xref: /openbmc/qemu/hw/i386/pc.c (revision df1f50c3)
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 const size_t pc_compat_8_0_len = G_N_ELEMENTS(pc_compat_8_0);
121 
122 GlobalProperty pc_compat_7_2[] = {
123     { "ICH9-LPC", "noreboot", "true" },
124 };
125 const size_t pc_compat_7_2_len = G_N_ELEMENTS(pc_compat_7_2);
126 
127 GlobalProperty pc_compat_7_1[] = {};
128 const size_t pc_compat_7_1_len = G_N_ELEMENTS(pc_compat_7_1);
129 
130 GlobalProperty pc_compat_7_0[] = {};
131 const size_t pc_compat_7_0_len = G_N_ELEMENTS(pc_compat_7_0);
132 
133 GlobalProperty pc_compat_6_2[] = {
134     { "virtio-mem", "unplugged-inaccessible", "off" },
135 };
136 const size_t pc_compat_6_2_len = G_N_ELEMENTS(pc_compat_6_2);
137 
138 GlobalProperty pc_compat_6_1[] = {
139     { TYPE_X86_CPU, "hv-version-id-build", "0x1bbc" },
140     { TYPE_X86_CPU, "hv-version-id-major", "0x0006" },
141     { TYPE_X86_CPU, "hv-version-id-minor", "0x0001" },
142     { "ICH9-LPC", "x-keep-pci-slot-hpc", "false" },
143 };
144 const size_t pc_compat_6_1_len = G_N_ELEMENTS(pc_compat_6_1);
145 
146 GlobalProperty pc_compat_6_0[] = {
147     { "qemu64" "-" TYPE_X86_CPU, "family", "6" },
148     { "qemu64" "-" TYPE_X86_CPU, "model", "6" },
149     { "qemu64" "-" TYPE_X86_CPU, "stepping", "3" },
150     { TYPE_X86_CPU, "x-vendor-cpuid-only", "off" },
151     { "ICH9-LPC", ACPI_PM_PROP_ACPI_PCIHP_BRIDGE, "off" },
152     { "ICH9-LPC", "x-keep-pci-slot-hpc", "true" },
153 };
154 const size_t pc_compat_6_0_len = G_N_ELEMENTS(pc_compat_6_0);
155 
156 GlobalProperty pc_compat_5_2[] = {
157     { "ICH9-LPC", "x-smi-cpu-hotunplug", "off" },
158 };
159 const size_t pc_compat_5_2_len = G_N_ELEMENTS(pc_compat_5_2);
160 
161 GlobalProperty pc_compat_5_1[] = {
162     { "ICH9-LPC", "x-smi-cpu-hotplug", "off" },
163     { TYPE_X86_CPU, "kvm-msi-ext-dest-id", "off" },
164 };
165 const size_t pc_compat_5_1_len = G_N_ELEMENTS(pc_compat_5_1);
166 
167 GlobalProperty pc_compat_5_0[] = {
168 };
169 const size_t pc_compat_5_0_len = G_N_ELEMENTS(pc_compat_5_0);
170 
171 GlobalProperty pc_compat_4_2[] = {
172     { "mch", "smbase-smram", "off" },
173 };
174 const size_t pc_compat_4_2_len = G_N_ELEMENTS(pc_compat_4_2);
175 
176 GlobalProperty pc_compat_4_1[] = {};
177 const size_t pc_compat_4_1_len = G_N_ELEMENTS(pc_compat_4_1);
178 
179 GlobalProperty pc_compat_4_0[] = {};
180 const size_t pc_compat_4_0_len = G_N_ELEMENTS(pc_compat_4_0);
181 
182 GlobalProperty pc_compat_3_1[] = {
183     { "intel-iommu", "dma-drain", "off" },
184     { "Opteron_G3" "-" TYPE_X86_CPU, "rdtscp", "off" },
185     { "Opteron_G4" "-" TYPE_X86_CPU, "rdtscp", "off" },
186     { "Opteron_G4" "-" TYPE_X86_CPU, "npt", "off" },
187     { "Opteron_G4" "-" TYPE_X86_CPU, "nrip-save", "off" },
188     { "Opteron_G5" "-" TYPE_X86_CPU, "rdtscp", "off" },
189     { "Opteron_G5" "-" TYPE_X86_CPU, "npt", "off" },
190     { "Opteron_G5" "-" TYPE_X86_CPU, "nrip-save", "off" },
191     { "EPYC" "-" TYPE_X86_CPU, "npt", "off" },
192     { "EPYC" "-" TYPE_X86_CPU, "nrip-save", "off" },
193     { "EPYC-IBPB" "-" TYPE_X86_CPU, "npt", "off" },
194     { "EPYC-IBPB" "-" TYPE_X86_CPU, "nrip-save", "off" },
195     { "Skylake-Client" "-" TYPE_X86_CPU,      "mpx", "on" },
196     { "Skylake-Client-IBRS" "-" TYPE_X86_CPU, "mpx", "on" },
197     { "Skylake-Server" "-" TYPE_X86_CPU,      "mpx", "on" },
198     { "Skylake-Server-IBRS" "-" TYPE_X86_CPU, "mpx", "on" },
199     { "Cascadelake-Server" "-" TYPE_X86_CPU,  "mpx", "on" },
200     { "Icelake-Client" "-" TYPE_X86_CPU,      "mpx", "on" },
201     { "Icelake-Server" "-" TYPE_X86_CPU,      "mpx", "on" },
202     { "Cascadelake-Server" "-" TYPE_X86_CPU, "stepping", "5" },
203     { TYPE_X86_CPU, "x-intel-pt-auto-level", "off" },
204 };
205 const size_t pc_compat_3_1_len = G_N_ELEMENTS(pc_compat_3_1);
206 
207 GlobalProperty pc_compat_3_0[] = {
208     { TYPE_X86_CPU, "x-hv-synic-kvm-only", "on" },
209     { "Skylake-Server" "-" TYPE_X86_CPU, "pku", "off" },
210     { "Skylake-Server-IBRS" "-" TYPE_X86_CPU, "pku", "off" },
211 };
212 const size_t pc_compat_3_0_len = G_N_ELEMENTS(pc_compat_3_0);
213 
214 GlobalProperty pc_compat_2_12[] = {
215     { TYPE_X86_CPU, "legacy-cache", "on" },
216     { TYPE_X86_CPU, "topoext", "off" },
217     { "EPYC-" TYPE_X86_CPU, "xlevel", "0x8000000a" },
218     { "EPYC-IBPB-" TYPE_X86_CPU, "xlevel", "0x8000000a" },
219 };
220 const size_t pc_compat_2_12_len = G_N_ELEMENTS(pc_compat_2_12);
221 
222 GlobalProperty pc_compat_2_11[] = {
223     { TYPE_X86_CPU, "x-migrate-smi-count", "off" },
224     { "Skylake-Server" "-" TYPE_X86_CPU, "clflushopt", "off" },
225 };
226 const size_t pc_compat_2_11_len = G_N_ELEMENTS(pc_compat_2_11);
227 
228 GlobalProperty pc_compat_2_10[] = {
229     { TYPE_X86_CPU, "x-hv-max-vps", "0x40" },
230     { "i440FX-pcihost", "x-pci-hole64-fix", "off" },
231     { "q35-pcihost", "x-pci-hole64-fix", "off" },
232 };
233 const size_t pc_compat_2_10_len = G_N_ELEMENTS(pc_compat_2_10);
234 
235 GlobalProperty pc_compat_2_9[] = {
236     { "mch", "extended-tseg-mbytes", "0" },
237 };
238 const size_t pc_compat_2_9_len = G_N_ELEMENTS(pc_compat_2_9);
239 
240 GlobalProperty pc_compat_2_8[] = {
241     { TYPE_X86_CPU, "tcg-cpuid", "off" },
242     { "kvmclock", "x-mach-use-reliable-get-clock", "off" },
243     { "ICH9-LPC", "x-smi-broadcast", "off" },
244     { TYPE_X86_CPU, "vmware-cpuid-freq", "off" },
245     { "Haswell-" TYPE_X86_CPU, "stepping", "1" },
246 };
247 const size_t pc_compat_2_8_len = G_N_ELEMENTS(pc_compat_2_8);
248 
249 GlobalProperty pc_compat_2_7[] = {
250     { TYPE_X86_CPU, "l3-cache", "off" },
251     { TYPE_X86_CPU, "full-cpuid-auto-level", "off" },
252     { "Opteron_G3" "-" TYPE_X86_CPU, "family", "15" },
253     { "Opteron_G3" "-" TYPE_X86_CPU, "model", "6" },
254     { "Opteron_G3" "-" TYPE_X86_CPU, "stepping", "1" },
255     { "isa-pcspk", "migrate", "off" },
256 };
257 const size_t pc_compat_2_7_len = G_N_ELEMENTS(pc_compat_2_7);
258 
259 GlobalProperty pc_compat_2_6[] = {
260     { TYPE_X86_CPU, "cpuid-0xb", "off" },
261     { "vmxnet3", "romfile", "" },
262     { TYPE_X86_CPU, "fill-mtrr-mask", "off" },
263     { "apic-common", "legacy-instance-id", "on", }
264 };
265 const size_t pc_compat_2_6_len = G_N_ELEMENTS(pc_compat_2_6);
266 
267 GlobalProperty pc_compat_2_5[] = {};
268 const size_t pc_compat_2_5_len = G_N_ELEMENTS(pc_compat_2_5);
269 
270 GlobalProperty pc_compat_2_4[] = {
271     PC_CPU_MODEL_IDS("2.4.0")
272     { "Haswell-" TYPE_X86_CPU, "abm", "off" },
273     { "Haswell-noTSX-" TYPE_X86_CPU, "abm", "off" },
274     { "Broadwell-" TYPE_X86_CPU, "abm", "off" },
275     { "Broadwell-noTSX-" TYPE_X86_CPU, "abm", "off" },
276     { "host" "-" TYPE_X86_CPU, "host-cache-info", "on" },
277     { TYPE_X86_CPU, "check", "off" },
278     { "qemu64" "-" TYPE_X86_CPU, "sse4a", "on" },
279     { "qemu64" "-" TYPE_X86_CPU, "abm", "on" },
280     { "qemu64" "-" TYPE_X86_CPU, "popcnt", "on" },
281     { "qemu32" "-" TYPE_X86_CPU, "popcnt", "on" },
282     { "Opteron_G2" "-" TYPE_X86_CPU, "rdtscp", "on" },
283     { "Opteron_G3" "-" TYPE_X86_CPU, "rdtscp", "on" },
284     { "Opteron_G4" "-" TYPE_X86_CPU, "rdtscp", "on" },
285     { "Opteron_G5" "-" TYPE_X86_CPU, "rdtscp", "on", }
286 };
287 const size_t pc_compat_2_4_len = G_N_ELEMENTS(pc_compat_2_4);
288 
289 GlobalProperty pc_compat_2_3[] = {
290     PC_CPU_MODEL_IDS("2.3.0")
291     { TYPE_X86_CPU, "arat", "off" },
292     { "qemu64" "-" TYPE_X86_CPU, "min-level", "4" },
293     { "kvm64" "-" TYPE_X86_CPU, "min-level", "5" },
294     { "pentium3" "-" TYPE_X86_CPU, "min-level", "2" },
295     { "n270" "-" TYPE_X86_CPU, "min-level", "5" },
296     { "Conroe" "-" TYPE_X86_CPU, "min-level", "4" },
297     { "Penryn" "-" TYPE_X86_CPU, "min-level", "4" },
298     { "Nehalem" "-" TYPE_X86_CPU, "min-level", "4" },
299     { "n270" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
300     { "Penryn" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
301     { "Conroe" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
302     { "Nehalem" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
303     { "Westmere" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
304     { "SandyBridge" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
305     { "IvyBridge" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
306     { "Haswell" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
307     { "Haswell-noTSX" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
308     { "Broadwell" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
309     { "Broadwell-noTSX" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
310     { TYPE_X86_CPU, "kvm-no-smi-migration", "on" },
311 };
312 const size_t pc_compat_2_3_len = G_N_ELEMENTS(pc_compat_2_3);
313 
314 GlobalProperty pc_compat_2_2[] = {
315     PC_CPU_MODEL_IDS("2.2.0")
316     { "kvm64" "-" TYPE_X86_CPU, "vme", "off" },
317     { "kvm32" "-" TYPE_X86_CPU, "vme", "off" },
318     { "Conroe" "-" TYPE_X86_CPU, "vme", "off" },
319     { "Penryn" "-" TYPE_X86_CPU, "vme", "off" },
320     { "Nehalem" "-" TYPE_X86_CPU, "vme", "off" },
321     { "Westmere" "-" TYPE_X86_CPU, "vme", "off" },
322     { "SandyBridge" "-" TYPE_X86_CPU, "vme", "off" },
323     { "Haswell" "-" TYPE_X86_CPU, "vme", "off" },
324     { "Broadwell" "-" TYPE_X86_CPU, "vme", "off" },
325     { "Opteron_G1" "-" TYPE_X86_CPU, "vme", "off" },
326     { "Opteron_G2" "-" TYPE_X86_CPU, "vme", "off" },
327     { "Opteron_G3" "-" TYPE_X86_CPU, "vme", "off" },
328     { "Opteron_G4" "-" TYPE_X86_CPU, "vme", "off" },
329     { "Opteron_G5" "-" TYPE_X86_CPU, "vme", "off" },
330     { "Haswell" "-" TYPE_X86_CPU, "f16c", "off" },
331     { "Haswell" "-" TYPE_X86_CPU, "rdrand", "off" },
332     { "Broadwell" "-" TYPE_X86_CPU, "f16c", "off" },
333     { "Broadwell" "-" TYPE_X86_CPU, "rdrand", "off" },
334 };
335 const size_t pc_compat_2_2_len = G_N_ELEMENTS(pc_compat_2_2);
336 
337 GlobalProperty pc_compat_2_1[] = {
338     PC_CPU_MODEL_IDS("2.1.0")
339     { "coreduo" "-" TYPE_X86_CPU, "vmx", "on" },
340     { "core2duo" "-" TYPE_X86_CPU, "vmx", "on" },
341 };
342 const size_t pc_compat_2_1_len = G_N_ELEMENTS(pc_compat_2_1);
343 
344 GlobalProperty pc_compat_2_0[] = {
345     PC_CPU_MODEL_IDS("2.0.0")
346     { "virtio-scsi-pci", "any_layout", "off" },
347     { "PIIX4_PM", "memory-hotplug-support", "off" },
348     { "apic", "version", "0x11" },
349     { "nec-usb-xhci", "superspeed-ports-first", "off" },
350     { "nec-usb-xhci", "force-pcie-endcap", "on" },
351     { "pci-serial", "prog_if", "0" },
352     { "pci-serial-2x", "prog_if", "0" },
353     { "pci-serial-4x", "prog_if", "0" },
354     { "virtio-net-pci", "guest_announce", "off" },
355     { "ICH9-LPC", "memory-hotplug-support", "off" },
356 };
357 const size_t pc_compat_2_0_len = G_N_ELEMENTS(pc_compat_2_0);
358 
359 GlobalProperty pc_compat_1_7[] = {
360     PC_CPU_MODEL_IDS("1.7.0")
361     { TYPE_USB_DEVICE, "msos-desc", "no" },
362     { "PIIX4_PM", ACPI_PM_PROP_ACPI_PCIHP_BRIDGE, "off" },
363     { "hpet", HPET_INTCAP, "4" },
364 };
365 const size_t pc_compat_1_7_len = G_N_ELEMENTS(pc_compat_1_7);
366 
367 GlobalProperty pc_compat_1_6[] = {
368     PC_CPU_MODEL_IDS("1.6.0")
369     { "e1000", "mitigation", "off" },
370     { "qemu64-" TYPE_X86_CPU, "model", "2" },
371     { "qemu32-" TYPE_X86_CPU, "model", "3" },
372     { "i440FX-pcihost", "short_root_bus", "1" },
373     { "q35-pcihost", "short_root_bus", "1" },
374 };
375 const size_t pc_compat_1_6_len = G_N_ELEMENTS(pc_compat_1_6);
376 
377 GlobalProperty pc_compat_1_5[] = {
378     PC_CPU_MODEL_IDS("1.5.0")
379     { "Conroe-" TYPE_X86_CPU, "model", "2" },
380     { "Conroe-" TYPE_X86_CPU, "min-level", "2" },
381     { "Penryn-" TYPE_X86_CPU, "model", "2" },
382     { "Penryn-" TYPE_X86_CPU, "min-level", "2" },
383     { "Nehalem-" TYPE_X86_CPU, "model", "2" },
384     { "Nehalem-" TYPE_X86_CPU, "min-level", "2" },
385     { "virtio-net-pci", "any_layout", "off" },
386     { TYPE_X86_CPU, "pmu", "on" },
387     { "i440FX-pcihost", "short_root_bus", "0" },
388     { "q35-pcihost", "short_root_bus", "0" },
389 };
390 const size_t pc_compat_1_5_len = G_N_ELEMENTS(pc_compat_1_5);
391 
392 GlobalProperty pc_compat_1_4[] = {
393     PC_CPU_MODEL_IDS("1.4.0")
394     { "scsi-hd", "discard_granularity", "0" },
395     { "scsi-cd", "discard_granularity", "0" },
396     { "ide-hd", "discard_granularity", "0" },
397     { "ide-cd", "discard_granularity", "0" },
398     { "virtio-blk-pci", "discard_granularity", "0" },
399     /* DEV_NVECTORS_UNSPECIFIED as a uint32_t string: */
400     { "virtio-serial-pci", "vectors", "0xFFFFFFFF" },
401     { "virtio-net-pci", "ctrl_guest_offloads", "off" },
402     { "e1000", "romfile", "pxe-e1000.rom" },
403     { "ne2k_pci", "romfile", "pxe-ne2k_pci.rom" },
404     { "pcnet", "romfile", "pxe-pcnet.rom" },
405     { "rtl8139", "romfile", "pxe-rtl8139.rom" },
406     { "virtio-net-pci", "romfile", "pxe-virtio.rom" },
407     { "486-" TYPE_X86_CPU, "model", "0" },
408     { "n270" "-" TYPE_X86_CPU, "movbe", "off" },
409     { "Westmere" "-" TYPE_X86_CPU, "pclmulqdq", "off" },
410 };
411 const size_t pc_compat_1_4_len = G_N_ELEMENTS(pc_compat_1_4);
412 
413 GSIState *pc_gsi_create(qemu_irq **irqs, bool pci_enabled)
414 {
415     GSIState *s;
416 
417     s = g_new0(GSIState, 1);
418     if (kvm_ioapic_in_kernel()) {
419         kvm_pc_setup_irq_routing(pci_enabled);
420     }
421     *irqs = qemu_allocate_irqs(gsi_handler, s, IOAPIC_NUM_PINS);
422 
423     return s;
424 }
425 
426 static void ioport80_write(void *opaque, hwaddr addr, uint64_t data,
427                            unsigned size)
428 {
429 }
430 
431 static uint64_t ioport80_read(void *opaque, hwaddr addr, unsigned size)
432 {
433     return 0xffffffffffffffffULL;
434 }
435 
436 /* MSDOS compatibility mode FPU exception support */
437 static void ioportF0_write(void *opaque, hwaddr addr, uint64_t data,
438                            unsigned size)
439 {
440     if (tcg_enabled()) {
441         cpu_set_ignne();
442     }
443 }
444 
445 static uint64_t ioportF0_read(void *opaque, hwaddr addr, unsigned size)
446 {
447     return 0xffffffffffffffffULL;
448 }
449 
450 /* PC cmos mappings */
451 
452 #define REG_EQUIPMENT_BYTE          0x14
453 
454 static void cmos_init_hd(MC146818RtcState *s, int type_ofs, int info_ofs,
455                          int16_t cylinders, int8_t heads, int8_t sectors)
456 {
457     mc146818rtc_set_cmos_data(s, type_ofs, 47);
458     mc146818rtc_set_cmos_data(s, info_ofs, cylinders);
459     mc146818rtc_set_cmos_data(s, info_ofs + 1, cylinders >> 8);
460     mc146818rtc_set_cmos_data(s, info_ofs + 2, heads);
461     mc146818rtc_set_cmos_data(s, info_ofs + 3, 0xff);
462     mc146818rtc_set_cmos_data(s, info_ofs + 4, 0xff);
463     mc146818rtc_set_cmos_data(s, info_ofs + 5, 0xc0 | ((heads > 8) << 3));
464     mc146818rtc_set_cmos_data(s, info_ofs + 6, cylinders);
465     mc146818rtc_set_cmos_data(s, info_ofs + 7, cylinders >> 8);
466     mc146818rtc_set_cmos_data(s, info_ofs + 8, sectors);
467 }
468 
469 /* convert boot_device letter to something recognizable by the bios */
470 static int boot_device2nibble(char boot_device)
471 {
472     switch(boot_device) {
473     case 'a':
474     case 'b':
475         return 0x01; /* floppy boot */
476     case 'c':
477         return 0x02; /* hard drive boot */
478     case 'd':
479         return 0x03; /* CD-ROM boot */
480     case 'n':
481         return 0x04; /* Network boot */
482     }
483     return 0;
484 }
485 
486 static void set_boot_dev(MC146818RtcState *s, const char *boot_device,
487                          Error **errp)
488 {
489 #define PC_MAX_BOOT_DEVICES 3
490     int nbds, bds[3] = { 0, };
491     int i;
492 
493     nbds = strlen(boot_device);
494     if (nbds > PC_MAX_BOOT_DEVICES) {
495         error_setg(errp, "Too many boot devices for PC");
496         return;
497     }
498     for (i = 0; i < nbds; i++) {
499         bds[i] = boot_device2nibble(boot_device[i]);
500         if (bds[i] == 0) {
501             error_setg(errp, "Invalid boot device for PC: '%c'",
502                        boot_device[i]);
503             return;
504         }
505     }
506     mc146818rtc_set_cmos_data(s, 0x3d, (bds[1] << 4) | bds[0]);
507     mc146818rtc_set_cmos_data(s, 0x38, (bds[2] << 4) | (fd_bootchk ? 0x0 : 0x1));
508 }
509 
510 static void pc_boot_set(void *opaque, const char *boot_device, Error **errp)
511 {
512     set_boot_dev(opaque, boot_device, errp);
513 }
514 
515 static void pc_cmos_init_floppy(MC146818RtcState *rtc_state, ISADevice *floppy)
516 {
517     int val, nb, i;
518     FloppyDriveType fd_type[2] = { FLOPPY_DRIVE_TYPE_NONE,
519                                    FLOPPY_DRIVE_TYPE_NONE };
520 
521     /* floppy type */
522     if (floppy) {
523         for (i = 0; i < 2; i++) {
524             fd_type[i] = isa_fdc_get_drive_type(floppy, i);
525         }
526     }
527     val = (cmos_get_fd_drive_type(fd_type[0]) << 4) |
528         cmos_get_fd_drive_type(fd_type[1]);
529     mc146818rtc_set_cmos_data(rtc_state, 0x10, val);
530 
531     val = mc146818rtc_get_cmos_data(rtc_state, REG_EQUIPMENT_BYTE);
532     nb = 0;
533     if (fd_type[0] != FLOPPY_DRIVE_TYPE_NONE) {
534         nb++;
535     }
536     if (fd_type[1] != FLOPPY_DRIVE_TYPE_NONE) {
537         nb++;
538     }
539     switch (nb) {
540     case 0:
541         break;
542     case 1:
543         val |= 0x01; /* 1 drive, ready for boot */
544         break;
545     case 2:
546         val |= 0x41; /* 2 drives, ready for boot */
547         break;
548     }
549     mc146818rtc_set_cmos_data(rtc_state, REG_EQUIPMENT_BYTE, val);
550 }
551 
552 typedef struct pc_cmos_init_late_arg {
553     MC146818RtcState *rtc_state;
554     BusState *idebus[2];
555 } pc_cmos_init_late_arg;
556 
557 typedef struct check_fdc_state {
558     ISADevice *floppy;
559     bool multiple;
560 } CheckFdcState;
561 
562 static int check_fdc(Object *obj, void *opaque)
563 {
564     CheckFdcState *state = opaque;
565     Object *fdc;
566     uint32_t iobase;
567     Error *local_err = NULL;
568 
569     fdc = object_dynamic_cast(obj, TYPE_ISA_FDC);
570     if (!fdc) {
571         return 0;
572     }
573 
574     iobase = object_property_get_uint(obj, "iobase", &local_err);
575     if (local_err || iobase != 0x3f0) {
576         error_free(local_err);
577         return 0;
578     }
579 
580     if (state->floppy) {
581         state->multiple = true;
582     } else {
583         state->floppy = ISA_DEVICE(obj);
584     }
585     return 0;
586 }
587 
588 static const char * const fdc_container_path[] = {
589     "/unattached", "/peripheral", "/peripheral-anon"
590 };
591 
592 /*
593  * Locate the FDC at IO address 0x3f0, in order to configure the CMOS registers
594  * and ACPI objects.
595  */
596 static ISADevice *pc_find_fdc0(void)
597 {
598     int i;
599     Object *container;
600     CheckFdcState state = { 0 };
601 
602     for (i = 0; i < ARRAY_SIZE(fdc_container_path); i++) {
603         container = container_get(qdev_get_machine(), fdc_container_path[i]);
604         object_child_foreach(container, check_fdc, &state);
605     }
606 
607     if (state.multiple) {
608         warn_report("multiple floppy disk controllers with "
609                     "iobase=0x3f0 have been found");
610         error_printf("the one being picked for CMOS setup might not reflect "
611                      "your intent");
612     }
613 
614     return state.floppy;
615 }
616 
617 static void pc_cmos_init_late(void *opaque)
618 {
619     pc_cmos_init_late_arg *arg = opaque;
620     MC146818RtcState *s = arg->rtc_state;
621     int16_t cylinders;
622     int8_t heads, sectors;
623     int val;
624     int i, trans;
625 
626     val = 0;
627     if (arg->idebus[0] && ide_get_geometry(arg->idebus[0], 0,
628                                            &cylinders, &heads, &sectors) >= 0) {
629         cmos_init_hd(s, 0x19, 0x1b, cylinders, heads, sectors);
630         val |= 0xf0;
631     }
632     if (arg->idebus[0] && ide_get_geometry(arg->idebus[0], 1,
633                                            &cylinders, &heads, &sectors) >= 0) {
634         cmos_init_hd(s, 0x1a, 0x24, cylinders, heads, sectors);
635         val |= 0x0f;
636     }
637     mc146818rtc_set_cmos_data(s, 0x12, val);
638 
639     val = 0;
640     for (i = 0; i < 4; i++) {
641         /* NOTE: ide_get_geometry() returns the physical
642            geometry.  It is always such that: 1 <= sects <= 63, 1
643            <= heads <= 16, 1 <= cylinders <= 16383. The BIOS
644            geometry can be different if a translation is done. */
645         if (arg->idebus[i / 2] &&
646             ide_get_geometry(arg->idebus[i / 2], i % 2,
647                              &cylinders, &heads, &sectors) >= 0) {
648             trans = ide_get_bios_chs_trans(arg->idebus[i / 2], i % 2) - 1;
649             assert((trans & ~3) == 0);
650             val |= trans << (i * 2);
651         }
652     }
653     mc146818rtc_set_cmos_data(s, 0x39, val);
654 
655     pc_cmos_init_floppy(s, pc_find_fdc0());
656 
657     qemu_unregister_reset(pc_cmos_init_late, opaque);
658 }
659 
660 void pc_cmos_init(PCMachineState *pcms,
661                   BusState *idebus0, BusState *idebus1,
662                   ISADevice *rtc)
663 {
664     int val;
665     static pc_cmos_init_late_arg arg;
666     X86MachineState *x86ms = X86_MACHINE(pcms);
667     MC146818RtcState *s = MC146818_RTC(rtc);
668 
669     /* various important CMOS locations needed by PC/Bochs bios */
670 
671     /* memory size */
672     /* base memory (first MiB) */
673     val = MIN(x86ms->below_4g_mem_size / KiB, 640);
674     mc146818rtc_set_cmos_data(s, 0x15, val);
675     mc146818rtc_set_cmos_data(s, 0x16, val >> 8);
676     /* extended memory (next 64MiB) */
677     if (x86ms->below_4g_mem_size > 1 * MiB) {
678         val = (x86ms->below_4g_mem_size - 1 * MiB) / KiB;
679     } else {
680         val = 0;
681     }
682     if (val > 65535)
683         val = 65535;
684     mc146818rtc_set_cmos_data(s, 0x17, val);
685     mc146818rtc_set_cmos_data(s, 0x18, val >> 8);
686     mc146818rtc_set_cmos_data(s, 0x30, val);
687     mc146818rtc_set_cmos_data(s, 0x31, val >> 8);
688     /* memory between 16MiB and 4GiB */
689     if (x86ms->below_4g_mem_size > 16 * MiB) {
690         val = (x86ms->below_4g_mem_size - 16 * MiB) / (64 * KiB);
691     } else {
692         val = 0;
693     }
694     if (val > 65535)
695         val = 65535;
696     mc146818rtc_set_cmos_data(s, 0x34, val);
697     mc146818rtc_set_cmos_data(s, 0x35, val >> 8);
698     /* memory above 4GiB */
699     val = x86ms->above_4g_mem_size / 65536;
700     mc146818rtc_set_cmos_data(s, 0x5b, val);
701     mc146818rtc_set_cmos_data(s, 0x5c, val >> 8);
702     mc146818rtc_set_cmos_data(s, 0x5d, val >> 16);
703 
704     object_property_add_link(OBJECT(pcms), "rtc_state",
705                              TYPE_ISA_DEVICE,
706                              (Object **)&x86ms->rtc,
707                              object_property_allow_set_link,
708                              OBJ_PROP_LINK_STRONG);
709     object_property_set_link(OBJECT(pcms), "rtc_state", OBJECT(s),
710                              &error_abort);
711 
712     set_boot_dev(s, MACHINE(pcms)->boot_config.order, &error_fatal);
713 
714     val = 0;
715     val |= 0x02; /* FPU is there */
716     val |= 0x04; /* PS/2 mouse installed */
717     mc146818rtc_set_cmos_data(s, REG_EQUIPMENT_BYTE, val);
718 
719     /* hard drives and FDC */
720     arg.rtc_state = s;
721     arg.idebus[0] = idebus0;
722     arg.idebus[1] = idebus1;
723     qemu_register_reset(pc_cmos_init_late, &arg);
724 }
725 
726 static void handle_a20_line_change(void *opaque, int irq, int level)
727 {
728     X86CPU *cpu = opaque;
729 
730     /* XXX: send to all CPUs ? */
731     /* XXX: add logic to handle multiple A20 line sources */
732     x86_cpu_set_a20(cpu, level);
733 }
734 
735 #define NE2000_NB_MAX 6
736 
737 static const int ne2000_io[NE2000_NB_MAX] = { 0x300, 0x320, 0x340, 0x360,
738                                               0x280, 0x380 };
739 static const int ne2000_irq[NE2000_NB_MAX] = { 9, 10, 11, 3, 4, 5 };
740 
741 static void pc_init_ne2k_isa(ISABus *bus, NICInfo *nd)
742 {
743     static int nb_ne2k = 0;
744 
745     if (nb_ne2k == NE2000_NB_MAX)
746         return;
747     isa_ne2000_init(bus, ne2000_io[nb_ne2k],
748                     ne2000_irq[nb_ne2k], nd);
749     nb_ne2k++;
750 }
751 
752 void pc_acpi_smi_interrupt(void *opaque, int irq, int level)
753 {
754     X86CPU *cpu = opaque;
755 
756     if (level) {
757         cpu_interrupt(CPU(cpu), CPU_INTERRUPT_SMI);
758     }
759 }
760 
761 static
762 void pc_machine_done(Notifier *notifier, void *data)
763 {
764     PCMachineState *pcms = container_of(notifier,
765                                         PCMachineState, machine_done);
766     X86MachineState *x86ms = X86_MACHINE(pcms);
767 
768     cxl_hook_up_pxb_registers(pcms->bus, &pcms->cxl_devices_state,
769                               &error_fatal);
770 
771     if (pcms->cxl_devices_state.is_enabled) {
772         cxl_fmws_link_targets(&pcms->cxl_devices_state, &error_fatal);
773     }
774 
775     /* set the number of CPUs */
776     x86_rtc_set_cpus_count(x86ms->rtc, x86ms->boot_cpus);
777 
778     fw_cfg_add_extra_pci_roots(pcms->bus, x86ms->fw_cfg);
779 
780     acpi_setup();
781     if (x86ms->fw_cfg) {
782         fw_cfg_build_smbios(MACHINE(pcms), x86ms->fw_cfg);
783         fw_cfg_build_feature_control(MACHINE(pcms), x86ms->fw_cfg);
784         /* update FW_CFG_NB_CPUS to account for -device added CPUs */
785         fw_cfg_modify_i16(x86ms->fw_cfg, FW_CFG_NB_CPUS, x86ms->boot_cpus);
786     }
787 }
788 
789 void pc_guest_info_init(PCMachineState *pcms)
790 {
791     X86MachineState *x86ms = X86_MACHINE(pcms);
792 
793     x86ms->apic_xrupt_override = true;
794     pcms->machine_done.notify = pc_machine_done;
795     qemu_add_machine_init_done_notifier(&pcms->machine_done);
796 }
797 
798 /* setup pci memory address space mapping into system address space */
799 void pc_pci_as_mapping_init(MemoryRegion *system_memory,
800                             MemoryRegion *pci_address_space)
801 {
802     /* Set to lower priority than RAM */
803     memory_region_add_subregion_overlap(system_memory, 0x0,
804                                         pci_address_space, -1);
805 }
806 
807 void xen_load_linux(PCMachineState *pcms)
808 {
809     int i;
810     FWCfgState *fw_cfg;
811     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
812     X86MachineState *x86ms = X86_MACHINE(pcms);
813 
814     assert(MACHINE(pcms)->kernel_filename != NULL);
815 
816     fw_cfg = fw_cfg_init_io(FW_CFG_IO_BASE);
817     fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, x86ms->boot_cpus);
818     rom_set_fw(fw_cfg);
819 
820     x86_load_linux(x86ms, fw_cfg, pcmc->acpi_data_size,
821                    pcmc->pvh_enabled);
822     for (i = 0; i < nb_option_roms; i++) {
823         assert(!strcmp(option_rom[i].name, "linuxboot.bin") ||
824                !strcmp(option_rom[i].name, "linuxboot_dma.bin") ||
825                !strcmp(option_rom[i].name, "pvh.bin") ||
826                !strcmp(option_rom[i].name, "multiboot.bin") ||
827                !strcmp(option_rom[i].name, "multiboot_dma.bin"));
828         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
829     }
830     x86ms->fw_cfg = fw_cfg;
831 }
832 
833 #define PC_ROM_MIN_VGA     0xc0000
834 #define PC_ROM_MIN_OPTION  0xc8000
835 #define PC_ROM_MAX         0xe0000
836 #define PC_ROM_ALIGN       0x800
837 #define PC_ROM_SIZE        (PC_ROM_MAX - PC_ROM_MIN_VGA)
838 
839 static hwaddr pc_above_4g_end(PCMachineState *pcms)
840 {
841     X86MachineState *x86ms = X86_MACHINE(pcms);
842 
843     if (pcms->sgx_epc.size != 0) {
844         return sgx_epc_above_4g_end(&pcms->sgx_epc);
845     }
846 
847     return x86ms->above_4g_mem_start + x86ms->above_4g_mem_size;
848 }
849 
850 static void pc_get_device_memory_range(PCMachineState *pcms,
851                                        hwaddr *base,
852                                        ram_addr_t *device_mem_size)
853 {
854     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
855     MachineState *machine = MACHINE(pcms);
856     ram_addr_t size;
857     hwaddr addr;
858 
859     size = machine->maxram_size - machine->ram_size;
860     addr = ROUND_UP(pc_above_4g_end(pcms), 1 * GiB);
861 
862     if (pcmc->enforce_aligned_dimm) {
863         /* size device region assuming 1G page max alignment per slot */
864         size += (1 * GiB) * machine->ram_slots;
865     }
866 
867     *base = addr;
868     *device_mem_size = size;
869 }
870 
871 static uint64_t pc_get_cxl_range_start(PCMachineState *pcms)
872 {
873     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
874     hwaddr cxl_base;
875     ram_addr_t size;
876 
877     if (pcmc->has_reserved_memory) {
878         pc_get_device_memory_range(pcms, &cxl_base, &size);
879         cxl_base += size;
880     } else {
881         cxl_base = pc_above_4g_end(pcms);
882     }
883 
884     return cxl_base;
885 }
886 
887 static uint64_t pc_get_cxl_range_end(PCMachineState *pcms)
888 {
889     uint64_t start = pc_get_cxl_range_start(pcms) + MiB;
890 
891     if (pcms->cxl_devices_state.fixed_windows) {
892         GList *it;
893 
894         start = ROUND_UP(start, 256 * MiB);
895         for (it = pcms->cxl_devices_state.fixed_windows; it; it = it->next) {
896             CXLFixedWindow *fw = it->data;
897             start += fw->size;
898         }
899     }
900 
901     return start;
902 }
903 
904 static hwaddr pc_max_used_gpa(PCMachineState *pcms, uint64_t pci_hole64_size)
905 {
906     X86CPU *cpu = X86_CPU(first_cpu);
907 
908     /* 32-bit systems don't have hole64 thus return max CPU address */
909     if (cpu->phys_bits <= 32) {
910         return ((hwaddr)1 << cpu->phys_bits) - 1;
911     }
912 
913     return pc_pci_hole64_start() + pci_hole64_size - 1;
914 }
915 
916 /*
917  * AMD systems with an IOMMU have an additional hole close to the
918  * 1Tb, which are special GPAs that cannot be DMA mapped. Depending
919  * on kernel version, VFIO may or may not let you DMA map those ranges.
920  * Starting Linux v5.4 we validate it, and can't create guests on AMD machines
921  * with certain memory sizes. It's also wrong to use those IOVA ranges
922  * in detriment of leading to IOMMU INVALID_DEVICE_REQUEST or worse.
923  * The ranges reserved for Hyper-Transport are:
924  *
925  * FD_0000_0000h - FF_FFFF_FFFFh
926  *
927  * The ranges represent the following:
928  *
929  * Base Address   Top Address  Use
930  *
931  * FD_0000_0000h FD_F7FF_FFFFh Reserved interrupt address space
932  * FD_F800_0000h FD_F8FF_FFFFh Interrupt/EOI IntCtl
933  * FD_F900_0000h FD_F90F_FFFFh Legacy PIC IACK
934  * FD_F910_0000h FD_F91F_FFFFh System Management
935  * FD_F920_0000h FD_FAFF_FFFFh Reserved Page Tables
936  * FD_FB00_0000h FD_FBFF_FFFFh Address Translation
937  * FD_FC00_0000h FD_FDFF_FFFFh I/O Space
938  * FD_FE00_0000h FD_FFFF_FFFFh Configuration
939  * FE_0000_0000h FE_1FFF_FFFFh Extended Configuration/Device Messages
940  * FE_2000_0000h FF_FFFF_FFFFh Reserved
941  *
942  * See AMD IOMMU spec, section 2.1.2 "IOMMU Logical Topology",
943  * Table 3: Special Address Controls (GPA) for more information.
944  */
945 #define AMD_HT_START         0xfd00000000UL
946 #define AMD_HT_END           0xffffffffffUL
947 #define AMD_ABOVE_1TB_START  (AMD_HT_END + 1)
948 #define AMD_HT_SIZE          (AMD_ABOVE_1TB_START - AMD_HT_START)
949 
950 void pc_memory_init(PCMachineState *pcms,
951                     MemoryRegion *system_memory,
952                     MemoryRegion *rom_memory,
953                     MemoryRegion **ram_memory,
954                     uint64_t pci_hole64_size)
955 {
956     int linux_boot, i;
957     MemoryRegion *option_rom_mr;
958     MemoryRegion *ram_below_4g, *ram_above_4g;
959     FWCfgState *fw_cfg;
960     MachineState *machine = MACHINE(pcms);
961     MachineClass *mc = MACHINE_GET_CLASS(machine);
962     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
963     X86MachineState *x86ms = X86_MACHINE(pcms);
964     hwaddr maxphysaddr, maxusedaddr;
965     hwaddr cxl_base, cxl_resv_end = 0;
966     X86CPU *cpu = X86_CPU(first_cpu);
967 
968     assert(machine->ram_size == x86ms->below_4g_mem_size +
969                                 x86ms->above_4g_mem_size);
970 
971     linux_boot = (machine->kernel_filename != NULL);
972 
973     /*
974      * The HyperTransport range close to the 1T boundary is unique to AMD
975      * hosts with IOMMUs enabled. Restrict the ram-above-4g relocation
976      * to above 1T to AMD vCPUs only. @enforce_amd_1tb_hole is only false in
977      * older machine types (<= 7.0) for compatibility purposes.
978      */
979     if (IS_AMD_CPU(&cpu->env) && pcmc->enforce_amd_1tb_hole) {
980         /* Bail out if max possible address does not cross HT range */
981         if (pc_max_used_gpa(pcms, pci_hole64_size) >= AMD_HT_START) {
982             x86ms->above_4g_mem_start = AMD_ABOVE_1TB_START;
983         }
984 
985         /*
986          * Advertise the HT region if address space covers the reserved
987          * region or if we relocate.
988          */
989         if (cpu->phys_bits >= 40) {
990             e820_add_entry(AMD_HT_START, AMD_HT_SIZE, E820_RESERVED);
991         }
992     }
993 
994     /*
995      * phys-bits is required to be appropriately configured
996      * to make sure max used GPA is reachable.
997      */
998     maxusedaddr = pc_max_used_gpa(pcms, pci_hole64_size);
999     maxphysaddr = ((hwaddr)1 << cpu->phys_bits) - 1;
1000     if (maxphysaddr < maxusedaddr) {
1001         error_report("Address space limit 0x%"PRIx64" < 0x%"PRIx64
1002                      " phys-bits too low (%u)",
1003                      maxphysaddr, maxusedaddr, cpu->phys_bits);
1004         exit(EXIT_FAILURE);
1005     }
1006 
1007     /*
1008      * Split single memory region and use aliases to address portions of it,
1009      * done for backwards compatibility with older qemus.
1010      */
1011     *ram_memory = machine->ram;
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     *rtc_state = ISA_DEVICE(mc146818_rtc_init(isa_bus, 2000, rtc_irq));
1322 
1323 #ifdef CONFIG_XEN_EMU
1324     if (xen_mode == XEN_EMULATE) {
1325         xen_evtchn_connect_gsis(gsi);
1326         if (pcms->bus) {
1327             pci_create_simple(pcms->bus, -1, "xen-platform");
1328         }
1329         xen_bus_init();
1330         xen_be_init();
1331     }
1332 #endif
1333 
1334     qemu_register_boot_set(pc_boot_set, *rtc_state);
1335 
1336     if (!xen_enabled() &&
1337         (x86ms->pit == ON_OFF_AUTO_AUTO || x86ms->pit == ON_OFF_AUTO_ON)) {
1338         if (kvm_pit_in_kernel()) {
1339             pit = kvm_pit_init(isa_bus, 0x40);
1340         } else {
1341             pit = i8254_pit_init(isa_bus, 0x40, pit_isa_irq, pit_alt_irq);
1342         }
1343         if (hpet) {
1344             /* connect PIT to output control line of the HPET */
1345             qdev_connect_gpio_out(hpet, 0, qdev_get_gpio_in(DEVICE(pit), 0));
1346         }
1347         pcspk_init(pcms->pcspk, isa_bus, pit);
1348     }
1349 
1350     /* Super I/O */
1351     pc_superio_init(isa_bus, create_fdctrl, pcms->i8042_enabled,
1352                     pcms->vmport != ON_OFF_AUTO_ON);
1353 }
1354 
1355 void pc_nic_init(PCMachineClass *pcmc, ISABus *isa_bus, PCIBus *pci_bus)
1356 {
1357     int i;
1358 
1359     rom_set_order_override(FW_CFG_ORDER_OVERRIDE_NIC);
1360     for (i = 0; i < nb_nics; i++) {
1361         NICInfo *nd = &nd_table[i];
1362         const char *model = nd->model ? nd->model : pcmc->default_nic_model;
1363 
1364         if (g_str_equal(model, "ne2k_isa")) {
1365             pc_init_ne2k_isa(isa_bus, nd);
1366         } else {
1367             pci_nic_init_nofail(nd, pci_bus, model, NULL);
1368         }
1369     }
1370     rom_reset_order_override();
1371 }
1372 
1373 void pc_i8259_create(ISABus *isa_bus, qemu_irq *i8259_irqs)
1374 {
1375     qemu_irq *i8259;
1376 
1377     if (kvm_pic_in_kernel()) {
1378         i8259 = kvm_i8259_init(isa_bus);
1379     } else if (xen_enabled()) {
1380         i8259 = xen_interrupt_controller_init();
1381     } else {
1382         i8259 = i8259_init(isa_bus, x86_allocate_cpu_irq());
1383     }
1384 
1385     for (size_t i = 0; i < ISA_NUM_IRQS; i++) {
1386         i8259_irqs[i] = i8259[i];
1387     }
1388 
1389     g_free(i8259);
1390 }
1391 
1392 static void pc_memory_pre_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
1393                                Error **errp)
1394 {
1395     const PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1396     const X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
1397     const PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1398     const MachineState *ms = MACHINE(hotplug_dev);
1399     const bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
1400     const uint64_t legacy_align = TARGET_PAGE_SIZE;
1401     Error *local_err = NULL;
1402 
1403     /*
1404      * When -no-acpi is used with Q35 machine type, no ACPI is built,
1405      * but pcms->acpi_dev is still created. Check !acpi_enabled in
1406      * addition to cover this case.
1407      */
1408     if (!x86ms->acpi_dev || !x86_machine_is_acpi_enabled(x86ms)) {
1409         error_setg(errp,
1410                    "memory hotplug is not enabled: missing acpi device or acpi disabled");
1411         return;
1412     }
1413 
1414     if (is_nvdimm && !ms->nvdimms_state->is_enabled) {
1415         error_setg(errp, "nvdimm is not enabled: missing 'nvdimm' in '-M'");
1416         return;
1417     }
1418 
1419     hotplug_handler_pre_plug(x86ms->acpi_dev, dev, &local_err);
1420     if (local_err) {
1421         error_propagate(errp, local_err);
1422         return;
1423     }
1424 
1425     pc_dimm_pre_plug(PC_DIMM(dev), MACHINE(hotplug_dev),
1426                      pcmc->enforce_aligned_dimm ? NULL : &legacy_align, errp);
1427 }
1428 
1429 static void pc_memory_plug(HotplugHandler *hotplug_dev,
1430                            DeviceState *dev, Error **errp)
1431 {
1432     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1433     X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
1434     MachineState *ms = MACHINE(hotplug_dev);
1435     bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
1436 
1437     pc_dimm_plug(PC_DIMM(dev), MACHINE(pcms));
1438 
1439     if (is_nvdimm) {
1440         nvdimm_plug(ms->nvdimms_state);
1441     }
1442 
1443     hotplug_handler_plug(x86ms->acpi_dev, dev, &error_abort);
1444 }
1445 
1446 static void pc_memory_unplug_request(HotplugHandler *hotplug_dev,
1447                                      DeviceState *dev, Error **errp)
1448 {
1449     X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
1450 
1451     /*
1452      * When -no-acpi is used with Q35 machine type, no ACPI is built,
1453      * but pcms->acpi_dev is still created. Check !acpi_enabled in
1454      * addition to cover this case.
1455      */
1456     if (!x86ms->acpi_dev || !x86_machine_is_acpi_enabled(x86ms)) {
1457         error_setg(errp,
1458                    "memory hotplug is not enabled: missing acpi device or acpi disabled");
1459         return;
1460     }
1461 
1462     if (object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM)) {
1463         error_setg(errp, "nvdimm device hot unplug is not supported yet.");
1464         return;
1465     }
1466 
1467     hotplug_handler_unplug_request(x86ms->acpi_dev, dev,
1468                                    errp);
1469 }
1470 
1471 static void pc_memory_unplug(HotplugHandler *hotplug_dev,
1472                              DeviceState *dev, Error **errp)
1473 {
1474     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1475     X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
1476     Error *local_err = NULL;
1477 
1478     hotplug_handler_unplug(x86ms->acpi_dev, dev, &local_err);
1479     if (local_err) {
1480         goto out;
1481     }
1482 
1483     pc_dimm_unplug(PC_DIMM(dev), MACHINE(pcms));
1484     qdev_unrealize(dev);
1485  out:
1486     error_propagate(errp, local_err);
1487 }
1488 
1489 static void pc_virtio_md_pci_pre_plug(HotplugHandler *hotplug_dev,
1490                                       DeviceState *dev, Error **errp)
1491 {
1492     HotplugHandler *hotplug_dev2 = qdev_get_bus_hotplug_handler(dev);
1493     Error *local_err = NULL;
1494 
1495     if (!hotplug_dev2 && dev->hotplugged) {
1496         /*
1497          * Without a bus hotplug handler, we cannot control the plug/unplug
1498          * order. We should never reach this point when hotplugging on x86,
1499          * however, better add a safety net.
1500          */
1501         error_setg(errp, "hotplug of virtio based memory devices not supported"
1502                    " on this bus.");
1503         return;
1504     }
1505     /*
1506      * First, see if we can plug this memory device at all. If that
1507      * succeeds, branch of to the actual hotplug handler.
1508      */
1509     memory_device_pre_plug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev), NULL,
1510                            &local_err);
1511     if (!local_err && hotplug_dev2) {
1512         hotplug_handler_pre_plug(hotplug_dev2, dev, &local_err);
1513     }
1514     error_propagate(errp, local_err);
1515 }
1516 
1517 static void pc_virtio_md_pci_plug(HotplugHandler *hotplug_dev,
1518                                   DeviceState *dev, Error **errp)
1519 {
1520     HotplugHandler *hotplug_dev2 = qdev_get_bus_hotplug_handler(dev);
1521     Error *local_err = NULL;
1522 
1523     /*
1524      * Plug the memory device first and then branch off to the actual
1525      * hotplug handler. If that one fails, we can easily undo the memory
1526      * device bits.
1527      */
1528     memory_device_plug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev));
1529     if (hotplug_dev2) {
1530         hotplug_handler_plug(hotplug_dev2, dev, &local_err);
1531         if (local_err) {
1532             memory_device_unplug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev));
1533         }
1534     }
1535     error_propagate(errp, local_err);
1536 }
1537 
1538 static void pc_virtio_md_pci_unplug_request(HotplugHandler *hotplug_dev,
1539                                             DeviceState *dev, Error **errp)
1540 {
1541     /* We don't support hot unplug of virtio based memory devices */
1542     error_setg(errp, "virtio based memory devices cannot be unplugged.");
1543 }
1544 
1545 static void pc_virtio_md_pci_unplug(HotplugHandler *hotplug_dev,
1546                                     DeviceState *dev, Error **errp)
1547 {
1548     /* We don't support hot unplug of virtio based memory devices */
1549 }
1550 
1551 static void pc_machine_device_pre_plug_cb(HotplugHandler *hotplug_dev,
1552                                           DeviceState *dev, Error **errp)
1553 {
1554     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1555         pc_memory_pre_plug(hotplug_dev, dev, errp);
1556     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1557         x86_cpu_pre_plug(hotplug_dev, dev, errp);
1558     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI) ||
1559                object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI)) {
1560         pc_virtio_md_pci_pre_plug(hotplug_dev, dev, errp);
1561     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
1562         /* Declare the APIC range as the reserved MSI region */
1563         char *resv_prop_str = g_strdup_printf("0xfee00000:0xfeefffff:%d",
1564                                               VIRTIO_IOMMU_RESV_MEM_T_MSI);
1565 
1566         object_property_set_uint(OBJECT(dev), "len-reserved-regions", 1, errp);
1567         object_property_set_str(OBJECT(dev), "reserved-regions[0]",
1568                                 resv_prop_str, errp);
1569         g_free(resv_prop_str);
1570     }
1571 
1572     if (object_dynamic_cast(OBJECT(dev), TYPE_X86_IOMMU_DEVICE) ||
1573         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
1574         PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1575 
1576         if (pcms->iommu) {
1577             error_setg(errp, "QEMU does not support multiple vIOMMUs "
1578                        "for x86 yet.");
1579             return;
1580         }
1581         pcms->iommu = dev;
1582     }
1583 }
1584 
1585 static void pc_machine_device_plug_cb(HotplugHandler *hotplug_dev,
1586                                       DeviceState *dev, Error **errp)
1587 {
1588     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1589         pc_memory_plug(hotplug_dev, dev, errp);
1590     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1591         x86_cpu_plug(hotplug_dev, dev, errp);
1592     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI) ||
1593                object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI)) {
1594         pc_virtio_md_pci_plug(hotplug_dev, dev, errp);
1595     }
1596 }
1597 
1598 static void pc_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
1599                                                 DeviceState *dev, Error **errp)
1600 {
1601     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1602         pc_memory_unplug_request(hotplug_dev, dev, errp);
1603     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1604         x86_cpu_unplug_request_cb(hotplug_dev, dev, errp);
1605     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI) ||
1606                object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI)) {
1607         pc_virtio_md_pci_unplug_request(hotplug_dev, dev, errp);
1608     } else {
1609         error_setg(errp, "acpi: device unplug request for not supported device"
1610                    " type: %s", object_get_typename(OBJECT(dev)));
1611     }
1612 }
1613 
1614 static void pc_machine_device_unplug_cb(HotplugHandler *hotplug_dev,
1615                                         DeviceState *dev, Error **errp)
1616 {
1617     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1618         pc_memory_unplug(hotplug_dev, dev, errp);
1619     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1620         x86_cpu_unplug_cb(hotplug_dev, dev, errp);
1621     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI) ||
1622                object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI)) {
1623         pc_virtio_md_pci_unplug(hotplug_dev, dev, errp);
1624     } else {
1625         error_setg(errp, "acpi: device unplug for not supported device"
1626                    " type: %s", object_get_typename(OBJECT(dev)));
1627     }
1628 }
1629 
1630 static HotplugHandler *pc_get_hotplug_handler(MachineState *machine,
1631                                              DeviceState *dev)
1632 {
1633     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) ||
1634         object_dynamic_cast(OBJECT(dev), TYPE_CPU) ||
1635         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI) ||
1636         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI) ||
1637         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI) ||
1638         object_dynamic_cast(OBJECT(dev), TYPE_X86_IOMMU_DEVICE)) {
1639         return HOTPLUG_HANDLER(machine);
1640     }
1641 
1642     return NULL;
1643 }
1644 
1645 static void
1646 pc_machine_get_device_memory_region_size(Object *obj, Visitor *v,
1647                                          const char *name, void *opaque,
1648                                          Error **errp)
1649 {
1650     MachineState *ms = MACHINE(obj);
1651     int64_t value = 0;
1652 
1653     if (ms->device_memory) {
1654         value = memory_region_size(&ms->device_memory->mr);
1655     }
1656 
1657     visit_type_int(v, name, &value, errp);
1658 }
1659 
1660 static void pc_machine_get_vmport(Object *obj, Visitor *v, const char *name,
1661                                   void *opaque, Error **errp)
1662 {
1663     PCMachineState *pcms = PC_MACHINE(obj);
1664     OnOffAuto vmport = pcms->vmport;
1665 
1666     visit_type_OnOffAuto(v, name, &vmport, errp);
1667 }
1668 
1669 static void pc_machine_set_vmport(Object *obj, Visitor *v, const char *name,
1670                                   void *opaque, Error **errp)
1671 {
1672     PCMachineState *pcms = PC_MACHINE(obj);
1673 
1674     visit_type_OnOffAuto(v, name, &pcms->vmport, errp);
1675 }
1676 
1677 static bool pc_machine_get_smbus(Object *obj, Error **errp)
1678 {
1679     PCMachineState *pcms = PC_MACHINE(obj);
1680 
1681     return pcms->smbus_enabled;
1682 }
1683 
1684 static void pc_machine_set_smbus(Object *obj, bool value, Error **errp)
1685 {
1686     PCMachineState *pcms = PC_MACHINE(obj);
1687 
1688     pcms->smbus_enabled = value;
1689 }
1690 
1691 static bool pc_machine_get_sata(Object *obj, Error **errp)
1692 {
1693     PCMachineState *pcms = PC_MACHINE(obj);
1694 
1695     return pcms->sata_enabled;
1696 }
1697 
1698 static void pc_machine_set_sata(Object *obj, bool value, Error **errp)
1699 {
1700     PCMachineState *pcms = PC_MACHINE(obj);
1701 
1702     pcms->sata_enabled = value;
1703 }
1704 
1705 static bool pc_machine_get_hpet(Object *obj, Error **errp)
1706 {
1707     PCMachineState *pcms = PC_MACHINE(obj);
1708 
1709     return pcms->hpet_enabled;
1710 }
1711 
1712 static void pc_machine_set_hpet(Object *obj, bool value, Error **errp)
1713 {
1714     PCMachineState *pcms = PC_MACHINE(obj);
1715 
1716     pcms->hpet_enabled = value;
1717 }
1718 
1719 static bool pc_machine_get_i8042(Object *obj, Error **errp)
1720 {
1721     PCMachineState *pcms = PC_MACHINE(obj);
1722 
1723     return pcms->i8042_enabled;
1724 }
1725 
1726 static void pc_machine_set_i8042(Object *obj, bool value, Error **errp)
1727 {
1728     PCMachineState *pcms = PC_MACHINE(obj);
1729 
1730     pcms->i8042_enabled = value;
1731 }
1732 
1733 static bool pc_machine_get_default_bus_bypass_iommu(Object *obj, Error **errp)
1734 {
1735     PCMachineState *pcms = PC_MACHINE(obj);
1736 
1737     return pcms->default_bus_bypass_iommu;
1738 }
1739 
1740 static void pc_machine_set_default_bus_bypass_iommu(Object *obj, bool value,
1741                                                     Error **errp)
1742 {
1743     PCMachineState *pcms = PC_MACHINE(obj);
1744 
1745     pcms->default_bus_bypass_iommu = value;
1746 }
1747 
1748 static void pc_machine_get_smbios_ep(Object *obj, Visitor *v, const char *name,
1749                                      void *opaque, Error **errp)
1750 {
1751     PCMachineState *pcms = PC_MACHINE(obj);
1752     SmbiosEntryPointType smbios_entry_point_type = pcms->smbios_entry_point_type;
1753 
1754     visit_type_SmbiosEntryPointType(v, name, &smbios_entry_point_type, errp);
1755 }
1756 
1757 static void pc_machine_set_smbios_ep(Object *obj, Visitor *v, const char *name,
1758                                      void *opaque, Error **errp)
1759 {
1760     PCMachineState *pcms = PC_MACHINE(obj);
1761 
1762     visit_type_SmbiosEntryPointType(v, name, &pcms->smbios_entry_point_type, errp);
1763 }
1764 
1765 static void pc_machine_get_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 = pcms->max_ram_below_4g;
1771 
1772     visit_type_size(v, name, &value, errp);
1773 }
1774 
1775 static void pc_machine_set_max_ram_below_4g(Object *obj, Visitor *v,
1776                                             const char *name, void *opaque,
1777                                             Error **errp)
1778 {
1779     PCMachineState *pcms = PC_MACHINE(obj);
1780     uint64_t value;
1781 
1782     if (!visit_type_size(v, name, &value, errp)) {
1783         return;
1784     }
1785     if (value > 4 * GiB) {
1786         error_setg(errp,
1787                    "Machine option 'max-ram-below-4g=%"PRIu64
1788                    "' expects size less than or equal to 4G", value);
1789         return;
1790     }
1791 
1792     if (value < 1 * MiB) {
1793         warn_report("Only %" PRIu64 " bytes of RAM below the 4GiB boundary,"
1794                     "BIOS may not work with less than 1MiB", value);
1795     }
1796 
1797     pcms->max_ram_below_4g = value;
1798 }
1799 
1800 static void pc_machine_get_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 = pcms->max_fw_size;
1806 
1807     visit_type_size(v, name, &value, errp);
1808 }
1809 
1810 static void pc_machine_set_max_fw_size(Object *obj, Visitor *v,
1811                                        const char *name, void *opaque,
1812                                        Error **errp)
1813 {
1814     PCMachineState *pcms = PC_MACHINE(obj);
1815     uint64_t value;
1816 
1817     if (!visit_type_size(v, name, &value, errp)) {
1818         return;
1819     }
1820 
1821     /*
1822     * We don't have a theoretically justifiable exact lower bound on the base
1823     * address of any flash mapping. In practice, the IO-APIC MMIO range is
1824     * [0xFEE00000..0xFEE01000] -- see IO_APIC_DEFAULT_ADDRESS --, leaving free
1825     * only 18MB-4KB below 4G. For now, restrict the cumulative mapping to 8MB in
1826     * size.
1827     */
1828     if (value > 16 * MiB) {
1829         error_setg(errp,
1830                    "User specified max allowed firmware size %" PRIu64 " is "
1831                    "greater than 16MiB. If combined firwmare size exceeds "
1832                    "16MiB the system may not boot, or experience intermittent"
1833                    "stability issues.",
1834                    value);
1835         return;
1836     }
1837 
1838     pcms->max_fw_size = value;
1839 }
1840 
1841 
1842 static void pc_machine_initfn(Object *obj)
1843 {
1844     PCMachineState *pcms = PC_MACHINE(obj);
1845 
1846 #ifdef CONFIG_VMPORT
1847     pcms->vmport = ON_OFF_AUTO_AUTO;
1848 #else
1849     pcms->vmport = ON_OFF_AUTO_OFF;
1850 #endif /* CONFIG_VMPORT */
1851     pcms->max_ram_below_4g = 0; /* use default */
1852     pcms->smbios_entry_point_type = SMBIOS_ENTRY_POINT_TYPE_32;
1853 
1854     /* acpi build is enabled by default if machine supports it */
1855     pcms->acpi_build_enabled = PC_MACHINE_GET_CLASS(pcms)->has_acpi_build;
1856     pcms->smbus_enabled = true;
1857     pcms->sata_enabled = true;
1858     pcms->i8042_enabled = true;
1859     pcms->max_fw_size = 8 * MiB;
1860 #ifdef CONFIG_HPET
1861     pcms->hpet_enabled = true;
1862 #endif
1863     pcms->default_bus_bypass_iommu = false;
1864 
1865     pc_system_flash_create(pcms);
1866     pcms->pcspk = isa_new(TYPE_PC_SPEAKER);
1867     object_property_add_alias(OBJECT(pcms), "pcspk-audiodev",
1868                               OBJECT(pcms->pcspk), "audiodev");
1869     cxl_machine_init(obj, &pcms->cxl_devices_state);
1870 }
1871 
1872 int pc_machine_kvm_type(MachineState *machine, const char *kvm_type)
1873 {
1874 #ifdef CONFIG_XEN_EMU
1875     if (xen_mode == XEN_EMULATE) {
1876         xen_overlay_create();
1877         xen_evtchn_create();
1878         xen_gnttab_create();
1879         xen_xenstore_create();
1880     }
1881 #endif
1882     return 0;
1883 }
1884 
1885 static void pc_machine_reset(MachineState *machine, ShutdownCause reason)
1886 {
1887     CPUState *cs;
1888     X86CPU *cpu;
1889 
1890     qemu_devices_reset(reason);
1891 
1892     /* Reset APIC after devices have been reset to cancel
1893      * any changes that qemu_devices_reset() might have done.
1894      */
1895     CPU_FOREACH(cs) {
1896         cpu = X86_CPU(cs);
1897 
1898         x86_cpu_after_reset(cpu);
1899     }
1900 }
1901 
1902 static void pc_machine_wakeup(MachineState *machine)
1903 {
1904     cpu_synchronize_all_states();
1905     pc_machine_reset(machine, SHUTDOWN_CAUSE_NONE);
1906     cpu_synchronize_all_post_reset();
1907 }
1908 
1909 static bool pc_hotplug_allowed(MachineState *ms, DeviceState *dev, Error **errp)
1910 {
1911     X86IOMMUState *iommu = x86_iommu_get_default();
1912     IntelIOMMUState *intel_iommu;
1913 
1914     if (iommu &&
1915         object_dynamic_cast((Object *)iommu, TYPE_INTEL_IOMMU_DEVICE) &&
1916         object_dynamic_cast((Object *)dev, "vfio-pci")) {
1917         intel_iommu = INTEL_IOMMU_DEVICE(iommu);
1918         if (!intel_iommu->caching_mode) {
1919             error_setg(errp, "Device assignment is not allowed without "
1920                        "enabling caching-mode=on for Intel IOMMU.");
1921             return false;
1922         }
1923     }
1924 
1925     return true;
1926 }
1927 
1928 static void pc_machine_class_init(ObjectClass *oc, void *data)
1929 {
1930     MachineClass *mc = MACHINE_CLASS(oc);
1931     PCMachineClass *pcmc = PC_MACHINE_CLASS(oc);
1932     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
1933 
1934     pcmc->pci_enabled = true;
1935     pcmc->has_acpi_build = true;
1936     pcmc->rsdp_in_ram = true;
1937     pcmc->smbios_defaults = true;
1938     pcmc->smbios_uuid_encoded = true;
1939     pcmc->gigabyte_align = true;
1940     pcmc->has_reserved_memory = true;
1941     pcmc->kvmclock_enabled = true;
1942     pcmc->enforce_aligned_dimm = true;
1943     pcmc->enforce_amd_1tb_hole = true;
1944     /* BIOS ACPI tables: 128K. Other BIOS datastructures: less than 4K reported
1945      * to be used at the moment, 32K should be enough for a while.  */
1946     pcmc->acpi_data_size = 0x20000 + 0x8000;
1947     pcmc->pvh_enabled = true;
1948     pcmc->kvmclock_create_always = true;
1949     pcmc->resizable_acpi_blob = true;
1950     assert(!mc->get_hotplug_handler);
1951     mc->get_hotplug_handler = pc_get_hotplug_handler;
1952     mc->hotplug_allowed = pc_hotplug_allowed;
1953     mc->cpu_index_to_instance_props = x86_cpu_index_to_props;
1954     mc->get_default_cpu_node_id = x86_get_default_cpu_node_id;
1955     mc->possible_cpu_arch_ids = x86_possible_cpu_arch_ids;
1956     mc->auto_enable_numa_with_memhp = true;
1957     mc->auto_enable_numa_with_memdev = true;
1958     mc->has_hotpluggable_cpus = true;
1959     mc->default_boot_order = "cad";
1960     mc->block_default_type = IF_IDE;
1961     mc->max_cpus = 255;
1962     mc->reset = pc_machine_reset;
1963     mc->wakeup = pc_machine_wakeup;
1964     hc->pre_plug = pc_machine_device_pre_plug_cb;
1965     hc->plug = pc_machine_device_plug_cb;
1966     hc->unplug_request = pc_machine_device_unplug_request_cb;
1967     hc->unplug = pc_machine_device_unplug_cb;
1968     mc->default_cpu_type = TARGET_DEFAULT_CPU_TYPE;
1969     mc->nvdimm_supported = true;
1970     mc->smp_props.dies_supported = true;
1971     mc->default_ram_id = "pc.ram";
1972 
1973     object_class_property_add(oc, PC_MACHINE_MAX_RAM_BELOW_4G, "size",
1974         pc_machine_get_max_ram_below_4g, pc_machine_set_max_ram_below_4g,
1975         NULL, NULL);
1976     object_class_property_set_description(oc, PC_MACHINE_MAX_RAM_BELOW_4G,
1977         "Maximum ram below the 4G boundary (32bit boundary)");
1978 
1979     object_class_property_add(oc, PC_MACHINE_DEVMEM_REGION_SIZE, "int",
1980         pc_machine_get_device_memory_region_size, NULL,
1981         NULL, NULL);
1982 
1983     object_class_property_add(oc, PC_MACHINE_VMPORT, "OnOffAuto",
1984         pc_machine_get_vmport, pc_machine_set_vmport,
1985         NULL, NULL);
1986     object_class_property_set_description(oc, PC_MACHINE_VMPORT,
1987         "Enable vmport (pc & q35)");
1988 
1989     object_class_property_add_bool(oc, PC_MACHINE_SMBUS,
1990         pc_machine_get_smbus, pc_machine_set_smbus);
1991     object_class_property_set_description(oc, PC_MACHINE_SMBUS,
1992         "Enable/disable system management bus");
1993 
1994     object_class_property_add_bool(oc, PC_MACHINE_SATA,
1995         pc_machine_get_sata, pc_machine_set_sata);
1996     object_class_property_set_description(oc, PC_MACHINE_SATA,
1997         "Enable/disable Serial ATA bus");
1998 
1999     object_class_property_add_bool(oc, "hpet",
2000         pc_machine_get_hpet, pc_machine_set_hpet);
2001     object_class_property_set_description(oc, "hpet",
2002         "Enable/disable high precision event timer emulation");
2003 
2004     object_class_property_add_bool(oc, PC_MACHINE_I8042,
2005         pc_machine_get_i8042, pc_machine_set_i8042);
2006 
2007     object_class_property_add_bool(oc, "default-bus-bypass-iommu",
2008         pc_machine_get_default_bus_bypass_iommu,
2009         pc_machine_set_default_bus_bypass_iommu);
2010 
2011     object_class_property_add(oc, PC_MACHINE_MAX_FW_SIZE, "size",
2012         pc_machine_get_max_fw_size, pc_machine_set_max_fw_size,
2013         NULL, NULL);
2014     object_class_property_set_description(oc, PC_MACHINE_MAX_FW_SIZE,
2015         "Maximum combined firmware size");
2016 
2017     object_class_property_add(oc, PC_MACHINE_SMBIOS_EP, "str",
2018         pc_machine_get_smbios_ep, pc_machine_set_smbios_ep,
2019         NULL, NULL);
2020     object_class_property_set_description(oc, PC_MACHINE_SMBIOS_EP,
2021         "SMBIOS Entry Point type [32, 64]");
2022 }
2023 
2024 static const TypeInfo pc_machine_info = {
2025     .name = TYPE_PC_MACHINE,
2026     .parent = TYPE_X86_MACHINE,
2027     .abstract = true,
2028     .instance_size = sizeof(PCMachineState),
2029     .instance_init = pc_machine_initfn,
2030     .class_size = sizeof(PCMachineClass),
2031     .class_init = pc_machine_class_init,
2032     .interfaces = (InterfaceInfo[]) {
2033          { TYPE_HOTPLUG_HANDLER },
2034          { }
2035     },
2036 };
2037 
2038 static void pc_machine_register_types(void)
2039 {
2040     type_register_static(&pc_machine_info);
2041 }
2042 
2043 type_init(pc_machine_register_types)
2044