xref: /openbmc/qemu/hw/i386/pc.c (revision 44602af8)
1 /*
2  * QEMU PC System Emulator
3  *
4  * Copyright (c) 2003-2004 Fabrice Bellard
5  *
6  * Permission is hereby granted, free of charge, to any person obtaining a copy
7  * of this software and associated documentation files (the "Software"), to deal
8  * in the Software without restriction, including without limitation the rights
9  * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
10  * copies of the Software, and to permit persons to whom the Software is
11  * furnished to do so, subject to the following conditions:
12  *
13  * The above copyright notice and this permission notice shall be included in
14  * all copies or substantial portions of the Software.
15  *
16  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
19  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
22  * THE SOFTWARE.
23  */
24 
25 #include "qemu/osdep.h"
26 #include "qemu/units.h"
27 #include "hw/i386/x86.h"
28 #include "hw/i386/pc.h"
29 #include "hw/char/serial.h"
30 #include "hw/char/parallel.h"
31 #include "hw/i386/apic.h"
32 #include "hw/i386/topology.h"
33 #include "hw/i386/fw_cfg.h"
34 #include "hw/i386/vmport.h"
35 #include "sysemu/cpus.h"
36 #include "hw/block/fdc.h"
37 #include "hw/ide.h"
38 #include "hw/pci/pci.h"
39 #include "hw/pci/pci_bus.h"
40 #include "hw/pci-bridge/pci_expander_bridge.h"
41 #include "hw/nvram/fw_cfg.h"
42 #include "hw/timer/hpet.h"
43 #include "hw/firmware/smbios.h"
44 #include "hw/loader.h"
45 #include "elf.h"
46 #include "migration/vmstate.h"
47 #include "multiboot.h"
48 #include "hw/rtc/mc146818rtc.h"
49 #include "hw/intc/i8259.h"
50 #include "hw/dma/i8257.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 "qapi/qmp/qerror.h"
96 #include "e820_memory_layout.h"
97 #include "fw_cfg.h"
98 #include "trace.h"
99 #include CONFIG_DEVICES
100 
101 /*
102  * Helper for setting model-id for CPU models that changed model-id
103  * depending on QEMU versions up to QEMU 2.4.
104  */
105 #define PC_CPU_MODEL_IDS(v) \
106     { "qemu32-" TYPE_X86_CPU, "model-id", "QEMU Virtual CPU version " v, },\
107     { "qemu64-" TYPE_X86_CPU, "model-id", "QEMU Virtual CPU version " v, },\
108     { "athlon-" TYPE_X86_CPU, "model-id", "QEMU Virtual CPU version " v, },
109 
110 GlobalProperty pc_compat_7_0[] = {};
111 const size_t pc_compat_7_0_len = G_N_ELEMENTS(pc_compat_7_0);
112 
113 GlobalProperty pc_compat_6_2[] = {
114     { "virtio-mem", "unplugged-inaccessible", "off" },
115 };
116 const size_t pc_compat_6_2_len = G_N_ELEMENTS(pc_compat_6_2);
117 
118 GlobalProperty pc_compat_6_1[] = {
119     { TYPE_X86_CPU, "hv-version-id-build", "0x1bbc" },
120     { TYPE_X86_CPU, "hv-version-id-major", "0x0006" },
121     { TYPE_X86_CPU, "hv-version-id-minor", "0x0001" },
122     { "ICH9-LPC", "x-keep-pci-slot-hpc", "false" },
123 };
124 const size_t pc_compat_6_1_len = G_N_ELEMENTS(pc_compat_6_1);
125 
126 GlobalProperty pc_compat_6_0[] = {
127     { "qemu64" "-" TYPE_X86_CPU, "family", "6" },
128     { "qemu64" "-" TYPE_X86_CPU, "model", "6" },
129     { "qemu64" "-" TYPE_X86_CPU, "stepping", "3" },
130     { TYPE_X86_CPU, "x-vendor-cpuid-only", "off" },
131     { "ICH9-LPC", ACPI_PM_PROP_ACPI_PCIHP_BRIDGE, "off" },
132     { "ICH9-LPC", "x-keep-pci-slot-hpc", "true" },
133 };
134 const size_t pc_compat_6_0_len = G_N_ELEMENTS(pc_compat_6_0);
135 
136 GlobalProperty pc_compat_5_2[] = {
137     { "ICH9-LPC", "x-smi-cpu-hotunplug", "off" },
138 };
139 const size_t pc_compat_5_2_len = G_N_ELEMENTS(pc_compat_5_2);
140 
141 GlobalProperty pc_compat_5_1[] = {
142     { "ICH9-LPC", "x-smi-cpu-hotplug", "off" },
143     { TYPE_X86_CPU, "kvm-msi-ext-dest-id", "off" },
144 };
145 const size_t pc_compat_5_1_len = G_N_ELEMENTS(pc_compat_5_1);
146 
147 GlobalProperty pc_compat_5_0[] = {
148 };
149 const size_t pc_compat_5_0_len = G_N_ELEMENTS(pc_compat_5_0);
150 
151 GlobalProperty pc_compat_4_2[] = {
152     { "mch", "smbase-smram", "off" },
153 };
154 const size_t pc_compat_4_2_len = G_N_ELEMENTS(pc_compat_4_2);
155 
156 GlobalProperty pc_compat_4_1[] = {};
157 const size_t pc_compat_4_1_len = G_N_ELEMENTS(pc_compat_4_1);
158 
159 GlobalProperty pc_compat_4_0[] = {};
160 const size_t pc_compat_4_0_len = G_N_ELEMENTS(pc_compat_4_0);
161 
162 GlobalProperty pc_compat_3_1[] = {
163     { "intel-iommu", "dma-drain", "off" },
164     { "Opteron_G3" "-" TYPE_X86_CPU, "rdtscp", "off" },
165     { "Opteron_G4" "-" TYPE_X86_CPU, "rdtscp", "off" },
166     { "Opteron_G4" "-" TYPE_X86_CPU, "npt", "off" },
167     { "Opteron_G4" "-" TYPE_X86_CPU, "nrip-save", "off" },
168     { "Opteron_G5" "-" TYPE_X86_CPU, "rdtscp", "off" },
169     { "Opteron_G5" "-" TYPE_X86_CPU, "npt", "off" },
170     { "Opteron_G5" "-" TYPE_X86_CPU, "nrip-save", "off" },
171     { "EPYC" "-" TYPE_X86_CPU, "npt", "off" },
172     { "EPYC" "-" TYPE_X86_CPU, "nrip-save", "off" },
173     { "EPYC-IBPB" "-" TYPE_X86_CPU, "npt", "off" },
174     { "EPYC-IBPB" "-" TYPE_X86_CPU, "nrip-save", "off" },
175     { "Skylake-Client" "-" TYPE_X86_CPU,      "mpx", "on" },
176     { "Skylake-Client-IBRS" "-" TYPE_X86_CPU, "mpx", "on" },
177     { "Skylake-Server" "-" TYPE_X86_CPU,      "mpx", "on" },
178     { "Skylake-Server-IBRS" "-" TYPE_X86_CPU, "mpx", "on" },
179     { "Cascadelake-Server" "-" TYPE_X86_CPU,  "mpx", "on" },
180     { "Icelake-Client" "-" TYPE_X86_CPU,      "mpx", "on" },
181     { "Icelake-Server" "-" TYPE_X86_CPU,      "mpx", "on" },
182     { "Cascadelake-Server" "-" TYPE_X86_CPU, "stepping", "5" },
183     { TYPE_X86_CPU, "x-intel-pt-auto-level", "off" },
184 };
185 const size_t pc_compat_3_1_len = G_N_ELEMENTS(pc_compat_3_1);
186 
187 GlobalProperty pc_compat_3_0[] = {
188     { TYPE_X86_CPU, "x-hv-synic-kvm-only", "on" },
189     { "Skylake-Server" "-" TYPE_X86_CPU, "pku", "off" },
190     { "Skylake-Server-IBRS" "-" TYPE_X86_CPU, "pku", "off" },
191 };
192 const size_t pc_compat_3_0_len = G_N_ELEMENTS(pc_compat_3_0);
193 
194 GlobalProperty pc_compat_2_12[] = {
195     { TYPE_X86_CPU, "legacy-cache", "on" },
196     { TYPE_X86_CPU, "topoext", "off" },
197     { "EPYC-" TYPE_X86_CPU, "xlevel", "0x8000000a" },
198     { "EPYC-IBPB-" TYPE_X86_CPU, "xlevel", "0x8000000a" },
199 };
200 const size_t pc_compat_2_12_len = G_N_ELEMENTS(pc_compat_2_12);
201 
202 GlobalProperty pc_compat_2_11[] = {
203     { TYPE_X86_CPU, "x-migrate-smi-count", "off" },
204     { "Skylake-Server" "-" TYPE_X86_CPU, "clflushopt", "off" },
205 };
206 const size_t pc_compat_2_11_len = G_N_ELEMENTS(pc_compat_2_11);
207 
208 GlobalProperty pc_compat_2_10[] = {
209     { TYPE_X86_CPU, "x-hv-max-vps", "0x40" },
210     { "i440FX-pcihost", "x-pci-hole64-fix", "off" },
211     { "q35-pcihost", "x-pci-hole64-fix", "off" },
212 };
213 const size_t pc_compat_2_10_len = G_N_ELEMENTS(pc_compat_2_10);
214 
215 GlobalProperty pc_compat_2_9[] = {
216     { "mch", "extended-tseg-mbytes", "0" },
217 };
218 const size_t pc_compat_2_9_len = G_N_ELEMENTS(pc_compat_2_9);
219 
220 GlobalProperty pc_compat_2_8[] = {
221     { TYPE_X86_CPU, "tcg-cpuid", "off" },
222     { "kvmclock", "x-mach-use-reliable-get-clock", "off" },
223     { "ICH9-LPC", "x-smi-broadcast", "off" },
224     { TYPE_X86_CPU, "vmware-cpuid-freq", "off" },
225     { "Haswell-" TYPE_X86_CPU, "stepping", "1" },
226 };
227 const size_t pc_compat_2_8_len = G_N_ELEMENTS(pc_compat_2_8);
228 
229 GlobalProperty pc_compat_2_7[] = {
230     { TYPE_X86_CPU, "l3-cache", "off" },
231     { TYPE_X86_CPU, "full-cpuid-auto-level", "off" },
232     { "Opteron_G3" "-" TYPE_X86_CPU, "family", "15" },
233     { "Opteron_G3" "-" TYPE_X86_CPU, "model", "6" },
234     { "Opteron_G3" "-" TYPE_X86_CPU, "stepping", "1" },
235     { "isa-pcspk", "migrate", "off" },
236 };
237 const size_t pc_compat_2_7_len = G_N_ELEMENTS(pc_compat_2_7);
238 
239 GlobalProperty pc_compat_2_6[] = {
240     { TYPE_X86_CPU, "cpuid-0xb", "off" },
241     { "vmxnet3", "romfile", "" },
242     { TYPE_X86_CPU, "fill-mtrr-mask", "off" },
243     { "apic-common", "legacy-instance-id", "on", }
244 };
245 const size_t pc_compat_2_6_len = G_N_ELEMENTS(pc_compat_2_6);
246 
247 GlobalProperty pc_compat_2_5[] = {};
248 const size_t pc_compat_2_5_len = G_N_ELEMENTS(pc_compat_2_5);
249 
250 GlobalProperty pc_compat_2_4[] = {
251     PC_CPU_MODEL_IDS("2.4.0")
252     { "Haswell-" TYPE_X86_CPU, "abm", "off" },
253     { "Haswell-noTSX-" TYPE_X86_CPU, "abm", "off" },
254     { "Broadwell-" TYPE_X86_CPU, "abm", "off" },
255     { "Broadwell-noTSX-" TYPE_X86_CPU, "abm", "off" },
256     { "host" "-" TYPE_X86_CPU, "host-cache-info", "on" },
257     { TYPE_X86_CPU, "check", "off" },
258     { "qemu64" "-" TYPE_X86_CPU, "sse4a", "on" },
259     { "qemu64" "-" TYPE_X86_CPU, "abm", "on" },
260     { "qemu64" "-" TYPE_X86_CPU, "popcnt", "on" },
261     { "qemu32" "-" TYPE_X86_CPU, "popcnt", "on" },
262     { "Opteron_G2" "-" TYPE_X86_CPU, "rdtscp", "on" },
263     { "Opteron_G3" "-" TYPE_X86_CPU, "rdtscp", "on" },
264     { "Opteron_G4" "-" TYPE_X86_CPU, "rdtscp", "on" },
265     { "Opteron_G5" "-" TYPE_X86_CPU, "rdtscp", "on", }
266 };
267 const size_t pc_compat_2_4_len = G_N_ELEMENTS(pc_compat_2_4);
268 
269 GlobalProperty pc_compat_2_3[] = {
270     PC_CPU_MODEL_IDS("2.3.0")
271     { TYPE_X86_CPU, "arat", "off" },
272     { "qemu64" "-" TYPE_X86_CPU, "min-level", "4" },
273     { "kvm64" "-" TYPE_X86_CPU, "min-level", "5" },
274     { "pentium3" "-" TYPE_X86_CPU, "min-level", "2" },
275     { "n270" "-" TYPE_X86_CPU, "min-level", "5" },
276     { "Conroe" "-" TYPE_X86_CPU, "min-level", "4" },
277     { "Penryn" "-" TYPE_X86_CPU, "min-level", "4" },
278     { "Nehalem" "-" TYPE_X86_CPU, "min-level", "4" },
279     { "n270" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
280     { "Penryn" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
281     { "Conroe" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
282     { "Nehalem" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
283     { "Westmere" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
284     { "SandyBridge" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
285     { "IvyBridge" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
286     { "Haswell" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
287     { "Haswell-noTSX" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
288     { "Broadwell" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
289     { "Broadwell-noTSX" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
290     { TYPE_X86_CPU, "kvm-no-smi-migration", "on" },
291 };
292 const size_t pc_compat_2_3_len = G_N_ELEMENTS(pc_compat_2_3);
293 
294 GlobalProperty pc_compat_2_2[] = {
295     PC_CPU_MODEL_IDS("2.2.0")
296     { "kvm64" "-" TYPE_X86_CPU, "vme", "off" },
297     { "kvm32" "-" TYPE_X86_CPU, "vme", "off" },
298     { "Conroe" "-" TYPE_X86_CPU, "vme", "off" },
299     { "Penryn" "-" TYPE_X86_CPU, "vme", "off" },
300     { "Nehalem" "-" TYPE_X86_CPU, "vme", "off" },
301     { "Westmere" "-" TYPE_X86_CPU, "vme", "off" },
302     { "SandyBridge" "-" TYPE_X86_CPU, "vme", "off" },
303     { "Haswell" "-" TYPE_X86_CPU, "vme", "off" },
304     { "Broadwell" "-" TYPE_X86_CPU, "vme", "off" },
305     { "Opteron_G1" "-" TYPE_X86_CPU, "vme", "off" },
306     { "Opteron_G2" "-" TYPE_X86_CPU, "vme", "off" },
307     { "Opteron_G3" "-" TYPE_X86_CPU, "vme", "off" },
308     { "Opteron_G4" "-" TYPE_X86_CPU, "vme", "off" },
309     { "Opteron_G5" "-" TYPE_X86_CPU, "vme", "off" },
310     { "Haswell" "-" TYPE_X86_CPU, "f16c", "off" },
311     { "Haswell" "-" TYPE_X86_CPU, "rdrand", "off" },
312     { "Broadwell" "-" TYPE_X86_CPU, "f16c", "off" },
313     { "Broadwell" "-" TYPE_X86_CPU, "rdrand", "off" },
314 };
315 const size_t pc_compat_2_2_len = G_N_ELEMENTS(pc_compat_2_2);
316 
317 GlobalProperty pc_compat_2_1[] = {
318     PC_CPU_MODEL_IDS("2.1.0")
319     { "coreduo" "-" TYPE_X86_CPU, "vmx", "on" },
320     { "core2duo" "-" TYPE_X86_CPU, "vmx", "on" },
321 };
322 const size_t pc_compat_2_1_len = G_N_ELEMENTS(pc_compat_2_1);
323 
324 GlobalProperty pc_compat_2_0[] = {
325     PC_CPU_MODEL_IDS("2.0.0")
326     { "virtio-scsi-pci", "any_layout", "off" },
327     { "PIIX4_PM", "memory-hotplug-support", "off" },
328     { "apic", "version", "0x11" },
329     { "nec-usb-xhci", "superspeed-ports-first", "off" },
330     { "nec-usb-xhci", "force-pcie-endcap", "on" },
331     { "pci-serial", "prog_if", "0" },
332     { "pci-serial-2x", "prog_if", "0" },
333     { "pci-serial-4x", "prog_if", "0" },
334     { "virtio-net-pci", "guest_announce", "off" },
335     { "ICH9-LPC", "memory-hotplug-support", "off" },
336 };
337 const size_t pc_compat_2_0_len = G_N_ELEMENTS(pc_compat_2_0);
338 
339 GlobalProperty pc_compat_1_7[] = {
340     PC_CPU_MODEL_IDS("1.7.0")
341     { TYPE_USB_DEVICE, "msos-desc", "no" },
342     { "PIIX4_PM", ACPI_PM_PROP_ACPI_PCIHP_BRIDGE, "off" },
343     { "hpet", HPET_INTCAP, "4" },
344 };
345 const size_t pc_compat_1_7_len = G_N_ELEMENTS(pc_compat_1_7);
346 
347 GlobalProperty pc_compat_1_6[] = {
348     PC_CPU_MODEL_IDS("1.6.0")
349     { "e1000", "mitigation", "off" },
350     { "qemu64-" TYPE_X86_CPU, "model", "2" },
351     { "qemu32-" TYPE_X86_CPU, "model", "3" },
352     { "i440FX-pcihost", "short_root_bus", "1" },
353     { "q35-pcihost", "short_root_bus", "1" },
354 };
355 const size_t pc_compat_1_6_len = G_N_ELEMENTS(pc_compat_1_6);
356 
357 GlobalProperty pc_compat_1_5[] = {
358     PC_CPU_MODEL_IDS("1.5.0")
359     { "Conroe-" TYPE_X86_CPU, "model", "2" },
360     { "Conroe-" TYPE_X86_CPU, "min-level", "2" },
361     { "Penryn-" TYPE_X86_CPU, "model", "2" },
362     { "Penryn-" TYPE_X86_CPU, "min-level", "2" },
363     { "Nehalem-" TYPE_X86_CPU, "model", "2" },
364     { "Nehalem-" TYPE_X86_CPU, "min-level", "2" },
365     { "virtio-net-pci", "any_layout", "off" },
366     { TYPE_X86_CPU, "pmu", "on" },
367     { "i440FX-pcihost", "short_root_bus", "0" },
368     { "q35-pcihost", "short_root_bus", "0" },
369 };
370 const size_t pc_compat_1_5_len = G_N_ELEMENTS(pc_compat_1_5);
371 
372 GlobalProperty pc_compat_1_4[] = {
373     PC_CPU_MODEL_IDS("1.4.0")
374     { "scsi-hd", "discard_granularity", "0" },
375     { "scsi-cd", "discard_granularity", "0" },
376     { "ide-hd", "discard_granularity", "0" },
377     { "ide-cd", "discard_granularity", "0" },
378     { "virtio-blk-pci", "discard_granularity", "0" },
379     /* DEV_NVECTORS_UNSPECIFIED as a uint32_t string: */
380     { "virtio-serial-pci", "vectors", "0xFFFFFFFF" },
381     { "virtio-net-pci", "ctrl_guest_offloads", "off" },
382     { "e1000", "romfile", "pxe-e1000.rom" },
383     { "ne2k_pci", "romfile", "pxe-ne2k_pci.rom" },
384     { "pcnet", "romfile", "pxe-pcnet.rom" },
385     { "rtl8139", "romfile", "pxe-rtl8139.rom" },
386     { "virtio-net-pci", "romfile", "pxe-virtio.rom" },
387     { "486-" TYPE_X86_CPU, "model", "0" },
388     { "n270" "-" TYPE_X86_CPU, "movbe", "off" },
389     { "Westmere" "-" TYPE_X86_CPU, "pclmulqdq", "off" },
390 };
391 const size_t pc_compat_1_4_len = G_N_ELEMENTS(pc_compat_1_4);
392 
393 GSIState *pc_gsi_create(qemu_irq **irqs, bool pci_enabled)
394 {
395     GSIState *s;
396 
397     s = g_new0(GSIState, 1);
398     if (kvm_ioapic_in_kernel()) {
399         kvm_pc_setup_irq_routing(pci_enabled);
400     }
401     *irqs = qemu_allocate_irqs(gsi_handler, s, GSI_NUM_PINS);
402 
403     return s;
404 }
405 
406 static void ioport80_write(void *opaque, hwaddr addr, uint64_t data,
407                            unsigned size)
408 {
409 }
410 
411 static uint64_t ioport80_read(void *opaque, hwaddr addr, unsigned size)
412 {
413     return 0xffffffffffffffffULL;
414 }
415 
416 /* MSDOS compatibility mode FPU exception support */
417 static void ioportF0_write(void *opaque, hwaddr addr, uint64_t data,
418                            unsigned size)
419 {
420     if (tcg_enabled()) {
421         cpu_set_ignne();
422     }
423 }
424 
425 static uint64_t ioportF0_read(void *opaque, hwaddr addr, unsigned size)
426 {
427     return 0xffffffffffffffffULL;
428 }
429 
430 /* PC cmos mappings */
431 
432 #define REG_EQUIPMENT_BYTE          0x14
433 
434 static void cmos_init_hd(ISADevice *s, int type_ofs, int info_ofs,
435                          int16_t cylinders, int8_t heads, int8_t sectors)
436 {
437     rtc_set_memory(s, type_ofs, 47);
438     rtc_set_memory(s, info_ofs, cylinders);
439     rtc_set_memory(s, info_ofs + 1, cylinders >> 8);
440     rtc_set_memory(s, info_ofs + 2, heads);
441     rtc_set_memory(s, info_ofs + 3, 0xff);
442     rtc_set_memory(s, info_ofs + 4, 0xff);
443     rtc_set_memory(s, info_ofs + 5, 0xc0 | ((heads > 8) << 3));
444     rtc_set_memory(s, info_ofs + 6, cylinders);
445     rtc_set_memory(s, info_ofs + 7, cylinders >> 8);
446     rtc_set_memory(s, info_ofs + 8, sectors);
447 }
448 
449 /* convert boot_device letter to something recognizable by the bios */
450 static int boot_device2nibble(char boot_device)
451 {
452     switch(boot_device) {
453     case 'a':
454     case 'b':
455         return 0x01; /* floppy boot */
456     case 'c':
457         return 0x02; /* hard drive boot */
458     case 'd':
459         return 0x03; /* CD-ROM boot */
460     case 'n':
461         return 0x04; /* Network boot */
462     }
463     return 0;
464 }
465 
466 static void set_boot_dev(ISADevice *s, const char *boot_device, Error **errp)
467 {
468 #define PC_MAX_BOOT_DEVICES 3
469     int nbds, bds[3] = { 0, };
470     int i;
471 
472     nbds = strlen(boot_device);
473     if (nbds > PC_MAX_BOOT_DEVICES) {
474         error_setg(errp, "Too many boot devices for PC");
475         return;
476     }
477     for (i = 0; i < nbds; i++) {
478         bds[i] = boot_device2nibble(boot_device[i]);
479         if (bds[i] == 0) {
480             error_setg(errp, "Invalid boot device for PC: '%c'",
481                        boot_device[i]);
482             return;
483         }
484     }
485     rtc_set_memory(s, 0x3d, (bds[1] << 4) | bds[0]);
486     rtc_set_memory(s, 0x38, (bds[2] << 4) | (fd_bootchk ? 0x0 : 0x1));
487 }
488 
489 static void pc_boot_set(void *opaque, const char *boot_device, Error **errp)
490 {
491     set_boot_dev(opaque, boot_device, errp);
492 }
493 
494 static void pc_cmos_init_floppy(ISADevice *rtc_state, ISADevice *floppy)
495 {
496     int val, nb, i;
497     FloppyDriveType fd_type[2] = { FLOPPY_DRIVE_TYPE_NONE,
498                                    FLOPPY_DRIVE_TYPE_NONE };
499 
500     /* floppy type */
501     if (floppy) {
502         for (i = 0; i < 2; i++) {
503             fd_type[i] = isa_fdc_get_drive_type(floppy, i);
504         }
505     }
506     val = (cmos_get_fd_drive_type(fd_type[0]) << 4) |
507         cmos_get_fd_drive_type(fd_type[1]);
508     rtc_set_memory(rtc_state, 0x10, val);
509 
510     val = rtc_get_memory(rtc_state, REG_EQUIPMENT_BYTE);
511     nb = 0;
512     if (fd_type[0] != FLOPPY_DRIVE_TYPE_NONE) {
513         nb++;
514     }
515     if (fd_type[1] != FLOPPY_DRIVE_TYPE_NONE) {
516         nb++;
517     }
518     switch (nb) {
519     case 0:
520         break;
521     case 1:
522         val |= 0x01; /* 1 drive, ready for boot */
523         break;
524     case 2:
525         val |= 0x41; /* 2 drives, ready for boot */
526         break;
527     }
528     rtc_set_memory(rtc_state, REG_EQUIPMENT_BYTE, val);
529 }
530 
531 typedef struct pc_cmos_init_late_arg {
532     ISADevice *rtc_state;
533     BusState *idebus[2];
534 } pc_cmos_init_late_arg;
535 
536 typedef struct check_fdc_state {
537     ISADevice *floppy;
538     bool multiple;
539 } CheckFdcState;
540 
541 static int check_fdc(Object *obj, void *opaque)
542 {
543     CheckFdcState *state = opaque;
544     Object *fdc;
545     uint32_t iobase;
546     Error *local_err = NULL;
547 
548     fdc = object_dynamic_cast(obj, TYPE_ISA_FDC);
549     if (!fdc) {
550         return 0;
551     }
552 
553     iobase = object_property_get_uint(obj, "iobase", &local_err);
554     if (local_err || iobase != 0x3f0) {
555         error_free(local_err);
556         return 0;
557     }
558 
559     if (state->floppy) {
560         state->multiple = true;
561     } else {
562         state->floppy = ISA_DEVICE(obj);
563     }
564     return 0;
565 }
566 
567 static const char * const fdc_container_path[] = {
568     "/unattached", "/peripheral", "/peripheral-anon"
569 };
570 
571 /*
572  * Locate the FDC at IO address 0x3f0, in order to configure the CMOS registers
573  * and ACPI objects.
574  */
575 static ISADevice *pc_find_fdc0(void)
576 {
577     int i;
578     Object *container;
579     CheckFdcState state = { 0 };
580 
581     for (i = 0; i < ARRAY_SIZE(fdc_container_path); i++) {
582         container = container_get(qdev_get_machine(), fdc_container_path[i]);
583         object_child_foreach(container, check_fdc, &state);
584     }
585 
586     if (state.multiple) {
587         warn_report("multiple floppy disk controllers with "
588                     "iobase=0x3f0 have been found");
589         error_printf("the one being picked for CMOS setup might not reflect "
590                      "your intent");
591     }
592 
593     return state.floppy;
594 }
595 
596 static void pc_cmos_init_late(void *opaque)
597 {
598     pc_cmos_init_late_arg *arg = opaque;
599     ISADevice *s = arg->rtc_state;
600     int16_t cylinders;
601     int8_t heads, sectors;
602     int val;
603     int i, trans;
604 
605     val = 0;
606     if (arg->idebus[0] && ide_get_geometry(arg->idebus[0], 0,
607                                            &cylinders, &heads, &sectors) >= 0) {
608         cmos_init_hd(s, 0x19, 0x1b, cylinders, heads, sectors);
609         val |= 0xf0;
610     }
611     if (arg->idebus[0] && ide_get_geometry(arg->idebus[0], 1,
612                                            &cylinders, &heads, &sectors) >= 0) {
613         cmos_init_hd(s, 0x1a, 0x24, cylinders, heads, sectors);
614         val |= 0x0f;
615     }
616     rtc_set_memory(s, 0x12, val);
617 
618     val = 0;
619     for (i = 0; i < 4; i++) {
620         /* NOTE: ide_get_geometry() returns the physical
621            geometry.  It is always such that: 1 <= sects <= 63, 1
622            <= heads <= 16, 1 <= cylinders <= 16383. The BIOS
623            geometry can be different if a translation is done. */
624         if (arg->idebus[i / 2] &&
625             ide_get_geometry(arg->idebus[i / 2], i % 2,
626                              &cylinders, &heads, &sectors) >= 0) {
627             trans = ide_get_bios_chs_trans(arg->idebus[i / 2], i % 2) - 1;
628             assert((trans & ~3) == 0);
629             val |= trans << (i * 2);
630         }
631     }
632     rtc_set_memory(s, 0x39, val);
633 
634     pc_cmos_init_floppy(s, pc_find_fdc0());
635 
636     qemu_unregister_reset(pc_cmos_init_late, opaque);
637 }
638 
639 void pc_cmos_init(PCMachineState *pcms,
640                   BusState *idebus0, BusState *idebus1,
641                   ISADevice *s)
642 {
643     int val;
644     static pc_cmos_init_late_arg arg;
645     X86MachineState *x86ms = X86_MACHINE(pcms);
646 
647     /* various important CMOS locations needed by PC/Bochs bios */
648 
649     /* memory size */
650     /* base memory (first MiB) */
651     val = MIN(x86ms->below_4g_mem_size / KiB, 640);
652     rtc_set_memory(s, 0x15, val);
653     rtc_set_memory(s, 0x16, val >> 8);
654     /* extended memory (next 64MiB) */
655     if (x86ms->below_4g_mem_size > 1 * MiB) {
656         val = (x86ms->below_4g_mem_size - 1 * MiB) / KiB;
657     } else {
658         val = 0;
659     }
660     if (val > 65535)
661         val = 65535;
662     rtc_set_memory(s, 0x17, val);
663     rtc_set_memory(s, 0x18, val >> 8);
664     rtc_set_memory(s, 0x30, val);
665     rtc_set_memory(s, 0x31, val >> 8);
666     /* memory between 16MiB and 4GiB */
667     if (x86ms->below_4g_mem_size > 16 * MiB) {
668         val = (x86ms->below_4g_mem_size - 16 * MiB) / (64 * KiB);
669     } else {
670         val = 0;
671     }
672     if (val > 65535)
673         val = 65535;
674     rtc_set_memory(s, 0x34, val);
675     rtc_set_memory(s, 0x35, val >> 8);
676     /* memory above 4GiB */
677     val = x86ms->above_4g_mem_size / 65536;
678     rtc_set_memory(s, 0x5b, val);
679     rtc_set_memory(s, 0x5c, val >> 8);
680     rtc_set_memory(s, 0x5d, val >> 16);
681 
682     object_property_add_link(OBJECT(pcms), "rtc_state",
683                              TYPE_ISA_DEVICE,
684                              (Object **)&x86ms->rtc,
685                              object_property_allow_set_link,
686                              OBJ_PROP_LINK_STRONG);
687     object_property_set_link(OBJECT(pcms), "rtc_state", OBJECT(s),
688                              &error_abort);
689 
690     set_boot_dev(s, MACHINE(pcms)->boot_config.order, &error_fatal);
691 
692     val = 0;
693     val |= 0x02; /* FPU is there */
694     val |= 0x04; /* PS/2 mouse installed */
695     rtc_set_memory(s, REG_EQUIPMENT_BYTE, val);
696 
697     /* hard drives and FDC */
698     arg.rtc_state = s;
699     arg.idebus[0] = idebus0;
700     arg.idebus[1] = idebus1;
701     qemu_register_reset(pc_cmos_init_late, &arg);
702 }
703 
704 static void handle_a20_line_change(void *opaque, int irq, int level)
705 {
706     X86CPU *cpu = opaque;
707 
708     /* XXX: send to all CPUs ? */
709     /* XXX: add logic to handle multiple A20 line sources */
710     x86_cpu_set_a20(cpu, level);
711 }
712 
713 #define NE2000_NB_MAX 6
714 
715 static const int ne2000_io[NE2000_NB_MAX] = { 0x300, 0x320, 0x340, 0x360,
716                                               0x280, 0x380 };
717 static const int ne2000_irq[NE2000_NB_MAX] = { 9, 10, 11, 3, 4, 5 };
718 
719 static void pc_init_ne2k_isa(ISABus *bus, NICInfo *nd)
720 {
721     static int nb_ne2k = 0;
722 
723     if (nb_ne2k == NE2000_NB_MAX)
724         return;
725     isa_ne2000_init(bus, ne2000_io[nb_ne2k],
726                     ne2000_irq[nb_ne2k], nd);
727     nb_ne2k++;
728 }
729 
730 void pc_acpi_smi_interrupt(void *opaque, int irq, int level)
731 {
732     X86CPU *cpu = opaque;
733 
734     if (level) {
735         cpu_interrupt(CPU(cpu), CPU_INTERRUPT_SMI);
736     }
737 }
738 
739 static
740 void pc_machine_done(Notifier *notifier, void *data)
741 {
742     PCMachineState *pcms = container_of(notifier,
743                                         PCMachineState, machine_done);
744     X86MachineState *x86ms = X86_MACHINE(pcms);
745 
746     cxl_hook_up_pxb_registers(pcms->bus, &pcms->cxl_devices_state,
747                               &error_fatal);
748 
749     if (pcms->cxl_devices_state.is_enabled) {
750         cxl_fmws_link_targets(&pcms->cxl_devices_state, &error_fatal);
751     }
752 
753     /* set the number of CPUs */
754     x86_rtc_set_cpus_count(x86ms->rtc, x86ms->boot_cpus);
755 
756     fw_cfg_add_extra_pci_roots(pcms->bus, x86ms->fw_cfg);
757 
758     acpi_setup();
759     if (x86ms->fw_cfg) {
760         fw_cfg_build_smbios(MACHINE(pcms), x86ms->fw_cfg);
761         fw_cfg_build_feature_control(MACHINE(pcms), x86ms->fw_cfg);
762         /* update FW_CFG_NB_CPUS to account for -device added CPUs */
763         fw_cfg_modify_i16(x86ms->fw_cfg, FW_CFG_NB_CPUS, x86ms->boot_cpus);
764     }
765 }
766 
767 void pc_guest_info_init(PCMachineState *pcms)
768 {
769     X86MachineState *x86ms = X86_MACHINE(pcms);
770 
771     x86ms->apic_xrupt_override = true;
772     pcms->machine_done.notify = pc_machine_done;
773     qemu_add_machine_init_done_notifier(&pcms->machine_done);
774 }
775 
776 /* setup pci memory address space mapping into system address space */
777 void pc_pci_as_mapping_init(Object *owner, MemoryRegion *system_memory,
778                             MemoryRegion *pci_address_space)
779 {
780     /* Set to lower priority than RAM */
781     memory_region_add_subregion_overlap(system_memory, 0x0,
782                                         pci_address_space, -1);
783 }
784 
785 void xen_load_linux(PCMachineState *pcms)
786 {
787     int i;
788     FWCfgState *fw_cfg;
789     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
790     X86MachineState *x86ms = X86_MACHINE(pcms);
791 
792     assert(MACHINE(pcms)->kernel_filename != NULL);
793 
794     fw_cfg = fw_cfg_init_io(FW_CFG_IO_BASE);
795     fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, x86ms->boot_cpus);
796     rom_set_fw(fw_cfg);
797 
798     x86_load_linux(x86ms, fw_cfg, pcmc->acpi_data_size,
799                    pcmc->pvh_enabled, pcmc->legacy_no_rng_seed);
800     for (i = 0; i < nb_option_roms; i++) {
801         assert(!strcmp(option_rom[i].name, "linuxboot.bin") ||
802                !strcmp(option_rom[i].name, "linuxboot_dma.bin") ||
803                !strcmp(option_rom[i].name, "pvh.bin") ||
804                !strcmp(option_rom[i].name, "multiboot.bin") ||
805                !strcmp(option_rom[i].name, "multiboot_dma.bin"));
806         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
807     }
808     x86ms->fw_cfg = fw_cfg;
809 }
810 
811 #define PC_ROM_MIN_VGA     0xc0000
812 #define PC_ROM_MIN_OPTION  0xc8000
813 #define PC_ROM_MAX         0xe0000
814 #define PC_ROM_ALIGN       0x800
815 #define PC_ROM_SIZE        (PC_ROM_MAX - PC_ROM_MIN_VGA)
816 
817 static hwaddr pc_above_4g_end(PCMachineState *pcms)
818 {
819     X86MachineState *x86ms = X86_MACHINE(pcms);
820 
821     if (pcms->sgx_epc.size != 0) {
822         return sgx_epc_above_4g_end(&pcms->sgx_epc);
823     }
824 
825     return x86ms->above_4g_mem_start + x86ms->above_4g_mem_size;
826 }
827 
828 static void pc_get_device_memory_range(PCMachineState *pcms,
829                                        hwaddr *base,
830                                        ram_addr_t *device_mem_size)
831 {
832     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
833     MachineState *machine = MACHINE(pcms);
834     ram_addr_t size;
835     hwaddr addr;
836 
837     size = machine->maxram_size - machine->ram_size;
838     addr = ROUND_UP(pc_above_4g_end(pcms), 1 * GiB);
839 
840     if (pcmc->enforce_aligned_dimm) {
841         /* size device region assuming 1G page max alignment per slot */
842         size += (1 * GiB) * machine->ram_slots;
843     }
844 
845     *base = addr;
846     *device_mem_size = size;
847 }
848 
849 static uint64_t pc_get_cxl_range_start(PCMachineState *pcms)
850 {
851     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
852     hwaddr cxl_base;
853     ram_addr_t size;
854 
855     if (pcmc->has_reserved_memory) {
856         pc_get_device_memory_range(pcms, &cxl_base, &size);
857         cxl_base += size;
858     } else {
859         cxl_base = pc_above_4g_end(pcms);
860     }
861 
862     return cxl_base;
863 }
864 
865 static uint64_t pc_get_cxl_range_end(PCMachineState *pcms)
866 {
867     uint64_t start = pc_get_cxl_range_start(pcms) + MiB;
868 
869     if (pcms->cxl_devices_state.fixed_windows) {
870         GList *it;
871 
872         start = ROUND_UP(start, 256 * MiB);
873         for (it = pcms->cxl_devices_state.fixed_windows; it; it = it->next) {
874             CXLFixedWindow *fw = it->data;
875             start += fw->size;
876         }
877     }
878 
879     return start;
880 }
881 
882 static hwaddr pc_max_used_gpa(PCMachineState *pcms, uint64_t pci_hole64_size)
883 {
884     X86CPU *cpu = X86_CPU(first_cpu);
885 
886     /* 32-bit systems don't have hole64 thus return max CPU address */
887     if (cpu->phys_bits <= 32) {
888         return ((hwaddr)1 << cpu->phys_bits) - 1;
889     }
890 
891     return pc_pci_hole64_start() + pci_hole64_size - 1;
892 }
893 
894 /*
895  * AMD systems with an IOMMU have an additional hole close to the
896  * 1Tb, which are special GPAs that cannot be DMA mapped. Depending
897  * on kernel version, VFIO may or may not let you DMA map those ranges.
898  * Starting Linux v5.4 we validate it, and can't create guests on AMD machines
899  * with certain memory sizes. It's also wrong to use those IOVA ranges
900  * in detriment of leading to IOMMU INVALID_DEVICE_REQUEST or worse.
901  * The ranges reserved for Hyper-Transport are:
902  *
903  * FD_0000_0000h - FF_FFFF_FFFFh
904  *
905  * The ranges represent the following:
906  *
907  * Base Address   Top Address  Use
908  *
909  * FD_0000_0000h FD_F7FF_FFFFh Reserved interrupt address space
910  * FD_F800_0000h FD_F8FF_FFFFh Interrupt/EOI IntCtl
911  * FD_F900_0000h FD_F90F_FFFFh Legacy PIC IACK
912  * FD_F910_0000h FD_F91F_FFFFh System Management
913  * FD_F920_0000h FD_FAFF_FFFFh Reserved Page Tables
914  * FD_FB00_0000h FD_FBFF_FFFFh Address Translation
915  * FD_FC00_0000h FD_FDFF_FFFFh I/O Space
916  * FD_FE00_0000h FD_FFFF_FFFFh Configuration
917  * FE_0000_0000h FE_1FFF_FFFFh Extended Configuration/Device Messages
918  * FE_2000_0000h FF_FFFF_FFFFh Reserved
919  *
920  * See AMD IOMMU spec, section 2.1.2 "IOMMU Logical Topology",
921  * Table 3: Special Address Controls (GPA) for more information.
922  */
923 #define AMD_HT_START         0xfd00000000UL
924 #define AMD_HT_END           0xffffffffffUL
925 #define AMD_ABOVE_1TB_START  (AMD_HT_END + 1)
926 #define AMD_HT_SIZE          (AMD_ABOVE_1TB_START - AMD_HT_START)
927 
928 void pc_memory_init(PCMachineState *pcms,
929                     MemoryRegion *system_memory,
930                     MemoryRegion *rom_memory,
931                     MemoryRegion **ram_memory,
932                     uint64_t pci_hole64_size)
933 {
934     int linux_boot, i;
935     MemoryRegion *option_rom_mr;
936     MemoryRegion *ram_below_4g, *ram_above_4g;
937     FWCfgState *fw_cfg;
938     MachineState *machine = MACHINE(pcms);
939     MachineClass *mc = MACHINE_GET_CLASS(machine);
940     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
941     X86MachineState *x86ms = X86_MACHINE(pcms);
942     hwaddr maxphysaddr, maxusedaddr;
943     hwaddr cxl_base, cxl_resv_end = 0;
944     X86CPU *cpu = X86_CPU(first_cpu);
945 
946     assert(machine->ram_size == x86ms->below_4g_mem_size +
947                                 x86ms->above_4g_mem_size);
948 
949     linux_boot = (machine->kernel_filename != NULL);
950 
951     /*
952      * The HyperTransport range close to the 1T boundary is unique to AMD
953      * hosts with IOMMUs enabled. Restrict the ram-above-4g relocation
954      * to above 1T to AMD vCPUs only. @enforce_amd_1tb_hole is only false in
955      * older machine types (<= 7.0) for compatibility purposes.
956      */
957     if (IS_AMD_CPU(&cpu->env) && pcmc->enforce_amd_1tb_hole) {
958         /* Bail out if max possible address does not cross HT range */
959         if (pc_max_used_gpa(pcms, pci_hole64_size) >= AMD_HT_START) {
960             x86ms->above_4g_mem_start = AMD_ABOVE_1TB_START;
961         }
962 
963         /*
964          * Advertise the HT region if address space covers the reserved
965          * region or if we relocate.
966          */
967         if (cpu->phys_bits >= 40) {
968             e820_add_entry(AMD_HT_START, AMD_HT_SIZE, E820_RESERVED);
969         }
970     }
971 
972     /*
973      * phys-bits is required to be appropriately configured
974      * to make sure max used GPA is reachable.
975      */
976     maxusedaddr = pc_max_used_gpa(pcms, pci_hole64_size);
977     maxphysaddr = ((hwaddr)1 << cpu->phys_bits) - 1;
978     if (maxphysaddr < maxusedaddr) {
979         error_report("Address space limit 0x%"PRIx64" < 0x%"PRIx64
980                      " phys-bits too low (%u)",
981                      maxphysaddr, maxusedaddr, cpu->phys_bits);
982         exit(EXIT_FAILURE);
983     }
984 
985     /*
986      * Split single memory region and use aliases to address portions of it,
987      * done for backwards compatibility with older qemus.
988      */
989     *ram_memory = machine->ram;
990     ram_below_4g = g_malloc(sizeof(*ram_below_4g));
991     memory_region_init_alias(ram_below_4g, NULL, "ram-below-4g", machine->ram,
992                              0, x86ms->below_4g_mem_size);
993     memory_region_add_subregion(system_memory, 0, ram_below_4g);
994     e820_add_entry(0, x86ms->below_4g_mem_size, E820_RAM);
995     if (x86ms->above_4g_mem_size > 0) {
996         ram_above_4g = g_malloc(sizeof(*ram_above_4g));
997         memory_region_init_alias(ram_above_4g, NULL, "ram-above-4g",
998                                  machine->ram,
999                                  x86ms->below_4g_mem_size,
1000                                  x86ms->above_4g_mem_size);
1001         memory_region_add_subregion(system_memory, x86ms->above_4g_mem_start,
1002                                     ram_above_4g);
1003         e820_add_entry(x86ms->above_4g_mem_start, x86ms->above_4g_mem_size,
1004                        E820_RAM);
1005     }
1006 
1007     if (pcms->sgx_epc.size != 0) {
1008         e820_add_entry(pcms->sgx_epc.base, pcms->sgx_epc.size, E820_RESERVED);
1009     }
1010 
1011     if (!pcmc->has_reserved_memory &&
1012         (machine->ram_slots ||
1013          (machine->maxram_size > machine->ram_size))) {
1014 
1015         error_report("\"-memory 'slots|maxmem'\" is not supported by: %s",
1016                      mc->name);
1017         exit(EXIT_FAILURE);
1018     }
1019 
1020     /* always allocate the device memory information */
1021     machine->device_memory = g_malloc0(sizeof(*machine->device_memory));
1022 
1023     /* initialize device memory address space */
1024     if (pcmc->has_reserved_memory &&
1025         (machine->ram_size < machine->maxram_size)) {
1026         ram_addr_t device_mem_size;
1027 
1028         if (machine->ram_slots > ACPI_MAX_RAM_SLOTS) {
1029             error_report("unsupported amount of memory slots: %"PRIu64,
1030                          machine->ram_slots);
1031             exit(EXIT_FAILURE);
1032         }
1033 
1034         if (QEMU_ALIGN_UP(machine->maxram_size,
1035                           TARGET_PAGE_SIZE) != machine->maxram_size) {
1036             error_report("maximum memory size must by aligned to multiple of "
1037                          "%d bytes", TARGET_PAGE_SIZE);
1038             exit(EXIT_FAILURE);
1039         }
1040 
1041         pc_get_device_memory_range(pcms, &machine->device_memory->base, &device_mem_size);
1042 
1043         if ((machine->device_memory->base + device_mem_size) <
1044             device_mem_size) {
1045             error_report("unsupported amount of maximum memory: " RAM_ADDR_FMT,
1046                          machine->maxram_size);
1047             exit(EXIT_FAILURE);
1048         }
1049 
1050         memory_region_init(&machine->device_memory->mr, OBJECT(pcms),
1051                            "device-memory", device_mem_size);
1052         memory_region_add_subregion(system_memory, machine->device_memory->base,
1053                                     &machine->device_memory->mr);
1054     }
1055 
1056     if (pcms->cxl_devices_state.is_enabled) {
1057         MemoryRegion *mr = &pcms->cxl_devices_state.host_mr;
1058         hwaddr cxl_size = MiB;
1059 
1060         cxl_base = pc_get_cxl_range_start(pcms);
1061         e820_add_entry(cxl_base, cxl_size, E820_RESERVED);
1062         memory_region_init(mr, OBJECT(machine), "cxl_host_reg", cxl_size);
1063         memory_region_add_subregion(system_memory, cxl_base, mr);
1064         cxl_resv_end = cxl_base + cxl_size;
1065         if (pcms->cxl_devices_state.fixed_windows) {
1066             hwaddr cxl_fmw_base;
1067             GList *it;
1068 
1069             cxl_fmw_base = ROUND_UP(cxl_base + cxl_size, 256 * MiB);
1070             for (it = pcms->cxl_devices_state.fixed_windows; it; it = it->next) {
1071                 CXLFixedWindow *fw = it->data;
1072 
1073                 fw->base = cxl_fmw_base;
1074                 memory_region_init_io(&fw->mr, OBJECT(machine), &cfmws_ops, fw,
1075                                       "cxl-fixed-memory-region", fw->size);
1076                 memory_region_add_subregion(system_memory, fw->base, &fw->mr);
1077                 e820_add_entry(fw->base, fw->size, E820_RESERVED);
1078                 cxl_fmw_base += fw->size;
1079                 cxl_resv_end = cxl_fmw_base;
1080             }
1081         }
1082     }
1083 
1084     /* Initialize PC system firmware */
1085     pc_system_firmware_init(pcms, rom_memory);
1086 
1087     option_rom_mr = g_malloc(sizeof(*option_rom_mr));
1088     memory_region_init_ram(option_rom_mr, NULL, "pc.rom", PC_ROM_SIZE,
1089                            &error_fatal);
1090     if (pcmc->pci_enabled) {
1091         memory_region_set_readonly(option_rom_mr, true);
1092     }
1093     memory_region_add_subregion_overlap(rom_memory,
1094                                         PC_ROM_MIN_VGA,
1095                                         option_rom_mr,
1096                                         1);
1097 
1098     fw_cfg = fw_cfg_arch_create(machine,
1099                                 x86ms->boot_cpus, x86ms->apic_id_limit);
1100 
1101     rom_set_fw(fw_cfg);
1102 
1103     if (pcmc->has_reserved_memory && machine->device_memory->base) {
1104         uint64_t *val = g_malloc(sizeof(*val));
1105         PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1106         uint64_t res_mem_end = machine->device_memory->base;
1107 
1108         if (!pcmc->broken_reserved_end) {
1109             res_mem_end += memory_region_size(&machine->device_memory->mr);
1110         }
1111 
1112         if (pcms->cxl_devices_state.is_enabled) {
1113             res_mem_end = cxl_resv_end;
1114         }
1115         *val = cpu_to_le64(ROUND_UP(res_mem_end, 1 * GiB));
1116         fw_cfg_add_file(fw_cfg, "etc/reserved-memory-end", val, sizeof(*val));
1117     }
1118 
1119     if (linux_boot) {
1120         x86_load_linux(x86ms, fw_cfg, pcmc->acpi_data_size,
1121                        pcmc->pvh_enabled, pcmc->legacy_no_rng_seed);
1122     }
1123 
1124     for (i = 0; i < nb_option_roms; i++) {
1125         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1126     }
1127     x86ms->fw_cfg = fw_cfg;
1128 
1129     /* Init default IOAPIC address space */
1130     x86ms->ioapic_as = &address_space_memory;
1131 
1132     /* Init ACPI memory hotplug IO base address */
1133     pcms->memhp_io_base = ACPI_MEMORY_HOTPLUG_BASE;
1134 }
1135 
1136 /*
1137  * The 64bit pci hole starts after "above 4G RAM" and
1138  * potentially the space reserved for memory hotplug.
1139  */
1140 uint64_t pc_pci_hole64_start(void)
1141 {
1142     PCMachineState *pcms = PC_MACHINE(qdev_get_machine());
1143     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1144     MachineState *ms = MACHINE(pcms);
1145     uint64_t hole64_start = 0;
1146     ram_addr_t size = 0;
1147 
1148     if (pcms->cxl_devices_state.is_enabled) {
1149         hole64_start = pc_get_cxl_range_end(pcms);
1150     } else if (pcmc->has_reserved_memory && (ms->ram_size < ms->maxram_size)) {
1151         pc_get_device_memory_range(pcms, &hole64_start, &size);
1152         if (!pcmc->broken_reserved_end) {
1153             hole64_start += size;
1154         }
1155     } else {
1156         hole64_start = pc_above_4g_end(pcms);
1157     }
1158 
1159     return ROUND_UP(hole64_start, 1 * GiB);
1160 }
1161 
1162 DeviceState *pc_vga_init(ISABus *isa_bus, PCIBus *pci_bus)
1163 {
1164     DeviceState *dev = NULL;
1165 
1166     rom_set_order_override(FW_CFG_ORDER_OVERRIDE_VGA);
1167     if (pci_bus) {
1168         PCIDevice *pcidev = pci_vga_init(pci_bus);
1169         dev = pcidev ? &pcidev->qdev : NULL;
1170     } else if (isa_bus) {
1171         ISADevice *isadev = isa_vga_init(isa_bus);
1172         dev = isadev ? DEVICE(isadev) : NULL;
1173     }
1174     rom_reset_order_override();
1175     return dev;
1176 }
1177 
1178 static const MemoryRegionOps ioport80_io_ops = {
1179     .write = ioport80_write,
1180     .read = ioport80_read,
1181     .endianness = DEVICE_NATIVE_ENDIAN,
1182     .impl = {
1183         .min_access_size = 1,
1184         .max_access_size = 1,
1185     },
1186 };
1187 
1188 static const MemoryRegionOps ioportF0_io_ops = {
1189     .write = ioportF0_write,
1190     .read = ioportF0_read,
1191     .endianness = DEVICE_NATIVE_ENDIAN,
1192     .impl = {
1193         .min_access_size = 1,
1194         .max_access_size = 1,
1195     },
1196 };
1197 
1198 static void pc_superio_init(ISABus *isa_bus, bool create_fdctrl,
1199                             bool create_i8042, bool no_vmport)
1200 {
1201     int i;
1202     DriveInfo *fd[MAX_FD];
1203     qemu_irq *a20_line;
1204     ISADevice *fdc, *i8042, *port92, *vmmouse;
1205 
1206     serial_hds_isa_init(isa_bus, 0, MAX_ISA_SERIAL_PORTS);
1207     parallel_hds_isa_init(isa_bus, MAX_PARALLEL_PORTS);
1208 
1209     for (i = 0; i < MAX_FD; i++) {
1210         fd[i] = drive_get(IF_FLOPPY, 0, i);
1211         create_fdctrl |= !!fd[i];
1212     }
1213     if (create_fdctrl) {
1214         fdc = isa_new(TYPE_ISA_FDC);
1215         if (fdc) {
1216             isa_realize_and_unref(fdc, isa_bus, &error_fatal);
1217             isa_fdc_init_drives(fdc, fd);
1218         }
1219     }
1220 
1221     if (!create_i8042) {
1222         return;
1223     }
1224 
1225     i8042 = isa_create_simple(isa_bus, TYPE_I8042);
1226     if (!no_vmport) {
1227         isa_create_simple(isa_bus, TYPE_VMPORT);
1228         vmmouse = isa_try_new("vmmouse");
1229     } else {
1230         vmmouse = NULL;
1231     }
1232     if (vmmouse) {
1233         object_property_set_link(OBJECT(vmmouse), TYPE_I8042, OBJECT(i8042),
1234                                  &error_abort);
1235         isa_realize_and_unref(vmmouse, isa_bus, &error_fatal);
1236     }
1237     port92 = isa_create_simple(isa_bus, TYPE_PORT92);
1238 
1239     a20_line = qemu_allocate_irqs(handle_a20_line_change, first_cpu, 2);
1240     i8042_setup_a20_line(i8042, a20_line[0]);
1241     qdev_connect_gpio_out_named(DEVICE(port92),
1242                                 PORT92_A20_LINE, 0, a20_line[1]);
1243     g_free(a20_line);
1244 }
1245 
1246 void pc_basic_device_init(struct PCMachineState *pcms,
1247                           ISABus *isa_bus, qemu_irq *gsi,
1248                           ISADevice **rtc_state,
1249                           bool create_fdctrl,
1250                           uint32_t hpet_irqs)
1251 {
1252     int i;
1253     DeviceState *hpet = NULL;
1254     int pit_isa_irq = 0;
1255     qemu_irq pit_alt_irq = NULL;
1256     qemu_irq rtc_irq = NULL;
1257     ISADevice *pit = NULL;
1258     MemoryRegion *ioport80_io = g_new(MemoryRegion, 1);
1259     MemoryRegion *ioportF0_io = g_new(MemoryRegion, 1);
1260     X86MachineState *x86ms = X86_MACHINE(pcms);
1261 
1262     memory_region_init_io(ioport80_io, NULL, &ioport80_io_ops, NULL, "ioport80", 1);
1263     memory_region_add_subregion(isa_bus->address_space_io, 0x80, ioport80_io);
1264 
1265     memory_region_init_io(ioportF0_io, NULL, &ioportF0_io_ops, NULL, "ioportF0", 1);
1266     memory_region_add_subregion(isa_bus->address_space_io, 0xf0, ioportF0_io);
1267 
1268     /*
1269      * Check if an HPET shall be created.
1270      *
1271      * Without KVM_CAP_PIT_STATE2, we cannot switch off the in-kernel PIT
1272      * when the HPET wants to take over. Thus we have to disable the latter.
1273      */
1274     if (pcms->hpet_enabled && (!kvm_irqchip_in_kernel() ||
1275                                kvm_has_pit_state2())) {
1276         hpet = qdev_try_new(TYPE_HPET);
1277         if (!hpet) {
1278             error_report("couldn't create HPET device");
1279             exit(1);
1280         }
1281         /*
1282          * For pc-piix-*, hpet's intcap is always IRQ2. For pc-q35-1.7 and
1283          * earlier, use IRQ2 for compat. Otherwise, use IRQ16~23, IRQ8 and
1284          * IRQ2.
1285          */
1286         uint8_t compat = object_property_get_uint(OBJECT(hpet),
1287                 HPET_INTCAP, NULL);
1288         if (!compat) {
1289             qdev_prop_set_uint32(hpet, HPET_INTCAP, hpet_irqs);
1290         }
1291         sysbus_realize_and_unref(SYS_BUS_DEVICE(hpet), &error_fatal);
1292         sysbus_mmio_map(SYS_BUS_DEVICE(hpet), 0, HPET_BASE);
1293 
1294         for (i = 0; i < GSI_NUM_PINS; i++) {
1295             sysbus_connect_irq(SYS_BUS_DEVICE(hpet), i, gsi[i]);
1296         }
1297         pit_isa_irq = -1;
1298         pit_alt_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_PIT_INT);
1299         rtc_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_RTC_INT);
1300     }
1301     *rtc_state = mc146818_rtc_init(isa_bus, 2000, rtc_irq);
1302 
1303     qemu_register_boot_set(pc_boot_set, *rtc_state);
1304 
1305     if (!xen_enabled() &&
1306         (x86ms->pit == ON_OFF_AUTO_AUTO || x86ms->pit == ON_OFF_AUTO_ON)) {
1307         if (kvm_pit_in_kernel()) {
1308             pit = kvm_pit_init(isa_bus, 0x40);
1309         } else {
1310             pit = i8254_pit_init(isa_bus, 0x40, pit_isa_irq, pit_alt_irq);
1311         }
1312         if (hpet) {
1313             /* connect PIT to output control line of the HPET */
1314             qdev_connect_gpio_out(hpet, 0, qdev_get_gpio_in(DEVICE(pit), 0));
1315         }
1316         pcspk_init(pcms->pcspk, isa_bus, pit);
1317     }
1318 
1319     i8257_dma_init(isa_bus, 0);
1320 
1321     /* Super I/O */
1322     pc_superio_init(isa_bus, create_fdctrl, pcms->i8042_enabled,
1323                     pcms->vmport != ON_OFF_AUTO_ON);
1324 }
1325 
1326 void pc_nic_init(PCMachineClass *pcmc, ISABus *isa_bus, PCIBus *pci_bus)
1327 {
1328     int i;
1329 
1330     rom_set_order_override(FW_CFG_ORDER_OVERRIDE_NIC);
1331     for (i = 0; i < nb_nics; i++) {
1332         NICInfo *nd = &nd_table[i];
1333         const char *model = nd->model ? nd->model : pcmc->default_nic_model;
1334 
1335         if (g_str_equal(model, "ne2k_isa")) {
1336             pc_init_ne2k_isa(isa_bus, nd);
1337         } else {
1338             pci_nic_init_nofail(nd, pci_bus, model, NULL);
1339         }
1340     }
1341     rom_reset_order_override();
1342 }
1343 
1344 void pc_i8259_create(ISABus *isa_bus, qemu_irq *i8259_irqs)
1345 {
1346     qemu_irq *i8259;
1347 
1348     if (kvm_pic_in_kernel()) {
1349         i8259 = kvm_i8259_init(isa_bus);
1350     } else if (xen_enabled()) {
1351         i8259 = xen_interrupt_controller_init();
1352     } else {
1353         i8259 = i8259_init(isa_bus, x86_allocate_cpu_irq());
1354     }
1355 
1356     for (size_t i = 0; i < ISA_NUM_IRQS; i++) {
1357         i8259_irqs[i] = i8259[i];
1358     }
1359 
1360     g_free(i8259);
1361 }
1362 
1363 static void pc_memory_pre_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
1364                                Error **errp)
1365 {
1366     const PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1367     const X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
1368     const PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1369     const MachineState *ms = MACHINE(hotplug_dev);
1370     const bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
1371     const uint64_t legacy_align = TARGET_PAGE_SIZE;
1372     Error *local_err = NULL;
1373 
1374     /*
1375      * When -no-acpi is used with Q35 machine type, no ACPI is built,
1376      * but pcms->acpi_dev is still created. Check !acpi_enabled in
1377      * addition to cover this case.
1378      */
1379     if (!x86ms->acpi_dev || !x86_machine_is_acpi_enabled(x86ms)) {
1380         error_setg(errp,
1381                    "memory hotplug is not enabled: missing acpi device or acpi disabled");
1382         return;
1383     }
1384 
1385     if (is_nvdimm && !ms->nvdimms_state->is_enabled) {
1386         error_setg(errp, "nvdimm is not enabled: missing 'nvdimm' in '-M'");
1387         return;
1388     }
1389 
1390     hotplug_handler_pre_plug(x86ms->acpi_dev, dev, &local_err);
1391     if (local_err) {
1392         error_propagate(errp, local_err);
1393         return;
1394     }
1395 
1396     pc_dimm_pre_plug(PC_DIMM(dev), MACHINE(hotplug_dev),
1397                      pcmc->enforce_aligned_dimm ? NULL : &legacy_align, errp);
1398 }
1399 
1400 static void pc_memory_plug(HotplugHandler *hotplug_dev,
1401                            DeviceState *dev, Error **errp)
1402 {
1403     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1404     X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
1405     MachineState *ms = MACHINE(hotplug_dev);
1406     bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
1407 
1408     pc_dimm_plug(PC_DIMM(dev), MACHINE(pcms));
1409 
1410     if (is_nvdimm) {
1411         nvdimm_plug(ms->nvdimms_state);
1412     }
1413 
1414     hotplug_handler_plug(x86ms->acpi_dev, dev, &error_abort);
1415 }
1416 
1417 static void pc_memory_unplug_request(HotplugHandler *hotplug_dev,
1418                                      DeviceState *dev, Error **errp)
1419 {
1420     X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
1421 
1422     /*
1423      * When -no-acpi is used with Q35 machine type, no ACPI is built,
1424      * but pcms->acpi_dev is still created. Check !acpi_enabled in
1425      * addition to cover this case.
1426      */
1427     if (!x86ms->acpi_dev || !x86_machine_is_acpi_enabled(x86ms)) {
1428         error_setg(errp,
1429                    "memory hotplug is not enabled: missing acpi device or acpi disabled");
1430         return;
1431     }
1432 
1433     if (object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM)) {
1434         error_setg(errp, "nvdimm device hot unplug is not supported yet.");
1435         return;
1436     }
1437 
1438     hotplug_handler_unplug_request(x86ms->acpi_dev, dev,
1439                                    errp);
1440 }
1441 
1442 static void pc_memory_unplug(HotplugHandler *hotplug_dev,
1443                              DeviceState *dev, Error **errp)
1444 {
1445     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1446     X86MachineState *x86ms = X86_MACHINE(hotplug_dev);
1447     Error *local_err = NULL;
1448 
1449     hotplug_handler_unplug(x86ms->acpi_dev, dev, &local_err);
1450     if (local_err) {
1451         goto out;
1452     }
1453 
1454     pc_dimm_unplug(PC_DIMM(dev), MACHINE(pcms));
1455     qdev_unrealize(dev);
1456  out:
1457     error_propagate(errp, local_err);
1458 }
1459 
1460 static void pc_virtio_md_pci_pre_plug(HotplugHandler *hotplug_dev,
1461                                       DeviceState *dev, Error **errp)
1462 {
1463     HotplugHandler *hotplug_dev2 = qdev_get_bus_hotplug_handler(dev);
1464     Error *local_err = NULL;
1465 
1466     if (!hotplug_dev2 && dev->hotplugged) {
1467         /*
1468          * Without a bus hotplug handler, we cannot control the plug/unplug
1469          * order. We should never reach this point when hotplugging on x86,
1470          * however, better add a safety net.
1471          */
1472         error_setg(errp, "hotplug of virtio based memory devices not supported"
1473                    " on this bus.");
1474         return;
1475     }
1476     /*
1477      * First, see if we can plug this memory device at all. If that
1478      * succeeds, branch of to the actual hotplug handler.
1479      */
1480     memory_device_pre_plug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev), NULL,
1481                            &local_err);
1482     if (!local_err && hotplug_dev2) {
1483         hotplug_handler_pre_plug(hotplug_dev2, dev, &local_err);
1484     }
1485     error_propagate(errp, local_err);
1486 }
1487 
1488 static void pc_virtio_md_pci_plug(HotplugHandler *hotplug_dev,
1489                                   DeviceState *dev, Error **errp)
1490 {
1491     HotplugHandler *hotplug_dev2 = qdev_get_bus_hotplug_handler(dev);
1492     Error *local_err = NULL;
1493 
1494     /*
1495      * Plug the memory device first and then branch off to the actual
1496      * hotplug handler. If that one fails, we can easily undo the memory
1497      * device bits.
1498      */
1499     memory_device_plug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev));
1500     if (hotplug_dev2) {
1501         hotplug_handler_plug(hotplug_dev2, dev, &local_err);
1502         if (local_err) {
1503             memory_device_unplug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev));
1504         }
1505     }
1506     error_propagate(errp, local_err);
1507 }
1508 
1509 static void pc_virtio_md_pci_unplug_request(HotplugHandler *hotplug_dev,
1510                                             DeviceState *dev, Error **errp)
1511 {
1512     /* We don't support hot unplug of virtio based memory devices */
1513     error_setg(errp, "virtio based memory devices cannot be unplugged.");
1514 }
1515 
1516 static void pc_virtio_md_pci_unplug(HotplugHandler *hotplug_dev,
1517                                     DeviceState *dev, Error **errp)
1518 {
1519     /* We don't support hot unplug of virtio based memory devices */
1520 }
1521 
1522 static void pc_machine_device_pre_plug_cb(HotplugHandler *hotplug_dev,
1523                                           DeviceState *dev, Error **errp)
1524 {
1525     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1526         pc_memory_pre_plug(hotplug_dev, dev, errp);
1527     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1528         x86_cpu_pre_plug(hotplug_dev, dev, errp);
1529     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI) ||
1530                object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI)) {
1531         pc_virtio_md_pci_pre_plug(hotplug_dev, dev, errp);
1532     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
1533         /* Declare the APIC range as the reserved MSI region */
1534         char *resv_prop_str = g_strdup_printf("0xfee00000:0xfeefffff:%d",
1535                                               VIRTIO_IOMMU_RESV_MEM_T_MSI);
1536 
1537         object_property_set_uint(OBJECT(dev), "len-reserved-regions", 1, errp);
1538         object_property_set_str(OBJECT(dev), "reserved-regions[0]",
1539                                 resv_prop_str, errp);
1540         g_free(resv_prop_str);
1541     }
1542 
1543     if (object_dynamic_cast(OBJECT(dev), TYPE_X86_IOMMU_DEVICE) ||
1544         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
1545         PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1546 
1547         if (pcms->iommu) {
1548             error_setg(errp, "QEMU does not support multiple vIOMMUs "
1549                        "for x86 yet.");
1550             return;
1551         }
1552         pcms->iommu = dev;
1553     }
1554 }
1555 
1556 static void pc_machine_device_plug_cb(HotplugHandler *hotplug_dev,
1557                                       DeviceState *dev, Error **errp)
1558 {
1559     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1560         pc_memory_plug(hotplug_dev, dev, errp);
1561     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1562         x86_cpu_plug(hotplug_dev, dev, errp);
1563     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI) ||
1564                object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI)) {
1565         pc_virtio_md_pci_plug(hotplug_dev, dev, errp);
1566     }
1567 }
1568 
1569 static void pc_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
1570                                                 DeviceState *dev, Error **errp)
1571 {
1572     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1573         pc_memory_unplug_request(hotplug_dev, dev, errp);
1574     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1575         x86_cpu_unplug_request_cb(hotplug_dev, dev, errp);
1576     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI) ||
1577                object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI)) {
1578         pc_virtio_md_pci_unplug_request(hotplug_dev, dev, errp);
1579     } else {
1580         error_setg(errp, "acpi: device unplug request for not supported device"
1581                    " type: %s", object_get_typename(OBJECT(dev)));
1582     }
1583 }
1584 
1585 static void pc_machine_device_unplug_cb(HotplugHandler *hotplug_dev,
1586                                         DeviceState *dev, Error **errp)
1587 {
1588     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1589         pc_memory_unplug(hotplug_dev, dev, errp);
1590     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1591         x86_cpu_unplug_cb(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_unplug(hotplug_dev, dev, errp);
1595     } else {
1596         error_setg(errp, "acpi: device unplug for not supported device"
1597                    " type: %s", object_get_typename(OBJECT(dev)));
1598     }
1599 }
1600 
1601 static HotplugHandler *pc_get_hotplug_handler(MachineState *machine,
1602                                              DeviceState *dev)
1603 {
1604     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) ||
1605         object_dynamic_cast(OBJECT(dev), TYPE_CPU) ||
1606         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI) ||
1607         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI) ||
1608         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI) ||
1609         object_dynamic_cast(OBJECT(dev), TYPE_X86_IOMMU_DEVICE)) {
1610         return HOTPLUG_HANDLER(machine);
1611     }
1612 
1613     return NULL;
1614 }
1615 
1616 static void
1617 pc_machine_get_device_memory_region_size(Object *obj, Visitor *v,
1618                                          const char *name, void *opaque,
1619                                          Error **errp)
1620 {
1621     MachineState *ms = MACHINE(obj);
1622     int64_t value = 0;
1623 
1624     if (ms->device_memory) {
1625         value = memory_region_size(&ms->device_memory->mr);
1626     }
1627 
1628     visit_type_int(v, name, &value, errp);
1629 }
1630 
1631 static void pc_machine_get_vmport(Object *obj, Visitor *v, const char *name,
1632                                   void *opaque, Error **errp)
1633 {
1634     PCMachineState *pcms = PC_MACHINE(obj);
1635     OnOffAuto vmport = pcms->vmport;
1636 
1637     visit_type_OnOffAuto(v, name, &vmport, errp);
1638 }
1639 
1640 static void pc_machine_set_vmport(Object *obj, Visitor *v, const char *name,
1641                                   void *opaque, Error **errp)
1642 {
1643     PCMachineState *pcms = PC_MACHINE(obj);
1644 
1645     visit_type_OnOffAuto(v, name, &pcms->vmport, errp);
1646 }
1647 
1648 static bool pc_machine_get_smbus(Object *obj, Error **errp)
1649 {
1650     PCMachineState *pcms = PC_MACHINE(obj);
1651 
1652     return pcms->smbus_enabled;
1653 }
1654 
1655 static void pc_machine_set_smbus(Object *obj, bool value, Error **errp)
1656 {
1657     PCMachineState *pcms = PC_MACHINE(obj);
1658 
1659     pcms->smbus_enabled = value;
1660 }
1661 
1662 static bool pc_machine_get_sata(Object *obj, Error **errp)
1663 {
1664     PCMachineState *pcms = PC_MACHINE(obj);
1665 
1666     return pcms->sata_enabled;
1667 }
1668 
1669 static void pc_machine_set_sata(Object *obj, bool value, Error **errp)
1670 {
1671     PCMachineState *pcms = PC_MACHINE(obj);
1672 
1673     pcms->sata_enabled = value;
1674 }
1675 
1676 static bool pc_machine_get_hpet(Object *obj, Error **errp)
1677 {
1678     PCMachineState *pcms = PC_MACHINE(obj);
1679 
1680     return pcms->hpet_enabled;
1681 }
1682 
1683 static void pc_machine_set_hpet(Object *obj, bool value, Error **errp)
1684 {
1685     PCMachineState *pcms = PC_MACHINE(obj);
1686 
1687     pcms->hpet_enabled = value;
1688 }
1689 
1690 static bool pc_machine_get_i8042(Object *obj, Error **errp)
1691 {
1692     PCMachineState *pcms = PC_MACHINE(obj);
1693 
1694     return pcms->i8042_enabled;
1695 }
1696 
1697 static void pc_machine_set_i8042(Object *obj, bool value, Error **errp)
1698 {
1699     PCMachineState *pcms = PC_MACHINE(obj);
1700 
1701     pcms->i8042_enabled = value;
1702 }
1703 
1704 static bool pc_machine_get_default_bus_bypass_iommu(Object *obj, Error **errp)
1705 {
1706     PCMachineState *pcms = PC_MACHINE(obj);
1707 
1708     return pcms->default_bus_bypass_iommu;
1709 }
1710 
1711 static void pc_machine_set_default_bus_bypass_iommu(Object *obj, bool value,
1712                                                     Error **errp)
1713 {
1714     PCMachineState *pcms = PC_MACHINE(obj);
1715 
1716     pcms->default_bus_bypass_iommu = value;
1717 }
1718 
1719 static void pc_machine_get_smbios_ep(Object *obj, Visitor *v, const char *name,
1720                                      void *opaque, Error **errp)
1721 {
1722     PCMachineState *pcms = PC_MACHINE(obj);
1723     SmbiosEntryPointType smbios_entry_point_type = pcms->smbios_entry_point_type;
1724 
1725     visit_type_SmbiosEntryPointType(v, name, &smbios_entry_point_type, errp);
1726 }
1727 
1728 static void pc_machine_set_smbios_ep(Object *obj, Visitor *v, const char *name,
1729                                      void *opaque, Error **errp)
1730 {
1731     PCMachineState *pcms = PC_MACHINE(obj);
1732 
1733     visit_type_SmbiosEntryPointType(v, name, &pcms->smbios_entry_point_type, errp);
1734 }
1735 
1736 static void pc_machine_get_max_ram_below_4g(Object *obj, Visitor *v,
1737                                             const char *name, void *opaque,
1738                                             Error **errp)
1739 {
1740     PCMachineState *pcms = PC_MACHINE(obj);
1741     uint64_t value = pcms->max_ram_below_4g;
1742 
1743     visit_type_size(v, name, &value, errp);
1744 }
1745 
1746 static void pc_machine_set_max_ram_below_4g(Object *obj, Visitor *v,
1747                                             const char *name, void *opaque,
1748                                             Error **errp)
1749 {
1750     PCMachineState *pcms = PC_MACHINE(obj);
1751     uint64_t value;
1752 
1753     if (!visit_type_size(v, name, &value, errp)) {
1754         return;
1755     }
1756     if (value > 4 * GiB) {
1757         error_setg(errp,
1758                    "Machine option 'max-ram-below-4g=%"PRIu64
1759                    "' expects size less than or equal to 4G", value);
1760         return;
1761     }
1762 
1763     if (value < 1 * MiB) {
1764         warn_report("Only %" PRIu64 " bytes of RAM below the 4GiB boundary,"
1765                     "BIOS may not work with less than 1MiB", value);
1766     }
1767 
1768     pcms->max_ram_below_4g = value;
1769 }
1770 
1771 static void pc_machine_get_max_fw_size(Object *obj, Visitor *v,
1772                                        const char *name, void *opaque,
1773                                        Error **errp)
1774 {
1775     PCMachineState *pcms = PC_MACHINE(obj);
1776     uint64_t value = pcms->max_fw_size;
1777 
1778     visit_type_size(v, name, &value, errp);
1779 }
1780 
1781 static void pc_machine_set_max_fw_size(Object *obj, Visitor *v,
1782                                        const char *name, void *opaque,
1783                                        Error **errp)
1784 {
1785     PCMachineState *pcms = PC_MACHINE(obj);
1786     Error *error = NULL;
1787     uint64_t value;
1788 
1789     visit_type_size(v, name, &value, &error);
1790     if (error) {
1791         error_propagate(errp, error);
1792         return;
1793     }
1794 
1795     /*
1796     * We don't have a theoretically justifiable exact lower bound on the base
1797     * address of any flash mapping. In practice, the IO-APIC MMIO range is
1798     * [0xFEE00000..0xFEE01000] -- see IO_APIC_DEFAULT_ADDRESS --, leaving free
1799     * only 18MB-4KB below 4G. For now, restrict the cumulative mapping to 8MB in
1800     * size.
1801     */
1802     if (value > 16 * MiB) {
1803         error_setg(errp,
1804                    "User specified max allowed firmware size %" PRIu64 " is "
1805                    "greater than 16MiB. If combined firwmare size exceeds "
1806                    "16MiB the system may not boot, or experience intermittent"
1807                    "stability issues.",
1808                    value);
1809         return;
1810     }
1811 
1812     pcms->max_fw_size = value;
1813 }
1814 
1815 
1816 static void pc_machine_initfn(Object *obj)
1817 {
1818     PCMachineState *pcms = PC_MACHINE(obj);
1819 
1820 #ifdef CONFIG_VMPORT
1821     pcms->vmport = ON_OFF_AUTO_AUTO;
1822 #else
1823     pcms->vmport = ON_OFF_AUTO_OFF;
1824 #endif /* CONFIG_VMPORT */
1825     pcms->max_ram_below_4g = 0; /* use default */
1826     pcms->smbios_entry_point_type = SMBIOS_ENTRY_POINT_TYPE_32;
1827 
1828     /* acpi build is enabled by default if machine supports it */
1829     pcms->acpi_build_enabled = PC_MACHINE_GET_CLASS(pcms)->has_acpi_build;
1830     pcms->smbus_enabled = true;
1831     pcms->sata_enabled = true;
1832     pcms->i8042_enabled = true;
1833     pcms->max_fw_size = 8 * MiB;
1834 #ifdef CONFIG_HPET
1835     pcms->hpet_enabled = true;
1836 #endif
1837     pcms->default_bus_bypass_iommu = false;
1838 
1839     pc_system_flash_create(pcms);
1840     pcms->pcspk = isa_new(TYPE_PC_SPEAKER);
1841     object_property_add_alias(OBJECT(pcms), "pcspk-audiodev",
1842                               OBJECT(pcms->pcspk), "audiodev");
1843     cxl_machine_init(obj, &pcms->cxl_devices_state);
1844 }
1845 
1846 static void pc_machine_reset(MachineState *machine)
1847 {
1848     CPUState *cs;
1849     X86CPU *cpu;
1850 
1851     qemu_devices_reset();
1852 
1853     /* Reset APIC after devices have been reset to cancel
1854      * any changes that qemu_devices_reset() might have done.
1855      */
1856     CPU_FOREACH(cs) {
1857         cpu = X86_CPU(cs);
1858 
1859         if (cpu->apic_state) {
1860             device_legacy_reset(cpu->apic_state);
1861         }
1862     }
1863 }
1864 
1865 static void pc_machine_wakeup(MachineState *machine)
1866 {
1867     cpu_synchronize_all_states();
1868     pc_machine_reset(machine);
1869     cpu_synchronize_all_post_reset();
1870 }
1871 
1872 static bool pc_hotplug_allowed(MachineState *ms, DeviceState *dev, Error **errp)
1873 {
1874     X86IOMMUState *iommu = x86_iommu_get_default();
1875     IntelIOMMUState *intel_iommu;
1876 
1877     if (iommu &&
1878         object_dynamic_cast((Object *)iommu, TYPE_INTEL_IOMMU_DEVICE) &&
1879         object_dynamic_cast((Object *)dev, "vfio-pci")) {
1880         intel_iommu = INTEL_IOMMU_DEVICE(iommu);
1881         if (!intel_iommu->caching_mode) {
1882             error_setg(errp, "Device assignment is not allowed without "
1883                        "enabling caching-mode=on for Intel IOMMU.");
1884             return false;
1885         }
1886     }
1887 
1888     return true;
1889 }
1890 
1891 static void pc_machine_class_init(ObjectClass *oc, void *data)
1892 {
1893     MachineClass *mc = MACHINE_CLASS(oc);
1894     PCMachineClass *pcmc = PC_MACHINE_CLASS(oc);
1895     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
1896 
1897     pcmc->pci_enabled = true;
1898     pcmc->has_acpi_build = true;
1899     pcmc->rsdp_in_ram = true;
1900     pcmc->smbios_defaults = true;
1901     pcmc->smbios_uuid_encoded = true;
1902     pcmc->gigabyte_align = true;
1903     pcmc->has_reserved_memory = true;
1904     pcmc->kvmclock_enabled = true;
1905     pcmc->enforce_aligned_dimm = true;
1906     pcmc->enforce_amd_1tb_hole = true;
1907     /* BIOS ACPI tables: 128K. Other BIOS datastructures: less than 4K reported
1908      * to be used at the moment, 32K should be enough for a while.  */
1909     pcmc->acpi_data_size = 0x20000 + 0x8000;
1910     pcmc->pvh_enabled = true;
1911     pcmc->kvmclock_create_always = true;
1912     assert(!mc->get_hotplug_handler);
1913     mc->get_hotplug_handler = pc_get_hotplug_handler;
1914     mc->hotplug_allowed = pc_hotplug_allowed;
1915     mc->cpu_index_to_instance_props = x86_cpu_index_to_props;
1916     mc->get_default_cpu_node_id = x86_get_default_cpu_node_id;
1917     mc->possible_cpu_arch_ids = x86_possible_cpu_arch_ids;
1918     mc->auto_enable_numa_with_memhp = true;
1919     mc->auto_enable_numa_with_memdev = true;
1920     mc->has_hotpluggable_cpus = true;
1921     mc->default_boot_order = "cad";
1922     mc->block_default_type = IF_IDE;
1923     mc->max_cpus = 255;
1924     mc->reset = pc_machine_reset;
1925     mc->wakeup = pc_machine_wakeup;
1926     hc->pre_plug = pc_machine_device_pre_plug_cb;
1927     hc->plug = pc_machine_device_plug_cb;
1928     hc->unplug_request = pc_machine_device_unplug_request_cb;
1929     hc->unplug = pc_machine_device_unplug_cb;
1930     mc->default_cpu_type = TARGET_DEFAULT_CPU_TYPE;
1931     mc->nvdimm_supported = true;
1932     mc->smp_props.dies_supported = true;
1933     mc->default_ram_id = "pc.ram";
1934 
1935     object_class_property_add(oc, PC_MACHINE_MAX_RAM_BELOW_4G, "size",
1936         pc_machine_get_max_ram_below_4g, pc_machine_set_max_ram_below_4g,
1937         NULL, NULL);
1938     object_class_property_set_description(oc, PC_MACHINE_MAX_RAM_BELOW_4G,
1939         "Maximum ram below the 4G boundary (32bit boundary)");
1940 
1941     object_class_property_add(oc, PC_MACHINE_DEVMEM_REGION_SIZE, "int",
1942         pc_machine_get_device_memory_region_size, NULL,
1943         NULL, NULL);
1944 
1945     object_class_property_add(oc, PC_MACHINE_VMPORT, "OnOffAuto",
1946         pc_machine_get_vmport, pc_machine_set_vmport,
1947         NULL, NULL);
1948     object_class_property_set_description(oc, PC_MACHINE_VMPORT,
1949         "Enable vmport (pc & q35)");
1950 
1951     object_class_property_add_bool(oc, PC_MACHINE_SMBUS,
1952         pc_machine_get_smbus, pc_machine_set_smbus);
1953     object_class_property_set_description(oc, PC_MACHINE_SMBUS,
1954         "Enable/disable system management bus");
1955 
1956     object_class_property_add_bool(oc, PC_MACHINE_SATA,
1957         pc_machine_get_sata, pc_machine_set_sata);
1958     object_class_property_set_description(oc, PC_MACHINE_SATA,
1959         "Enable/disable Serial ATA bus");
1960 
1961     object_class_property_add_bool(oc, "hpet",
1962         pc_machine_get_hpet, pc_machine_set_hpet);
1963     object_class_property_set_description(oc, "hpet",
1964         "Enable/disable high precision event timer emulation");
1965 
1966     object_class_property_add_bool(oc, PC_MACHINE_I8042,
1967         pc_machine_get_i8042, pc_machine_set_i8042);
1968 
1969     object_class_property_add_bool(oc, "default-bus-bypass-iommu",
1970         pc_machine_get_default_bus_bypass_iommu,
1971         pc_machine_set_default_bus_bypass_iommu);
1972 
1973     object_class_property_add(oc, PC_MACHINE_MAX_FW_SIZE, "size",
1974         pc_machine_get_max_fw_size, pc_machine_set_max_fw_size,
1975         NULL, NULL);
1976     object_class_property_set_description(oc, PC_MACHINE_MAX_FW_SIZE,
1977         "Maximum combined firmware size");
1978 
1979     object_class_property_add(oc, PC_MACHINE_SMBIOS_EP, "str",
1980         pc_machine_get_smbios_ep, pc_machine_set_smbios_ep,
1981         NULL, NULL);
1982     object_class_property_set_description(oc, PC_MACHINE_SMBIOS_EP,
1983         "SMBIOS Entry Point type [32, 64]");
1984 }
1985 
1986 static const TypeInfo pc_machine_info = {
1987     .name = TYPE_PC_MACHINE,
1988     .parent = TYPE_X86_MACHINE,
1989     .abstract = true,
1990     .instance_size = sizeof(PCMachineState),
1991     .instance_init = pc_machine_initfn,
1992     .class_size = sizeof(PCMachineClass),
1993     .class_init = pc_machine_class_init,
1994     .interfaces = (InterfaceInfo[]) {
1995          { TYPE_HOTPLUG_HANDLER },
1996          { }
1997     },
1998 };
1999 
2000 static void pc_machine_register_types(void)
2001 {
2002     type_register_static(&pc_machine_info);
2003 }
2004 
2005 type_init(pc_machine_register_types)
2006