xref: /openbmc/qemu/hw/i386/pc.c (revision 243705aa)
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 #ifdef CONFIG_XEN_EMU
106 #include "hw/xen/xen-legacy-backend.h"
107 #include "hw/xen/xen-bus.h"
108 #endif
109 
110 /*
111  * Helper for setting model-id for CPU models that changed model-id
112  * depending on QEMU versions up to QEMU 2.4.
113  */
114 #define PC_CPU_MODEL_IDS(v) \
115     { "qemu32-" TYPE_X86_CPU, "model-id", "QEMU Virtual CPU version " v, },\
116     { "qemu64-" TYPE_X86_CPU, "model-id", "QEMU Virtual CPU version " v, },\
117     { "athlon-" TYPE_X86_CPU, "model-id", "QEMU Virtual CPU version " v, },
118 
119 GlobalProperty pc_compat_8_0[] = {
120     { "virtio-mem", "unplugged-inaccessible", "auto" },
121 };
122 const size_t pc_compat_8_0_len = G_N_ELEMENTS(pc_compat_8_0);
123 
124 GlobalProperty pc_compat_7_2[] = {
125     { "ICH9-LPC", "noreboot", "true" },
126 };
127 const size_t pc_compat_7_2_len = G_N_ELEMENTS(pc_compat_7_2);
128 
129 GlobalProperty pc_compat_7_1[] = {};
130 const size_t pc_compat_7_1_len = G_N_ELEMENTS(pc_compat_7_1);
131 
132 GlobalProperty pc_compat_7_0[] = {};
133 const size_t pc_compat_7_0_len = G_N_ELEMENTS(pc_compat_7_0);
134 
135 GlobalProperty pc_compat_6_2[] = {
136     { "virtio-mem", "unplugged-inaccessible", "off" },
137 };
138 const size_t pc_compat_6_2_len = G_N_ELEMENTS(pc_compat_6_2);
139 
140 GlobalProperty pc_compat_6_1[] = {
141     { TYPE_X86_CPU, "hv-version-id-build", "0x1bbc" },
142     { TYPE_X86_CPU, "hv-version-id-major", "0x0006" },
143     { TYPE_X86_CPU, "hv-version-id-minor", "0x0001" },
144     { "ICH9-LPC", "x-keep-pci-slot-hpc", "false" },
145 };
146 const size_t pc_compat_6_1_len = G_N_ELEMENTS(pc_compat_6_1);
147 
148 GlobalProperty pc_compat_6_0[] = {
149     { "qemu64" "-" TYPE_X86_CPU, "family", "6" },
150     { "qemu64" "-" TYPE_X86_CPU, "model", "6" },
151     { "qemu64" "-" TYPE_X86_CPU, "stepping", "3" },
152     { TYPE_X86_CPU, "x-vendor-cpuid-only", "off" },
153     { "ICH9-LPC", ACPI_PM_PROP_ACPI_PCIHP_BRIDGE, "off" },
154     { "ICH9-LPC", "x-keep-pci-slot-hpc", "true" },
155 };
156 const size_t pc_compat_6_0_len = G_N_ELEMENTS(pc_compat_6_0);
157 
158 GlobalProperty pc_compat_5_2[] = {
159     { "ICH9-LPC", "x-smi-cpu-hotunplug", "off" },
160 };
161 const size_t pc_compat_5_2_len = G_N_ELEMENTS(pc_compat_5_2);
162 
163 GlobalProperty pc_compat_5_1[] = {
164     { "ICH9-LPC", "x-smi-cpu-hotplug", "off" },
165     { TYPE_X86_CPU, "kvm-msi-ext-dest-id", "off" },
166 };
167 const size_t pc_compat_5_1_len = G_N_ELEMENTS(pc_compat_5_1);
168 
169 GlobalProperty pc_compat_5_0[] = {
170 };
171 const size_t pc_compat_5_0_len = G_N_ELEMENTS(pc_compat_5_0);
172 
173 GlobalProperty pc_compat_4_2[] = {
174     { "mch", "smbase-smram", "off" },
175 };
176 const size_t pc_compat_4_2_len = G_N_ELEMENTS(pc_compat_4_2);
177 
178 GlobalProperty pc_compat_4_1[] = {};
179 const size_t pc_compat_4_1_len = G_N_ELEMENTS(pc_compat_4_1);
180 
181 GlobalProperty pc_compat_4_0[] = {};
182 const size_t pc_compat_4_0_len = G_N_ELEMENTS(pc_compat_4_0);
183 
184 GlobalProperty pc_compat_3_1[] = {
185     { "intel-iommu", "dma-drain", "off" },
186     { "Opteron_G3" "-" TYPE_X86_CPU, "rdtscp", "off" },
187     { "Opteron_G4" "-" TYPE_X86_CPU, "rdtscp", "off" },
188     { "Opteron_G4" "-" TYPE_X86_CPU, "npt", "off" },
189     { "Opteron_G4" "-" TYPE_X86_CPU, "nrip-save", "off" },
190     { "Opteron_G5" "-" TYPE_X86_CPU, "rdtscp", "off" },
191     { "Opteron_G5" "-" TYPE_X86_CPU, "npt", "off" },
192     { "Opteron_G5" "-" TYPE_X86_CPU, "nrip-save", "off" },
193     { "EPYC" "-" TYPE_X86_CPU, "npt", "off" },
194     { "EPYC" "-" TYPE_X86_CPU, "nrip-save", "off" },
195     { "EPYC-IBPB" "-" TYPE_X86_CPU, "npt", "off" },
196     { "EPYC-IBPB" "-" TYPE_X86_CPU, "nrip-save", "off" },
197     { "Skylake-Client" "-" TYPE_X86_CPU,      "mpx", "on" },
198     { "Skylake-Client-IBRS" "-" TYPE_X86_CPU, "mpx", "on" },
199     { "Skylake-Server" "-" TYPE_X86_CPU,      "mpx", "on" },
200     { "Skylake-Server-IBRS" "-" TYPE_X86_CPU, "mpx", "on" },
201     { "Cascadelake-Server" "-" TYPE_X86_CPU,  "mpx", "on" },
202     { "Icelake-Client" "-" TYPE_X86_CPU,      "mpx", "on" },
203     { "Icelake-Server" "-" TYPE_X86_CPU,      "mpx", "on" },
204     { "Cascadelake-Server" "-" TYPE_X86_CPU, "stepping", "5" },
205     { TYPE_X86_CPU, "x-intel-pt-auto-level", "off" },
206 };
207 const size_t pc_compat_3_1_len = G_N_ELEMENTS(pc_compat_3_1);
208 
209 GlobalProperty pc_compat_3_0[] = {
210     { TYPE_X86_CPU, "x-hv-synic-kvm-only", "on" },
211     { "Skylake-Server" "-" TYPE_X86_CPU, "pku", "off" },
212     { "Skylake-Server-IBRS" "-" TYPE_X86_CPU, "pku", "off" },
213 };
214 const size_t pc_compat_3_0_len = G_N_ELEMENTS(pc_compat_3_0);
215 
216 GlobalProperty pc_compat_2_12[] = {
217     { TYPE_X86_CPU, "legacy-cache", "on" },
218     { TYPE_X86_CPU, "topoext", "off" },
219     { "EPYC-" TYPE_X86_CPU, "xlevel", "0x8000000a" },
220     { "EPYC-IBPB-" TYPE_X86_CPU, "xlevel", "0x8000000a" },
221 };
222 const size_t pc_compat_2_12_len = G_N_ELEMENTS(pc_compat_2_12);
223 
224 GlobalProperty pc_compat_2_11[] = {
225     { TYPE_X86_CPU, "x-migrate-smi-count", "off" },
226     { "Skylake-Server" "-" TYPE_X86_CPU, "clflushopt", "off" },
227 };
228 const size_t pc_compat_2_11_len = G_N_ELEMENTS(pc_compat_2_11);
229 
230 GlobalProperty pc_compat_2_10[] = {
231     { TYPE_X86_CPU, "x-hv-max-vps", "0x40" },
232     { "i440FX-pcihost", "x-pci-hole64-fix", "off" },
233     { "q35-pcihost", "x-pci-hole64-fix", "off" },
234 };
235 const size_t pc_compat_2_10_len = G_N_ELEMENTS(pc_compat_2_10);
236 
237 GlobalProperty pc_compat_2_9[] = {
238     { "mch", "extended-tseg-mbytes", "0" },
239 };
240 const size_t pc_compat_2_9_len = G_N_ELEMENTS(pc_compat_2_9);
241 
242 GlobalProperty pc_compat_2_8[] = {
243     { TYPE_X86_CPU, "tcg-cpuid", "off" },
244     { "kvmclock", "x-mach-use-reliable-get-clock", "off" },
245     { "ICH9-LPC", "x-smi-broadcast", "off" },
246     { TYPE_X86_CPU, "vmware-cpuid-freq", "off" },
247     { "Haswell-" TYPE_X86_CPU, "stepping", "1" },
248 };
249 const size_t pc_compat_2_8_len = G_N_ELEMENTS(pc_compat_2_8);
250 
251 GlobalProperty pc_compat_2_7[] = {
252     { TYPE_X86_CPU, "l3-cache", "off" },
253     { TYPE_X86_CPU, "full-cpuid-auto-level", "off" },
254     { "Opteron_G3" "-" TYPE_X86_CPU, "family", "15" },
255     { "Opteron_G3" "-" TYPE_X86_CPU, "model", "6" },
256     { "Opteron_G3" "-" TYPE_X86_CPU, "stepping", "1" },
257     { "isa-pcspk", "migrate", "off" },
258 };
259 const size_t pc_compat_2_7_len = G_N_ELEMENTS(pc_compat_2_7);
260 
261 GlobalProperty pc_compat_2_6[] = {
262     { TYPE_X86_CPU, "cpuid-0xb", "off" },
263     { "vmxnet3", "romfile", "" },
264     { TYPE_X86_CPU, "fill-mtrr-mask", "off" },
265     { "apic-common", "legacy-instance-id", "on", }
266 };
267 const size_t pc_compat_2_6_len = G_N_ELEMENTS(pc_compat_2_6);
268 
269 GlobalProperty pc_compat_2_5[] = {};
270 const size_t pc_compat_2_5_len = G_N_ELEMENTS(pc_compat_2_5);
271 
272 GlobalProperty pc_compat_2_4[] = {
273     PC_CPU_MODEL_IDS("2.4.0")
274     { "Haswell-" TYPE_X86_CPU, "abm", "off" },
275     { "Haswell-noTSX-" TYPE_X86_CPU, "abm", "off" },
276     { "Broadwell-" TYPE_X86_CPU, "abm", "off" },
277     { "Broadwell-noTSX-" TYPE_X86_CPU, "abm", "off" },
278     { "host" "-" TYPE_X86_CPU, "host-cache-info", "on" },
279     { TYPE_X86_CPU, "check", "off" },
280     { "qemu64" "-" TYPE_X86_CPU, "sse4a", "on" },
281     { "qemu64" "-" TYPE_X86_CPU, "abm", "on" },
282     { "qemu64" "-" TYPE_X86_CPU, "popcnt", "on" },
283     { "qemu32" "-" TYPE_X86_CPU, "popcnt", "on" },
284     { "Opteron_G2" "-" TYPE_X86_CPU, "rdtscp", "on" },
285     { "Opteron_G3" "-" TYPE_X86_CPU, "rdtscp", "on" },
286     { "Opteron_G4" "-" TYPE_X86_CPU, "rdtscp", "on" },
287     { "Opteron_G5" "-" TYPE_X86_CPU, "rdtscp", "on", }
288 };
289 const size_t pc_compat_2_4_len = G_N_ELEMENTS(pc_compat_2_4);
290 
291 GlobalProperty pc_compat_2_3[] = {
292     PC_CPU_MODEL_IDS("2.3.0")
293     { TYPE_X86_CPU, "arat", "off" },
294     { "qemu64" "-" TYPE_X86_CPU, "min-level", "4" },
295     { "kvm64" "-" TYPE_X86_CPU, "min-level", "5" },
296     { "pentium3" "-" TYPE_X86_CPU, "min-level", "2" },
297     { "n270" "-" TYPE_X86_CPU, "min-level", "5" },
298     { "Conroe" "-" TYPE_X86_CPU, "min-level", "4" },
299     { "Penryn" "-" TYPE_X86_CPU, "min-level", "4" },
300     { "Nehalem" "-" TYPE_X86_CPU, "min-level", "4" },
301     { "n270" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
302     { "Penryn" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
303     { "Conroe" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
304     { "Nehalem" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
305     { "Westmere" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
306     { "SandyBridge" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
307     { "IvyBridge" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
308     { "Haswell" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
309     { "Haswell-noTSX" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
310     { "Broadwell" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
311     { "Broadwell-noTSX" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
312     { TYPE_X86_CPU, "kvm-no-smi-migration", "on" },
313 };
314 const size_t pc_compat_2_3_len = G_N_ELEMENTS(pc_compat_2_3);
315 
316 GlobalProperty pc_compat_2_2[] = {
317     PC_CPU_MODEL_IDS("2.2.0")
318     { "kvm64" "-" TYPE_X86_CPU, "vme", "off" },
319     { "kvm32" "-" TYPE_X86_CPU, "vme", "off" },
320     { "Conroe" "-" TYPE_X86_CPU, "vme", "off" },
321     { "Penryn" "-" TYPE_X86_CPU, "vme", "off" },
322     { "Nehalem" "-" TYPE_X86_CPU, "vme", "off" },
323     { "Westmere" "-" TYPE_X86_CPU, "vme", "off" },
324     { "SandyBridge" "-" TYPE_X86_CPU, "vme", "off" },
325     { "Haswell" "-" TYPE_X86_CPU, "vme", "off" },
326     { "Broadwell" "-" TYPE_X86_CPU, "vme", "off" },
327     { "Opteron_G1" "-" TYPE_X86_CPU, "vme", "off" },
328     { "Opteron_G2" "-" TYPE_X86_CPU, "vme", "off" },
329     { "Opteron_G3" "-" TYPE_X86_CPU, "vme", "off" },
330     { "Opteron_G4" "-" TYPE_X86_CPU, "vme", "off" },
331     { "Opteron_G5" "-" TYPE_X86_CPU, "vme", "off" },
332     { "Haswell" "-" TYPE_X86_CPU, "f16c", "off" },
333     { "Haswell" "-" TYPE_X86_CPU, "rdrand", "off" },
334     { "Broadwell" "-" TYPE_X86_CPU, "f16c", "off" },
335     { "Broadwell" "-" TYPE_X86_CPU, "rdrand", "off" },
336 };
337 const size_t pc_compat_2_2_len = G_N_ELEMENTS(pc_compat_2_2);
338 
339 GlobalProperty pc_compat_2_1[] = {
340     PC_CPU_MODEL_IDS("2.1.0")
341     { "coreduo" "-" TYPE_X86_CPU, "vmx", "on" },
342     { "core2duo" "-" TYPE_X86_CPU, "vmx", "on" },
343 };
344 const size_t pc_compat_2_1_len = G_N_ELEMENTS(pc_compat_2_1);
345 
346 GlobalProperty pc_compat_2_0[] = {
347     PC_CPU_MODEL_IDS("2.0.0")
348     { "virtio-scsi-pci", "any_layout", "off" },
349     { "PIIX4_PM", "memory-hotplug-support", "off" },
350     { "apic", "version", "0x11" },
351     { "nec-usb-xhci", "superspeed-ports-first", "off" },
352     { "nec-usb-xhci", "force-pcie-endcap", "on" },
353     { "pci-serial", "prog_if", "0" },
354     { "pci-serial-2x", "prog_if", "0" },
355     { "pci-serial-4x", "prog_if", "0" },
356     { "virtio-net-pci", "guest_announce", "off" },
357     { "ICH9-LPC", "memory-hotplug-support", "off" },
358 };
359 const size_t pc_compat_2_0_len = G_N_ELEMENTS(pc_compat_2_0);
360 
361 GlobalProperty pc_compat_1_7[] = {
362     PC_CPU_MODEL_IDS("1.7.0")
363     { TYPE_USB_DEVICE, "msos-desc", "no" },
364     { "PIIX4_PM", ACPI_PM_PROP_ACPI_PCIHP_BRIDGE, "off" },
365     { "hpet", HPET_INTCAP, "4" },
366 };
367 const size_t pc_compat_1_7_len = G_N_ELEMENTS(pc_compat_1_7);
368 
369 GlobalProperty pc_compat_1_6[] = {
370     PC_CPU_MODEL_IDS("1.6.0")
371     { "e1000", "mitigation", "off" },
372     { "qemu64-" TYPE_X86_CPU, "model", "2" },
373     { "qemu32-" TYPE_X86_CPU, "model", "3" },
374     { "i440FX-pcihost", "short_root_bus", "1" },
375     { "q35-pcihost", "short_root_bus", "1" },
376 };
377 const size_t pc_compat_1_6_len = G_N_ELEMENTS(pc_compat_1_6);
378 
379 GlobalProperty pc_compat_1_5[] = {
380     PC_CPU_MODEL_IDS("1.5.0")
381     { "Conroe-" TYPE_X86_CPU, "model", "2" },
382     { "Conroe-" TYPE_X86_CPU, "min-level", "2" },
383     { "Penryn-" TYPE_X86_CPU, "model", "2" },
384     { "Penryn-" TYPE_X86_CPU, "min-level", "2" },
385     { "Nehalem-" TYPE_X86_CPU, "model", "2" },
386     { "Nehalem-" TYPE_X86_CPU, "min-level", "2" },
387     { "virtio-net-pci", "any_layout", "off" },
388     { TYPE_X86_CPU, "pmu", "on" },
389     { "i440FX-pcihost", "short_root_bus", "0" },
390     { "q35-pcihost", "short_root_bus", "0" },
391 };
392 const size_t pc_compat_1_5_len = G_N_ELEMENTS(pc_compat_1_5);
393 
394 GlobalProperty pc_compat_1_4[] = {
395     PC_CPU_MODEL_IDS("1.4.0")
396     { "scsi-hd", "discard_granularity", "0" },
397     { "scsi-cd", "discard_granularity", "0" },
398     { "ide-hd", "discard_granularity", "0" },
399     { "ide-cd", "discard_granularity", "0" },
400     { "virtio-blk-pci", "discard_granularity", "0" },
401     /* DEV_NVECTORS_UNSPECIFIED as a uint32_t string: */
402     { "virtio-serial-pci", "vectors", "0xFFFFFFFF" },
403     { "virtio-net-pci", "ctrl_guest_offloads", "off" },
404     { "e1000", "romfile", "pxe-e1000.rom" },
405     { "ne2k_pci", "romfile", "pxe-ne2k_pci.rom" },
406     { "pcnet", "romfile", "pxe-pcnet.rom" },
407     { "rtl8139", "romfile", "pxe-rtl8139.rom" },
408     { "virtio-net-pci", "romfile", "pxe-virtio.rom" },
409     { "486-" TYPE_X86_CPU, "model", "0" },
410     { "n270" "-" TYPE_X86_CPU, "movbe", "off" },
411     { "Westmere" "-" TYPE_X86_CPU, "pclmulqdq", "off" },
412 };
413 const size_t pc_compat_1_4_len = G_N_ELEMENTS(pc_compat_1_4);
414 
415 GSIState *pc_gsi_create(qemu_irq **irqs, bool pci_enabled)
416 {
417     GSIState *s;
418 
419     s = g_new0(GSIState, 1);
420     if (kvm_ioapic_in_kernel()) {
421         kvm_pc_setup_irq_routing(pci_enabled);
422     }
423     *irqs = qemu_allocate_irqs(gsi_handler, s, IOAPIC_NUM_PINS);
424 
425     return s;
426 }
427 
428 static void ioport80_write(void *opaque, hwaddr addr, uint64_t data,
429                            unsigned size)
430 {
431 }
432 
433 static uint64_t ioport80_read(void *opaque, hwaddr addr, unsigned size)
434 {
435     return 0xffffffffffffffffULL;
436 }
437 
438 /* MSDOS compatibility mode FPU exception support */
439 static void ioportF0_write(void *opaque, hwaddr addr, uint64_t data,
440                            unsigned size)
441 {
442     if (tcg_enabled()) {
443         cpu_set_ignne();
444     }
445 }
446 
447 static uint64_t ioportF0_read(void *opaque, hwaddr addr, unsigned size)
448 {
449     return 0xffffffffffffffffULL;
450 }
451 
452 /* PC cmos mappings */
453 
454 #define REG_EQUIPMENT_BYTE          0x14
455 
456 static void cmos_init_hd(MC146818RtcState *s, int type_ofs, int info_ofs,
457                          int16_t cylinders, int8_t heads, int8_t sectors)
458 {
459     mc146818rtc_set_cmos_data(s, type_ofs, 47);
460     mc146818rtc_set_cmos_data(s, info_ofs, cylinders);
461     mc146818rtc_set_cmos_data(s, info_ofs + 1, cylinders >> 8);
462     mc146818rtc_set_cmos_data(s, info_ofs + 2, heads);
463     mc146818rtc_set_cmos_data(s, info_ofs + 3, 0xff);
464     mc146818rtc_set_cmos_data(s, info_ofs + 4, 0xff);
465     mc146818rtc_set_cmos_data(s, info_ofs + 5, 0xc0 | ((heads > 8) << 3));
466     mc146818rtc_set_cmos_data(s, info_ofs + 6, cylinders);
467     mc146818rtc_set_cmos_data(s, info_ofs + 7, cylinders >> 8);
468     mc146818rtc_set_cmos_data(s, info_ofs + 8, sectors);
469 }
470 
471 /* convert boot_device letter to something recognizable by the bios */
472 static int boot_device2nibble(char boot_device)
473 {
474     switch(boot_device) {
475     case 'a':
476     case 'b':
477         return 0x01; /* floppy boot */
478     case 'c':
479         return 0x02; /* hard drive boot */
480     case 'd':
481         return 0x03; /* CD-ROM boot */
482     case 'n':
483         return 0x04; /* Network boot */
484     }
485     return 0;
486 }
487 
488 static void set_boot_dev(MC146818RtcState *s, const char *boot_device,
489                          Error **errp)
490 {
491 #define PC_MAX_BOOT_DEVICES 3
492     int nbds, bds[3] = { 0, };
493     int i;
494 
495     nbds = strlen(boot_device);
496     if (nbds > PC_MAX_BOOT_DEVICES) {
497         error_setg(errp, "Too many boot devices for PC");
498         return;
499     }
500     for (i = 0; i < nbds; i++) {
501         bds[i] = boot_device2nibble(boot_device[i]);
502         if (bds[i] == 0) {
503             error_setg(errp, "Invalid boot device for PC: '%c'",
504                        boot_device[i]);
505             return;
506         }
507     }
508     mc146818rtc_set_cmos_data(s, 0x3d, (bds[1] << 4) | bds[0]);
509     mc146818rtc_set_cmos_data(s, 0x38, (bds[2] << 4) | (fd_bootchk ? 0x0 : 0x1));
510 }
511 
512 static void pc_boot_set(void *opaque, const char *boot_device, Error **errp)
513 {
514     set_boot_dev(opaque, boot_device, errp);
515 }
516 
517 static void pc_cmos_init_floppy(MC146818RtcState *rtc_state, ISADevice *floppy)
518 {
519     int val, nb, i;
520     FloppyDriveType fd_type[2] = { FLOPPY_DRIVE_TYPE_NONE,
521                                    FLOPPY_DRIVE_TYPE_NONE };
522 
523     /* floppy type */
524     if (floppy) {
525         for (i = 0; i < 2; i++) {
526             fd_type[i] = isa_fdc_get_drive_type(floppy, i);
527         }
528     }
529     val = (cmos_get_fd_drive_type(fd_type[0]) << 4) |
530         cmos_get_fd_drive_type(fd_type[1]);
531     mc146818rtc_set_cmos_data(rtc_state, 0x10, val);
532 
533     val = mc146818rtc_get_cmos_data(rtc_state, REG_EQUIPMENT_BYTE);
534     nb = 0;
535     if (fd_type[0] != FLOPPY_DRIVE_TYPE_NONE) {
536         nb++;
537     }
538     if (fd_type[1] != FLOPPY_DRIVE_TYPE_NONE) {
539         nb++;
540     }
541     switch (nb) {
542     case 0:
543         break;
544     case 1:
545         val |= 0x01; /* 1 drive, ready for boot */
546         break;
547     case 2:
548         val |= 0x41; /* 2 drives, ready for boot */
549         break;
550     }
551     mc146818rtc_set_cmos_data(rtc_state, REG_EQUIPMENT_BYTE, val);
552 }
553 
554 typedef struct pc_cmos_init_late_arg {
555     MC146818RtcState *rtc_state;
556     BusState *idebus[2];
557 } pc_cmos_init_late_arg;
558 
559 typedef struct check_fdc_state {
560     ISADevice *floppy;
561     bool multiple;
562 } CheckFdcState;
563 
564 static int check_fdc(Object *obj, void *opaque)
565 {
566     CheckFdcState *state = opaque;
567     Object *fdc;
568     uint32_t iobase;
569     Error *local_err = NULL;
570 
571     fdc = object_dynamic_cast(obj, TYPE_ISA_FDC);
572     if (!fdc) {
573         return 0;
574     }
575 
576     iobase = object_property_get_uint(obj, "iobase", &local_err);
577     if (local_err || iobase != 0x3f0) {
578         error_free(local_err);
579         return 0;
580     }
581 
582     if (state->floppy) {
583         state->multiple = true;
584     } else {
585         state->floppy = ISA_DEVICE(obj);
586     }
587     return 0;
588 }
589 
590 static const char * const fdc_container_path[] = {
591     "/unattached", "/peripheral", "/peripheral-anon"
592 };
593 
594 /*
595  * Locate the FDC at IO address 0x3f0, in order to configure the CMOS registers
596  * and ACPI objects.
597  */
598 static ISADevice *pc_find_fdc0(void)
599 {
600     int i;
601     Object *container;
602     CheckFdcState state = { 0 };
603 
604     for (i = 0; i < ARRAY_SIZE(fdc_container_path); i++) {
605         container = container_get(qdev_get_machine(), fdc_container_path[i]);
606         object_child_foreach(container, check_fdc, &state);
607     }
608 
609     if (state.multiple) {
610         warn_report("multiple floppy disk controllers with "
611                     "iobase=0x3f0 have been found");
612         error_printf("the one being picked for CMOS setup might not reflect "
613                      "your intent");
614     }
615 
616     return state.floppy;
617 }
618 
619 static void pc_cmos_init_late(void *opaque)
620 {
621     pc_cmos_init_late_arg *arg = opaque;
622     MC146818RtcState *s = arg->rtc_state;
623     int16_t cylinders;
624     int8_t heads, sectors;
625     int val;
626     int i, trans;
627 
628     val = 0;
629     if (arg->idebus[0] && ide_get_geometry(arg->idebus[0], 0,
630                                            &cylinders, &heads, &sectors) >= 0) {
631         cmos_init_hd(s, 0x19, 0x1b, cylinders, heads, sectors);
632         val |= 0xf0;
633     }
634     if (arg->idebus[0] && ide_get_geometry(arg->idebus[0], 1,
635                                            &cylinders, &heads, &sectors) >= 0) {
636         cmos_init_hd(s, 0x1a, 0x24, cylinders, heads, sectors);
637         val |= 0x0f;
638     }
639     mc146818rtc_set_cmos_data(s, 0x12, val);
640 
641     val = 0;
642     for (i = 0; i < 4; i++) {
643         /* NOTE: ide_get_geometry() returns the physical
644            geometry.  It is always such that: 1 <= sects <= 63, 1
645            <= heads <= 16, 1 <= cylinders <= 16383. The BIOS
646            geometry can be different if a translation is done. */
647         if (arg->idebus[i / 2] &&
648             ide_get_geometry(arg->idebus[i / 2], i % 2,
649                              &cylinders, &heads, &sectors) >= 0) {
650             trans = ide_get_bios_chs_trans(arg->idebus[i / 2], i % 2) - 1;
651             assert((trans & ~3) == 0);
652             val |= trans << (i * 2);
653         }
654     }
655     mc146818rtc_set_cmos_data(s, 0x39, val);
656 
657     pc_cmos_init_floppy(s, pc_find_fdc0());
658 
659     qemu_unregister_reset(pc_cmos_init_late, opaque);
660 }
661 
662 void pc_cmos_init(PCMachineState *pcms,
663                   BusState *idebus0, BusState *idebus1,
664                   ISADevice *rtc)
665 {
666     int val;
667     static pc_cmos_init_late_arg arg;
668     X86MachineState *x86ms = X86_MACHINE(pcms);
669     MC146818RtcState *s = MC146818_RTC(rtc);
670 
671     /* various important CMOS locations needed by PC/Bochs bios */
672 
673     /* memory size */
674     /* base memory (first MiB) */
675     val = MIN(x86ms->below_4g_mem_size / KiB, 640);
676     mc146818rtc_set_cmos_data(s, 0x15, val);
677     mc146818rtc_set_cmos_data(s, 0x16, val >> 8);
678     /* extended memory (next 64MiB) */
679     if (x86ms->below_4g_mem_size > 1 * MiB) {
680         val = (x86ms->below_4g_mem_size - 1 * MiB) / KiB;
681     } else {
682         val = 0;
683     }
684     if (val > 65535)
685         val = 65535;
686     mc146818rtc_set_cmos_data(s, 0x17, val);
687     mc146818rtc_set_cmos_data(s, 0x18, val >> 8);
688     mc146818rtc_set_cmos_data(s, 0x30, val);
689     mc146818rtc_set_cmos_data(s, 0x31, val >> 8);
690     /* memory between 16MiB and 4GiB */
691     if (x86ms->below_4g_mem_size > 16 * MiB) {
692         val = (x86ms->below_4g_mem_size - 16 * MiB) / (64 * KiB);
693     } else {
694         val = 0;
695     }
696     if (val > 65535)
697         val = 65535;
698     mc146818rtc_set_cmos_data(s, 0x34, val);
699     mc146818rtc_set_cmos_data(s, 0x35, val >> 8);
700     /* memory above 4GiB */
701     val = x86ms->above_4g_mem_size / 65536;
702     mc146818rtc_set_cmos_data(s, 0x5b, val);
703     mc146818rtc_set_cmos_data(s, 0x5c, val >> 8);
704     mc146818rtc_set_cmos_data(s, 0x5d, val >> 16);
705 
706     object_property_add_link(OBJECT(pcms), "rtc_state",
707                              TYPE_ISA_DEVICE,
708                              (Object **)&x86ms->rtc,
709                              object_property_allow_set_link,
710                              OBJ_PROP_LINK_STRONG);
711     object_property_set_link(OBJECT(pcms), "rtc_state", OBJECT(s),
712                              &error_abort);
713 
714     set_boot_dev(s, MACHINE(pcms)->boot_config.order, &error_fatal);
715 
716     val = 0;
717     val |= 0x02; /* FPU is there */
718     val |= 0x04; /* PS/2 mouse installed */
719     mc146818rtc_set_cmos_data(s, REG_EQUIPMENT_BYTE, val);
720 
721     /* hard drives and FDC */
722     arg.rtc_state = s;
723     arg.idebus[0] = idebus0;
724     arg.idebus[1] = idebus1;
725     qemu_register_reset(pc_cmos_init_late, &arg);
726 }
727 
728 static void handle_a20_line_change(void *opaque, int irq, int level)
729 {
730     X86CPU *cpu = opaque;
731 
732     /* XXX: send to all CPUs ? */
733     /* XXX: add logic to handle multiple A20 line sources */
734     x86_cpu_set_a20(cpu, level);
735 }
736 
737 #define NE2000_NB_MAX 6
738 
739 static const int ne2000_io[NE2000_NB_MAX] = { 0x300, 0x320, 0x340, 0x360,
740                                               0x280, 0x380 };
741 static const int ne2000_irq[NE2000_NB_MAX] = { 9, 10, 11, 3, 4, 5 };
742 
743 static void pc_init_ne2k_isa(ISABus *bus, NICInfo *nd)
744 {
745     static int nb_ne2k = 0;
746 
747     if (nb_ne2k == NE2000_NB_MAX)
748         return;
749     isa_ne2000_init(bus, ne2000_io[nb_ne2k],
750                     ne2000_irq[nb_ne2k], nd);
751     nb_ne2k++;
752 }
753 
754 void pc_acpi_smi_interrupt(void *opaque, int irq, int level)
755 {
756     X86CPU *cpu = opaque;
757 
758     if (level) {
759         cpu_interrupt(CPU(cpu), CPU_INTERRUPT_SMI);
760     }
761 }
762 
763 static
764 void pc_machine_done(Notifier *notifier, void *data)
765 {
766     PCMachineState *pcms = container_of(notifier,
767                                         PCMachineState, machine_done);
768     X86MachineState *x86ms = X86_MACHINE(pcms);
769 
770     cxl_hook_up_pxb_registers(pcms->bus, &pcms->cxl_devices_state,
771                               &error_fatal);
772 
773     if (pcms->cxl_devices_state.is_enabled) {
774         cxl_fmws_link_targets(&pcms->cxl_devices_state, &error_fatal);
775     }
776 
777     /* set the number of CPUs */
778     x86_rtc_set_cpus_count(x86ms->rtc, x86ms->boot_cpus);
779 
780     fw_cfg_add_extra_pci_roots(pcms->bus, x86ms->fw_cfg);
781 
782     acpi_setup();
783     if (x86ms->fw_cfg) {
784         fw_cfg_build_smbios(MACHINE(pcms), x86ms->fw_cfg);
785         fw_cfg_build_feature_control(MACHINE(pcms), x86ms->fw_cfg);
786         /* update FW_CFG_NB_CPUS to account for -device added CPUs */
787         fw_cfg_modify_i16(x86ms->fw_cfg, FW_CFG_NB_CPUS, x86ms->boot_cpus);
788     }
789 }
790 
791 void pc_guest_info_init(PCMachineState *pcms)
792 {
793     X86MachineState *x86ms = X86_MACHINE(pcms);
794 
795     x86ms->apic_xrupt_override = true;
796     pcms->machine_done.notify = pc_machine_done;
797     qemu_add_machine_init_done_notifier(&pcms->machine_done);
798 }
799 
800 /* setup pci memory address space mapping into system address space */
801 void pc_pci_as_mapping_init(MemoryRegion *system_memory,
802                             MemoryRegion *pci_address_space)
803 {
804     /* Set to lower priority than RAM */
805     memory_region_add_subregion_overlap(system_memory, 0x0,
806                                         pci_address_space, -1);
807 }
808 
809 void xen_load_linux(PCMachineState *pcms)
810 {
811     int i;
812     FWCfgState *fw_cfg;
813     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
814     X86MachineState *x86ms = X86_MACHINE(pcms);
815 
816     assert(MACHINE(pcms)->kernel_filename != NULL);
817 
818     fw_cfg = fw_cfg_init_io(FW_CFG_IO_BASE);
819     fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, x86ms->boot_cpus);
820     rom_set_fw(fw_cfg);
821 
822     x86_load_linux(x86ms, fw_cfg, pcmc->acpi_data_size,
823                    pcmc->pvh_enabled);
824     for (i = 0; i < nb_option_roms; i++) {
825         assert(!strcmp(option_rom[i].name, "linuxboot.bin") ||
826                !strcmp(option_rom[i].name, "linuxboot_dma.bin") ||
827                !strcmp(option_rom[i].name, "pvh.bin") ||
828                !strcmp(option_rom[i].name, "multiboot.bin") ||
829                !strcmp(option_rom[i].name, "multiboot_dma.bin"));
830         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
831     }
832     x86ms->fw_cfg = fw_cfg;
833 }
834 
835 #define PC_ROM_MIN_VGA     0xc0000
836 #define PC_ROM_MIN_OPTION  0xc8000
837 #define PC_ROM_MAX         0xe0000
838 #define PC_ROM_ALIGN       0x800
839 #define PC_ROM_SIZE        (PC_ROM_MAX - PC_ROM_MIN_VGA)
840 
841 static hwaddr pc_above_4g_end(PCMachineState *pcms)
842 {
843     X86MachineState *x86ms = X86_MACHINE(pcms);
844 
845     if (pcms->sgx_epc.size != 0) {
846         return sgx_epc_above_4g_end(&pcms->sgx_epc);
847     }
848 
849     return x86ms->above_4g_mem_start + x86ms->above_4g_mem_size;
850 }
851 
852 static void pc_get_device_memory_range(PCMachineState *pcms,
853                                        hwaddr *base,
854                                        ram_addr_t *device_mem_size)
855 {
856     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
857     MachineState *machine = MACHINE(pcms);
858     ram_addr_t size;
859     hwaddr addr;
860 
861     size = machine->maxram_size - machine->ram_size;
862     addr = ROUND_UP(pc_above_4g_end(pcms), 1 * GiB);
863 
864     if (pcmc->enforce_aligned_dimm) {
865         /* size device region assuming 1G page max alignment per slot */
866         size += (1 * GiB) * machine->ram_slots;
867     }
868 
869     *base = addr;
870     *device_mem_size = size;
871 }
872 
873 static uint64_t pc_get_cxl_range_start(PCMachineState *pcms)
874 {
875     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
876     hwaddr cxl_base;
877     ram_addr_t size;
878 
879     if (pcmc->has_reserved_memory) {
880         pc_get_device_memory_range(pcms, &cxl_base, &size);
881         cxl_base += size;
882     } else {
883         cxl_base = pc_above_4g_end(pcms);
884     }
885 
886     return cxl_base;
887 }
888 
889 static uint64_t pc_get_cxl_range_end(PCMachineState *pcms)
890 {
891     uint64_t start = pc_get_cxl_range_start(pcms) + MiB;
892 
893     if (pcms->cxl_devices_state.fixed_windows) {
894         GList *it;
895 
896         start = ROUND_UP(start, 256 * MiB);
897         for (it = pcms->cxl_devices_state.fixed_windows; it; it = it->next) {
898             CXLFixedWindow *fw = it->data;
899             start += fw->size;
900         }
901     }
902 
903     return start;
904 }
905 
906 static hwaddr pc_max_used_gpa(PCMachineState *pcms, uint64_t pci_hole64_size)
907 {
908     X86CPU *cpu = X86_CPU(first_cpu);
909 
910     /* 32-bit systems don't have hole64 thus return max CPU address */
911     if (cpu->phys_bits <= 32) {
912         return ((hwaddr)1 << cpu->phys_bits) - 1;
913     }
914 
915     return pc_pci_hole64_start() + pci_hole64_size - 1;
916 }
917 
918 /*
919  * AMD systems with an IOMMU have an additional hole close to the
920  * 1Tb, which are special GPAs that cannot be DMA mapped. Depending
921  * on kernel version, VFIO may or may not let you DMA map those ranges.
922  * Starting Linux v5.4 we validate it, and can't create guests on AMD machines
923  * with certain memory sizes. It's also wrong to use those IOVA ranges
924  * in detriment of leading to IOMMU INVALID_DEVICE_REQUEST or worse.
925  * The ranges reserved for Hyper-Transport are:
926  *
927  * FD_0000_0000h - FF_FFFF_FFFFh
928  *
929  * The ranges represent the following:
930  *
931  * Base Address   Top Address  Use
932  *
933  * FD_0000_0000h FD_F7FF_FFFFh Reserved interrupt address space
934  * FD_F800_0000h FD_F8FF_FFFFh Interrupt/EOI IntCtl
935  * FD_F900_0000h FD_F90F_FFFFh Legacy PIC IACK
936  * FD_F910_0000h FD_F91F_FFFFh System Management
937  * FD_F920_0000h FD_FAFF_FFFFh Reserved Page Tables
938  * FD_FB00_0000h FD_FBFF_FFFFh Address Translation
939  * FD_FC00_0000h FD_FDFF_FFFFh I/O Space
940  * FD_FE00_0000h FD_FFFF_FFFFh Configuration
941  * FE_0000_0000h FE_1FFF_FFFFh Extended Configuration/Device Messages
942  * FE_2000_0000h FF_FFFF_FFFFh Reserved
943  *
944  * See AMD IOMMU spec, section 2.1.2 "IOMMU Logical Topology",
945  * Table 3: Special Address Controls (GPA) for more information.
946  */
947 #define AMD_HT_START         0xfd00000000UL
948 #define AMD_HT_END           0xffffffffffUL
949 #define AMD_ABOVE_1TB_START  (AMD_HT_END + 1)
950 #define AMD_HT_SIZE          (AMD_ABOVE_1TB_START - AMD_HT_START)
951 
952 void pc_memory_init(PCMachineState *pcms,
953                     MemoryRegion *system_memory,
954                     MemoryRegion *rom_memory,
955                     uint64_t pci_hole64_size)
956 {
957     int linux_boot, i;
958     MemoryRegion *option_rom_mr;
959     MemoryRegion *ram_below_4g, *ram_above_4g;
960     FWCfgState *fw_cfg;
961     MachineState *machine = MACHINE(pcms);
962     MachineClass *mc = MACHINE_GET_CLASS(machine);
963     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
964     X86MachineState *x86ms = X86_MACHINE(pcms);
965     hwaddr maxphysaddr, maxusedaddr;
966     hwaddr cxl_base, cxl_resv_end = 0;
967     X86CPU *cpu = X86_CPU(first_cpu);
968 
969     assert(machine->ram_size == x86ms->below_4g_mem_size +
970                                 x86ms->above_4g_mem_size);
971 
972     linux_boot = (machine->kernel_filename != NULL);
973 
974     /*
975      * The HyperTransport range close to the 1T boundary is unique to AMD
976      * hosts with IOMMUs enabled. Restrict the ram-above-4g relocation
977      * to above 1T to AMD vCPUs only. @enforce_amd_1tb_hole is only false in
978      * older machine types (<= 7.0) for compatibility purposes.
979      */
980     if (IS_AMD_CPU(&cpu->env) && pcmc->enforce_amd_1tb_hole) {
981         /* Bail out if max possible address does not cross HT range */
982         if (pc_max_used_gpa(pcms, pci_hole64_size) >= AMD_HT_START) {
983             x86ms->above_4g_mem_start = AMD_ABOVE_1TB_START;
984         }
985 
986         /*
987          * Advertise the HT region if address space covers the reserved
988          * region or if we relocate.
989          */
990         if (cpu->phys_bits >= 40) {
991             e820_add_entry(AMD_HT_START, AMD_HT_SIZE, E820_RESERVED);
992         }
993     }
994 
995     /*
996      * phys-bits is required to be appropriately configured
997      * to make sure max used GPA is reachable.
998      */
999     maxusedaddr = pc_max_used_gpa(pcms, pci_hole64_size);
1000     maxphysaddr = ((hwaddr)1 << cpu->phys_bits) - 1;
1001     if (maxphysaddr < maxusedaddr) {
1002         error_report("Address space limit 0x%"PRIx64" < 0x%"PRIx64
1003                      " phys-bits too low (%u)",
1004                      maxphysaddr, maxusedaddr, cpu->phys_bits);
1005         exit(EXIT_FAILURE);
1006     }
1007 
1008     /*
1009      * Split single memory region and use aliases to address portions of it,
1010      * done for backwards compatibility with older qemus.
1011      */
1012     ram_below_4g = g_malloc(sizeof(*ram_below_4g));
1013     memory_region_init_alias(ram_below_4g, NULL, "ram-below-4g", machine->ram,
1014                              0, x86ms->below_4g_mem_size);
1015     memory_region_add_subregion(system_memory, 0, ram_below_4g);
1016     e820_add_entry(0, x86ms->below_4g_mem_size, E820_RAM);
1017     if (x86ms->above_4g_mem_size > 0) {
1018         ram_above_4g = g_malloc(sizeof(*ram_above_4g));
1019         memory_region_init_alias(ram_above_4g, NULL, "ram-above-4g",
1020                                  machine->ram,
1021                                  x86ms->below_4g_mem_size,
1022                                  x86ms->above_4g_mem_size);
1023         memory_region_add_subregion(system_memory, x86ms->above_4g_mem_start,
1024                                     ram_above_4g);
1025         e820_add_entry(x86ms->above_4g_mem_start, x86ms->above_4g_mem_size,
1026                        E820_RAM);
1027     }
1028 
1029     if (pcms->sgx_epc.size != 0) {
1030         e820_add_entry(pcms->sgx_epc.base, pcms->sgx_epc.size, E820_RESERVED);
1031     }
1032 
1033     if (!pcmc->has_reserved_memory &&
1034         (machine->ram_slots ||
1035          (machine->maxram_size > machine->ram_size))) {
1036 
1037         error_report("\"-memory 'slots|maxmem'\" is not supported by: %s",
1038                      mc->name);
1039         exit(EXIT_FAILURE);
1040     }
1041 
1042     /* always allocate the device memory information */
1043     machine->device_memory = g_malloc0(sizeof(*machine->device_memory));
1044 
1045     /* initialize device memory address space */
1046     if (pcmc->has_reserved_memory &&
1047         (machine->ram_size < machine->maxram_size)) {
1048         ram_addr_t device_mem_size;
1049 
1050         if (machine->ram_slots > ACPI_MAX_RAM_SLOTS) {
1051             error_report("unsupported amount of memory slots: %"PRIu64,
1052                          machine->ram_slots);
1053             exit(EXIT_FAILURE);
1054         }
1055 
1056         if (QEMU_ALIGN_UP(machine->maxram_size,
1057                           TARGET_PAGE_SIZE) != machine->maxram_size) {
1058             error_report("maximum memory size must by aligned to multiple of "
1059                          "%d bytes", TARGET_PAGE_SIZE);
1060             exit(EXIT_FAILURE);
1061         }
1062 
1063         pc_get_device_memory_range(pcms, &machine->device_memory->base, &device_mem_size);
1064 
1065         if ((machine->device_memory->base + device_mem_size) <
1066             device_mem_size) {
1067             error_report("unsupported amount of maximum memory: " RAM_ADDR_FMT,
1068                          machine->maxram_size);
1069             exit(EXIT_FAILURE);
1070         }
1071 
1072         memory_region_init(&machine->device_memory->mr, OBJECT(pcms),
1073                            "device-memory", device_mem_size);
1074         memory_region_add_subregion(system_memory, machine->device_memory->base,
1075                                     &machine->device_memory->mr);
1076     }
1077 
1078     if (pcms->cxl_devices_state.is_enabled) {
1079         MemoryRegion *mr = &pcms->cxl_devices_state.host_mr;
1080         hwaddr cxl_size = MiB;
1081 
1082         cxl_base = pc_get_cxl_range_start(pcms);
1083         memory_region_init(mr, OBJECT(machine), "cxl_host_reg", cxl_size);
1084         memory_region_add_subregion(system_memory, cxl_base, mr);
1085         cxl_resv_end = cxl_base + cxl_size;
1086         if (pcms->cxl_devices_state.fixed_windows) {
1087             hwaddr cxl_fmw_base;
1088             GList *it;
1089 
1090             cxl_fmw_base = ROUND_UP(cxl_base + cxl_size, 256 * MiB);
1091             for (it = pcms->cxl_devices_state.fixed_windows; it; it = it->next) {
1092                 CXLFixedWindow *fw = it->data;
1093 
1094                 fw->base = cxl_fmw_base;
1095                 memory_region_init_io(&fw->mr, OBJECT(machine), &cfmws_ops, fw,
1096                                       "cxl-fixed-memory-region", fw->size);
1097                 memory_region_add_subregion(system_memory, fw->base, &fw->mr);
1098                 cxl_fmw_base += fw->size;
1099                 cxl_resv_end = cxl_fmw_base;
1100             }
1101         }
1102     }
1103 
1104     /* Initialize PC system firmware */
1105     pc_system_firmware_init(pcms, rom_memory);
1106 
1107     option_rom_mr = g_malloc(sizeof(*option_rom_mr));
1108     memory_region_init_ram(option_rom_mr, NULL, "pc.rom", PC_ROM_SIZE,
1109                            &error_fatal);
1110     if (pcmc->pci_enabled) {
1111         memory_region_set_readonly(option_rom_mr, true);
1112     }
1113     memory_region_add_subregion_overlap(rom_memory,
1114                                         PC_ROM_MIN_VGA,
1115                                         option_rom_mr,
1116                                         1);
1117 
1118     fw_cfg = fw_cfg_arch_create(machine,
1119                                 x86ms->boot_cpus, x86ms->apic_id_limit);
1120 
1121     rom_set_fw(fw_cfg);
1122 
1123     if (pcmc->has_reserved_memory && machine->device_memory->base) {
1124         uint64_t *val = g_malloc(sizeof(*val));
1125         PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1126         uint64_t res_mem_end = machine->device_memory->base;
1127 
1128         if (!pcmc->broken_reserved_end) {
1129             res_mem_end += memory_region_size(&machine->device_memory->mr);
1130         }
1131 
1132         if (pcms->cxl_devices_state.is_enabled) {
1133             res_mem_end = cxl_resv_end;
1134         }
1135         *val = cpu_to_le64(ROUND_UP(res_mem_end, 1 * GiB));
1136         fw_cfg_add_file(fw_cfg, "etc/reserved-memory-end", val, sizeof(*val));
1137     }
1138 
1139     if (linux_boot) {
1140         x86_load_linux(x86ms, fw_cfg, pcmc->acpi_data_size,
1141                        pcmc->pvh_enabled);
1142     }
1143 
1144     for (i = 0; i < nb_option_roms; i++) {
1145         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1146     }
1147     x86ms->fw_cfg = fw_cfg;
1148 
1149     /* Init default IOAPIC address space */
1150     x86ms->ioapic_as = &address_space_memory;
1151 
1152     /* Init ACPI memory hotplug IO base address */
1153     pcms->memhp_io_base = ACPI_MEMORY_HOTPLUG_BASE;
1154 }
1155 
1156 /*
1157  * The 64bit pci hole starts after "above 4G RAM" and
1158  * potentially the space reserved for memory hotplug.
1159  */
1160 uint64_t pc_pci_hole64_start(void)
1161 {
1162     PCMachineState *pcms = PC_MACHINE(qdev_get_machine());
1163     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1164     MachineState *ms = MACHINE(pcms);
1165     uint64_t hole64_start = 0;
1166     ram_addr_t size = 0;
1167 
1168     if (pcms->cxl_devices_state.is_enabled) {
1169         hole64_start = pc_get_cxl_range_end(pcms);
1170     } else if (pcmc->has_reserved_memory && (ms->ram_size < ms->maxram_size)) {
1171         pc_get_device_memory_range(pcms, &hole64_start, &size);
1172         if (!pcmc->broken_reserved_end) {
1173             hole64_start += size;
1174         }
1175     } else {
1176         hole64_start = pc_above_4g_end(pcms);
1177     }
1178 
1179     return ROUND_UP(hole64_start, 1 * GiB);
1180 }
1181 
1182 DeviceState *pc_vga_init(ISABus *isa_bus, PCIBus *pci_bus)
1183 {
1184     DeviceState *dev = NULL;
1185 
1186     rom_set_order_override(FW_CFG_ORDER_OVERRIDE_VGA);
1187     if (pci_bus) {
1188         PCIDevice *pcidev = pci_vga_init(pci_bus);
1189         dev = pcidev ? &pcidev->qdev : NULL;
1190     } else if (isa_bus) {
1191         ISADevice *isadev = isa_vga_init(isa_bus);
1192         dev = isadev ? DEVICE(isadev) : NULL;
1193     }
1194     rom_reset_order_override();
1195     return dev;
1196 }
1197 
1198 static const MemoryRegionOps ioport80_io_ops = {
1199     .write = ioport80_write,
1200     .read = ioport80_read,
1201     .endianness = DEVICE_NATIVE_ENDIAN,
1202     .impl = {
1203         .min_access_size = 1,
1204         .max_access_size = 1,
1205     },
1206 };
1207 
1208 static const MemoryRegionOps ioportF0_io_ops = {
1209     .write = ioportF0_write,
1210     .read = ioportF0_read,
1211     .endianness = DEVICE_NATIVE_ENDIAN,
1212     .impl = {
1213         .min_access_size = 1,
1214         .max_access_size = 1,
1215     },
1216 };
1217 
1218 static void pc_superio_init(ISABus *isa_bus, bool create_fdctrl,
1219                             bool create_i8042, bool no_vmport)
1220 {
1221     int i;
1222     DriveInfo *fd[MAX_FD];
1223     qemu_irq *a20_line;
1224     ISADevice *fdc, *i8042, *port92, *vmmouse;
1225 
1226     serial_hds_isa_init(isa_bus, 0, MAX_ISA_SERIAL_PORTS);
1227     parallel_hds_isa_init(isa_bus, MAX_PARALLEL_PORTS);
1228 
1229     for (i = 0; i < MAX_FD; i++) {
1230         fd[i] = drive_get(IF_FLOPPY, 0, i);
1231         create_fdctrl |= !!fd[i];
1232     }
1233     if (create_fdctrl) {
1234         fdc = isa_new(TYPE_ISA_FDC);
1235         if (fdc) {
1236             isa_realize_and_unref(fdc, isa_bus, &error_fatal);
1237             isa_fdc_init_drives(fdc, fd);
1238         }
1239     }
1240 
1241     if (!create_i8042) {
1242         return;
1243     }
1244 
1245     i8042 = isa_create_simple(isa_bus, TYPE_I8042);
1246     if (!no_vmport) {
1247         isa_create_simple(isa_bus, TYPE_VMPORT);
1248         vmmouse = isa_try_new("vmmouse");
1249     } else {
1250         vmmouse = NULL;
1251     }
1252     if (vmmouse) {
1253         object_property_set_link(OBJECT(vmmouse), TYPE_I8042, OBJECT(i8042),
1254                                  &error_abort);
1255         isa_realize_and_unref(vmmouse, isa_bus, &error_fatal);
1256     }
1257     port92 = isa_create_simple(isa_bus, TYPE_PORT92);
1258 
1259     a20_line = qemu_allocate_irqs(handle_a20_line_change, first_cpu, 2);
1260     i8042_setup_a20_line(i8042, a20_line[0]);
1261     qdev_connect_gpio_out_named(DEVICE(port92),
1262                                 PORT92_A20_LINE, 0, a20_line[1]);
1263     g_free(a20_line);
1264 }
1265 
1266 void pc_basic_device_init(struct PCMachineState *pcms,
1267                           ISABus *isa_bus, qemu_irq *gsi,
1268                           ISADevice *rtc_state,
1269                           bool create_fdctrl,
1270                           uint32_t hpet_irqs)
1271 {
1272     int i;
1273     DeviceState *hpet = NULL;
1274     int pit_isa_irq = 0;
1275     qemu_irq pit_alt_irq = NULL;
1276     qemu_irq rtc_irq = NULL;
1277     ISADevice *pit = NULL;
1278     MemoryRegion *ioport80_io = g_new(MemoryRegion, 1);
1279     MemoryRegion *ioportF0_io = g_new(MemoryRegion, 1);
1280     X86MachineState *x86ms = X86_MACHINE(pcms);
1281 
1282     memory_region_init_io(ioport80_io, NULL, &ioport80_io_ops, NULL, "ioport80", 1);
1283     memory_region_add_subregion(isa_bus->address_space_io, 0x80, ioport80_io);
1284 
1285     memory_region_init_io(ioportF0_io, NULL, &ioportF0_io_ops, NULL, "ioportF0", 1);
1286     memory_region_add_subregion(isa_bus->address_space_io, 0xf0, ioportF0_io);
1287 
1288     /*
1289      * Check if an HPET shall be created.
1290      *
1291      * Without KVM_CAP_PIT_STATE2, we cannot switch off the in-kernel PIT
1292      * when the HPET wants to take over. Thus we have to disable the latter.
1293      */
1294     if (pcms->hpet_enabled && (!kvm_irqchip_in_kernel() ||
1295                                kvm_has_pit_state2())) {
1296         hpet = qdev_try_new(TYPE_HPET);
1297         if (!hpet) {
1298             error_report("couldn't create HPET device");
1299             exit(1);
1300         }
1301         /*
1302          * For pc-piix-*, hpet's intcap is always IRQ2. For pc-q35-1.7 and
1303          * earlier, use IRQ2 for compat. Otherwise, use IRQ16~23, IRQ8 and
1304          * IRQ2.
1305          */
1306         uint8_t compat = object_property_get_uint(OBJECT(hpet),
1307                 HPET_INTCAP, NULL);
1308         if (!compat) {
1309             qdev_prop_set_uint32(hpet, HPET_INTCAP, hpet_irqs);
1310         }
1311         sysbus_realize_and_unref(SYS_BUS_DEVICE(hpet), &error_fatal);
1312         sysbus_mmio_map(SYS_BUS_DEVICE(hpet), 0, HPET_BASE);
1313 
1314         for (i = 0; i < IOAPIC_NUM_PINS; i++) {
1315             sysbus_connect_irq(SYS_BUS_DEVICE(hpet), i, gsi[i]);
1316         }
1317         pit_isa_irq = -1;
1318         pit_alt_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_PIT_INT);
1319         rtc_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_RTC_INT);
1320     }
1321 
1322     if (rtc_irq) {
1323         qdev_connect_gpio_out(DEVICE(rtc_state), 0, rtc_irq);
1324     } else {
1325         uint32_t irq = object_property_get_uint(OBJECT(rtc_state),
1326                                                 "irq",
1327                                                 &error_fatal);
1328         isa_connect_gpio_out(rtc_state, 0, irq);
1329     }
1330     object_property_add_alias(OBJECT(pcms), "rtc-time", OBJECT(rtc_state),
1331                               "date");
1332 
1333 #ifdef CONFIG_XEN_EMU
1334     if (xen_mode == XEN_EMULATE) {
1335         xen_overlay_create();
1336         xen_evtchn_create(IOAPIC_NUM_PINS, gsi);
1337         xen_gnttab_create();
1338         xen_xenstore_create();
1339         if (pcms->bus) {
1340             pci_create_simple(pcms->bus, -1, "xen-platform");
1341         }
1342         xen_bus_init();
1343         xen_be_init();
1344     }
1345 #endif
1346 
1347     qemu_register_boot_set(pc_boot_set, rtc_state);
1348 
1349     if (!xen_enabled() &&
1350         (x86ms->pit == ON_OFF_AUTO_AUTO || x86ms->pit == ON_OFF_AUTO_ON)) {
1351         if (kvm_pit_in_kernel()) {
1352             pit = kvm_pit_init(isa_bus, 0x40);
1353         } else {
1354             pit = i8254_pit_init(isa_bus, 0x40, pit_isa_irq, pit_alt_irq);
1355         }
1356         if (hpet) {
1357             /* connect PIT to output control line of the HPET */
1358             qdev_connect_gpio_out(hpet, 0, qdev_get_gpio_in(DEVICE(pit), 0));
1359         }
1360         pcspk_init(pcms->pcspk, isa_bus, pit);
1361     }
1362 
1363     /* Super I/O */
1364     pc_superio_init(isa_bus, create_fdctrl, pcms->i8042_enabled,
1365                     pcms->vmport != ON_OFF_AUTO_ON);
1366 }
1367 
1368 void pc_nic_init(PCMachineClass *pcmc, ISABus *isa_bus, PCIBus *pci_bus)
1369 {
1370     MachineClass *mc = MACHINE_CLASS(pcmc);
1371     int i;
1372 
1373     rom_set_order_override(FW_CFG_ORDER_OVERRIDE_NIC);
1374     for (i = 0; i < nb_nics; i++) {
1375         NICInfo *nd = &nd_table[i];
1376         const char *model = nd->model ? nd->model : mc->default_nic;
1377 
1378         if (g_str_equal(model, "ne2k_isa")) {
1379             pc_init_ne2k_isa(isa_bus, nd);
1380         } else {
1381             pci_nic_init_nofail(nd, pci_bus, model, NULL);
1382         }
1383     }
1384     rom_reset_order_override();
1385 }
1386 
1387 void pc_i8259_create(ISABus *isa_bus, qemu_irq *i8259_irqs)
1388 {
1389     qemu_irq *i8259;
1390 
1391     if (kvm_pic_in_kernel()) {
1392         i8259 = kvm_i8259_init(isa_bus);
1393     } else if (xen_enabled()) {
1394         i8259 = xen_interrupt_controller_init();
1395     } else {
1396         i8259 = i8259_init(isa_bus, x86_allocate_cpu_irq());
1397     }
1398 
1399     for (size_t i = 0; i < ISA_NUM_IRQS; i++) {
1400         i8259_irqs[i] = i8259[i];
1401     }
1402 
1403     g_free(i8259);
1404 }
1405 
1406 static void pc_memory_pre_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
1407                                Error **errp)
1408 {
1409     const PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1410     const X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
1411     const PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1412     const MachineState *ms = MACHINE(hotplug_dev);
1413     const bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
1414     const uint64_t legacy_align = TARGET_PAGE_SIZE;
1415     Error *local_err = NULL;
1416 
1417     /*
1418      * When -no-acpi is used with Q35 machine type, no ACPI is built,
1419      * but pcms->acpi_dev is still created. Check !acpi_enabled in
1420      * addition to cover this case.
1421      */
1422     if (!x86ms->acpi_dev || !x86_machine_is_acpi_enabled(x86ms)) {
1423         error_setg(errp,
1424                    "memory hotplug is not enabled: missing acpi device or acpi disabled");
1425         return;
1426     }
1427 
1428     if (is_nvdimm && !ms->nvdimms_state->is_enabled) {
1429         error_setg(errp, "nvdimm is not enabled: missing 'nvdimm' in '-M'");
1430         return;
1431     }
1432 
1433     hotplug_handler_pre_plug(x86ms->acpi_dev, dev, &local_err);
1434     if (local_err) {
1435         error_propagate(errp, local_err);
1436         return;
1437     }
1438 
1439     pc_dimm_pre_plug(PC_DIMM(dev), MACHINE(hotplug_dev),
1440                      pcmc->enforce_aligned_dimm ? NULL : &legacy_align, errp);
1441 }
1442 
1443 static void pc_memory_plug(HotplugHandler *hotplug_dev,
1444                            DeviceState *dev, Error **errp)
1445 {
1446     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1447     X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
1448     MachineState *ms = MACHINE(hotplug_dev);
1449     bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
1450 
1451     pc_dimm_plug(PC_DIMM(dev), MACHINE(pcms));
1452 
1453     if (is_nvdimm) {
1454         nvdimm_plug(ms->nvdimms_state);
1455     }
1456 
1457     hotplug_handler_plug(x86ms->acpi_dev, dev, &error_abort);
1458 }
1459 
1460 static void pc_memory_unplug_request(HotplugHandler *hotplug_dev,
1461                                      DeviceState *dev, Error **errp)
1462 {
1463     X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
1464 
1465     /*
1466      * When -no-acpi is used with Q35 machine type, no ACPI is built,
1467      * but pcms->acpi_dev is still created. Check !acpi_enabled in
1468      * addition to cover this case.
1469      */
1470     if (!x86ms->acpi_dev || !x86_machine_is_acpi_enabled(x86ms)) {
1471         error_setg(errp,
1472                    "memory hotplug is not enabled: missing acpi device or acpi disabled");
1473         return;
1474     }
1475 
1476     if (object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM)) {
1477         error_setg(errp, "nvdimm device hot unplug is not supported yet.");
1478         return;
1479     }
1480 
1481     hotplug_handler_unplug_request(x86ms->acpi_dev, dev,
1482                                    errp);
1483 }
1484 
1485 static void pc_memory_unplug(HotplugHandler *hotplug_dev,
1486                              DeviceState *dev, Error **errp)
1487 {
1488     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1489     X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
1490     Error *local_err = NULL;
1491 
1492     hotplug_handler_unplug(x86ms->acpi_dev, dev, &local_err);
1493     if (local_err) {
1494         goto out;
1495     }
1496 
1497     pc_dimm_unplug(PC_DIMM(dev), MACHINE(pcms));
1498     qdev_unrealize(dev);
1499  out:
1500     error_propagate(errp, local_err);
1501 }
1502 
1503 static void pc_virtio_md_pci_pre_plug(HotplugHandler *hotplug_dev,
1504                                       DeviceState *dev, Error **errp)
1505 {
1506     HotplugHandler *hotplug_dev2 = qdev_get_bus_hotplug_handler(dev);
1507     Error *local_err = NULL;
1508 
1509     if (!hotplug_dev2 && dev->hotplugged) {
1510         /*
1511          * Without a bus hotplug handler, we cannot control the plug/unplug
1512          * order. We should never reach this point when hotplugging on x86,
1513          * however, better add a safety net.
1514          */
1515         error_setg(errp, "hotplug of virtio based memory devices not supported"
1516                    " on this bus.");
1517         return;
1518     }
1519     /*
1520      * First, see if we can plug this memory device at all. If that
1521      * succeeds, branch of to the actual hotplug handler.
1522      */
1523     memory_device_pre_plug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev), NULL,
1524                            &local_err);
1525     if (!local_err && hotplug_dev2) {
1526         hotplug_handler_pre_plug(hotplug_dev2, dev, &local_err);
1527     }
1528     error_propagate(errp, local_err);
1529 }
1530 
1531 static void pc_virtio_md_pci_plug(HotplugHandler *hotplug_dev,
1532                                   DeviceState *dev, Error **errp)
1533 {
1534     HotplugHandler *hotplug_dev2 = qdev_get_bus_hotplug_handler(dev);
1535     Error *local_err = NULL;
1536 
1537     /*
1538      * Plug the memory device first and then branch off to the actual
1539      * hotplug handler. If that one fails, we can easily undo the memory
1540      * device bits.
1541      */
1542     memory_device_plug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev));
1543     if (hotplug_dev2) {
1544         hotplug_handler_plug(hotplug_dev2, dev, &local_err);
1545         if (local_err) {
1546             memory_device_unplug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev));
1547         }
1548     }
1549     error_propagate(errp, local_err);
1550 }
1551 
1552 static void pc_virtio_md_pci_unplug_request(HotplugHandler *hotplug_dev,
1553                                             DeviceState *dev, Error **errp)
1554 {
1555     /* We don't support hot unplug of virtio based memory devices */
1556     error_setg(errp, "virtio based memory devices cannot be unplugged.");
1557 }
1558 
1559 static void pc_virtio_md_pci_unplug(HotplugHandler *hotplug_dev,
1560                                     DeviceState *dev, Error **errp)
1561 {
1562     /* We don't support hot unplug of virtio based memory devices */
1563 }
1564 
1565 static void pc_machine_device_pre_plug_cb(HotplugHandler *hotplug_dev,
1566                                           DeviceState *dev, Error **errp)
1567 {
1568     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1569         pc_memory_pre_plug(hotplug_dev, dev, errp);
1570     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1571         x86_cpu_pre_plug(hotplug_dev, dev, errp);
1572     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI) ||
1573                object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI)) {
1574         pc_virtio_md_pci_pre_plug(hotplug_dev, dev, errp);
1575     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
1576         /* Declare the APIC range as the reserved MSI region */
1577         char *resv_prop_str = g_strdup_printf("0xfee00000:0xfeefffff:%d",
1578                                               VIRTIO_IOMMU_RESV_MEM_T_MSI);
1579 
1580         object_property_set_uint(OBJECT(dev), "len-reserved-regions", 1, errp);
1581         object_property_set_str(OBJECT(dev), "reserved-regions[0]",
1582                                 resv_prop_str, errp);
1583         g_free(resv_prop_str);
1584     }
1585 
1586     if (object_dynamic_cast(OBJECT(dev), TYPE_X86_IOMMU_DEVICE) ||
1587         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
1588         PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1589 
1590         if (pcms->iommu) {
1591             error_setg(errp, "QEMU does not support multiple vIOMMUs "
1592                        "for x86 yet.");
1593             return;
1594         }
1595         pcms->iommu = dev;
1596     }
1597 }
1598 
1599 static void pc_machine_device_plug_cb(HotplugHandler *hotplug_dev,
1600                                       DeviceState *dev, Error **errp)
1601 {
1602     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1603         pc_memory_plug(hotplug_dev, dev, errp);
1604     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1605         x86_cpu_plug(hotplug_dev, dev, errp);
1606     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI) ||
1607                object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI)) {
1608         pc_virtio_md_pci_plug(hotplug_dev, dev, errp);
1609     }
1610 }
1611 
1612 static void pc_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
1613                                                 DeviceState *dev, Error **errp)
1614 {
1615     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1616         pc_memory_unplug_request(hotplug_dev, dev, errp);
1617     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1618         x86_cpu_unplug_request_cb(hotplug_dev, dev, errp);
1619     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI) ||
1620                object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI)) {
1621         pc_virtio_md_pci_unplug_request(hotplug_dev, dev, errp);
1622     } else {
1623         error_setg(errp, "acpi: device unplug request for not supported device"
1624                    " type: %s", object_get_typename(OBJECT(dev)));
1625     }
1626 }
1627 
1628 static void pc_machine_device_unplug_cb(HotplugHandler *hotplug_dev,
1629                                         DeviceState *dev, Error **errp)
1630 {
1631     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1632         pc_memory_unplug(hotplug_dev, dev, errp);
1633     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1634         x86_cpu_unplug_cb(hotplug_dev, dev, errp);
1635     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI) ||
1636                object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI)) {
1637         pc_virtio_md_pci_unplug(hotplug_dev, dev, errp);
1638     } else {
1639         error_setg(errp, "acpi: device unplug for not supported device"
1640                    " type: %s", object_get_typename(OBJECT(dev)));
1641     }
1642 }
1643 
1644 static HotplugHandler *pc_get_hotplug_handler(MachineState *machine,
1645                                              DeviceState *dev)
1646 {
1647     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) ||
1648         object_dynamic_cast(OBJECT(dev), TYPE_CPU) ||
1649         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI) ||
1650         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI) ||
1651         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI) ||
1652         object_dynamic_cast(OBJECT(dev), TYPE_X86_IOMMU_DEVICE)) {
1653         return HOTPLUG_HANDLER(machine);
1654     }
1655 
1656     return NULL;
1657 }
1658 
1659 static void
1660 pc_machine_get_device_memory_region_size(Object *obj, Visitor *v,
1661                                          const char *name, void *opaque,
1662                                          Error **errp)
1663 {
1664     MachineState *ms = MACHINE(obj);
1665     int64_t value = 0;
1666 
1667     if (ms->device_memory) {
1668         value = memory_region_size(&ms->device_memory->mr);
1669     }
1670 
1671     visit_type_int(v, name, &value, errp);
1672 }
1673 
1674 static void pc_machine_get_vmport(Object *obj, Visitor *v, const char *name,
1675                                   void *opaque, Error **errp)
1676 {
1677     PCMachineState *pcms = PC_MACHINE(obj);
1678     OnOffAuto vmport = pcms->vmport;
1679 
1680     visit_type_OnOffAuto(v, name, &vmport, errp);
1681 }
1682 
1683 static void pc_machine_set_vmport(Object *obj, Visitor *v, const char *name,
1684                                   void *opaque, Error **errp)
1685 {
1686     PCMachineState *pcms = PC_MACHINE(obj);
1687 
1688     visit_type_OnOffAuto(v, name, &pcms->vmport, errp);
1689 }
1690 
1691 static bool pc_machine_get_smbus(Object *obj, Error **errp)
1692 {
1693     PCMachineState *pcms = PC_MACHINE(obj);
1694 
1695     return pcms->smbus_enabled;
1696 }
1697 
1698 static void pc_machine_set_smbus(Object *obj, bool value, Error **errp)
1699 {
1700     PCMachineState *pcms = PC_MACHINE(obj);
1701 
1702     pcms->smbus_enabled = value;
1703 }
1704 
1705 static bool pc_machine_get_sata(Object *obj, Error **errp)
1706 {
1707     PCMachineState *pcms = PC_MACHINE(obj);
1708 
1709     return pcms->sata_enabled;
1710 }
1711 
1712 static void pc_machine_set_sata(Object *obj, bool value, Error **errp)
1713 {
1714     PCMachineState *pcms = PC_MACHINE(obj);
1715 
1716     pcms->sata_enabled = value;
1717 }
1718 
1719 static bool pc_machine_get_hpet(Object *obj, Error **errp)
1720 {
1721     PCMachineState *pcms = PC_MACHINE(obj);
1722 
1723     return pcms->hpet_enabled;
1724 }
1725 
1726 static void pc_machine_set_hpet(Object *obj, bool value, Error **errp)
1727 {
1728     PCMachineState *pcms = PC_MACHINE(obj);
1729 
1730     pcms->hpet_enabled = value;
1731 }
1732 
1733 static bool pc_machine_get_i8042(Object *obj, Error **errp)
1734 {
1735     PCMachineState *pcms = PC_MACHINE(obj);
1736 
1737     return pcms->i8042_enabled;
1738 }
1739 
1740 static void pc_machine_set_i8042(Object *obj, bool value, Error **errp)
1741 {
1742     PCMachineState *pcms = PC_MACHINE(obj);
1743 
1744     pcms->i8042_enabled = value;
1745 }
1746 
1747 static bool pc_machine_get_default_bus_bypass_iommu(Object *obj, Error **errp)
1748 {
1749     PCMachineState *pcms = PC_MACHINE(obj);
1750 
1751     return pcms->default_bus_bypass_iommu;
1752 }
1753 
1754 static void pc_machine_set_default_bus_bypass_iommu(Object *obj, bool value,
1755                                                     Error **errp)
1756 {
1757     PCMachineState *pcms = PC_MACHINE(obj);
1758 
1759     pcms->default_bus_bypass_iommu = value;
1760 }
1761 
1762 static void pc_machine_get_smbios_ep(Object *obj, Visitor *v, const char *name,
1763                                      void *opaque, Error **errp)
1764 {
1765     PCMachineState *pcms = PC_MACHINE(obj);
1766     SmbiosEntryPointType smbios_entry_point_type = pcms->smbios_entry_point_type;
1767 
1768     visit_type_SmbiosEntryPointType(v, name, &smbios_entry_point_type, errp);
1769 }
1770 
1771 static void pc_machine_set_smbios_ep(Object *obj, Visitor *v, const char *name,
1772                                      void *opaque, Error **errp)
1773 {
1774     PCMachineState *pcms = PC_MACHINE(obj);
1775 
1776     visit_type_SmbiosEntryPointType(v, name, &pcms->smbios_entry_point_type, errp);
1777 }
1778 
1779 static void pc_machine_get_max_ram_below_4g(Object *obj, Visitor *v,
1780                                             const char *name, void *opaque,
1781                                             Error **errp)
1782 {
1783     PCMachineState *pcms = PC_MACHINE(obj);
1784     uint64_t value = pcms->max_ram_below_4g;
1785 
1786     visit_type_size(v, name, &value, errp);
1787 }
1788 
1789 static void pc_machine_set_max_ram_below_4g(Object *obj, Visitor *v,
1790                                             const char *name, void *opaque,
1791                                             Error **errp)
1792 {
1793     PCMachineState *pcms = PC_MACHINE(obj);
1794     uint64_t value;
1795 
1796     if (!visit_type_size(v, name, &value, errp)) {
1797         return;
1798     }
1799     if (value > 4 * GiB) {
1800         error_setg(errp,
1801                    "Machine option 'max-ram-below-4g=%"PRIu64
1802                    "' expects size less than or equal to 4G", value);
1803         return;
1804     }
1805 
1806     if (value < 1 * MiB) {
1807         warn_report("Only %" PRIu64 " bytes of RAM below the 4GiB boundary,"
1808                     "BIOS may not work with less than 1MiB", value);
1809     }
1810 
1811     pcms->max_ram_below_4g = value;
1812 }
1813 
1814 static void pc_machine_get_max_fw_size(Object *obj, Visitor *v,
1815                                        const char *name, void *opaque,
1816                                        Error **errp)
1817 {
1818     PCMachineState *pcms = PC_MACHINE(obj);
1819     uint64_t value = pcms->max_fw_size;
1820 
1821     visit_type_size(v, name, &value, errp);
1822 }
1823 
1824 static void pc_machine_set_max_fw_size(Object *obj, Visitor *v,
1825                                        const char *name, void *opaque,
1826                                        Error **errp)
1827 {
1828     PCMachineState *pcms = PC_MACHINE(obj);
1829     uint64_t value;
1830 
1831     if (!visit_type_size(v, name, &value, errp)) {
1832         return;
1833     }
1834 
1835     /*
1836     * We don't have a theoretically justifiable exact lower bound on the base
1837     * address of any flash mapping. In practice, the IO-APIC MMIO range is
1838     * [0xFEE00000..0xFEE01000] -- see IO_APIC_DEFAULT_ADDRESS --, leaving free
1839     * only 18MB-4KB below 4G. For now, restrict the cumulative mapping to 8MB in
1840     * size.
1841     */
1842     if (value > 16 * MiB) {
1843         error_setg(errp,
1844                    "User specified max allowed firmware size %" PRIu64 " is "
1845                    "greater than 16MiB. If combined firwmare size exceeds "
1846                    "16MiB the system may not boot, or experience intermittent"
1847                    "stability issues.",
1848                    value);
1849         return;
1850     }
1851 
1852     pcms->max_fw_size = value;
1853 }
1854 
1855 
1856 static void pc_machine_initfn(Object *obj)
1857 {
1858     PCMachineState *pcms = PC_MACHINE(obj);
1859 
1860 #ifdef CONFIG_VMPORT
1861     pcms->vmport = ON_OFF_AUTO_AUTO;
1862 #else
1863     pcms->vmport = ON_OFF_AUTO_OFF;
1864 #endif /* CONFIG_VMPORT */
1865     pcms->max_ram_below_4g = 0; /* use default */
1866     pcms->smbios_entry_point_type = SMBIOS_ENTRY_POINT_TYPE_32;
1867 
1868     /* acpi build is enabled by default if machine supports it */
1869     pcms->acpi_build_enabled = PC_MACHINE_GET_CLASS(pcms)->has_acpi_build;
1870     pcms->smbus_enabled = true;
1871     pcms->sata_enabled = true;
1872     pcms->i8042_enabled = true;
1873     pcms->max_fw_size = 8 * MiB;
1874 #ifdef CONFIG_HPET
1875     pcms->hpet_enabled = true;
1876 #endif
1877     pcms->default_bus_bypass_iommu = false;
1878 
1879     pc_system_flash_create(pcms);
1880     pcms->pcspk = isa_new(TYPE_PC_SPEAKER);
1881     object_property_add_alias(OBJECT(pcms), "pcspk-audiodev",
1882                               OBJECT(pcms->pcspk), "audiodev");
1883     cxl_machine_init(obj, &pcms->cxl_devices_state);
1884 }
1885 
1886 int pc_machine_kvm_type(MachineState *machine, const char *kvm_type)
1887 {
1888     return 0;
1889 }
1890 
1891 static void pc_machine_reset(MachineState *machine, ShutdownCause reason)
1892 {
1893     CPUState *cs;
1894     X86CPU *cpu;
1895 
1896     qemu_devices_reset(reason);
1897 
1898     /* Reset APIC after devices have been reset to cancel
1899      * any changes that qemu_devices_reset() might have done.
1900      */
1901     CPU_FOREACH(cs) {
1902         cpu = X86_CPU(cs);
1903 
1904         x86_cpu_after_reset(cpu);
1905     }
1906 }
1907 
1908 static void pc_machine_wakeup(MachineState *machine)
1909 {
1910     cpu_synchronize_all_states();
1911     pc_machine_reset(machine, SHUTDOWN_CAUSE_NONE);
1912     cpu_synchronize_all_post_reset();
1913 }
1914 
1915 static bool pc_hotplug_allowed(MachineState *ms, DeviceState *dev, Error **errp)
1916 {
1917     X86IOMMUState *iommu = x86_iommu_get_default();
1918     IntelIOMMUState *intel_iommu;
1919 
1920     if (iommu &&
1921         object_dynamic_cast((Object *)iommu, TYPE_INTEL_IOMMU_DEVICE) &&
1922         object_dynamic_cast((Object *)dev, "vfio-pci")) {
1923         intel_iommu = INTEL_IOMMU_DEVICE(iommu);
1924         if (!intel_iommu->caching_mode) {
1925             error_setg(errp, "Device assignment is not allowed without "
1926                        "enabling caching-mode=on for Intel IOMMU.");
1927             return false;
1928         }
1929     }
1930 
1931     return true;
1932 }
1933 
1934 static void pc_machine_class_init(ObjectClass *oc, void *data)
1935 {
1936     MachineClass *mc = MACHINE_CLASS(oc);
1937     PCMachineClass *pcmc = PC_MACHINE_CLASS(oc);
1938     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
1939 
1940     pcmc->pci_enabled = true;
1941     pcmc->has_acpi_build = true;
1942     pcmc->rsdp_in_ram = true;
1943     pcmc->smbios_defaults = true;
1944     pcmc->smbios_uuid_encoded = true;
1945     pcmc->gigabyte_align = true;
1946     pcmc->has_reserved_memory = true;
1947     pcmc->kvmclock_enabled = true;
1948     pcmc->enforce_aligned_dimm = true;
1949     pcmc->enforce_amd_1tb_hole = true;
1950     /* BIOS ACPI tables: 128K. Other BIOS datastructures: less than 4K reported
1951      * to be used at the moment, 32K should be enough for a while.  */
1952     pcmc->acpi_data_size = 0x20000 + 0x8000;
1953     pcmc->pvh_enabled = true;
1954     pcmc->kvmclock_create_always = true;
1955     pcmc->resizable_acpi_blob = true;
1956     assert(!mc->get_hotplug_handler);
1957     mc->get_hotplug_handler = pc_get_hotplug_handler;
1958     mc->hotplug_allowed = pc_hotplug_allowed;
1959     mc->cpu_index_to_instance_props = x86_cpu_index_to_props;
1960     mc->get_default_cpu_node_id = x86_get_default_cpu_node_id;
1961     mc->possible_cpu_arch_ids = x86_possible_cpu_arch_ids;
1962     mc->auto_enable_numa_with_memhp = true;
1963     mc->auto_enable_numa_with_memdev = true;
1964     mc->has_hotpluggable_cpus = true;
1965     mc->default_boot_order = "cad";
1966     mc->block_default_type = IF_IDE;
1967     mc->max_cpus = 255;
1968     mc->reset = pc_machine_reset;
1969     mc->wakeup = pc_machine_wakeup;
1970     hc->pre_plug = pc_machine_device_pre_plug_cb;
1971     hc->plug = pc_machine_device_plug_cb;
1972     hc->unplug_request = pc_machine_device_unplug_request_cb;
1973     hc->unplug = pc_machine_device_unplug_cb;
1974     mc->default_cpu_type = TARGET_DEFAULT_CPU_TYPE;
1975     mc->nvdimm_supported = true;
1976     mc->smp_props.dies_supported = true;
1977     mc->default_ram_id = "pc.ram";
1978 
1979     object_class_property_add(oc, PC_MACHINE_MAX_RAM_BELOW_4G, "size",
1980         pc_machine_get_max_ram_below_4g, pc_machine_set_max_ram_below_4g,
1981         NULL, NULL);
1982     object_class_property_set_description(oc, PC_MACHINE_MAX_RAM_BELOW_4G,
1983         "Maximum ram below the 4G boundary (32bit boundary)");
1984 
1985     object_class_property_add(oc, PC_MACHINE_DEVMEM_REGION_SIZE, "int",
1986         pc_machine_get_device_memory_region_size, NULL,
1987         NULL, NULL);
1988 
1989     object_class_property_add(oc, PC_MACHINE_VMPORT, "OnOffAuto",
1990         pc_machine_get_vmport, pc_machine_set_vmport,
1991         NULL, NULL);
1992     object_class_property_set_description(oc, PC_MACHINE_VMPORT,
1993         "Enable vmport (pc & q35)");
1994 
1995     object_class_property_add_bool(oc, PC_MACHINE_SMBUS,
1996         pc_machine_get_smbus, pc_machine_set_smbus);
1997     object_class_property_set_description(oc, PC_MACHINE_SMBUS,
1998         "Enable/disable system management bus");
1999 
2000     object_class_property_add_bool(oc, PC_MACHINE_SATA,
2001         pc_machine_get_sata, pc_machine_set_sata);
2002     object_class_property_set_description(oc, PC_MACHINE_SATA,
2003         "Enable/disable Serial ATA bus");
2004 
2005     object_class_property_add_bool(oc, "hpet",
2006         pc_machine_get_hpet, pc_machine_set_hpet);
2007     object_class_property_set_description(oc, "hpet",
2008         "Enable/disable high precision event timer emulation");
2009 
2010     object_class_property_add_bool(oc, PC_MACHINE_I8042,
2011         pc_machine_get_i8042, pc_machine_set_i8042);
2012 
2013     object_class_property_add_bool(oc, "default-bus-bypass-iommu",
2014         pc_machine_get_default_bus_bypass_iommu,
2015         pc_machine_set_default_bus_bypass_iommu);
2016 
2017     object_class_property_add(oc, PC_MACHINE_MAX_FW_SIZE, "size",
2018         pc_machine_get_max_fw_size, pc_machine_set_max_fw_size,
2019         NULL, NULL);
2020     object_class_property_set_description(oc, PC_MACHINE_MAX_FW_SIZE,
2021         "Maximum combined firmware size");
2022 
2023     object_class_property_add(oc, PC_MACHINE_SMBIOS_EP, "str",
2024         pc_machine_get_smbios_ep, pc_machine_set_smbios_ep,
2025         NULL, NULL);
2026     object_class_property_set_description(oc, PC_MACHINE_SMBIOS_EP,
2027         "SMBIOS Entry Point type [32, 64]");
2028 }
2029 
2030 static const TypeInfo pc_machine_info = {
2031     .name = TYPE_PC_MACHINE,
2032     .parent = TYPE_X86_MACHINE,
2033     .abstract = true,
2034     .instance_size = sizeof(PCMachineState),
2035     .instance_init = pc_machine_initfn,
2036     .class_size = sizeof(PCMachineClass),
2037     .class_init = pc_machine_class_init,
2038     .interfaces = (InterfaceInfo[]) {
2039          { TYPE_HOTPLUG_HANDLER },
2040          { }
2041     },
2042 };
2043 
2044 static void pc_machine_register_types(void)
2045 {
2046     type_register_static(&pc_machine_info);
2047 }
2048 
2049 type_init(pc_machine_register_types)
2050