xref: /openbmc/qemu/hw/i386/pc.c (revision 73ec0ead)
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/mem/memory-device.h"
94 #include "sysemu/replay.h"
95 #include "target/i386/cpu.h"
96 #include "e820_memory_layout.h"
97 #include "fw_cfg.h"
98 #include "trace.h"
99 #include CONFIG_DEVICES
100 
101 /*
102  * Helper for setting model-id for CPU models that changed model-id
103  * depending on QEMU versions up to QEMU 2.4.
104  */
105 #define PC_CPU_MODEL_IDS(v) \
106     { "qemu32-" TYPE_X86_CPU, "model-id", "QEMU Virtual CPU version " v, },\
107     { "qemu64-" TYPE_X86_CPU, "model-id", "QEMU Virtual CPU version " v, },\
108     { "athlon-" TYPE_X86_CPU, "model-id", "QEMU Virtual CPU version " v, },
109 
110 GlobalProperty pc_compat_7_2[] = {
111     { "ICH9-LPC", "noreboot", "true" },
112 };
113 const size_t pc_compat_7_2_len = G_N_ELEMENTS(pc_compat_7_2);
114 
115 GlobalProperty pc_compat_7_1[] = {};
116 const size_t pc_compat_7_1_len = G_N_ELEMENTS(pc_compat_7_1);
117 
118 GlobalProperty pc_compat_7_0[] = {};
119 const size_t pc_compat_7_0_len = G_N_ELEMENTS(pc_compat_7_0);
120 
121 GlobalProperty pc_compat_6_2[] = {
122     { "virtio-mem", "unplugged-inaccessible", "off" },
123 };
124 const size_t pc_compat_6_2_len = G_N_ELEMENTS(pc_compat_6_2);
125 
126 GlobalProperty pc_compat_6_1[] = {
127     { TYPE_X86_CPU, "hv-version-id-build", "0x1bbc" },
128     { TYPE_X86_CPU, "hv-version-id-major", "0x0006" },
129     { TYPE_X86_CPU, "hv-version-id-minor", "0x0001" },
130     { "ICH9-LPC", "x-keep-pci-slot-hpc", "false" },
131 };
132 const size_t pc_compat_6_1_len = G_N_ELEMENTS(pc_compat_6_1);
133 
134 GlobalProperty pc_compat_6_0[] = {
135     { "qemu64" "-" TYPE_X86_CPU, "family", "6" },
136     { "qemu64" "-" TYPE_X86_CPU, "model", "6" },
137     { "qemu64" "-" TYPE_X86_CPU, "stepping", "3" },
138     { TYPE_X86_CPU, "x-vendor-cpuid-only", "off" },
139     { "ICH9-LPC", ACPI_PM_PROP_ACPI_PCIHP_BRIDGE, "off" },
140     { "ICH9-LPC", "x-keep-pci-slot-hpc", "true" },
141 };
142 const size_t pc_compat_6_0_len = G_N_ELEMENTS(pc_compat_6_0);
143 
144 GlobalProperty pc_compat_5_2[] = {
145     { "ICH9-LPC", "x-smi-cpu-hotunplug", "off" },
146 };
147 const size_t pc_compat_5_2_len = G_N_ELEMENTS(pc_compat_5_2);
148 
149 GlobalProperty pc_compat_5_1[] = {
150     { "ICH9-LPC", "x-smi-cpu-hotplug", "off" },
151     { TYPE_X86_CPU, "kvm-msi-ext-dest-id", "off" },
152 };
153 const size_t pc_compat_5_1_len = G_N_ELEMENTS(pc_compat_5_1);
154 
155 GlobalProperty pc_compat_5_0[] = {
156 };
157 const size_t pc_compat_5_0_len = G_N_ELEMENTS(pc_compat_5_0);
158 
159 GlobalProperty pc_compat_4_2[] = {
160     { "mch", "smbase-smram", "off" },
161 };
162 const size_t pc_compat_4_2_len = G_N_ELEMENTS(pc_compat_4_2);
163 
164 GlobalProperty pc_compat_4_1[] = {};
165 const size_t pc_compat_4_1_len = G_N_ELEMENTS(pc_compat_4_1);
166 
167 GlobalProperty pc_compat_4_0[] = {};
168 const size_t pc_compat_4_0_len = G_N_ELEMENTS(pc_compat_4_0);
169 
170 GlobalProperty pc_compat_3_1[] = {
171     { "intel-iommu", "dma-drain", "off" },
172     { "Opteron_G3" "-" TYPE_X86_CPU, "rdtscp", "off" },
173     { "Opteron_G4" "-" TYPE_X86_CPU, "rdtscp", "off" },
174     { "Opteron_G4" "-" TYPE_X86_CPU, "npt", "off" },
175     { "Opteron_G4" "-" TYPE_X86_CPU, "nrip-save", "off" },
176     { "Opteron_G5" "-" TYPE_X86_CPU, "rdtscp", "off" },
177     { "Opteron_G5" "-" TYPE_X86_CPU, "npt", "off" },
178     { "Opteron_G5" "-" TYPE_X86_CPU, "nrip-save", "off" },
179     { "EPYC" "-" TYPE_X86_CPU, "npt", "off" },
180     { "EPYC" "-" TYPE_X86_CPU, "nrip-save", "off" },
181     { "EPYC-IBPB" "-" TYPE_X86_CPU, "npt", "off" },
182     { "EPYC-IBPB" "-" TYPE_X86_CPU, "nrip-save", "off" },
183     { "Skylake-Client" "-" TYPE_X86_CPU,      "mpx", "on" },
184     { "Skylake-Client-IBRS" "-" TYPE_X86_CPU, "mpx", "on" },
185     { "Skylake-Server" "-" TYPE_X86_CPU,      "mpx", "on" },
186     { "Skylake-Server-IBRS" "-" TYPE_X86_CPU, "mpx", "on" },
187     { "Cascadelake-Server" "-" TYPE_X86_CPU,  "mpx", "on" },
188     { "Icelake-Client" "-" TYPE_X86_CPU,      "mpx", "on" },
189     { "Icelake-Server" "-" TYPE_X86_CPU,      "mpx", "on" },
190     { "Cascadelake-Server" "-" TYPE_X86_CPU, "stepping", "5" },
191     { TYPE_X86_CPU, "x-intel-pt-auto-level", "off" },
192 };
193 const size_t pc_compat_3_1_len = G_N_ELEMENTS(pc_compat_3_1);
194 
195 GlobalProperty pc_compat_3_0[] = {
196     { TYPE_X86_CPU, "x-hv-synic-kvm-only", "on" },
197     { "Skylake-Server" "-" TYPE_X86_CPU, "pku", "off" },
198     { "Skylake-Server-IBRS" "-" TYPE_X86_CPU, "pku", "off" },
199 };
200 const size_t pc_compat_3_0_len = G_N_ELEMENTS(pc_compat_3_0);
201 
202 GlobalProperty pc_compat_2_12[] = {
203     { TYPE_X86_CPU, "legacy-cache", "on" },
204     { TYPE_X86_CPU, "topoext", "off" },
205     { "EPYC-" TYPE_X86_CPU, "xlevel", "0x8000000a" },
206     { "EPYC-IBPB-" TYPE_X86_CPU, "xlevel", "0x8000000a" },
207 };
208 const size_t pc_compat_2_12_len = G_N_ELEMENTS(pc_compat_2_12);
209 
210 GlobalProperty pc_compat_2_11[] = {
211     { TYPE_X86_CPU, "x-migrate-smi-count", "off" },
212     { "Skylake-Server" "-" TYPE_X86_CPU, "clflushopt", "off" },
213 };
214 const size_t pc_compat_2_11_len = G_N_ELEMENTS(pc_compat_2_11);
215 
216 GlobalProperty pc_compat_2_10[] = {
217     { TYPE_X86_CPU, "x-hv-max-vps", "0x40" },
218     { "i440FX-pcihost", "x-pci-hole64-fix", "off" },
219     { "q35-pcihost", "x-pci-hole64-fix", "off" },
220 };
221 const size_t pc_compat_2_10_len = G_N_ELEMENTS(pc_compat_2_10);
222 
223 GlobalProperty pc_compat_2_9[] = {
224     { "mch", "extended-tseg-mbytes", "0" },
225 };
226 const size_t pc_compat_2_9_len = G_N_ELEMENTS(pc_compat_2_9);
227 
228 GlobalProperty pc_compat_2_8[] = {
229     { TYPE_X86_CPU, "tcg-cpuid", "off" },
230     { "kvmclock", "x-mach-use-reliable-get-clock", "off" },
231     { "ICH9-LPC", "x-smi-broadcast", "off" },
232     { TYPE_X86_CPU, "vmware-cpuid-freq", "off" },
233     { "Haswell-" TYPE_X86_CPU, "stepping", "1" },
234 };
235 const size_t pc_compat_2_8_len = G_N_ELEMENTS(pc_compat_2_8);
236 
237 GlobalProperty pc_compat_2_7[] = {
238     { TYPE_X86_CPU, "l3-cache", "off" },
239     { TYPE_X86_CPU, "full-cpuid-auto-level", "off" },
240     { "Opteron_G3" "-" TYPE_X86_CPU, "family", "15" },
241     { "Opteron_G3" "-" TYPE_X86_CPU, "model", "6" },
242     { "Opteron_G3" "-" TYPE_X86_CPU, "stepping", "1" },
243     { "isa-pcspk", "migrate", "off" },
244 };
245 const size_t pc_compat_2_7_len = G_N_ELEMENTS(pc_compat_2_7);
246 
247 GlobalProperty pc_compat_2_6[] = {
248     { TYPE_X86_CPU, "cpuid-0xb", "off" },
249     { "vmxnet3", "romfile", "" },
250     { TYPE_X86_CPU, "fill-mtrr-mask", "off" },
251     { "apic-common", "legacy-instance-id", "on", }
252 };
253 const size_t pc_compat_2_6_len = G_N_ELEMENTS(pc_compat_2_6);
254 
255 GlobalProperty pc_compat_2_5[] = {};
256 const size_t pc_compat_2_5_len = G_N_ELEMENTS(pc_compat_2_5);
257 
258 GlobalProperty pc_compat_2_4[] = {
259     PC_CPU_MODEL_IDS("2.4.0")
260     { "Haswell-" TYPE_X86_CPU, "abm", "off" },
261     { "Haswell-noTSX-" TYPE_X86_CPU, "abm", "off" },
262     { "Broadwell-" TYPE_X86_CPU, "abm", "off" },
263     { "Broadwell-noTSX-" TYPE_X86_CPU, "abm", "off" },
264     { "host" "-" TYPE_X86_CPU, "host-cache-info", "on" },
265     { TYPE_X86_CPU, "check", "off" },
266     { "qemu64" "-" TYPE_X86_CPU, "sse4a", "on" },
267     { "qemu64" "-" TYPE_X86_CPU, "abm", "on" },
268     { "qemu64" "-" TYPE_X86_CPU, "popcnt", "on" },
269     { "qemu32" "-" TYPE_X86_CPU, "popcnt", "on" },
270     { "Opteron_G2" "-" TYPE_X86_CPU, "rdtscp", "on" },
271     { "Opteron_G3" "-" TYPE_X86_CPU, "rdtscp", "on" },
272     { "Opteron_G4" "-" TYPE_X86_CPU, "rdtscp", "on" },
273     { "Opteron_G5" "-" TYPE_X86_CPU, "rdtscp", "on", }
274 };
275 const size_t pc_compat_2_4_len = G_N_ELEMENTS(pc_compat_2_4);
276 
277 GlobalProperty pc_compat_2_3[] = {
278     PC_CPU_MODEL_IDS("2.3.0")
279     { TYPE_X86_CPU, "arat", "off" },
280     { "qemu64" "-" TYPE_X86_CPU, "min-level", "4" },
281     { "kvm64" "-" TYPE_X86_CPU, "min-level", "5" },
282     { "pentium3" "-" TYPE_X86_CPU, "min-level", "2" },
283     { "n270" "-" TYPE_X86_CPU, "min-level", "5" },
284     { "Conroe" "-" TYPE_X86_CPU, "min-level", "4" },
285     { "Penryn" "-" TYPE_X86_CPU, "min-level", "4" },
286     { "Nehalem" "-" TYPE_X86_CPU, "min-level", "4" },
287     { "n270" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
288     { "Penryn" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
289     { "Conroe" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
290     { "Nehalem" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
291     { "Westmere" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
292     { "SandyBridge" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
293     { "IvyBridge" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
294     { "Haswell" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
295     { "Haswell-noTSX" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
296     { "Broadwell" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
297     { "Broadwell-noTSX" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
298     { TYPE_X86_CPU, "kvm-no-smi-migration", "on" },
299 };
300 const size_t pc_compat_2_3_len = G_N_ELEMENTS(pc_compat_2_3);
301 
302 GlobalProperty pc_compat_2_2[] = {
303     PC_CPU_MODEL_IDS("2.2.0")
304     { "kvm64" "-" TYPE_X86_CPU, "vme", "off" },
305     { "kvm32" "-" TYPE_X86_CPU, "vme", "off" },
306     { "Conroe" "-" TYPE_X86_CPU, "vme", "off" },
307     { "Penryn" "-" TYPE_X86_CPU, "vme", "off" },
308     { "Nehalem" "-" TYPE_X86_CPU, "vme", "off" },
309     { "Westmere" "-" TYPE_X86_CPU, "vme", "off" },
310     { "SandyBridge" "-" TYPE_X86_CPU, "vme", "off" },
311     { "Haswell" "-" TYPE_X86_CPU, "vme", "off" },
312     { "Broadwell" "-" TYPE_X86_CPU, "vme", "off" },
313     { "Opteron_G1" "-" TYPE_X86_CPU, "vme", "off" },
314     { "Opteron_G2" "-" TYPE_X86_CPU, "vme", "off" },
315     { "Opteron_G3" "-" TYPE_X86_CPU, "vme", "off" },
316     { "Opteron_G4" "-" TYPE_X86_CPU, "vme", "off" },
317     { "Opteron_G5" "-" TYPE_X86_CPU, "vme", "off" },
318     { "Haswell" "-" TYPE_X86_CPU, "f16c", "off" },
319     { "Haswell" "-" TYPE_X86_CPU, "rdrand", "off" },
320     { "Broadwell" "-" TYPE_X86_CPU, "f16c", "off" },
321     { "Broadwell" "-" TYPE_X86_CPU, "rdrand", "off" },
322 };
323 const size_t pc_compat_2_2_len = G_N_ELEMENTS(pc_compat_2_2);
324 
325 GlobalProperty pc_compat_2_1[] = {
326     PC_CPU_MODEL_IDS("2.1.0")
327     { "coreduo" "-" TYPE_X86_CPU, "vmx", "on" },
328     { "core2duo" "-" TYPE_X86_CPU, "vmx", "on" },
329 };
330 const size_t pc_compat_2_1_len = G_N_ELEMENTS(pc_compat_2_1);
331 
332 GlobalProperty pc_compat_2_0[] = {
333     PC_CPU_MODEL_IDS("2.0.0")
334     { "virtio-scsi-pci", "any_layout", "off" },
335     { "PIIX4_PM", "memory-hotplug-support", "off" },
336     { "apic", "version", "0x11" },
337     { "nec-usb-xhci", "superspeed-ports-first", "off" },
338     { "nec-usb-xhci", "force-pcie-endcap", "on" },
339     { "pci-serial", "prog_if", "0" },
340     { "pci-serial-2x", "prog_if", "0" },
341     { "pci-serial-4x", "prog_if", "0" },
342     { "virtio-net-pci", "guest_announce", "off" },
343     { "ICH9-LPC", "memory-hotplug-support", "off" },
344 };
345 const size_t pc_compat_2_0_len = G_N_ELEMENTS(pc_compat_2_0);
346 
347 GlobalProperty pc_compat_1_7[] = {
348     PC_CPU_MODEL_IDS("1.7.0")
349     { TYPE_USB_DEVICE, "msos-desc", "no" },
350     { "PIIX4_PM", ACPI_PM_PROP_ACPI_PCIHP_BRIDGE, "off" },
351     { "hpet", HPET_INTCAP, "4" },
352 };
353 const size_t pc_compat_1_7_len = G_N_ELEMENTS(pc_compat_1_7);
354 
355 GlobalProperty pc_compat_1_6[] = {
356     PC_CPU_MODEL_IDS("1.6.0")
357     { "e1000", "mitigation", "off" },
358     { "qemu64-" TYPE_X86_CPU, "model", "2" },
359     { "qemu32-" TYPE_X86_CPU, "model", "3" },
360     { "i440FX-pcihost", "short_root_bus", "1" },
361     { "q35-pcihost", "short_root_bus", "1" },
362 };
363 const size_t pc_compat_1_6_len = G_N_ELEMENTS(pc_compat_1_6);
364 
365 GlobalProperty pc_compat_1_5[] = {
366     PC_CPU_MODEL_IDS("1.5.0")
367     { "Conroe-" TYPE_X86_CPU, "model", "2" },
368     { "Conroe-" TYPE_X86_CPU, "min-level", "2" },
369     { "Penryn-" TYPE_X86_CPU, "model", "2" },
370     { "Penryn-" TYPE_X86_CPU, "min-level", "2" },
371     { "Nehalem-" TYPE_X86_CPU, "model", "2" },
372     { "Nehalem-" TYPE_X86_CPU, "min-level", "2" },
373     { "virtio-net-pci", "any_layout", "off" },
374     { TYPE_X86_CPU, "pmu", "on" },
375     { "i440FX-pcihost", "short_root_bus", "0" },
376     { "q35-pcihost", "short_root_bus", "0" },
377 };
378 const size_t pc_compat_1_5_len = G_N_ELEMENTS(pc_compat_1_5);
379 
380 GlobalProperty pc_compat_1_4[] = {
381     PC_CPU_MODEL_IDS("1.4.0")
382     { "scsi-hd", "discard_granularity", "0" },
383     { "scsi-cd", "discard_granularity", "0" },
384     { "ide-hd", "discard_granularity", "0" },
385     { "ide-cd", "discard_granularity", "0" },
386     { "virtio-blk-pci", "discard_granularity", "0" },
387     /* DEV_NVECTORS_UNSPECIFIED as a uint32_t string: */
388     { "virtio-serial-pci", "vectors", "0xFFFFFFFF" },
389     { "virtio-net-pci", "ctrl_guest_offloads", "off" },
390     { "e1000", "romfile", "pxe-e1000.rom" },
391     { "ne2k_pci", "romfile", "pxe-ne2k_pci.rom" },
392     { "pcnet", "romfile", "pxe-pcnet.rom" },
393     { "rtl8139", "romfile", "pxe-rtl8139.rom" },
394     { "virtio-net-pci", "romfile", "pxe-virtio.rom" },
395     { "486-" TYPE_X86_CPU, "model", "0" },
396     { "n270" "-" TYPE_X86_CPU, "movbe", "off" },
397     { "Westmere" "-" TYPE_X86_CPU, "pclmulqdq", "off" },
398 };
399 const size_t pc_compat_1_4_len = G_N_ELEMENTS(pc_compat_1_4);
400 
401 GSIState *pc_gsi_create(qemu_irq **irqs, bool pci_enabled)
402 {
403     GSIState *s;
404 
405     s = g_new0(GSIState, 1);
406     if (kvm_ioapic_in_kernel()) {
407         kvm_pc_setup_irq_routing(pci_enabled);
408     }
409     *irqs = qemu_allocate_irqs(gsi_handler, s, IOAPIC_NUM_PINS);
410 
411     return s;
412 }
413 
414 static void ioport80_write(void *opaque, hwaddr addr, uint64_t data,
415                            unsigned size)
416 {
417 }
418 
419 static uint64_t ioport80_read(void *opaque, hwaddr addr, unsigned size)
420 {
421     return 0xffffffffffffffffULL;
422 }
423 
424 /* MSDOS compatibility mode FPU exception support */
425 static void ioportF0_write(void *opaque, hwaddr addr, uint64_t data,
426                            unsigned size)
427 {
428     if (tcg_enabled()) {
429         cpu_set_ignne();
430     }
431 }
432 
433 static uint64_t ioportF0_read(void *opaque, hwaddr addr, unsigned size)
434 {
435     return 0xffffffffffffffffULL;
436 }
437 
438 /* PC cmos mappings */
439 
440 #define REG_EQUIPMENT_BYTE          0x14
441 
442 static void cmos_init_hd(MC146818RtcState *s, int type_ofs, int info_ofs,
443                          int16_t cylinders, int8_t heads, int8_t sectors)
444 {
445     mc146818rtc_set_cmos_data(s, type_ofs, 47);
446     mc146818rtc_set_cmos_data(s, info_ofs, cylinders);
447     mc146818rtc_set_cmos_data(s, info_ofs + 1, cylinders >> 8);
448     mc146818rtc_set_cmos_data(s, info_ofs + 2, heads);
449     mc146818rtc_set_cmos_data(s, info_ofs + 3, 0xff);
450     mc146818rtc_set_cmos_data(s, info_ofs + 4, 0xff);
451     mc146818rtc_set_cmos_data(s, info_ofs + 5, 0xc0 | ((heads > 8) << 3));
452     mc146818rtc_set_cmos_data(s, info_ofs + 6, cylinders);
453     mc146818rtc_set_cmos_data(s, info_ofs + 7, cylinders >> 8);
454     mc146818rtc_set_cmos_data(s, info_ofs + 8, sectors);
455 }
456 
457 /* convert boot_device letter to something recognizable by the bios */
458 static int boot_device2nibble(char boot_device)
459 {
460     switch(boot_device) {
461     case 'a':
462     case 'b':
463         return 0x01; /* floppy boot */
464     case 'c':
465         return 0x02; /* hard drive boot */
466     case 'd':
467         return 0x03; /* CD-ROM boot */
468     case 'n':
469         return 0x04; /* Network boot */
470     }
471     return 0;
472 }
473 
474 static void set_boot_dev(MC146818RtcState *s, const char *boot_device,
475                          Error **errp)
476 {
477 #define PC_MAX_BOOT_DEVICES 3
478     int nbds, bds[3] = { 0, };
479     int i;
480 
481     nbds = strlen(boot_device);
482     if (nbds > PC_MAX_BOOT_DEVICES) {
483         error_setg(errp, "Too many boot devices for PC");
484         return;
485     }
486     for (i = 0; i < nbds; i++) {
487         bds[i] = boot_device2nibble(boot_device[i]);
488         if (bds[i] == 0) {
489             error_setg(errp, "Invalid boot device for PC: '%c'",
490                        boot_device[i]);
491             return;
492         }
493     }
494     mc146818rtc_set_cmos_data(s, 0x3d, (bds[1] << 4) | bds[0]);
495     mc146818rtc_set_cmos_data(s, 0x38, (bds[2] << 4) | (fd_bootchk ? 0x0 : 0x1));
496 }
497 
498 static void pc_boot_set(void *opaque, const char *boot_device, Error **errp)
499 {
500     set_boot_dev(opaque, boot_device, errp);
501 }
502 
503 static void pc_cmos_init_floppy(MC146818RtcState *rtc_state, ISADevice *floppy)
504 {
505     int val, nb, i;
506     FloppyDriveType fd_type[2] = { FLOPPY_DRIVE_TYPE_NONE,
507                                    FLOPPY_DRIVE_TYPE_NONE };
508 
509     /* floppy type */
510     if (floppy) {
511         for (i = 0; i < 2; i++) {
512             fd_type[i] = isa_fdc_get_drive_type(floppy, i);
513         }
514     }
515     val = (cmos_get_fd_drive_type(fd_type[0]) << 4) |
516         cmos_get_fd_drive_type(fd_type[1]);
517     mc146818rtc_set_cmos_data(rtc_state, 0x10, val);
518 
519     val = mc146818rtc_get_cmos_data(rtc_state, REG_EQUIPMENT_BYTE);
520     nb = 0;
521     if (fd_type[0] != FLOPPY_DRIVE_TYPE_NONE) {
522         nb++;
523     }
524     if (fd_type[1] != FLOPPY_DRIVE_TYPE_NONE) {
525         nb++;
526     }
527     switch (nb) {
528     case 0:
529         break;
530     case 1:
531         val |= 0x01; /* 1 drive, ready for boot */
532         break;
533     case 2:
534         val |= 0x41; /* 2 drives, ready for boot */
535         break;
536     }
537     mc146818rtc_set_cmos_data(rtc_state, REG_EQUIPMENT_BYTE, val);
538 }
539 
540 typedef struct pc_cmos_init_late_arg {
541     MC146818RtcState *rtc_state;
542     BusState *idebus[2];
543 } pc_cmos_init_late_arg;
544 
545 typedef struct check_fdc_state {
546     ISADevice *floppy;
547     bool multiple;
548 } CheckFdcState;
549 
550 static int check_fdc(Object *obj, void *opaque)
551 {
552     CheckFdcState *state = opaque;
553     Object *fdc;
554     uint32_t iobase;
555     Error *local_err = NULL;
556 
557     fdc = object_dynamic_cast(obj, TYPE_ISA_FDC);
558     if (!fdc) {
559         return 0;
560     }
561 
562     iobase = object_property_get_uint(obj, "iobase", &local_err);
563     if (local_err || iobase != 0x3f0) {
564         error_free(local_err);
565         return 0;
566     }
567 
568     if (state->floppy) {
569         state->multiple = true;
570     } else {
571         state->floppy = ISA_DEVICE(obj);
572     }
573     return 0;
574 }
575 
576 static const char * const fdc_container_path[] = {
577     "/unattached", "/peripheral", "/peripheral-anon"
578 };
579 
580 /*
581  * Locate the FDC at IO address 0x3f0, in order to configure the CMOS registers
582  * and ACPI objects.
583  */
584 static ISADevice *pc_find_fdc0(void)
585 {
586     int i;
587     Object *container;
588     CheckFdcState state = { 0 };
589 
590     for (i = 0; i < ARRAY_SIZE(fdc_container_path); i++) {
591         container = container_get(qdev_get_machine(), fdc_container_path[i]);
592         object_child_foreach(container, check_fdc, &state);
593     }
594 
595     if (state.multiple) {
596         warn_report("multiple floppy disk controllers with "
597                     "iobase=0x3f0 have been found");
598         error_printf("the one being picked for CMOS setup might not reflect "
599                      "your intent");
600     }
601 
602     return state.floppy;
603 }
604 
605 static void pc_cmos_init_late(void *opaque)
606 {
607     pc_cmos_init_late_arg *arg = opaque;
608     MC146818RtcState *s = arg->rtc_state;
609     int16_t cylinders;
610     int8_t heads, sectors;
611     int val;
612     int i, trans;
613 
614     val = 0;
615     if (arg->idebus[0] && ide_get_geometry(arg->idebus[0], 0,
616                                            &cylinders, &heads, &sectors) >= 0) {
617         cmos_init_hd(s, 0x19, 0x1b, cylinders, heads, sectors);
618         val |= 0xf0;
619     }
620     if (arg->idebus[0] && ide_get_geometry(arg->idebus[0], 1,
621                                            &cylinders, &heads, &sectors) >= 0) {
622         cmos_init_hd(s, 0x1a, 0x24, cylinders, heads, sectors);
623         val |= 0x0f;
624     }
625     mc146818rtc_set_cmos_data(s, 0x12, val);
626 
627     val = 0;
628     for (i = 0; i < 4; i++) {
629         /* NOTE: ide_get_geometry() returns the physical
630            geometry.  It is always such that: 1 <= sects <= 63, 1
631            <= heads <= 16, 1 <= cylinders <= 16383. The BIOS
632            geometry can be different if a translation is done. */
633         if (arg->idebus[i / 2] &&
634             ide_get_geometry(arg->idebus[i / 2], i % 2,
635                              &cylinders, &heads, &sectors) >= 0) {
636             trans = ide_get_bios_chs_trans(arg->idebus[i / 2], i % 2) - 1;
637             assert((trans & ~3) == 0);
638             val |= trans << (i * 2);
639         }
640     }
641     mc146818rtc_set_cmos_data(s, 0x39, val);
642 
643     pc_cmos_init_floppy(s, pc_find_fdc0());
644 
645     qemu_unregister_reset(pc_cmos_init_late, opaque);
646 }
647 
648 void pc_cmos_init(PCMachineState *pcms,
649                   BusState *idebus0, BusState *idebus1,
650                   ISADevice *rtc)
651 {
652     int val;
653     static pc_cmos_init_late_arg arg;
654     X86MachineState *x86ms = X86_MACHINE(pcms);
655     MC146818RtcState *s = MC146818_RTC(rtc);
656 
657     /* various important CMOS locations needed by PC/Bochs bios */
658 
659     /* memory size */
660     /* base memory (first MiB) */
661     val = MIN(x86ms->below_4g_mem_size / KiB, 640);
662     mc146818rtc_set_cmos_data(s, 0x15, val);
663     mc146818rtc_set_cmos_data(s, 0x16, val >> 8);
664     /* extended memory (next 64MiB) */
665     if (x86ms->below_4g_mem_size > 1 * MiB) {
666         val = (x86ms->below_4g_mem_size - 1 * MiB) / KiB;
667     } else {
668         val = 0;
669     }
670     if (val > 65535)
671         val = 65535;
672     mc146818rtc_set_cmos_data(s, 0x17, val);
673     mc146818rtc_set_cmos_data(s, 0x18, val >> 8);
674     mc146818rtc_set_cmos_data(s, 0x30, val);
675     mc146818rtc_set_cmos_data(s, 0x31, val >> 8);
676     /* memory between 16MiB and 4GiB */
677     if (x86ms->below_4g_mem_size > 16 * MiB) {
678         val = (x86ms->below_4g_mem_size - 16 * MiB) / (64 * KiB);
679     } else {
680         val = 0;
681     }
682     if (val > 65535)
683         val = 65535;
684     mc146818rtc_set_cmos_data(s, 0x34, val);
685     mc146818rtc_set_cmos_data(s, 0x35, val >> 8);
686     /* memory above 4GiB */
687     val = x86ms->above_4g_mem_size / 65536;
688     mc146818rtc_set_cmos_data(s, 0x5b, val);
689     mc146818rtc_set_cmos_data(s, 0x5c, val >> 8);
690     mc146818rtc_set_cmos_data(s, 0x5d, val >> 16);
691 
692     object_property_add_link(OBJECT(pcms), "rtc_state",
693                              TYPE_ISA_DEVICE,
694                              (Object **)&x86ms->rtc,
695                              object_property_allow_set_link,
696                              OBJ_PROP_LINK_STRONG);
697     object_property_set_link(OBJECT(pcms), "rtc_state", OBJECT(s),
698                              &error_abort);
699 
700     set_boot_dev(s, MACHINE(pcms)->boot_config.order, &error_fatal);
701 
702     val = 0;
703     val |= 0x02; /* FPU is there */
704     val |= 0x04; /* PS/2 mouse installed */
705     mc146818rtc_set_cmos_data(s, REG_EQUIPMENT_BYTE, val);
706 
707     /* hard drives and FDC */
708     arg.rtc_state = s;
709     arg.idebus[0] = idebus0;
710     arg.idebus[1] = idebus1;
711     qemu_register_reset(pc_cmos_init_late, &arg);
712 }
713 
714 static void handle_a20_line_change(void *opaque, int irq, int level)
715 {
716     X86CPU *cpu = opaque;
717 
718     /* XXX: send to all CPUs ? */
719     /* XXX: add logic to handle multiple A20 line sources */
720     x86_cpu_set_a20(cpu, level);
721 }
722 
723 #define NE2000_NB_MAX 6
724 
725 static const int ne2000_io[NE2000_NB_MAX] = { 0x300, 0x320, 0x340, 0x360,
726                                               0x280, 0x380 };
727 static const int ne2000_irq[NE2000_NB_MAX] = { 9, 10, 11, 3, 4, 5 };
728 
729 static void pc_init_ne2k_isa(ISABus *bus, NICInfo *nd)
730 {
731     static int nb_ne2k = 0;
732 
733     if (nb_ne2k == NE2000_NB_MAX)
734         return;
735     isa_ne2000_init(bus, ne2000_io[nb_ne2k],
736                     ne2000_irq[nb_ne2k], nd);
737     nb_ne2k++;
738 }
739 
740 void pc_acpi_smi_interrupt(void *opaque, int irq, int level)
741 {
742     X86CPU *cpu = opaque;
743 
744     if (level) {
745         cpu_interrupt(CPU(cpu), CPU_INTERRUPT_SMI);
746     }
747 }
748 
749 static
750 void pc_machine_done(Notifier *notifier, void *data)
751 {
752     PCMachineState *pcms = container_of(notifier,
753                                         PCMachineState, machine_done);
754     X86MachineState *x86ms = X86_MACHINE(pcms);
755 
756     cxl_hook_up_pxb_registers(pcms->bus, &pcms->cxl_devices_state,
757                               &error_fatal);
758 
759     if (pcms->cxl_devices_state.is_enabled) {
760         cxl_fmws_link_targets(&pcms->cxl_devices_state, &error_fatal);
761     }
762 
763     /* set the number of CPUs */
764     x86_rtc_set_cpus_count(x86ms->rtc, x86ms->boot_cpus);
765 
766     fw_cfg_add_extra_pci_roots(pcms->bus, x86ms->fw_cfg);
767 
768     acpi_setup();
769     if (x86ms->fw_cfg) {
770         fw_cfg_build_smbios(MACHINE(pcms), x86ms->fw_cfg);
771         fw_cfg_build_feature_control(MACHINE(pcms), x86ms->fw_cfg);
772         /* update FW_CFG_NB_CPUS to account for -device added CPUs */
773         fw_cfg_modify_i16(x86ms->fw_cfg, FW_CFG_NB_CPUS, x86ms->boot_cpus);
774     }
775 }
776 
777 void pc_guest_info_init(PCMachineState *pcms)
778 {
779     X86MachineState *x86ms = X86_MACHINE(pcms);
780 
781     x86ms->apic_xrupt_override = true;
782     pcms->machine_done.notify = pc_machine_done;
783     qemu_add_machine_init_done_notifier(&pcms->machine_done);
784 }
785 
786 /* setup pci memory address space mapping into system address space */
787 void pc_pci_as_mapping_init(MemoryRegion *system_memory,
788                             MemoryRegion *pci_address_space)
789 {
790     /* Set to lower priority than RAM */
791     memory_region_add_subregion_overlap(system_memory, 0x0,
792                                         pci_address_space, -1);
793 }
794 
795 void xen_load_linux(PCMachineState *pcms)
796 {
797     int i;
798     FWCfgState *fw_cfg;
799     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
800     X86MachineState *x86ms = X86_MACHINE(pcms);
801 
802     assert(MACHINE(pcms)->kernel_filename != NULL);
803 
804     fw_cfg = fw_cfg_init_io(FW_CFG_IO_BASE);
805     fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, x86ms->boot_cpus);
806     rom_set_fw(fw_cfg);
807 
808     x86_load_linux(x86ms, fw_cfg, pcmc->acpi_data_size,
809                    pcmc->pvh_enabled, pcmc->legacy_no_rng_seed);
810     for (i = 0; i < nb_option_roms; i++) {
811         assert(!strcmp(option_rom[i].name, "linuxboot.bin") ||
812                !strcmp(option_rom[i].name, "linuxboot_dma.bin") ||
813                !strcmp(option_rom[i].name, "pvh.bin") ||
814                !strcmp(option_rom[i].name, "multiboot.bin") ||
815                !strcmp(option_rom[i].name, "multiboot_dma.bin"));
816         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
817     }
818     x86ms->fw_cfg = fw_cfg;
819 }
820 
821 #define PC_ROM_MIN_VGA     0xc0000
822 #define PC_ROM_MIN_OPTION  0xc8000
823 #define PC_ROM_MAX         0xe0000
824 #define PC_ROM_ALIGN       0x800
825 #define PC_ROM_SIZE        (PC_ROM_MAX - PC_ROM_MIN_VGA)
826 
827 static hwaddr pc_above_4g_end(PCMachineState *pcms)
828 {
829     X86MachineState *x86ms = X86_MACHINE(pcms);
830 
831     if (pcms->sgx_epc.size != 0) {
832         return sgx_epc_above_4g_end(&pcms->sgx_epc);
833     }
834 
835     return x86ms->above_4g_mem_start + x86ms->above_4g_mem_size;
836 }
837 
838 static void pc_get_device_memory_range(PCMachineState *pcms,
839                                        hwaddr *base,
840                                        ram_addr_t *device_mem_size)
841 {
842     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
843     MachineState *machine = MACHINE(pcms);
844     ram_addr_t size;
845     hwaddr addr;
846 
847     size = machine->maxram_size - machine->ram_size;
848     addr = ROUND_UP(pc_above_4g_end(pcms), 1 * GiB);
849 
850     if (pcmc->enforce_aligned_dimm) {
851         /* size device region assuming 1G page max alignment per slot */
852         size += (1 * GiB) * machine->ram_slots;
853     }
854 
855     *base = addr;
856     *device_mem_size = size;
857 }
858 
859 static uint64_t pc_get_cxl_range_start(PCMachineState *pcms)
860 {
861     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
862     hwaddr cxl_base;
863     ram_addr_t size;
864 
865     if (pcmc->has_reserved_memory) {
866         pc_get_device_memory_range(pcms, &cxl_base, &size);
867         cxl_base += size;
868     } else {
869         cxl_base = pc_above_4g_end(pcms);
870     }
871 
872     return cxl_base;
873 }
874 
875 static uint64_t pc_get_cxl_range_end(PCMachineState *pcms)
876 {
877     uint64_t start = pc_get_cxl_range_start(pcms) + MiB;
878 
879     if (pcms->cxl_devices_state.fixed_windows) {
880         GList *it;
881 
882         start = ROUND_UP(start, 256 * MiB);
883         for (it = pcms->cxl_devices_state.fixed_windows; it; it = it->next) {
884             CXLFixedWindow *fw = it->data;
885             start += fw->size;
886         }
887     }
888 
889     return start;
890 }
891 
892 static hwaddr pc_max_used_gpa(PCMachineState *pcms, uint64_t pci_hole64_size)
893 {
894     X86CPU *cpu = X86_CPU(first_cpu);
895 
896     /* 32-bit systems don't have hole64 thus return max CPU address */
897     if (cpu->phys_bits <= 32) {
898         return ((hwaddr)1 << cpu->phys_bits) - 1;
899     }
900 
901     return pc_pci_hole64_start() + pci_hole64_size - 1;
902 }
903 
904 /*
905  * AMD systems with an IOMMU have an additional hole close to the
906  * 1Tb, which are special GPAs that cannot be DMA mapped. Depending
907  * on kernel version, VFIO may or may not let you DMA map those ranges.
908  * Starting Linux v5.4 we validate it, and can't create guests on AMD machines
909  * with certain memory sizes. It's also wrong to use those IOVA ranges
910  * in detriment of leading to IOMMU INVALID_DEVICE_REQUEST or worse.
911  * The ranges reserved for Hyper-Transport are:
912  *
913  * FD_0000_0000h - FF_FFFF_FFFFh
914  *
915  * The ranges represent the following:
916  *
917  * Base Address   Top Address  Use
918  *
919  * FD_0000_0000h FD_F7FF_FFFFh Reserved interrupt address space
920  * FD_F800_0000h FD_F8FF_FFFFh Interrupt/EOI IntCtl
921  * FD_F900_0000h FD_F90F_FFFFh Legacy PIC IACK
922  * FD_F910_0000h FD_F91F_FFFFh System Management
923  * FD_F920_0000h FD_FAFF_FFFFh Reserved Page Tables
924  * FD_FB00_0000h FD_FBFF_FFFFh Address Translation
925  * FD_FC00_0000h FD_FDFF_FFFFh I/O Space
926  * FD_FE00_0000h FD_FFFF_FFFFh Configuration
927  * FE_0000_0000h FE_1FFF_FFFFh Extended Configuration/Device Messages
928  * FE_2000_0000h FF_FFFF_FFFFh Reserved
929  *
930  * See AMD IOMMU spec, section 2.1.2 "IOMMU Logical Topology",
931  * Table 3: Special Address Controls (GPA) for more information.
932  */
933 #define AMD_HT_START         0xfd00000000UL
934 #define AMD_HT_END           0xffffffffffUL
935 #define AMD_ABOVE_1TB_START  (AMD_HT_END + 1)
936 #define AMD_HT_SIZE          (AMD_ABOVE_1TB_START - AMD_HT_START)
937 
938 void pc_memory_init(PCMachineState *pcms,
939                     MemoryRegion *system_memory,
940                     MemoryRegion *rom_memory,
941                     MemoryRegion **ram_memory,
942                     uint64_t pci_hole64_size)
943 {
944     int linux_boot, i;
945     MemoryRegion *option_rom_mr;
946     MemoryRegion *ram_below_4g, *ram_above_4g;
947     FWCfgState *fw_cfg;
948     MachineState *machine = MACHINE(pcms);
949     MachineClass *mc = MACHINE_GET_CLASS(machine);
950     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
951     X86MachineState *x86ms = X86_MACHINE(pcms);
952     hwaddr maxphysaddr, maxusedaddr;
953     hwaddr cxl_base, cxl_resv_end = 0;
954     X86CPU *cpu = X86_CPU(first_cpu);
955 
956     assert(machine->ram_size == x86ms->below_4g_mem_size +
957                                 x86ms->above_4g_mem_size);
958 
959     linux_boot = (machine->kernel_filename != NULL);
960 
961     /*
962      * The HyperTransport range close to the 1T boundary is unique to AMD
963      * hosts with IOMMUs enabled. Restrict the ram-above-4g relocation
964      * to above 1T to AMD vCPUs only. @enforce_amd_1tb_hole is only false in
965      * older machine types (<= 7.0) for compatibility purposes.
966      */
967     if (IS_AMD_CPU(&cpu->env) && pcmc->enforce_amd_1tb_hole) {
968         /* Bail out if max possible address does not cross HT range */
969         if (pc_max_used_gpa(pcms, pci_hole64_size) >= AMD_HT_START) {
970             x86ms->above_4g_mem_start = AMD_ABOVE_1TB_START;
971         }
972 
973         /*
974          * Advertise the HT region if address space covers the reserved
975          * region or if we relocate.
976          */
977         if (cpu->phys_bits >= 40) {
978             e820_add_entry(AMD_HT_START, AMD_HT_SIZE, E820_RESERVED);
979         }
980     }
981 
982     /*
983      * phys-bits is required to be appropriately configured
984      * to make sure max used GPA is reachable.
985      */
986     maxusedaddr = pc_max_used_gpa(pcms, pci_hole64_size);
987     maxphysaddr = ((hwaddr)1 << cpu->phys_bits) - 1;
988     if (maxphysaddr < maxusedaddr) {
989         error_report("Address space limit 0x%"PRIx64" < 0x%"PRIx64
990                      " phys-bits too low (%u)",
991                      maxphysaddr, maxusedaddr, cpu->phys_bits);
992         exit(EXIT_FAILURE);
993     }
994 
995     /*
996      * Split single memory region and use aliases to address portions of it,
997      * done for backwards compatibility with older qemus.
998      */
999     *ram_memory = machine->ram;
1000     ram_below_4g = g_malloc(sizeof(*ram_below_4g));
1001     memory_region_init_alias(ram_below_4g, NULL, "ram-below-4g", machine->ram,
1002                              0, x86ms->below_4g_mem_size);
1003     memory_region_add_subregion(system_memory, 0, ram_below_4g);
1004     e820_add_entry(0, x86ms->below_4g_mem_size, E820_RAM);
1005     if (x86ms->above_4g_mem_size > 0) {
1006         ram_above_4g = g_malloc(sizeof(*ram_above_4g));
1007         memory_region_init_alias(ram_above_4g, NULL, "ram-above-4g",
1008                                  machine->ram,
1009                                  x86ms->below_4g_mem_size,
1010                                  x86ms->above_4g_mem_size);
1011         memory_region_add_subregion(system_memory, x86ms->above_4g_mem_start,
1012                                     ram_above_4g);
1013         e820_add_entry(x86ms->above_4g_mem_start, x86ms->above_4g_mem_size,
1014                        E820_RAM);
1015     }
1016 
1017     if (pcms->sgx_epc.size != 0) {
1018         e820_add_entry(pcms->sgx_epc.base, pcms->sgx_epc.size, E820_RESERVED);
1019     }
1020 
1021     if (!pcmc->has_reserved_memory &&
1022         (machine->ram_slots ||
1023          (machine->maxram_size > machine->ram_size))) {
1024 
1025         error_report("\"-memory 'slots|maxmem'\" is not supported by: %s",
1026                      mc->name);
1027         exit(EXIT_FAILURE);
1028     }
1029 
1030     /* always allocate the device memory information */
1031     machine->device_memory = g_malloc0(sizeof(*machine->device_memory));
1032 
1033     /* initialize device memory address space */
1034     if (pcmc->has_reserved_memory &&
1035         (machine->ram_size < machine->maxram_size)) {
1036         ram_addr_t device_mem_size;
1037 
1038         if (machine->ram_slots > ACPI_MAX_RAM_SLOTS) {
1039             error_report("unsupported amount of memory slots: %"PRIu64,
1040                          machine->ram_slots);
1041             exit(EXIT_FAILURE);
1042         }
1043 
1044         if (QEMU_ALIGN_UP(machine->maxram_size,
1045                           TARGET_PAGE_SIZE) != machine->maxram_size) {
1046             error_report("maximum memory size must by aligned to multiple of "
1047                          "%d bytes", TARGET_PAGE_SIZE);
1048             exit(EXIT_FAILURE);
1049         }
1050 
1051         pc_get_device_memory_range(pcms, &machine->device_memory->base, &device_mem_size);
1052 
1053         if ((machine->device_memory->base + device_mem_size) <
1054             device_mem_size) {
1055             error_report("unsupported amount of maximum memory: " RAM_ADDR_FMT,
1056                          machine->maxram_size);
1057             exit(EXIT_FAILURE);
1058         }
1059 
1060         memory_region_init(&machine->device_memory->mr, OBJECT(pcms),
1061                            "device-memory", device_mem_size);
1062         memory_region_add_subregion(system_memory, machine->device_memory->base,
1063                                     &machine->device_memory->mr);
1064     }
1065 
1066     if (pcms->cxl_devices_state.is_enabled) {
1067         MemoryRegion *mr = &pcms->cxl_devices_state.host_mr;
1068         hwaddr cxl_size = MiB;
1069 
1070         cxl_base = pc_get_cxl_range_start(pcms);
1071         memory_region_init(mr, OBJECT(machine), "cxl_host_reg", cxl_size);
1072         memory_region_add_subregion(system_memory, cxl_base, mr);
1073         cxl_resv_end = cxl_base + cxl_size;
1074         if (pcms->cxl_devices_state.fixed_windows) {
1075             hwaddr cxl_fmw_base;
1076             GList *it;
1077 
1078             cxl_fmw_base = ROUND_UP(cxl_base + cxl_size, 256 * MiB);
1079             for (it = pcms->cxl_devices_state.fixed_windows; it; it = it->next) {
1080                 CXLFixedWindow *fw = it->data;
1081 
1082                 fw->base = cxl_fmw_base;
1083                 memory_region_init_io(&fw->mr, OBJECT(machine), &cfmws_ops, fw,
1084                                       "cxl-fixed-memory-region", fw->size);
1085                 memory_region_add_subregion(system_memory, fw->base, &fw->mr);
1086                 cxl_fmw_base += fw->size;
1087                 cxl_resv_end = cxl_fmw_base;
1088             }
1089         }
1090     }
1091 
1092     /* Initialize PC system firmware */
1093     pc_system_firmware_init(pcms, rom_memory);
1094 
1095     option_rom_mr = g_malloc(sizeof(*option_rom_mr));
1096     memory_region_init_ram(option_rom_mr, NULL, "pc.rom", PC_ROM_SIZE,
1097                            &error_fatal);
1098     if (pcmc->pci_enabled) {
1099         memory_region_set_readonly(option_rom_mr, true);
1100     }
1101     memory_region_add_subregion_overlap(rom_memory,
1102                                         PC_ROM_MIN_VGA,
1103                                         option_rom_mr,
1104                                         1);
1105 
1106     fw_cfg = fw_cfg_arch_create(machine,
1107                                 x86ms->boot_cpus, x86ms->apic_id_limit);
1108 
1109     rom_set_fw(fw_cfg);
1110 
1111     if (pcmc->has_reserved_memory && machine->device_memory->base) {
1112         uint64_t *val = g_malloc(sizeof(*val));
1113         PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1114         uint64_t res_mem_end = machine->device_memory->base;
1115 
1116         if (!pcmc->broken_reserved_end) {
1117             res_mem_end += memory_region_size(&machine->device_memory->mr);
1118         }
1119 
1120         if (pcms->cxl_devices_state.is_enabled) {
1121             res_mem_end = cxl_resv_end;
1122         }
1123         *val = cpu_to_le64(ROUND_UP(res_mem_end, 1 * GiB));
1124         fw_cfg_add_file(fw_cfg, "etc/reserved-memory-end", val, sizeof(*val));
1125     }
1126 
1127     if (linux_boot) {
1128         x86_load_linux(x86ms, fw_cfg, pcmc->acpi_data_size,
1129                        pcmc->pvh_enabled, pcmc->legacy_no_rng_seed);
1130     }
1131 
1132     for (i = 0; i < nb_option_roms; i++) {
1133         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1134     }
1135     x86ms->fw_cfg = fw_cfg;
1136 
1137     /* Init default IOAPIC address space */
1138     x86ms->ioapic_as = &address_space_memory;
1139 
1140     /* Init ACPI memory hotplug IO base address */
1141     pcms->memhp_io_base = ACPI_MEMORY_HOTPLUG_BASE;
1142 }
1143 
1144 /*
1145  * The 64bit pci hole starts after "above 4G RAM" and
1146  * potentially the space reserved for memory hotplug.
1147  */
1148 uint64_t pc_pci_hole64_start(void)
1149 {
1150     PCMachineState *pcms = PC_MACHINE(qdev_get_machine());
1151     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1152     MachineState *ms = MACHINE(pcms);
1153     uint64_t hole64_start = 0;
1154     ram_addr_t size = 0;
1155 
1156     if (pcms->cxl_devices_state.is_enabled) {
1157         hole64_start = pc_get_cxl_range_end(pcms);
1158     } else if (pcmc->has_reserved_memory && (ms->ram_size < ms->maxram_size)) {
1159         pc_get_device_memory_range(pcms, &hole64_start, &size);
1160         if (!pcmc->broken_reserved_end) {
1161             hole64_start += size;
1162         }
1163     } else {
1164         hole64_start = pc_above_4g_end(pcms);
1165     }
1166 
1167     return ROUND_UP(hole64_start, 1 * GiB);
1168 }
1169 
1170 DeviceState *pc_vga_init(ISABus *isa_bus, PCIBus *pci_bus)
1171 {
1172     DeviceState *dev = NULL;
1173 
1174     rom_set_order_override(FW_CFG_ORDER_OVERRIDE_VGA);
1175     if (pci_bus) {
1176         PCIDevice *pcidev = pci_vga_init(pci_bus);
1177         dev = pcidev ? &pcidev->qdev : NULL;
1178     } else if (isa_bus) {
1179         ISADevice *isadev = isa_vga_init(isa_bus);
1180         dev = isadev ? DEVICE(isadev) : NULL;
1181     }
1182     rom_reset_order_override();
1183     return dev;
1184 }
1185 
1186 static const MemoryRegionOps ioport80_io_ops = {
1187     .write = ioport80_write,
1188     .read = ioport80_read,
1189     .endianness = DEVICE_NATIVE_ENDIAN,
1190     .impl = {
1191         .min_access_size = 1,
1192         .max_access_size = 1,
1193     },
1194 };
1195 
1196 static const MemoryRegionOps ioportF0_io_ops = {
1197     .write = ioportF0_write,
1198     .read = ioportF0_read,
1199     .endianness = DEVICE_NATIVE_ENDIAN,
1200     .impl = {
1201         .min_access_size = 1,
1202         .max_access_size = 1,
1203     },
1204 };
1205 
1206 static void pc_superio_init(ISABus *isa_bus, bool create_fdctrl,
1207                             bool create_i8042, bool no_vmport)
1208 {
1209     int i;
1210     DriveInfo *fd[MAX_FD];
1211     qemu_irq *a20_line;
1212     ISADevice *fdc, *i8042, *port92, *vmmouse;
1213 
1214     serial_hds_isa_init(isa_bus, 0, MAX_ISA_SERIAL_PORTS);
1215     parallel_hds_isa_init(isa_bus, MAX_PARALLEL_PORTS);
1216 
1217     for (i = 0; i < MAX_FD; i++) {
1218         fd[i] = drive_get(IF_FLOPPY, 0, i);
1219         create_fdctrl |= !!fd[i];
1220     }
1221     if (create_fdctrl) {
1222         fdc = isa_new(TYPE_ISA_FDC);
1223         if (fdc) {
1224             isa_realize_and_unref(fdc, isa_bus, &error_fatal);
1225             isa_fdc_init_drives(fdc, fd);
1226         }
1227     }
1228 
1229     if (!create_i8042) {
1230         return;
1231     }
1232 
1233     i8042 = isa_create_simple(isa_bus, TYPE_I8042);
1234     if (!no_vmport) {
1235         isa_create_simple(isa_bus, TYPE_VMPORT);
1236         vmmouse = isa_try_new("vmmouse");
1237     } else {
1238         vmmouse = NULL;
1239     }
1240     if (vmmouse) {
1241         object_property_set_link(OBJECT(vmmouse), TYPE_I8042, OBJECT(i8042),
1242                                  &error_abort);
1243         isa_realize_and_unref(vmmouse, isa_bus, &error_fatal);
1244     }
1245     port92 = isa_create_simple(isa_bus, TYPE_PORT92);
1246 
1247     a20_line = qemu_allocate_irqs(handle_a20_line_change, first_cpu, 2);
1248     i8042_setup_a20_line(i8042, a20_line[0]);
1249     qdev_connect_gpio_out_named(DEVICE(port92),
1250                                 PORT92_A20_LINE, 0, a20_line[1]);
1251     g_free(a20_line);
1252 }
1253 
1254 void pc_basic_device_init(struct PCMachineState *pcms,
1255                           ISABus *isa_bus, qemu_irq *gsi,
1256                           ISADevice **rtc_state,
1257                           bool create_fdctrl,
1258                           uint32_t hpet_irqs)
1259 {
1260     int i;
1261     DeviceState *hpet = NULL;
1262     int pit_isa_irq = 0;
1263     qemu_irq pit_alt_irq = NULL;
1264     qemu_irq rtc_irq = NULL;
1265     ISADevice *pit = NULL;
1266     MemoryRegion *ioport80_io = g_new(MemoryRegion, 1);
1267     MemoryRegion *ioportF0_io = g_new(MemoryRegion, 1);
1268     X86MachineState *x86ms = X86_MACHINE(pcms);
1269 
1270     memory_region_init_io(ioport80_io, NULL, &ioport80_io_ops, NULL, "ioport80", 1);
1271     memory_region_add_subregion(isa_bus->address_space_io, 0x80, ioport80_io);
1272 
1273     memory_region_init_io(ioportF0_io, NULL, &ioportF0_io_ops, NULL, "ioportF0", 1);
1274     memory_region_add_subregion(isa_bus->address_space_io, 0xf0, ioportF0_io);
1275 
1276     /*
1277      * Check if an HPET shall be created.
1278      *
1279      * Without KVM_CAP_PIT_STATE2, we cannot switch off the in-kernel PIT
1280      * when the HPET wants to take over. Thus we have to disable the latter.
1281      */
1282     if (pcms->hpet_enabled && (!kvm_irqchip_in_kernel() ||
1283                                kvm_has_pit_state2())) {
1284         hpet = qdev_try_new(TYPE_HPET);
1285         if (!hpet) {
1286             error_report("couldn't create HPET device");
1287             exit(1);
1288         }
1289         /*
1290          * For pc-piix-*, hpet's intcap is always IRQ2. For pc-q35-1.7 and
1291          * earlier, use IRQ2 for compat. Otherwise, use IRQ16~23, IRQ8 and
1292          * IRQ2.
1293          */
1294         uint8_t compat = object_property_get_uint(OBJECT(hpet),
1295                 HPET_INTCAP, NULL);
1296         if (!compat) {
1297             qdev_prop_set_uint32(hpet, HPET_INTCAP, hpet_irqs);
1298         }
1299         sysbus_realize_and_unref(SYS_BUS_DEVICE(hpet), &error_fatal);
1300         sysbus_mmio_map(SYS_BUS_DEVICE(hpet), 0, HPET_BASE);
1301 
1302         for (i = 0; i < IOAPIC_NUM_PINS; i++) {
1303             sysbus_connect_irq(SYS_BUS_DEVICE(hpet), i, gsi[i]);
1304         }
1305         pit_isa_irq = -1;
1306         pit_alt_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_PIT_INT);
1307         rtc_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_RTC_INT);
1308     }
1309     *rtc_state = ISA_DEVICE(mc146818_rtc_init(isa_bus, 2000, rtc_irq));
1310 
1311     qemu_register_boot_set(pc_boot_set, *rtc_state);
1312 
1313     if (!xen_enabled() &&
1314         (x86ms->pit == ON_OFF_AUTO_AUTO || x86ms->pit == ON_OFF_AUTO_ON)) {
1315         if (kvm_pit_in_kernel()) {
1316             pit = kvm_pit_init(isa_bus, 0x40);
1317         } else {
1318             pit = i8254_pit_init(isa_bus, 0x40, pit_isa_irq, pit_alt_irq);
1319         }
1320         if (hpet) {
1321             /* connect PIT to output control line of the HPET */
1322             qdev_connect_gpio_out(hpet, 0, qdev_get_gpio_in(DEVICE(pit), 0));
1323         }
1324         pcspk_init(pcms->pcspk, isa_bus, pit);
1325     }
1326 
1327     /* Super I/O */
1328     pc_superio_init(isa_bus, create_fdctrl, pcms->i8042_enabled,
1329                     pcms->vmport != ON_OFF_AUTO_ON);
1330 }
1331 
1332 void pc_nic_init(PCMachineClass *pcmc, ISABus *isa_bus, PCIBus *pci_bus)
1333 {
1334     int i;
1335 
1336     rom_set_order_override(FW_CFG_ORDER_OVERRIDE_NIC);
1337     for (i = 0; i < nb_nics; i++) {
1338         NICInfo *nd = &nd_table[i];
1339         const char *model = nd->model ? nd->model : pcmc->default_nic_model;
1340 
1341         if (g_str_equal(model, "ne2k_isa")) {
1342             pc_init_ne2k_isa(isa_bus, nd);
1343         } else {
1344             pci_nic_init_nofail(nd, pci_bus, model, NULL);
1345         }
1346     }
1347     rom_reset_order_override();
1348 }
1349 
1350 void pc_i8259_create(ISABus *isa_bus, qemu_irq *i8259_irqs)
1351 {
1352     qemu_irq *i8259;
1353 
1354     if (kvm_pic_in_kernel()) {
1355         i8259 = kvm_i8259_init(isa_bus);
1356     } else if (xen_enabled()) {
1357         i8259 = xen_interrupt_controller_init();
1358     } else {
1359         i8259 = i8259_init(isa_bus, x86_allocate_cpu_irq());
1360     }
1361 
1362     for (size_t i = 0; i < ISA_NUM_IRQS; i++) {
1363         i8259_irqs[i] = i8259[i];
1364     }
1365 
1366     g_free(i8259);
1367 }
1368 
1369 static void pc_memory_pre_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
1370                                Error **errp)
1371 {
1372     const PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1373     const X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
1374     const PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1375     const MachineState *ms = MACHINE(hotplug_dev);
1376     const bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
1377     const uint64_t legacy_align = TARGET_PAGE_SIZE;
1378     Error *local_err = NULL;
1379 
1380     /*
1381      * When -no-acpi is used with Q35 machine type, no ACPI is built,
1382      * but pcms->acpi_dev is still created. Check !acpi_enabled in
1383      * addition to cover this case.
1384      */
1385     if (!x86ms->acpi_dev || !x86_machine_is_acpi_enabled(x86ms)) {
1386         error_setg(errp,
1387                    "memory hotplug is not enabled: missing acpi device or acpi disabled");
1388         return;
1389     }
1390 
1391     if (is_nvdimm && !ms->nvdimms_state->is_enabled) {
1392         error_setg(errp, "nvdimm is not enabled: missing 'nvdimm' in '-M'");
1393         return;
1394     }
1395 
1396     hotplug_handler_pre_plug(x86ms->acpi_dev, dev, &local_err);
1397     if (local_err) {
1398         error_propagate(errp, local_err);
1399         return;
1400     }
1401 
1402     pc_dimm_pre_plug(PC_DIMM(dev), MACHINE(hotplug_dev),
1403                      pcmc->enforce_aligned_dimm ? NULL : &legacy_align, errp);
1404 }
1405 
1406 static void pc_memory_plug(HotplugHandler *hotplug_dev,
1407                            DeviceState *dev, Error **errp)
1408 {
1409     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1410     X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
1411     MachineState *ms = MACHINE(hotplug_dev);
1412     bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
1413 
1414     pc_dimm_plug(PC_DIMM(dev), MACHINE(pcms));
1415 
1416     if (is_nvdimm) {
1417         nvdimm_plug(ms->nvdimms_state);
1418     }
1419 
1420     hotplug_handler_plug(x86ms->acpi_dev, dev, &error_abort);
1421 }
1422 
1423 static void pc_memory_unplug_request(HotplugHandler *hotplug_dev,
1424                                      DeviceState *dev, Error **errp)
1425 {
1426     X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
1427 
1428     /*
1429      * When -no-acpi is used with Q35 machine type, no ACPI is built,
1430      * but pcms->acpi_dev is still created. Check !acpi_enabled in
1431      * addition to cover this case.
1432      */
1433     if (!x86ms->acpi_dev || !x86_machine_is_acpi_enabled(x86ms)) {
1434         error_setg(errp,
1435                    "memory hotplug is not enabled: missing acpi device or acpi disabled");
1436         return;
1437     }
1438 
1439     if (object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM)) {
1440         error_setg(errp, "nvdimm device hot unplug is not supported yet.");
1441         return;
1442     }
1443 
1444     hotplug_handler_unplug_request(x86ms->acpi_dev, dev,
1445                                    errp);
1446 }
1447 
1448 static void pc_memory_unplug(HotplugHandler *hotplug_dev,
1449                              DeviceState *dev, Error **errp)
1450 {
1451     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1452     X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
1453     Error *local_err = NULL;
1454 
1455     hotplug_handler_unplug(x86ms->acpi_dev, dev, &local_err);
1456     if (local_err) {
1457         goto out;
1458     }
1459 
1460     pc_dimm_unplug(PC_DIMM(dev), MACHINE(pcms));
1461     qdev_unrealize(dev);
1462  out:
1463     error_propagate(errp, local_err);
1464 }
1465 
1466 static void pc_virtio_md_pci_pre_plug(HotplugHandler *hotplug_dev,
1467                                       DeviceState *dev, Error **errp)
1468 {
1469     HotplugHandler *hotplug_dev2 = qdev_get_bus_hotplug_handler(dev);
1470     Error *local_err = NULL;
1471 
1472     if (!hotplug_dev2 && dev->hotplugged) {
1473         /*
1474          * Without a bus hotplug handler, we cannot control the plug/unplug
1475          * order. We should never reach this point when hotplugging on x86,
1476          * however, better add a safety net.
1477          */
1478         error_setg(errp, "hotplug of virtio based memory devices not supported"
1479                    " on this bus.");
1480         return;
1481     }
1482     /*
1483      * First, see if we can plug this memory device at all. If that
1484      * succeeds, branch of to the actual hotplug handler.
1485      */
1486     memory_device_pre_plug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev), NULL,
1487                            &local_err);
1488     if (!local_err && hotplug_dev2) {
1489         hotplug_handler_pre_plug(hotplug_dev2, dev, &local_err);
1490     }
1491     error_propagate(errp, local_err);
1492 }
1493 
1494 static void pc_virtio_md_pci_plug(HotplugHandler *hotplug_dev,
1495                                   DeviceState *dev, Error **errp)
1496 {
1497     HotplugHandler *hotplug_dev2 = qdev_get_bus_hotplug_handler(dev);
1498     Error *local_err = NULL;
1499 
1500     /*
1501      * Plug the memory device first and then branch off to the actual
1502      * hotplug handler. If that one fails, we can easily undo the memory
1503      * device bits.
1504      */
1505     memory_device_plug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev));
1506     if (hotplug_dev2) {
1507         hotplug_handler_plug(hotplug_dev2, dev, &local_err);
1508         if (local_err) {
1509             memory_device_unplug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev));
1510         }
1511     }
1512     error_propagate(errp, local_err);
1513 }
1514 
1515 static void pc_virtio_md_pci_unplug_request(HotplugHandler *hotplug_dev,
1516                                             DeviceState *dev, Error **errp)
1517 {
1518     /* We don't support hot unplug of virtio based memory devices */
1519     error_setg(errp, "virtio based memory devices cannot be unplugged.");
1520 }
1521 
1522 static void pc_virtio_md_pci_unplug(HotplugHandler *hotplug_dev,
1523                                     DeviceState *dev, Error **errp)
1524 {
1525     /* We don't support hot unplug of virtio based memory devices */
1526 }
1527 
1528 static void pc_machine_device_pre_plug_cb(HotplugHandler *hotplug_dev,
1529                                           DeviceState *dev, Error **errp)
1530 {
1531     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1532         pc_memory_pre_plug(hotplug_dev, dev, errp);
1533     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1534         x86_cpu_pre_plug(hotplug_dev, dev, errp);
1535     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI) ||
1536                object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI)) {
1537         pc_virtio_md_pci_pre_plug(hotplug_dev, dev, errp);
1538     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
1539         /* Declare the APIC range as the reserved MSI region */
1540         char *resv_prop_str = g_strdup_printf("0xfee00000:0xfeefffff:%d",
1541                                               VIRTIO_IOMMU_RESV_MEM_T_MSI);
1542 
1543         object_property_set_uint(OBJECT(dev), "len-reserved-regions", 1, errp);
1544         object_property_set_str(OBJECT(dev), "reserved-regions[0]",
1545                                 resv_prop_str, errp);
1546         g_free(resv_prop_str);
1547     }
1548 
1549     if (object_dynamic_cast(OBJECT(dev), TYPE_X86_IOMMU_DEVICE) ||
1550         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
1551         PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1552 
1553         if (pcms->iommu) {
1554             error_setg(errp, "QEMU does not support multiple vIOMMUs "
1555                        "for x86 yet.");
1556             return;
1557         }
1558         pcms->iommu = dev;
1559     }
1560 }
1561 
1562 static void pc_machine_device_plug_cb(HotplugHandler *hotplug_dev,
1563                                       DeviceState *dev, Error **errp)
1564 {
1565     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1566         pc_memory_plug(hotplug_dev, dev, errp);
1567     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1568         x86_cpu_plug(hotplug_dev, dev, errp);
1569     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI) ||
1570                object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI)) {
1571         pc_virtio_md_pci_plug(hotplug_dev, dev, errp);
1572     }
1573 }
1574 
1575 static void pc_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
1576                                                 DeviceState *dev, Error **errp)
1577 {
1578     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1579         pc_memory_unplug_request(hotplug_dev, dev, errp);
1580     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1581         x86_cpu_unplug_request_cb(hotplug_dev, dev, errp);
1582     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI) ||
1583                object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI)) {
1584         pc_virtio_md_pci_unplug_request(hotplug_dev, dev, errp);
1585     } else {
1586         error_setg(errp, "acpi: device unplug request for not supported device"
1587                    " type: %s", object_get_typename(OBJECT(dev)));
1588     }
1589 }
1590 
1591 static void pc_machine_device_unplug_cb(HotplugHandler *hotplug_dev,
1592                                         DeviceState *dev, Error **errp)
1593 {
1594     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1595         pc_memory_unplug(hotplug_dev, dev, errp);
1596     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1597         x86_cpu_unplug_cb(hotplug_dev, dev, errp);
1598     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI) ||
1599                object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI)) {
1600         pc_virtio_md_pci_unplug(hotplug_dev, dev, errp);
1601     } else {
1602         error_setg(errp, "acpi: device unplug for not supported device"
1603                    " type: %s", object_get_typename(OBJECT(dev)));
1604     }
1605 }
1606 
1607 static HotplugHandler *pc_get_hotplug_handler(MachineState *machine,
1608                                              DeviceState *dev)
1609 {
1610     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) ||
1611         object_dynamic_cast(OBJECT(dev), TYPE_CPU) ||
1612         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI) ||
1613         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI) ||
1614         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI) ||
1615         object_dynamic_cast(OBJECT(dev), TYPE_X86_IOMMU_DEVICE)) {
1616         return HOTPLUG_HANDLER(machine);
1617     }
1618 
1619     return NULL;
1620 }
1621 
1622 static void
1623 pc_machine_get_device_memory_region_size(Object *obj, Visitor *v,
1624                                          const char *name, void *opaque,
1625                                          Error **errp)
1626 {
1627     MachineState *ms = MACHINE(obj);
1628     int64_t value = 0;
1629 
1630     if (ms->device_memory) {
1631         value = memory_region_size(&ms->device_memory->mr);
1632     }
1633 
1634     visit_type_int(v, name, &value, errp);
1635 }
1636 
1637 static void pc_machine_get_vmport(Object *obj, Visitor *v, const char *name,
1638                                   void *opaque, Error **errp)
1639 {
1640     PCMachineState *pcms = PC_MACHINE(obj);
1641     OnOffAuto vmport = pcms->vmport;
1642 
1643     visit_type_OnOffAuto(v, name, &vmport, errp);
1644 }
1645 
1646 static void pc_machine_set_vmport(Object *obj, Visitor *v, const char *name,
1647                                   void *opaque, Error **errp)
1648 {
1649     PCMachineState *pcms = PC_MACHINE(obj);
1650 
1651     visit_type_OnOffAuto(v, name, &pcms->vmport, errp);
1652 }
1653 
1654 static bool pc_machine_get_smbus(Object *obj, Error **errp)
1655 {
1656     PCMachineState *pcms = PC_MACHINE(obj);
1657 
1658     return pcms->smbus_enabled;
1659 }
1660 
1661 static void pc_machine_set_smbus(Object *obj, bool value, Error **errp)
1662 {
1663     PCMachineState *pcms = PC_MACHINE(obj);
1664 
1665     pcms->smbus_enabled = value;
1666 }
1667 
1668 static bool pc_machine_get_sata(Object *obj, Error **errp)
1669 {
1670     PCMachineState *pcms = PC_MACHINE(obj);
1671 
1672     return pcms->sata_enabled;
1673 }
1674 
1675 static void pc_machine_set_sata(Object *obj, bool value, Error **errp)
1676 {
1677     PCMachineState *pcms = PC_MACHINE(obj);
1678 
1679     pcms->sata_enabled = value;
1680 }
1681 
1682 static bool pc_machine_get_hpet(Object *obj, Error **errp)
1683 {
1684     PCMachineState *pcms = PC_MACHINE(obj);
1685 
1686     return pcms->hpet_enabled;
1687 }
1688 
1689 static void pc_machine_set_hpet(Object *obj, bool value, Error **errp)
1690 {
1691     PCMachineState *pcms = PC_MACHINE(obj);
1692 
1693     pcms->hpet_enabled = value;
1694 }
1695 
1696 static bool pc_machine_get_i8042(Object *obj, Error **errp)
1697 {
1698     PCMachineState *pcms = PC_MACHINE(obj);
1699 
1700     return pcms->i8042_enabled;
1701 }
1702 
1703 static void pc_machine_set_i8042(Object *obj, bool value, Error **errp)
1704 {
1705     PCMachineState *pcms = PC_MACHINE(obj);
1706 
1707     pcms->i8042_enabled = value;
1708 }
1709 
1710 static bool pc_machine_get_default_bus_bypass_iommu(Object *obj, Error **errp)
1711 {
1712     PCMachineState *pcms = PC_MACHINE(obj);
1713 
1714     return pcms->default_bus_bypass_iommu;
1715 }
1716 
1717 static void pc_machine_set_default_bus_bypass_iommu(Object *obj, bool value,
1718                                                     Error **errp)
1719 {
1720     PCMachineState *pcms = PC_MACHINE(obj);
1721 
1722     pcms->default_bus_bypass_iommu = value;
1723 }
1724 
1725 static void pc_machine_get_smbios_ep(Object *obj, Visitor *v, const char *name,
1726                                      void *opaque, Error **errp)
1727 {
1728     PCMachineState *pcms = PC_MACHINE(obj);
1729     SmbiosEntryPointType smbios_entry_point_type = pcms->smbios_entry_point_type;
1730 
1731     visit_type_SmbiosEntryPointType(v, name, &smbios_entry_point_type, errp);
1732 }
1733 
1734 static void pc_machine_set_smbios_ep(Object *obj, Visitor *v, const char *name,
1735                                      void *opaque, Error **errp)
1736 {
1737     PCMachineState *pcms = PC_MACHINE(obj);
1738 
1739     visit_type_SmbiosEntryPointType(v, name, &pcms->smbios_entry_point_type, errp);
1740 }
1741 
1742 static void pc_machine_get_max_ram_below_4g(Object *obj, Visitor *v,
1743                                             const char *name, void *opaque,
1744                                             Error **errp)
1745 {
1746     PCMachineState *pcms = PC_MACHINE(obj);
1747     uint64_t value = pcms->max_ram_below_4g;
1748 
1749     visit_type_size(v, name, &value, errp);
1750 }
1751 
1752 static void pc_machine_set_max_ram_below_4g(Object *obj, Visitor *v,
1753                                             const char *name, void *opaque,
1754                                             Error **errp)
1755 {
1756     PCMachineState *pcms = PC_MACHINE(obj);
1757     uint64_t value;
1758 
1759     if (!visit_type_size(v, name, &value, errp)) {
1760         return;
1761     }
1762     if (value > 4 * GiB) {
1763         error_setg(errp,
1764                    "Machine option 'max-ram-below-4g=%"PRIu64
1765                    "' expects size less than or equal to 4G", value);
1766         return;
1767     }
1768 
1769     if (value < 1 * MiB) {
1770         warn_report("Only %" PRIu64 " bytes of RAM below the 4GiB boundary,"
1771                     "BIOS may not work with less than 1MiB", value);
1772     }
1773 
1774     pcms->max_ram_below_4g = value;
1775 }
1776 
1777 static void pc_machine_get_max_fw_size(Object *obj, Visitor *v,
1778                                        const char *name, void *opaque,
1779                                        Error **errp)
1780 {
1781     PCMachineState *pcms = PC_MACHINE(obj);
1782     uint64_t value = pcms->max_fw_size;
1783 
1784     visit_type_size(v, name, &value, errp);
1785 }
1786 
1787 static void pc_machine_set_max_fw_size(Object *obj, Visitor *v,
1788                                        const char *name, void *opaque,
1789                                        Error **errp)
1790 {
1791     PCMachineState *pcms = PC_MACHINE(obj);
1792     uint64_t value;
1793 
1794     if (!visit_type_size(v, name, &value, errp)) {
1795         return;
1796     }
1797 
1798     /*
1799     * We don't have a theoretically justifiable exact lower bound on the base
1800     * address of any flash mapping. In practice, the IO-APIC MMIO range is
1801     * [0xFEE00000..0xFEE01000] -- see IO_APIC_DEFAULT_ADDRESS --, leaving free
1802     * only 18MB-4KB below 4G. For now, restrict the cumulative mapping to 8MB in
1803     * size.
1804     */
1805     if (value > 16 * MiB) {
1806         error_setg(errp,
1807                    "User specified max allowed firmware size %" PRIu64 " is "
1808                    "greater than 16MiB. If combined firwmare size exceeds "
1809                    "16MiB the system may not boot, or experience intermittent"
1810                    "stability issues.",
1811                    value);
1812         return;
1813     }
1814 
1815     pcms->max_fw_size = value;
1816 }
1817 
1818 
1819 static void pc_machine_initfn(Object *obj)
1820 {
1821     PCMachineState *pcms = PC_MACHINE(obj);
1822 
1823 #ifdef CONFIG_VMPORT
1824     pcms->vmport = ON_OFF_AUTO_AUTO;
1825 #else
1826     pcms->vmport = ON_OFF_AUTO_OFF;
1827 #endif /* CONFIG_VMPORT */
1828     pcms->max_ram_below_4g = 0; /* use default */
1829     pcms->smbios_entry_point_type = SMBIOS_ENTRY_POINT_TYPE_32;
1830 
1831     /* acpi build is enabled by default if machine supports it */
1832     pcms->acpi_build_enabled = PC_MACHINE_GET_CLASS(pcms)->has_acpi_build;
1833     pcms->smbus_enabled = true;
1834     pcms->sata_enabled = true;
1835     pcms->i8042_enabled = true;
1836     pcms->max_fw_size = 8 * MiB;
1837 #ifdef CONFIG_HPET
1838     pcms->hpet_enabled = true;
1839 #endif
1840     pcms->default_bus_bypass_iommu = false;
1841 
1842     pc_system_flash_create(pcms);
1843     pcms->pcspk = isa_new(TYPE_PC_SPEAKER);
1844     object_property_add_alias(OBJECT(pcms), "pcspk-audiodev",
1845                               OBJECT(pcms->pcspk), "audiodev");
1846     cxl_machine_init(obj, &pcms->cxl_devices_state);
1847 }
1848 
1849 static void pc_machine_reset(MachineState *machine, ShutdownCause reason)
1850 {
1851     CPUState *cs;
1852     X86CPU *cpu;
1853 
1854     qemu_devices_reset(reason);
1855 
1856     /* Reset APIC after devices have been reset to cancel
1857      * any changes that qemu_devices_reset() might have done.
1858      */
1859     CPU_FOREACH(cs) {
1860         cpu = X86_CPU(cs);
1861 
1862         x86_cpu_after_reset(cpu);
1863     }
1864 }
1865 
1866 static void pc_machine_wakeup(MachineState *machine)
1867 {
1868     cpu_synchronize_all_states();
1869     pc_machine_reset(machine, SHUTDOWN_CAUSE_NONE);
1870     cpu_synchronize_all_post_reset();
1871 }
1872 
1873 static bool pc_hotplug_allowed(MachineState *ms, DeviceState *dev, Error **errp)
1874 {
1875     X86IOMMUState *iommu = x86_iommu_get_default();
1876     IntelIOMMUState *intel_iommu;
1877 
1878     if (iommu &&
1879         object_dynamic_cast((Object *)iommu, TYPE_INTEL_IOMMU_DEVICE) &&
1880         object_dynamic_cast((Object *)dev, "vfio-pci")) {
1881         intel_iommu = INTEL_IOMMU_DEVICE(iommu);
1882         if (!intel_iommu->caching_mode) {
1883             error_setg(errp, "Device assignment is not allowed without "
1884                        "enabling caching-mode=on for Intel IOMMU.");
1885             return false;
1886         }
1887     }
1888 
1889     return true;
1890 }
1891 
1892 static void pc_machine_class_init(ObjectClass *oc, void *data)
1893 {
1894     MachineClass *mc = MACHINE_CLASS(oc);
1895     PCMachineClass *pcmc = PC_MACHINE_CLASS(oc);
1896     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
1897 
1898     pcmc->pci_enabled = true;
1899     pcmc->has_acpi_build = true;
1900     pcmc->rsdp_in_ram = true;
1901     pcmc->smbios_defaults = true;
1902     pcmc->smbios_uuid_encoded = true;
1903     pcmc->gigabyte_align = true;
1904     pcmc->has_reserved_memory = true;
1905     pcmc->kvmclock_enabled = true;
1906     pcmc->enforce_aligned_dimm = true;
1907     pcmc->enforce_amd_1tb_hole = true;
1908     /* BIOS ACPI tables: 128K. Other BIOS datastructures: less than 4K reported
1909      * to be used at the moment, 32K should be enough for a while.  */
1910     pcmc->acpi_data_size = 0x20000 + 0x8000;
1911     pcmc->pvh_enabled = true;
1912     pcmc->kvmclock_create_always = true;
1913     assert(!mc->get_hotplug_handler);
1914     mc->get_hotplug_handler = pc_get_hotplug_handler;
1915     mc->hotplug_allowed = pc_hotplug_allowed;
1916     mc->cpu_index_to_instance_props = x86_cpu_index_to_props;
1917     mc->get_default_cpu_node_id = x86_get_default_cpu_node_id;
1918     mc->possible_cpu_arch_ids = x86_possible_cpu_arch_ids;
1919     mc->auto_enable_numa_with_memhp = true;
1920     mc->auto_enable_numa_with_memdev = true;
1921     mc->has_hotpluggable_cpus = true;
1922     mc->default_boot_order = "cad";
1923     mc->block_default_type = IF_IDE;
1924     mc->max_cpus = 255;
1925     mc->reset = pc_machine_reset;
1926     mc->wakeup = pc_machine_wakeup;
1927     hc->pre_plug = pc_machine_device_pre_plug_cb;
1928     hc->plug = pc_machine_device_plug_cb;
1929     hc->unplug_request = pc_machine_device_unplug_request_cb;
1930     hc->unplug = pc_machine_device_unplug_cb;
1931     mc->default_cpu_type = TARGET_DEFAULT_CPU_TYPE;
1932     mc->nvdimm_supported = true;
1933     mc->smp_props.dies_supported = true;
1934     mc->default_ram_id = "pc.ram";
1935 
1936     object_class_property_add(oc, PC_MACHINE_MAX_RAM_BELOW_4G, "size",
1937         pc_machine_get_max_ram_below_4g, pc_machine_set_max_ram_below_4g,
1938         NULL, NULL);
1939     object_class_property_set_description(oc, PC_MACHINE_MAX_RAM_BELOW_4G,
1940         "Maximum ram below the 4G boundary (32bit boundary)");
1941 
1942     object_class_property_add(oc, PC_MACHINE_DEVMEM_REGION_SIZE, "int",
1943         pc_machine_get_device_memory_region_size, NULL,
1944         NULL, NULL);
1945 
1946     object_class_property_add(oc, PC_MACHINE_VMPORT, "OnOffAuto",
1947         pc_machine_get_vmport, pc_machine_set_vmport,
1948         NULL, NULL);
1949     object_class_property_set_description(oc, PC_MACHINE_VMPORT,
1950         "Enable vmport (pc & q35)");
1951 
1952     object_class_property_add_bool(oc, PC_MACHINE_SMBUS,
1953         pc_machine_get_smbus, pc_machine_set_smbus);
1954     object_class_property_set_description(oc, PC_MACHINE_SMBUS,
1955         "Enable/disable system management bus");
1956 
1957     object_class_property_add_bool(oc, PC_MACHINE_SATA,
1958         pc_machine_get_sata, pc_machine_set_sata);
1959     object_class_property_set_description(oc, PC_MACHINE_SATA,
1960         "Enable/disable Serial ATA bus");
1961 
1962     object_class_property_add_bool(oc, "hpet",
1963         pc_machine_get_hpet, pc_machine_set_hpet);
1964     object_class_property_set_description(oc, "hpet",
1965         "Enable/disable high precision event timer emulation");
1966 
1967     object_class_property_add_bool(oc, PC_MACHINE_I8042,
1968         pc_machine_get_i8042, pc_machine_set_i8042);
1969 
1970     object_class_property_add_bool(oc, "default-bus-bypass-iommu",
1971         pc_machine_get_default_bus_bypass_iommu,
1972         pc_machine_set_default_bus_bypass_iommu);
1973 
1974     object_class_property_add(oc, PC_MACHINE_MAX_FW_SIZE, "size",
1975         pc_machine_get_max_fw_size, pc_machine_set_max_fw_size,
1976         NULL, NULL);
1977     object_class_property_set_description(oc, PC_MACHINE_MAX_FW_SIZE,
1978         "Maximum combined firmware size");
1979 
1980     object_class_property_add(oc, PC_MACHINE_SMBIOS_EP, "str",
1981         pc_machine_get_smbios_ep, pc_machine_set_smbios_ep,
1982         NULL, NULL);
1983     object_class_property_set_description(oc, PC_MACHINE_SMBIOS_EP,
1984         "SMBIOS Entry Point type [32, 64]");
1985 }
1986 
1987 static const TypeInfo pc_machine_info = {
1988     .name = TYPE_PC_MACHINE,
1989     .parent = TYPE_X86_MACHINE,
1990     .abstract = true,
1991     .instance_size = sizeof(PCMachineState),
1992     .instance_init = pc_machine_initfn,
1993     .class_size = sizeof(PCMachineClass),
1994     .class_init = pc_machine_class_init,
1995     .interfaces = (InterfaceInfo[]) {
1996          { TYPE_HOTPLUG_HANDLER },
1997          { }
1998     },
1999 };
2000 
2001 static void pc_machine_register_types(void)
2002 {
2003     type_register_static(&pc_machine_info);
2004 }
2005 
2006 type_init(pc_machine_register_types)
2007