj, Visitor *v, const char *name,
1660                                   void *opaque, Error **errp)
1661 {
1662     PCMachineState *pcms = PC_MACHINE(obj);
1663 
1664     visit_type_OnOffAuto(v, name, &pcms->vmport, errp);
1665 }
1666 
1667 static bool pc_machine_get_smbus(Object *obj, Error **errp)
1668 {
1669     PCMachineState *pcms = PC_MACHINE(obj);
1670 
1671     return pcms->smbus_enabled;
1672 }
1673 
1674 static void pc_machine_set_smbus(Object *obj, bool value, Error **errp)
1675 {
1676     PCMachineState *pcms = PC_MACHINE(obj);
1677 
1678     pcms->smbus_enabled = value;
1679 }
1680 
1681 static bool pc_machine_get_sata(Object *obj, Error **errp)
1682 {
1683     PCMachineState *pcms = PC_MACHINE(obj);
1684 
1685     return pcms->sata_enabled;
1686 }
1687 
1688 static void pc_machine_set_sata(Object *obj, bool value, Error **errp)
1689 {
1690     PCMachineState *pcms = PC_MACHINE(obj);
1691 
1692     pcms->sata_enabled = value;
1693 }
1694 
1695 static bool pc_machine_get_hpet(Object *obj, Error **errp)
1696 {
1697     PCMachineState *pcms = PC_MACHINE(obj);
1698 
1699     return pcms->hpet_enabled;
1700 }
1701 
1702 static void pc_machine_set_hpet(Object *obj, bool value, Error **errp)
1703 {
1704     PCMachineState *pcms = PC_MACHINE(obj);
1705 
1706     pcms->hpet_enabled = value;
1707 }
1708 
1709 static bool pc_machine_get_i8042(Object *obj, Error **errp)
1710 {
1711     PCMachineState *pcms = PC_MACHINE(obj);
1712 
1713     return pcms->i8042_enabled;
1714 }
1715 
1716 static void pc_machine_set_i8042(Object *obj, bool value, Error **errp)
1717 {
1718     PCMachineState *pcms = PC_MACHINE(obj);
1719 
1720     pcms->i8042_enabled = value;
1721 }
1722 
1723 static bool pc_machine_get_default_bus_bypass_iommu(Object *obj, Error **errp)
1724 {
1725     PCMachineState *pcms = PC_MACHINE(obj);
1726 
1727     return pcms->default_bus_bypass_iommu;
1728 }
1729 
1730 static void pc_machine_set_default_bus_bypass_iommu(Object *obj, bool value,
1731                                                     Error **errp)
1732 {
1733     PCMachineState *pcms = PC_MACHINE(obj);
1734 
1735     pcms->default_bus_bypass_iommu = value;
1736 }
1737 
1738 static void pc_machine_get_smbios_ep(Object *obj, Visitor *v, const char *name,
1739                                      void *opaque, Error **errp)
1740 {
1741     PCMachineState *pcms = PC_MACHINE(obj);
1742     SmbiosEntryPointType smbios_entry_point_type = pcms->smbios_entry_point_type;
1743 
1744     visit_type_SmbiosEntryPointType(v, name, &smbios_entry_point_type, errp);
1745 }
1746 
1747 static void pc_machine_set_smbios_ep(Object *obj, Visitor *v, const char *name,
1748                                      void *opaque, Error **errp)
1749 {
1750     PCMachineState *pcms = PC_MACHINE(obj);
1751 
1752     visit_type_SmbiosEntryPointType(v, name, &pcms->smbios_entry_point_type, errp);
1753 }
1754 
1755 static void pc_machine_get_max_ram_below_4g(Object *obj, Visitor *v,
1756                                             const char *name, void *opaque,
1757                                             Error **errp)
1758 {
1759     PCMachineState *pcms = PC_MACHINE(obj);
1760     uint64_t value = pcms->max_ram_below_4g;
1761 
1762     visit_type_size(v, name, &value, errp);
1763 }
1764 
1765 static void pc_machine_set_max_ram_below_4g(Object *obj, Visitor *v,
1766                                             const char *name, void *opaque,
1767                                             Error **errp)
1768 {
1769     PCMachineState *pcms = PC_MACHINE(obj);
1770     uint64_t value;
1771 
1772     if (!visit_type_size(v, name, &value, errp)) {
1773         return;
1774     }
1775     if (value > 4 * GiB) {
1776         error_setg(errp,
1777                    "Machine option 'max-ram-below-4g=%"PRIu64
1778                    "' expects size less than or equal to 4G", value);
1779         return;
1780     }
1781 
1782     if (value < 1 * MiB) {
1783         warn_report("Only %" PRIu64 " bytes of RAM below the 4GiB boundary,"
1784                     "BIOS may not work with less than 1MiB", value);
1785     }
1786 
1787     pcms->max_ram_below_4g = value;
1788 }
1789 
1790 static void pc_machine_get_max_fw_size(Object *obj, Visitor *v,
1791                                        const char *name, void *opaque,
1792                                        Error **errp)
1793 {
1794     PCMachineState *pcms = PC_MACHINE(obj);
1795     uint64_t value = pcms->max_fw_size;
1796 
1797     visit_type_size(v, name, &value, errp);
1798 }
1799 
1800 static void pc_machine_set_max_fw_size(Object *obj, Visitor *v,
1801                                        const char *name, void *opaque,
1802                                        Error **errp)
1803 {
1804     PCMachineState *pcms = PC_MACHINE(obj);
1805     uint64_t value;
1806 
1807     if (!visit_type_size(v, name, &value, errp)) {
1808         return;
1809     }
1810 
1811     /*
1812     * We don't have a theoretically justifiable exact lower bound on the base
1813     * address of any flash mapping. In practice, the IO-APIC MMIO range is
1814     * [0xFEE00000..0xFEE01000] -- see IO_APIC_DEFAULT_ADDRESS --, leaving free
1815     * only 18MB-4KB below 4G. For now, restrict the cumulative mapping to 8MB in
1816     * size.
1817     */
1818     if (value > 16 * MiB) {
1819         error_setg(errp,
1820                    "User specified max allowed firmware size %" PRIu64 " is "
1821                    "greater than 16MiB. If combined firwmare size exceeds "
1822                    "16MiB the system may not boot, or experience intermittent"
1823                    "stability issues.",
1824                    value);
1825         return;
1826     }
1827 
1828     pcms->max_fw_size = value;
1829 }
1830 
1831 
1832 static void pc_machine_initfn(Object *obj)
1833 {
1834     PCMachineState *pcms = PC_MACHINE(obj);
1835 
1836 #ifdef CONFIG_VMPORT
1837     pcms->vmport = ON_OFF_AUTO_AUTO;
1838 #else
1839     pcms->vmport = ON_OFF_AUTO_OFF;
1840 #endif /* CONFIG_VMPORT */
1841     pcms->max_ram_below_4g = 0; /* use default */
1842     pcms->smbios_entry_point_type = SMBIOS_ENTRY_POINT_TYPE_32;
1843 
1844     /* acpi build is enabled by default if machine supports it */
1845     pcms->acpi_build_enabled = PC_MACHINE_GET_CLASS(pcms)->has_acpi_build;
1846     pcms->smbus_enabled = true;
1847     pcms->sata_enabled = true;
1848     pcms->i8042_enabled = true;
1849     pcms->max_fw_size = 8 * MiB;
1850 #ifdef CONFIG_HPET
1851     pcms->hpet_enabled = true;
1852 #endif
1853     pcms->default_bus_bypass_iommu = false;
1854 
1855     pc_system_flash_create(pcms);
1856     pcms->pcspk = isa_new(TYPE_PC_SPEAKER);
1857     object_property_add_alias(OBJECT(pcms), "pcspk-audiodev",
1858                               OBJECT(pcms->pcspk), "audiodev");
1859     cxl_machine_init(obj, &pcms->cxl_devices_state);
1860 }
1861 
1862 int pc_machine_kvm_type(MachineState *machine, const char *kvm_type)
1863 {
1864 #ifdef CONFIG_XEN_EMU
1865     if (xen_mode == XEN_EMULATE) {
1866         xen_overlay_create();
1867         xen_evtchn_create();
1868         xen_gnttab_create();
1869         xen_xenstore_create();
1870     }
1871 #endif
1872     return 0;
1873 }
1874 
1875 static void pc_machine_reset(MachineState *machine, ShutdownCause reason)
1876 {
1877     CPUState *cs;
1878     X86CPU *cpu;
1879 
1880     qemu_devices_reset(reason);
1881 
1882     /* Reset APIC after devices have been reset to cancel
1883      * any changes that qemu_devices_reset() might have done.
1884      */
1885     CPU_FOREACH(cs) {
1886         cpu = X86_CPU(cs);
1887 
1888         x86_cpu_after_reset(cpu);
1889     }
1890 }
1891 
1892 static void pc_machine_wakeup(MachineState *machine)
1893 {
1894     cpu_synchronize_all_states();
1895     pc_machine_reset(machine, SHUTDOWN_CAUSE_NONE);
1896     cpu_synchronize_all_post_reset();
1897 }
1898 
1899 static bool pc_hotplug_allowed(MachineState *ms, DeviceState *dev, Error **errp)
1900 {
1901     X86IOMMUState *iommu = x86_iommu_get_default();
1902     IntelIOMMUState *intel_iommu;
1903 
1904     if (iommu &&
1905         object_dynamic_cast((Object *)iommu, TYPE_INTEL_IOMMU_DEVICE) &&
1906         object_dynamic_cast((Object *)dev, "vfio-pci")) {
1907         intel_iommu = INTEL_IOMMU_DEVICE(iommu);
1908         if (!intel_iommu->caching_mode) {
1909             error_setg(errp, "Device assignment is not allowed without "
1910                        "enabling caching-mode=on for Intel IOMMU.");
1911             return false;
1912         }
1913     }
1914 
1915     return true;
1916 }
1917 
1918 static void pc_machine_class_init(ObjectClass *oc, void *data)
1919 {
1920     MachineClass *mc = MACHINE_CLASS(oc);
1921     PCMachineClass *pcmc = PC_MACHINE_CLASS(oc);
1922     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
1923 
1924     pcmc->pci_enabled = true;
1925     pcmc->has_acpi_build = true;
1926     pcmc->rsdp_in_ram = true;
1927     pcmc->smbios_defaults = true;
1928     pcmc->smbios_uuid_encoded = true;
1929     pcmc->gigabyte_align = true;
1930     pcmc->has_reserved_memory = true;
1931     pcmc->kvmclock_enabled = true;
1932     pcmc->enforce_aligned_dimm = true;
1933     pcmc->enforce_amd_1tb_hole = true;
1934     /* BIOS ACPI tables: 128K. Other BIOS datastructures: less than 4K reported
1935      * to be used at the moment, 32K should be enough for a while.  */
1936     pcmc->acpi_data_size = 0x20000 + 0x8000;
1937     pcmc->pvh_enabled = true;
1938     pcmc->kvmclock_create_always = true;
1939     assert(!mc->get_hotplug_handler);
1940     mc->get_hotplug_handler = pc_get_hotplug_handler;
1941     mc->hotplug_allowed = pc_hotplug_allowed;
1942     mc->cpu_index_to_instance_props = x86_cpu_index_to_props;
1943     mc->get_default_cpu_node_id = x86_get_default_cpu_node_id;
1944     mc->possible_cpu_arch_ids = x86_possible_cpu_arch_ids;
1945     mc->auto_enable_numa_with_memhp = true;
1946     mc->auto_enable_numa_with_memdev = true;
1947     mc->has_hotpluggable_cpus = true;
1948     mc->default_boot_order = "cad";
1949     mc->block_default_type = IF_IDE;
1950     mc->max_cpus = 255;
1951     mc->reset = pc_machine_reset;
1952     mc->wakeup = pc_machine_wakeup;
1953     hc->pre_plug = pc_machine_device_pre_plug_cb;
1954     hc->plug = pc_machine_device_plug_cb;
1955     hc->unplug_request = pc_machine_device_unplug_request_cb;
1956     hc->unplug = pc_machine_device_unplug_cb;
1957     mc->default_cpu_type = TARGET_DEFAULT_CPU_TYPE;
1958     mc->nvdimm_supported = true;
1959     mc->smp_props.dies_supported = true;
1960     mc->default_ram_id = "pc.ram";
1961 
1962     object_class_property_add(oc, PC_MACHINE_MAX_RAM_BELOW_4G, "size",
1963         pc_machine_get_max_ram_below_4g, pc_machine_set_max_ram_below_4g,
1964         NULL, NULL);
1965     object_class_property_set_description(oc, PC_MACHINE_MAX_RAM_BELOW_4G,
1966         "Maximum ram below the 4G boundary (32bit boundary)");
1967 
1968     object_class_property_add(oc, PC_MACHINE_DEVMEM_REGION_SIZE, "int",
1969         pc_machine_get_device_memory_region_size, NULL,
1970         NULL, NULL);
1971 
1972     object_class_property_add(oc, PC_MACHINE_VMPORT, "OnOffAuto",
1973         pc_machine_get_vmport, pc_machine_set_vmport,
1974         NULL, NULL);
1975     object_class_property_set_description(oc, PC_MACHINE_VMPORT,
1976         "Enable vmport (pc & q35)");
1977 
1978     object_class_property_add_bool(oc, PC_MACHINE_SMBUS,
1979         pc_machine_get_smbus, pc_machine_set_smbus);
1980     object_class_property_set_description(oc, PC_MACHINE_SMBUS,
1981         "Enable/disable system management bus");
1982 
1983     object_class_property_add_bool(oc, PC_MACHINE_SATA,
1984         pc_machine_get_sata, pc_machine_set_sata);
1985     object_class_property_set_description(oc, PC_MACHINE_SATA,
1986         "Enable/disable Serial ATA bus");
1987 
1988     object_class_property_add_bool(oc, "hpet",
1989         pc_machine_get_hpet, pc_machine_set_hpet);
1990     object_class_property_set_description(oc, "hpet",
1991         "Enable/disable high precision event timer emulation");
1992 
1993     object_class_property_add_bool(oc, PC_MACHINE_I8042,
1994         pc_machine_get_i8042, pc_machine_set_i8042);
1995 
1996     object_class_property_add_bool(oc, "default-bus-bypass-iommu",
1997         pc_machine_get_default_bus_bypass_iommu,
1998         pc_machine_set_default_bus_bypass_iommu);
1999 
2000     object_class_property_add(oc, PC_MACHINE_MAX_FW_SIZE, "size",
2001         pc_machine_get_max_fw_size, pc_machine_set_max_fw_size,
2002         NULL, NULL);
2003     object_class_property_set_description(oc, PC_MACHINE_MAX_FW_SIZE,
2004         "Maximum combined firmware size");
2005 
2006     object_class_property_add(oc, PC_MACHINE_SMBIOS_EP, "str",
2007         pc_machine_get_smbios_ep, pc_machine_set_smbios_ep,
2008         NULL, NULL);
2009     object_class_property_set_description(oc, PC_MACHINE_SMBIOS_EP,
2010         "SMBIOS Entry Point type [32, 64]");
2011 }
2012 
2013 static const TypeInfo pc_machine_info = {
2014     .name = TYPE_PC_MACHINE,
2015     .parent = TYPE_X86_MACHINE,
2016     .abstract = true,
2017     .instance_size = sizeof(PCMachineState),
2018     .instance_init = pc_machine_initfn,
2019     .class_size = sizeof(PCMachineClass),
2020     .class_init = pc_machine_class_init,
2021     .interfaces = (InterfaceInfo[]) {
2022          { TYPE_HOTPLUG_HANDLER },
2023          { }
2024     },
2025 };
2026 
2027 static void pc_machine_register_types(void)
2028 {
2029     type_register_static(&pc_machine_info);
2030 }
2031 
2032 type_init(pc_machine_register_types)
2033