xref: /openbmc/qemu/hw/i386/pc.c (revision ad9e5aa2)
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/nvram/fw_cfg.h"
41 #include "hw/timer/hpet.h"
42 #include "hw/firmware/smbios.h"
43 #include "hw/loader.h"
44 #include "elf.h"
45 #include "migration/vmstate.h"
46 #include "multiboot.h"
47 #include "hw/rtc/mc146818rtc.h"
48 #include "hw/intc/i8259.h"
49 #include "hw/dma/i8257.h"
50 #include "hw/timer/i8254.h"
51 #include "hw/input/i8042.h"
52 #include "hw/irq.h"
53 #include "hw/audio/pcspk.h"
54 #include "hw/pci/msi.h"
55 #include "hw/sysbus.h"
56 #include "sysemu/sysemu.h"
57 #include "sysemu/tcg.h"
58 #include "sysemu/numa.h"
59 #include "sysemu/kvm.h"
60 #include "sysemu/xen.h"
61 #include "sysemu/qtest.h"
62 #include "sysemu/reset.h"
63 #include "sysemu/runstate.h"
64 #include "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 "exec/address-spaces.h"
70 #include "sysemu/arch_init.h"
71 #include "qemu/bitmap.h"
72 #include "qemu/config-file.h"
73 #include "qemu/error-report.h"
74 #include "qemu/option.h"
75 #include "qemu/cutils.h"
76 #include "hw/acpi/acpi.h"
77 #include "hw/acpi/cpu_hotplug.h"
78 #include "hw/boards.h"
79 #include "acpi-build.h"
80 #include "hw/mem/pc-dimm.h"
81 #include "hw/mem/nvdimm.h"
82 #include "qapi/error.h"
83 #include "qapi/qapi-visit-common.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-pmem-pci.h"
91 #include "hw/mem/memory-device.h"
92 #include "sysemu/replay.h"
93 #include "qapi/qmp/qerror.h"
94 #include "config-devices.h"
95 #include "e820_memory_layout.h"
96 #include "fw_cfg.h"
97 #include "trace.h"
98 
99 GlobalProperty pc_compat_5_0[] = {};
100 const size_t pc_compat_5_0_len = G_N_ELEMENTS(pc_compat_5_0);
101 
102 GlobalProperty pc_compat_4_2[] = {
103     { "mch", "smbase-smram", "off" },
104 };
105 const size_t pc_compat_4_2_len = G_N_ELEMENTS(pc_compat_4_2);
106 
107 GlobalProperty pc_compat_4_1[] = {};
108 const size_t pc_compat_4_1_len = G_N_ELEMENTS(pc_compat_4_1);
109 
110 GlobalProperty pc_compat_4_0[] = {};
111 const size_t pc_compat_4_0_len = G_N_ELEMENTS(pc_compat_4_0);
112 
113 GlobalProperty pc_compat_3_1[] = {
114     { "intel-iommu", "dma-drain", "off" },
115     { "Opteron_G3" "-" TYPE_X86_CPU, "rdtscp", "off" },
116     { "Opteron_G4" "-" TYPE_X86_CPU, "rdtscp", "off" },
117     { "Opteron_G4" "-" TYPE_X86_CPU, "npt", "off" },
118     { "Opteron_G4" "-" TYPE_X86_CPU, "nrip-save", "off" },
119     { "Opteron_G5" "-" TYPE_X86_CPU, "rdtscp", "off" },
120     { "Opteron_G5" "-" TYPE_X86_CPU, "npt", "off" },
121     { "Opteron_G5" "-" TYPE_X86_CPU, "nrip-save", "off" },
122     { "EPYC" "-" TYPE_X86_CPU, "npt", "off" },
123     { "EPYC" "-" TYPE_X86_CPU, "nrip-save", "off" },
124     { "EPYC-IBPB" "-" TYPE_X86_CPU, "npt", "off" },
125     { "EPYC-IBPB" "-" TYPE_X86_CPU, "nrip-save", "off" },
126     { "Skylake-Client" "-" TYPE_X86_CPU,      "mpx", "on" },
127     { "Skylake-Client-IBRS" "-" TYPE_X86_CPU, "mpx", "on" },
128     { "Skylake-Server" "-" TYPE_X86_CPU,      "mpx", "on" },
129     { "Skylake-Server-IBRS" "-" TYPE_X86_CPU, "mpx", "on" },
130     { "Cascadelake-Server" "-" TYPE_X86_CPU,  "mpx", "on" },
131     { "Icelake-Client" "-" TYPE_X86_CPU,      "mpx", "on" },
132     { "Icelake-Server" "-" TYPE_X86_CPU,      "mpx", "on" },
133     { "Cascadelake-Server" "-" TYPE_X86_CPU, "stepping", "5" },
134     { TYPE_X86_CPU, "x-intel-pt-auto-level", "off" },
135 };
136 const size_t pc_compat_3_1_len = G_N_ELEMENTS(pc_compat_3_1);
137 
138 GlobalProperty pc_compat_3_0[] = {
139     { TYPE_X86_CPU, "x-hv-synic-kvm-only", "on" },
140     { "Skylake-Server" "-" TYPE_X86_CPU, "pku", "off" },
141     { "Skylake-Server-IBRS" "-" TYPE_X86_CPU, "pku", "off" },
142 };
143 const size_t pc_compat_3_0_len = G_N_ELEMENTS(pc_compat_3_0);
144 
145 GlobalProperty pc_compat_2_12[] = {
146     { TYPE_X86_CPU, "legacy-cache", "on" },
147     { TYPE_X86_CPU, "topoext", "off" },
148     { "EPYC-" TYPE_X86_CPU, "xlevel", "0x8000000a" },
149     { "EPYC-IBPB-" TYPE_X86_CPU, "xlevel", "0x8000000a" },
150 };
151 const size_t pc_compat_2_12_len = G_N_ELEMENTS(pc_compat_2_12);
152 
153 GlobalProperty pc_compat_2_11[] = {
154     { TYPE_X86_CPU, "x-migrate-smi-count", "off" },
155     { "Skylake-Server" "-" TYPE_X86_CPU, "clflushopt", "off" },
156 };
157 const size_t pc_compat_2_11_len = G_N_ELEMENTS(pc_compat_2_11);
158 
159 GlobalProperty pc_compat_2_10[] = {
160     { TYPE_X86_CPU, "x-hv-max-vps", "0x40" },
161     { "i440FX-pcihost", "x-pci-hole64-fix", "off" },
162     { "q35-pcihost", "x-pci-hole64-fix", "off" },
163 };
164 const size_t pc_compat_2_10_len = G_N_ELEMENTS(pc_compat_2_10);
165 
166 GlobalProperty pc_compat_2_9[] = {
167     { "mch", "extended-tseg-mbytes", "0" },
168 };
169 const size_t pc_compat_2_9_len = G_N_ELEMENTS(pc_compat_2_9);
170 
171 GlobalProperty pc_compat_2_8[] = {
172     { TYPE_X86_CPU, "tcg-cpuid", "off" },
173     { "kvmclock", "x-mach-use-reliable-get-clock", "off" },
174     { "ICH9-LPC", "x-smi-broadcast", "off" },
175     { TYPE_X86_CPU, "vmware-cpuid-freq", "off" },
176     { "Haswell-" TYPE_X86_CPU, "stepping", "1" },
177 };
178 const size_t pc_compat_2_8_len = G_N_ELEMENTS(pc_compat_2_8);
179 
180 GlobalProperty pc_compat_2_7[] = {
181     { TYPE_X86_CPU, "l3-cache", "off" },
182     { TYPE_X86_CPU, "full-cpuid-auto-level", "off" },
183     { "Opteron_G3" "-" TYPE_X86_CPU, "family", "15" },
184     { "Opteron_G3" "-" TYPE_X86_CPU, "model", "6" },
185     { "Opteron_G3" "-" TYPE_X86_CPU, "stepping", "1" },
186     { "isa-pcspk", "migrate", "off" },
187 };
188 const size_t pc_compat_2_7_len = G_N_ELEMENTS(pc_compat_2_7);
189 
190 GlobalProperty pc_compat_2_6[] = {
191     { TYPE_X86_CPU, "cpuid-0xb", "off" },
192     { "vmxnet3", "romfile", "" },
193     { TYPE_X86_CPU, "fill-mtrr-mask", "off" },
194     { "apic-common", "legacy-instance-id", "on", }
195 };
196 const size_t pc_compat_2_6_len = G_N_ELEMENTS(pc_compat_2_6);
197 
198 GlobalProperty pc_compat_2_5[] = {};
199 const size_t pc_compat_2_5_len = G_N_ELEMENTS(pc_compat_2_5);
200 
201 GlobalProperty pc_compat_2_4[] = {
202     PC_CPU_MODEL_IDS("2.4.0")
203     { "Haswell-" TYPE_X86_CPU, "abm", "off" },
204     { "Haswell-noTSX-" TYPE_X86_CPU, "abm", "off" },
205     { "Broadwell-" TYPE_X86_CPU, "abm", "off" },
206     { "Broadwell-noTSX-" TYPE_X86_CPU, "abm", "off" },
207     { "host" "-" TYPE_X86_CPU, "host-cache-info", "on" },
208     { TYPE_X86_CPU, "check", "off" },
209     { "qemu64" "-" TYPE_X86_CPU, "sse4a", "on" },
210     { "qemu64" "-" TYPE_X86_CPU, "abm", "on" },
211     { "qemu64" "-" TYPE_X86_CPU, "popcnt", "on" },
212     { "qemu32" "-" TYPE_X86_CPU, "popcnt", "on" },
213     { "Opteron_G2" "-" TYPE_X86_CPU, "rdtscp", "on" },
214     { "Opteron_G3" "-" TYPE_X86_CPU, "rdtscp", "on" },
215     { "Opteron_G4" "-" TYPE_X86_CPU, "rdtscp", "on" },
216     { "Opteron_G5" "-" TYPE_X86_CPU, "rdtscp", "on", }
217 };
218 const size_t pc_compat_2_4_len = G_N_ELEMENTS(pc_compat_2_4);
219 
220 GlobalProperty pc_compat_2_3[] = {
221     PC_CPU_MODEL_IDS("2.3.0")
222     { TYPE_X86_CPU, "arat", "off" },
223     { "qemu64" "-" TYPE_X86_CPU, "min-level", "4" },
224     { "kvm64" "-" TYPE_X86_CPU, "min-level", "5" },
225     { "pentium3" "-" TYPE_X86_CPU, "min-level", "2" },
226     { "n270" "-" TYPE_X86_CPU, "min-level", "5" },
227     { "Conroe" "-" TYPE_X86_CPU, "min-level", "4" },
228     { "Penryn" "-" TYPE_X86_CPU, "min-level", "4" },
229     { "Nehalem" "-" TYPE_X86_CPU, "min-level", "4" },
230     { "n270" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
231     { "Penryn" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
232     { "Conroe" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
233     { "Nehalem" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
234     { "Westmere" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
235     { "SandyBridge" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
236     { "IvyBridge" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
237     { "Haswell" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
238     { "Haswell-noTSX" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
239     { "Broadwell" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
240     { "Broadwell-noTSX" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
241     { TYPE_X86_CPU, "kvm-no-smi-migration", "on" },
242 };
243 const size_t pc_compat_2_3_len = G_N_ELEMENTS(pc_compat_2_3);
244 
245 GlobalProperty pc_compat_2_2[] = {
246     PC_CPU_MODEL_IDS("2.2.0")
247     { "kvm64" "-" TYPE_X86_CPU, "vme", "off" },
248     { "kvm32" "-" TYPE_X86_CPU, "vme", "off" },
249     { "Conroe" "-" TYPE_X86_CPU, "vme", "off" },
250     { "Penryn" "-" TYPE_X86_CPU, "vme", "off" },
251     { "Nehalem" "-" TYPE_X86_CPU, "vme", "off" },
252     { "Westmere" "-" TYPE_X86_CPU, "vme", "off" },
253     { "SandyBridge" "-" TYPE_X86_CPU, "vme", "off" },
254     { "Haswell" "-" TYPE_X86_CPU, "vme", "off" },
255     { "Broadwell" "-" TYPE_X86_CPU, "vme", "off" },
256     { "Opteron_G1" "-" TYPE_X86_CPU, "vme", "off" },
257     { "Opteron_G2" "-" TYPE_X86_CPU, "vme", "off" },
258     { "Opteron_G3" "-" TYPE_X86_CPU, "vme", "off" },
259     { "Opteron_G4" "-" TYPE_X86_CPU, "vme", "off" },
260     { "Opteron_G5" "-" TYPE_X86_CPU, "vme", "off" },
261     { "Haswell" "-" TYPE_X86_CPU, "f16c", "off" },
262     { "Haswell" "-" TYPE_X86_CPU, "rdrand", "off" },
263     { "Broadwell" "-" TYPE_X86_CPU, "f16c", "off" },
264     { "Broadwell" "-" TYPE_X86_CPU, "rdrand", "off" },
265 };
266 const size_t pc_compat_2_2_len = G_N_ELEMENTS(pc_compat_2_2);
267 
268 GlobalProperty pc_compat_2_1[] = {
269     PC_CPU_MODEL_IDS("2.1.0")
270     { "coreduo" "-" TYPE_X86_CPU, "vmx", "on" },
271     { "core2duo" "-" TYPE_X86_CPU, "vmx", "on" },
272 };
273 const size_t pc_compat_2_1_len = G_N_ELEMENTS(pc_compat_2_1);
274 
275 GlobalProperty pc_compat_2_0[] = {
276     PC_CPU_MODEL_IDS("2.0.0")
277     { "virtio-scsi-pci", "any_layout", "off" },
278     { "PIIX4_PM", "memory-hotplug-support", "off" },
279     { "apic", "version", "0x11" },
280     { "nec-usb-xhci", "superspeed-ports-first", "off" },
281     { "nec-usb-xhci", "force-pcie-endcap", "on" },
282     { "pci-serial", "prog_if", "0" },
283     { "pci-serial-2x", "prog_if", "0" },
284     { "pci-serial-4x", "prog_if", "0" },
285     { "virtio-net-pci", "guest_announce", "off" },
286     { "ICH9-LPC", "memory-hotplug-support", "off" },
287     { "xio3130-downstream", COMPAT_PROP_PCP, "off" },
288     { "ioh3420", COMPAT_PROP_PCP, "off" },
289 };
290 const size_t pc_compat_2_0_len = G_N_ELEMENTS(pc_compat_2_0);
291 
292 GlobalProperty pc_compat_1_7[] = {
293     PC_CPU_MODEL_IDS("1.7.0")
294     { TYPE_USB_DEVICE, "msos-desc", "no" },
295     { "PIIX4_PM", "acpi-pci-hotplug-with-bridge-support", "off" },
296     { "hpet", HPET_INTCAP, "4" },
297 };
298 const size_t pc_compat_1_7_len = G_N_ELEMENTS(pc_compat_1_7);
299 
300 GlobalProperty pc_compat_1_6[] = {
301     PC_CPU_MODEL_IDS("1.6.0")
302     { "e1000", "mitigation", "off" },
303     { "qemu64-" TYPE_X86_CPU, "model", "2" },
304     { "qemu32-" TYPE_X86_CPU, "model", "3" },
305     { "i440FX-pcihost", "short_root_bus", "1" },
306     { "q35-pcihost", "short_root_bus", "1" },
307 };
308 const size_t pc_compat_1_6_len = G_N_ELEMENTS(pc_compat_1_6);
309 
310 GlobalProperty pc_compat_1_5[] = {
311     PC_CPU_MODEL_IDS("1.5.0")
312     { "Conroe-" TYPE_X86_CPU, "model", "2" },
313     { "Conroe-" TYPE_X86_CPU, "min-level", "2" },
314     { "Penryn-" TYPE_X86_CPU, "model", "2" },
315     { "Penryn-" TYPE_X86_CPU, "min-level", "2" },
316     { "Nehalem-" TYPE_X86_CPU, "model", "2" },
317     { "Nehalem-" TYPE_X86_CPU, "min-level", "2" },
318     { "virtio-net-pci", "any_layout", "off" },
319     { TYPE_X86_CPU, "pmu", "on" },
320     { "i440FX-pcihost", "short_root_bus", "0" },
321     { "q35-pcihost", "short_root_bus", "0" },
322 };
323 const size_t pc_compat_1_5_len = G_N_ELEMENTS(pc_compat_1_5);
324 
325 GlobalProperty pc_compat_1_4[] = {
326     PC_CPU_MODEL_IDS("1.4.0")
327     { "scsi-hd", "discard_granularity", "0" },
328     { "scsi-cd", "discard_granularity", "0" },
329     { "scsi-disk", "discard_granularity", "0" },
330     { "ide-hd", "discard_granularity", "0" },
331     { "ide-cd", "discard_granularity", "0" },
332     { "ide-drive", "discard_granularity", "0" },
333     { "virtio-blk-pci", "discard_granularity", "0" },
334     /* DEV_NVECTORS_UNSPECIFIED as a uint32_t string: */
335     { "virtio-serial-pci", "vectors", "0xFFFFFFFF" },
336     { "virtio-net-pci", "ctrl_guest_offloads", "off" },
337     { "e1000", "romfile", "pxe-e1000.rom" },
338     { "ne2k_pci", "romfile", "pxe-ne2k_pci.rom" },
339     { "pcnet", "romfile", "pxe-pcnet.rom" },
340     { "rtl8139", "romfile", "pxe-rtl8139.rom" },
341     { "virtio-net-pci", "romfile", "pxe-virtio.rom" },
342     { "486-" TYPE_X86_CPU, "model", "0" },
343     { "n270" "-" TYPE_X86_CPU, "movbe", "off" },
344     { "Westmere" "-" TYPE_X86_CPU, "pclmulqdq", "off" },
345 };
346 const size_t pc_compat_1_4_len = G_N_ELEMENTS(pc_compat_1_4);
347 
348 GSIState *pc_gsi_create(qemu_irq **irqs, bool pci_enabled)
349 {
350     GSIState *s;
351 
352     s = g_new0(GSIState, 1);
353     if (kvm_ioapic_in_kernel()) {
354         kvm_pc_setup_irq_routing(pci_enabled);
355     }
356     *irqs = qemu_allocate_irqs(gsi_handler, s, GSI_NUM_PINS);
357 
358     return s;
359 }
360 
361 static void ioport80_write(void *opaque, hwaddr addr, uint64_t data,
362                            unsigned size)
363 {
364 }
365 
366 static uint64_t ioport80_read(void *opaque, hwaddr addr, unsigned size)
367 {
368     return 0xffffffffffffffffULL;
369 }
370 
371 /* MSDOS compatibility mode FPU exception support */
372 static void ioportF0_write(void *opaque, hwaddr addr, uint64_t data,
373                            unsigned size)
374 {
375     if (tcg_enabled()) {
376         cpu_set_ignne();
377     }
378 }
379 
380 static uint64_t ioportF0_read(void *opaque, hwaddr addr, unsigned size)
381 {
382     return 0xffffffffffffffffULL;
383 }
384 
385 /* PC cmos mappings */
386 
387 #define REG_EQUIPMENT_BYTE          0x14
388 
389 static void cmos_init_hd(ISADevice *s, int type_ofs, int info_ofs,
390                          int16_t cylinders, int8_t heads, int8_t sectors)
391 {
392     rtc_set_memory(s, type_ofs, 47);
393     rtc_set_memory(s, info_ofs, cylinders);
394     rtc_set_memory(s, info_ofs + 1, cylinders >> 8);
395     rtc_set_memory(s, info_ofs + 2, heads);
396     rtc_set_memory(s, info_ofs + 3, 0xff);
397     rtc_set_memory(s, info_ofs + 4, 0xff);
398     rtc_set_memory(s, info_ofs + 5, 0xc0 | ((heads > 8) << 3));
399     rtc_set_memory(s, info_ofs + 6, cylinders);
400     rtc_set_memory(s, info_ofs + 7, cylinders >> 8);
401     rtc_set_memory(s, info_ofs + 8, sectors);
402 }
403 
404 /* convert boot_device letter to something recognizable by the bios */
405 static int boot_device2nibble(char boot_device)
406 {
407     switch(boot_device) {
408     case 'a':
409     case 'b':
410         return 0x01; /* floppy boot */
411     case 'c':
412         return 0x02; /* hard drive boot */
413     case 'd':
414         return 0x03; /* CD-ROM boot */
415     case 'n':
416         return 0x04; /* Network boot */
417     }
418     return 0;
419 }
420 
421 static void set_boot_dev(ISADevice *s, const char *boot_device, Error **errp)
422 {
423 #define PC_MAX_BOOT_DEVICES 3
424     int nbds, bds[3] = { 0, };
425     int i;
426 
427     nbds = strlen(boot_device);
428     if (nbds > PC_MAX_BOOT_DEVICES) {
429         error_setg(errp, "Too many boot devices for PC");
430         return;
431     }
432     for (i = 0; i < nbds; i++) {
433         bds[i] = boot_device2nibble(boot_device[i]);
434         if (bds[i] == 0) {
435             error_setg(errp, "Invalid boot device for PC: '%c'",
436                        boot_device[i]);
437             return;
438         }
439     }
440     rtc_set_memory(s, 0x3d, (bds[1] << 4) | bds[0]);
441     rtc_set_memory(s, 0x38, (bds[2] << 4) | (fd_bootchk ? 0x0 : 0x1));
442 }
443 
444 static void pc_boot_set(void *opaque, const char *boot_device, Error **errp)
445 {
446     set_boot_dev(opaque, boot_device, errp);
447 }
448 
449 static void pc_cmos_init_floppy(ISADevice *rtc_state, ISADevice *floppy)
450 {
451     int val, nb, i;
452     FloppyDriveType fd_type[2] = { FLOPPY_DRIVE_TYPE_NONE,
453                                    FLOPPY_DRIVE_TYPE_NONE };
454 
455     /* floppy type */
456     if (floppy) {
457         for (i = 0; i < 2; i++) {
458             fd_type[i] = isa_fdc_get_drive_type(floppy, i);
459         }
460     }
461     val = (cmos_get_fd_drive_type(fd_type[0]) << 4) |
462         cmos_get_fd_drive_type(fd_type[1]);
463     rtc_set_memory(rtc_state, 0x10, val);
464 
465     val = rtc_get_memory(rtc_state, REG_EQUIPMENT_BYTE);
466     nb = 0;
467     if (fd_type[0] != FLOPPY_DRIVE_TYPE_NONE) {
468         nb++;
469     }
470     if (fd_type[1] != FLOPPY_DRIVE_TYPE_NONE) {
471         nb++;
472     }
473     switch (nb) {
474     case 0:
475         break;
476     case 1:
477         val |= 0x01; /* 1 drive, ready for boot */
478         break;
479     case 2:
480         val |= 0x41; /* 2 drives, ready for boot */
481         break;
482     }
483     rtc_set_memory(rtc_state, REG_EQUIPMENT_BYTE, val);
484 }
485 
486 typedef struct pc_cmos_init_late_arg {
487     ISADevice *rtc_state;
488     BusState *idebus[2];
489 } pc_cmos_init_late_arg;
490 
491 typedef struct check_fdc_state {
492     ISADevice *floppy;
493     bool multiple;
494 } CheckFdcState;
495 
496 static int check_fdc(Object *obj, void *opaque)
497 {
498     CheckFdcState *state = opaque;
499     Object *fdc;
500     uint32_t iobase;
501     Error *local_err = NULL;
502 
503     fdc = object_dynamic_cast(obj, TYPE_ISA_FDC);
504     if (!fdc) {
505         return 0;
506     }
507 
508     iobase = object_property_get_uint(obj, "iobase", &local_err);
509     if (local_err || iobase != 0x3f0) {
510         error_free(local_err);
511         return 0;
512     }
513 
514     if (state->floppy) {
515         state->multiple = true;
516     } else {
517         state->floppy = ISA_DEVICE(obj);
518     }
519     return 0;
520 }
521 
522 static const char * const fdc_container_path[] = {
523     "/unattached", "/peripheral", "/peripheral-anon"
524 };
525 
526 /*
527  * Locate the FDC at IO address 0x3f0, in order to configure the CMOS registers
528  * and ACPI objects.
529  */
530 ISADevice *pc_find_fdc0(void)
531 {
532     int i;
533     Object *container;
534     CheckFdcState state = { 0 };
535 
536     for (i = 0; i < ARRAY_SIZE(fdc_container_path); i++) {
537         container = container_get(qdev_get_machine(), fdc_container_path[i]);
538         object_child_foreach(container, check_fdc, &state);
539     }
540 
541     if (state.multiple) {
542         warn_report("multiple floppy disk controllers with "
543                     "iobase=0x3f0 have been found");
544         error_printf("the one being picked for CMOS setup might not reflect "
545                      "your intent");
546     }
547 
548     return state.floppy;
549 }
550 
551 static void pc_cmos_init_late(void *opaque)
552 {
553     pc_cmos_init_late_arg *arg = opaque;
554     ISADevice *s = arg->rtc_state;
555     int16_t cylinders;
556     int8_t heads, sectors;
557     int val;
558     int i, trans;
559 
560     val = 0;
561     if (arg->idebus[0] && ide_get_geometry(arg->idebus[0], 0,
562                                            &cylinders, &heads, &sectors) >= 0) {
563         cmos_init_hd(s, 0x19, 0x1b, cylinders, heads, sectors);
564         val |= 0xf0;
565     }
566     if (arg->idebus[0] && ide_get_geometry(arg->idebus[0], 1,
567                                            &cylinders, &heads, &sectors) >= 0) {
568         cmos_init_hd(s, 0x1a, 0x24, cylinders, heads, sectors);
569         val |= 0x0f;
570     }
571     rtc_set_memory(s, 0x12, val);
572 
573     val = 0;
574     for (i = 0; i < 4; i++) {
575         /* NOTE: ide_get_geometry() returns the physical
576            geometry.  It is always such that: 1 <= sects <= 63, 1
577            <= heads <= 16, 1 <= cylinders <= 16383. The BIOS
578            geometry can be different if a translation is done. */
579         if (arg->idebus[i / 2] &&
580             ide_get_geometry(arg->idebus[i / 2], i % 2,
581                              &cylinders, &heads, &sectors) >= 0) {
582             trans = ide_get_bios_chs_trans(arg->idebus[i / 2], i % 2) - 1;
583             assert((trans & ~3) == 0);
584             val |= trans << (i * 2);
585         }
586     }
587     rtc_set_memory(s, 0x39, val);
588 
589     pc_cmos_init_floppy(s, pc_find_fdc0());
590 
591     qemu_unregister_reset(pc_cmos_init_late, opaque);
592 }
593 
594 void pc_cmos_init(PCMachineState *pcms,
595                   BusState *idebus0, BusState *idebus1,
596                   ISADevice *s)
597 {
598     int val;
599     static pc_cmos_init_late_arg arg;
600     X86MachineState *x86ms = X86_MACHINE(pcms);
601 
602     /* various important CMOS locations needed by PC/Bochs bios */
603 
604     /* memory size */
605     /* base memory (first MiB) */
606     val = MIN(x86ms->below_4g_mem_size / KiB, 640);
607     rtc_set_memory(s, 0x15, val);
608     rtc_set_memory(s, 0x16, val >> 8);
609     /* extended memory (next 64MiB) */
610     if (x86ms->below_4g_mem_size > 1 * MiB) {
611         val = (x86ms->below_4g_mem_size - 1 * MiB) / KiB;
612     } else {
613         val = 0;
614     }
615     if (val > 65535)
616         val = 65535;
617     rtc_set_memory(s, 0x17, val);
618     rtc_set_memory(s, 0x18, val >> 8);
619     rtc_set_memory(s, 0x30, val);
620     rtc_set_memory(s, 0x31, val >> 8);
621     /* memory between 16MiB and 4GiB */
622     if (x86ms->below_4g_mem_size > 16 * MiB) {
623         val = (x86ms->below_4g_mem_size - 16 * MiB) / (64 * KiB);
624     } else {
625         val = 0;
626     }
627     if (val > 65535)
628         val = 65535;
629     rtc_set_memory(s, 0x34, val);
630     rtc_set_memory(s, 0x35, val >> 8);
631     /* memory above 4GiB */
632     val = x86ms->above_4g_mem_size / 65536;
633     rtc_set_memory(s, 0x5b, val);
634     rtc_set_memory(s, 0x5c, val >> 8);
635     rtc_set_memory(s, 0x5d, val >> 16);
636 
637     object_property_add_link(OBJECT(pcms), "rtc_state",
638                              TYPE_ISA_DEVICE,
639                              (Object **)&x86ms->rtc,
640                              object_property_allow_set_link,
641                              OBJ_PROP_LINK_STRONG);
642     object_property_set_link(OBJECT(pcms), OBJECT(s),
643                              "rtc_state", &error_abort);
644 
645     set_boot_dev(s, MACHINE(pcms)->boot_order, &error_fatal);
646 
647     val = 0;
648     val |= 0x02; /* FPU is there */
649     val |= 0x04; /* PS/2 mouse installed */
650     rtc_set_memory(s, REG_EQUIPMENT_BYTE, val);
651 
652     /* hard drives and FDC */
653     arg.rtc_state = s;
654     arg.idebus[0] = idebus0;
655     arg.idebus[1] = idebus1;
656     qemu_register_reset(pc_cmos_init_late, &arg);
657 }
658 
659 static void handle_a20_line_change(void *opaque, int irq, int level)
660 {
661     X86CPU *cpu = opaque;
662 
663     /* XXX: send to all CPUs ? */
664     /* XXX: add logic to handle multiple A20 line sources */
665     x86_cpu_set_a20(cpu, level);
666 }
667 
668 #define NE2000_NB_MAX 6
669 
670 static const int ne2000_io[NE2000_NB_MAX] = { 0x300, 0x320, 0x340, 0x360,
671                                               0x280, 0x380 };
672 static const int ne2000_irq[NE2000_NB_MAX] = { 9, 10, 11, 3, 4, 5 };
673 
674 void pc_init_ne2k_isa(ISABus *bus, NICInfo *nd)
675 {
676     static int nb_ne2k = 0;
677 
678     if (nb_ne2k == NE2000_NB_MAX)
679         return;
680     isa_ne2000_init(bus, ne2000_io[nb_ne2k],
681                     ne2000_irq[nb_ne2k], nd);
682     nb_ne2k++;
683 }
684 
685 void pc_acpi_smi_interrupt(void *opaque, int irq, int level)
686 {
687     X86CPU *cpu = opaque;
688 
689     if (level) {
690         cpu_interrupt(CPU(cpu), CPU_INTERRUPT_SMI);
691     }
692 }
693 
694 /*
695  * This function is very similar to smp_parse()
696  * in hw/core/machine.c but includes CPU die support.
697  */
698 void pc_smp_parse(MachineState *ms, QemuOpts *opts)
699 {
700     X86MachineState *x86ms = X86_MACHINE(ms);
701 
702     if (opts) {
703         unsigned cpus    = qemu_opt_get_number(opts, "cpus", 0);
704         unsigned sockets = qemu_opt_get_number(opts, "sockets", 0);
705         unsigned dies = qemu_opt_get_number(opts, "dies", 1);
706         unsigned cores   = qemu_opt_get_number(opts, "cores", 0);
707         unsigned threads = qemu_opt_get_number(opts, "threads", 0);
708 
709         /* compute missing values, prefer sockets over cores over threads */
710         if (cpus == 0 || sockets == 0) {
711             cores = cores > 0 ? cores : 1;
712             threads = threads > 0 ? threads : 1;
713             if (cpus == 0) {
714                 sockets = sockets > 0 ? sockets : 1;
715                 cpus = cores * threads * dies * sockets;
716             } else {
717                 ms->smp.max_cpus =
718                         qemu_opt_get_number(opts, "maxcpus", cpus);
719                 sockets = ms->smp.max_cpus / (cores * threads * dies);
720             }
721         } else if (cores == 0) {
722             threads = threads > 0 ? threads : 1;
723             cores = cpus / (sockets * dies * threads);
724             cores = cores > 0 ? cores : 1;
725         } else if (threads == 0) {
726             threads = cpus / (cores * dies * sockets);
727             threads = threads > 0 ? threads : 1;
728         } else if (sockets * dies * cores * threads < cpus) {
729             error_report("cpu topology: "
730                          "sockets (%u) * dies (%u) * cores (%u) * threads (%u) < "
731                          "smp_cpus (%u)",
732                          sockets, dies, cores, threads, cpus);
733             exit(1);
734         }
735 
736         ms->smp.max_cpus =
737                 qemu_opt_get_number(opts, "maxcpus", cpus);
738 
739         if (ms->smp.max_cpus < cpus) {
740             error_report("maxcpus must be equal to or greater than smp");
741             exit(1);
742         }
743 
744         if (sockets * dies * cores * threads > ms->smp.max_cpus) {
745             error_report("cpu topology: "
746                          "sockets (%u) * dies (%u) * cores (%u) * threads (%u) > "
747                          "maxcpus (%u)",
748                          sockets, dies, cores, threads,
749                          ms->smp.max_cpus);
750             exit(1);
751         }
752 
753         if (sockets * dies * cores * threads != ms->smp.max_cpus) {
754             warn_report("Invalid CPU topology deprecated: "
755                         "sockets (%u) * dies (%u) * cores (%u) * threads (%u) "
756                         "!= maxcpus (%u)",
757                         sockets, dies, cores, threads,
758                         ms->smp.max_cpus);
759         }
760 
761         ms->smp.cpus = cpus;
762         ms->smp.cores = cores;
763         ms->smp.threads = threads;
764         ms->smp.sockets = sockets;
765         x86ms->smp_dies = dies;
766     }
767 
768     if (ms->smp.cpus > 1) {
769         Error *blocker = NULL;
770         error_setg(&blocker, QERR_REPLAY_NOT_SUPPORTED, "smp");
771         replay_add_blocker(blocker);
772     }
773 }
774 
775 void pc_hot_add_cpu(MachineState *ms, const int64_t id, Error **errp)
776 {
777     X86MachineState *x86ms = X86_MACHINE(ms);
778     int64_t apic_id = x86_cpu_apic_id_from_index(x86ms, id);
779     Error *local_err = NULL;
780 
781     if (id < 0) {
782         error_setg(errp, "Invalid CPU id: %" PRIi64, id);
783         return;
784     }
785 
786     if (apic_id >= ACPI_CPU_HOTPLUG_ID_LIMIT) {
787         error_setg(errp, "Unable to add CPU: %" PRIi64
788                    ", resulting APIC ID (%" PRIi64 ") is too large",
789                    id, apic_id);
790         return;
791     }
792 
793 
794     x86_cpu_new(X86_MACHINE(ms), apic_id, &local_err);
795     if (local_err) {
796         error_propagate(errp, local_err);
797         return;
798     }
799 }
800 
801 static void rtc_set_cpus_count(ISADevice *rtc, uint16_t cpus_count)
802 {
803     if (cpus_count > 0xff) {
804         /* If the number of CPUs can't be represented in 8 bits, the
805          * BIOS must use "FW_CFG_NB_CPUS". Set RTC field to 0 just
806          * to make old BIOSes fail more predictably.
807          */
808         rtc_set_memory(rtc, 0x5f, 0);
809     } else {
810         rtc_set_memory(rtc, 0x5f, cpus_count - 1);
811     }
812 }
813 
814 static
815 void pc_machine_done(Notifier *notifier, void *data)
816 {
817     PCMachineState *pcms = container_of(notifier,
818                                         PCMachineState, machine_done);
819     X86MachineState *x86ms = X86_MACHINE(pcms);
820     PCIBus *bus = pcms->bus;
821 
822     /* set the number of CPUs */
823     rtc_set_cpus_count(x86ms->rtc, x86ms->boot_cpus);
824 
825     if (bus) {
826         int extra_hosts = 0;
827 
828         QLIST_FOREACH(bus, &bus->child, sibling) {
829             /* look for expander root buses */
830             if (pci_bus_is_root(bus)) {
831                 extra_hosts++;
832             }
833         }
834         if (extra_hosts && x86ms->fw_cfg) {
835             uint64_t *val = g_malloc(sizeof(*val));
836             *val = cpu_to_le64(extra_hosts);
837             fw_cfg_add_file(x86ms->fw_cfg,
838                     "etc/extra-pci-roots", val, sizeof(*val));
839         }
840     }
841 
842     acpi_setup();
843     if (x86ms->fw_cfg) {
844         fw_cfg_build_smbios(MACHINE(pcms), x86ms->fw_cfg);
845         fw_cfg_build_feature_control(MACHINE(pcms), x86ms->fw_cfg);
846         /* update FW_CFG_NB_CPUS to account for -device added CPUs */
847         fw_cfg_modify_i16(x86ms->fw_cfg, FW_CFG_NB_CPUS, x86ms->boot_cpus);
848     }
849 
850     if (x86ms->apic_id_limit > 255 && !xen_enabled()) {
851         IntelIOMMUState *iommu = INTEL_IOMMU_DEVICE(x86_iommu_get_default());
852 
853         if (!iommu || !x86_iommu_ir_supported(X86_IOMMU_DEVICE(iommu)) ||
854             iommu->intr_eim != ON_OFF_AUTO_ON) {
855             error_report("current -smp configuration requires "
856                          "Extended Interrupt Mode enabled. "
857                          "You can add an IOMMU using: "
858                          "-device intel-iommu,intremap=on,eim=on");
859             exit(EXIT_FAILURE);
860         }
861     }
862 }
863 
864 void pc_guest_info_init(PCMachineState *pcms)
865 {
866     int i;
867     MachineState *ms = MACHINE(pcms);
868     X86MachineState *x86ms = X86_MACHINE(pcms);
869 
870     x86ms->apic_xrupt_override = kvm_allows_irq0_override();
871     pcms->numa_nodes = ms->numa_state->num_nodes;
872     pcms->node_mem = g_malloc0(pcms->numa_nodes *
873                                     sizeof *pcms->node_mem);
874     for (i = 0; i < ms->numa_state->num_nodes; i++) {
875         pcms->node_mem[i] = ms->numa_state->nodes[i].node_mem;
876     }
877 
878     pcms->machine_done.notify = pc_machine_done;
879     qemu_add_machine_init_done_notifier(&pcms->machine_done);
880 }
881 
882 /* setup pci memory address space mapping into system address space */
883 void pc_pci_as_mapping_init(Object *owner, MemoryRegion *system_memory,
884                             MemoryRegion *pci_address_space)
885 {
886     /* Set to lower priority than RAM */
887     memory_region_add_subregion_overlap(system_memory, 0x0,
888                                         pci_address_space, -1);
889 }
890 
891 void xen_load_linux(PCMachineState *pcms)
892 {
893     int i;
894     FWCfgState *fw_cfg;
895     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
896     X86MachineState *x86ms = X86_MACHINE(pcms);
897 
898     assert(MACHINE(pcms)->kernel_filename != NULL);
899 
900     fw_cfg = fw_cfg_init_io(FW_CFG_IO_BASE);
901     fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, x86ms->boot_cpus);
902     rom_set_fw(fw_cfg);
903 
904     x86_load_linux(x86ms, fw_cfg, pcmc->acpi_data_size,
905                    pcmc->pvh_enabled, pcmc->linuxboot_dma_enabled);
906     for (i = 0; i < nb_option_roms; i++) {
907         assert(!strcmp(option_rom[i].name, "linuxboot.bin") ||
908                !strcmp(option_rom[i].name, "linuxboot_dma.bin") ||
909                !strcmp(option_rom[i].name, "pvh.bin") ||
910                !strcmp(option_rom[i].name, "multiboot.bin"));
911         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
912     }
913     x86ms->fw_cfg = fw_cfg;
914 }
915 
916 void pc_memory_init(PCMachineState *pcms,
917                     MemoryRegion *system_memory,
918                     MemoryRegion *rom_memory,
919                     MemoryRegion **ram_memory)
920 {
921     int linux_boot, i;
922     MemoryRegion *option_rom_mr;
923     MemoryRegion *ram_below_4g, *ram_above_4g;
924     FWCfgState *fw_cfg;
925     MachineState *machine = MACHINE(pcms);
926     MachineClass *mc = MACHINE_GET_CLASS(machine);
927     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
928     X86MachineState *x86ms = X86_MACHINE(pcms);
929 
930     assert(machine->ram_size == x86ms->below_4g_mem_size +
931                                 x86ms->above_4g_mem_size);
932 
933     linux_boot = (machine->kernel_filename != NULL);
934 
935     /*
936      * Split single memory region and use aliases to address portions of it,
937      * done for backwards compatibility with older qemus.
938      */
939     *ram_memory = machine->ram;
940     ram_below_4g = g_malloc(sizeof(*ram_below_4g));
941     memory_region_init_alias(ram_below_4g, NULL, "ram-below-4g", machine->ram,
942                              0, x86ms->below_4g_mem_size);
943     memory_region_add_subregion(system_memory, 0, ram_below_4g);
944     e820_add_entry(0, x86ms->below_4g_mem_size, E820_RAM);
945     if (x86ms->above_4g_mem_size > 0) {
946         ram_above_4g = g_malloc(sizeof(*ram_above_4g));
947         memory_region_init_alias(ram_above_4g, NULL, "ram-above-4g",
948                                  machine->ram,
949                                  x86ms->below_4g_mem_size,
950                                  x86ms->above_4g_mem_size);
951         memory_region_add_subregion(system_memory, 0x100000000ULL,
952                                     ram_above_4g);
953         e820_add_entry(0x100000000ULL, x86ms->above_4g_mem_size, E820_RAM);
954     }
955 
956     if (!pcmc->has_reserved_memory &&
957         (machine->ram_slots ||
958          (machine->maxram_size > machine->ram_size))) {
959 
960         error_report("\"-memory 'slots|maxmem'\" is not supported by: %s",
961                      mc->name);
962         exit(EXIT_FAILURE);
963     }
964 
965     /* always allocate the device memory information */
966     machine->device_memory = g_malloc0(sizeof(*machine->device_memory));
967 
968     /* initialize device memory address space */
969     if (pcmc->has_reserved_memory &&
970         (machine->ram_size < machine->maxram_size)) {
971         ram_addr_t device_mem_size = machine->maxram_size - machine->ram_size;
972 
973         if (machine->ram_slots > ACPI_MAX_RAM_SLOTS) {
974             error_report("unsupported amount of memory slots: %"PRIu64,
975                          machine->ram_slots);
976             exit(EXIT_FAILURE);
977         }
978 
979         if (QEMU_ALIGN_UP(machine->maxram_size,
980                           TARGET_PAGE_SIZE) != machine->maxram_size) {
981             error_report("maximum memory size must by aligned to multiple of "
982                          "%d bytes", TARGET_PAGE_SIZE);
983             exit(EXIT_FAILURE);
984         }
985 
986         machine->device_memory->base =
987             ROUND_UP(0x100000000ULL + x86ms->above_4g_mem_size, 1 * GiB);
988 
989         if (pcmc->enforce_aligned_dimm) {
990             /* size device region assuming 1G page max alignment per slot */
991             device_mem_size += (1 * GiB) * machine->ram_slots;
992         }
993 
994         if ((machine->device_memory->base + device_mem_size) <
995             device_mem_size) {
996             error_report("unsupported amount of maximum memory: " RAM_ADDR_FMT,
997                          machine->maxram_size);
998             exit(EXIT_FAILURE);
999         }
1000 
1001         memory_region_init(&machine->device_memory->mr, OBJECT(pcms),
1002                            "device-memory", device_mem_size);
1003         memory_region_add_subregion(system_memory, machine->device_memory->base,
1004                                     &machine->device_memory->mr);
1005     }
1006 
1007     /* Initialize PC system firmware */
1008     pc_system_firmware_init(pcms, rom_memory);
1009 
1010     option_rom_mr = g_malloc(sizeof(*option_rom_mr));
1011     memory_region_init_ram(option_rom_mr, NULL, "pc.rom", PC_ROM_SIZE,
1012                            &error_fatal);
1013     if (pcmc->pci_enabled) {
1014         memory_region_set_readonly(option_rom_mr, true);
1015     }
1016     memory_region_add_subregion_overlap(rom_memory,
1017                                         PC_ROM_MIN_VGA,
1018                                         option_rom_mr,
1019                                         1);
1020 
1021     fw_cfg = fw_cfg_arch_create(machine,
1022                                 x86ms->boot_cpus, x86ms->apic_id_limit);
1023 
1024     rom_set_fw(fw_cfg);
1025 
1026     if (pcmc->has_reserved_memory && machine->device_memory->base) {
1027         uint64_t *val = g_malloc(sizeof(*val));
1028         PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1029         uint64_t res_mem_end = machine->device_memory->base;
1030 
1031         if (!pcmc->broken_reserved_end) {
1032             res_mem_end += memory_region_size(&machine->device_memory->mr);
1033         }
1034         *val = cpu_to_le64(ROUND_UP(res_mem_end, 1 * GiB));
1035         fw_cfg_add_file(fw_cfg, "etc/reserved-memory-end", val, sizeof(*val));
1036     }
1037 
1038     if (linux_boot) {
1039         x86_load_linux(x86ms, fw_cfg, pcmc->acpi_data_size,
1040                        pcmc->pvh_enabled, pcmc->linuxboot_dma_enabled);
1041     }
1042 
1043     for (i = 0; i < nb_option_roms; i++) {
1044         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1045     }
1046     x86ms->fw_cfg = fw_cfg;
1047 
1048     /* Init default IOAPIC address space */
1049     x86ms->ioapic_as = &address_space_memory;
1050 
1051     /* Init ACPI memory hotplug IO base address */
1052     pcms->memhp_io_base = ACPI_MEMORY_HOTPLUG_BASE;
1053 }
1054 
1055 /*
1056  * The 64bit pci hole starts after "above 4G RAM" and
1057  * potentially the space reserved for memory hotplug.
1058  */
1059 uint64_t pc_pci_hole64_start(void)
1060 {
1061     PCMachineState *pcms = PC_MACHINE(qdev_get_machine());
1062     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1063     MachineState *ms = MACHINE(pcms);
1064     X86MachineState *x86ms = X86_MACHINE(pcms);
1065     uint64_t hole64_start = 0;
1066 
1067     if (pcmc->has_reserved_memory && ms->device_memory->base) {
1068         hole64_start = ms->device_memory->base;
1069         if (!pcmc->broken_reserved_end) {
1070             hole64_start += memory_region_size(&ms->device_memory->mr);
1071         }
1072     } else {
1073         hole64_start = 0x100000000ULL + x86ms->above_4g_mem_size;
1074     }
1075 
1076     return ROUND_UP(hole64_start, 1 * GiB);
1077 }
1078 
1079 DeviceState *pc_vga_init(ISABus *isa_bus, PCIBus *pci_bus)
1080 {
1081     DeviceState *dev = NULL;
1082 
1083     rom_set_order_override(FW_CFG_ORDER_OVERRIDE_VGA);
1084     if (pci_bus) {
1085         PCIDevice *pcidev = pci_vga_init(pci_bus);
1086         dev = pcidev ? &pcidev->qdev : NULL;
1087     } else if (isa_bus) {
1088         ISADevice *isadev = isa_vga_init(isa_bus);
1089         dev = isadev ? DEVICE(isadev) : NULL;
1090     }
1091     rom_reset_order_override();
1092     return dev;
1093 }
1094 
1095 static const MemoryRegionOps ioport80_io_ops = {
1096     .write = ioport80_write,
1097     .read = ioport80_read,
1098     .endianness = DEVICE_NATIVE_ENDIAN,
1099     .impl = {
1100         .min_access_size = 1,
1101         .max_access_size = 1,
1102     },
1103 };
1104 
1105 static const MemoryRegionOps ioportF0_io_ops = {
1106     .write = ioportF0_write,
1107     .read = ioportF0_read,
1108     .endianness = DEVICE_NATIVE_ENDIAN,
1109     .impl = {
1110         .min_access_size = 1,
1111         .max_access_size = 1,
1112     },
1113 };
1114 
1115 static void pc_superio_init(ISABus *isa_bus, bool create_fdctrl, bool no_vmport)
1116 {
1117     int i;
1118     DriveInfo *fd[MAX_FD];
1119     qemu_irq *a20_line;
1120     ISADevice *fdc, *i8042, *port92, *vmmouse;
1121 
1122     serial_hds_isa_init(isa_bus, 0, MAX_ISA_SERIAL_PORTS);
1123     parallel_hds_isa_init(isa_bus, MAX_PARALLEL_PORTS);
1124 
1125     for (i = 0; i < MAX_FD; i++) {
1126         fd[i] = drive_get(IF_FLOPPY, 0, i);
1127         create_fdctrl |= !!fd[i];
1128     }
1129     if (create_fdctrl) {
1130         fdc = isa_new(TYPE_ISA_FDC);
1131         if (fdc) {
1132             isa_realize_and_unref(fdc, isa_bus, &error_fatal);
1133             isa_fdc_init_drives(fdc, fd);
1134         }
1135     }
1136 
1137     i8042 = isa_create_simple(isa_bus, "i8042");
1138     if (!no_vmport) {
1139         isa_create_simple(isa_bus, TYPE_VMPORT);
1140         vmmouse = isa_try_new("vmmouse");
1141     } else {
1142         vmmouse = NULL;
1143     }
1144     if (vmmouse) {
1145         object_property_set_link(OBJECT(vmmouse), OBJECT(i8042),
1146                                  "i8042", &error_abort);
1147         isa_realize_and_unref(vmmouse, isa_bus, &error_fatal);
1148     }
1149     port92 = isa_create_simple(isa_bus, TYPE_PORT92);
1150 
1151     a20_line = qemu_allocate_irqs(handle_a20_line_change, first_cpu, 2);
1152     i8042_setup_a20_line(i8042, a20_line[0]);
1153     qdev_connect_gpio_out_named(DEVICE(port92),
1154                                 PORT92_A20_LINE, 0, a20_line[1]);
1155     g_free(a20_line);
1156 }
1157 
1158 void pc_basic_device_init(ISABus *isa_bus, qemu_irq *gsi,
1159                           ISADevice **rtc_state,
1160                           bool create_fdctrl,
1161                           bool no_vmport,
1162                           bool has_pit,
1163                           uint32_t hpet_irqs)
1164 {
1165     int i;
1166     DeviceState *hpet = NULL;
1167     int pit_isa_irq = 0;
1168     qemu_irq pit_alt_irq = NULL;
1169     qemu_irq rtc_irq = NULL;
1170     ISADevice *pit = NULL;
1171     MemoryRegion *ioport80_io = g_new(MemoryRegion, 1);
1172     MemoryRegion *ioportF0_io = g_new(MemoryRegion, 1);
1173 
1174     memory_region_init_io(ioport80_io, NULL, &ioport80_io_ops, NULL, "ioport80", 1);
1175     memory_region_add_subregion(isa_bus->address_space_io, 0x80, ioport80_io);
1176 
1177     memory_region_init_io(ioportF0_io, NULL, &ioportF0_io_ops, NULL, "ioportF0", 1);
1178     memory_region_add_subregion(isa_bus->address_space_io, 0xf0, ioportF0_io);
1179 
1180     /*
1181      * Check if an HPET shall be created.
1182      *
1183      * Without KVM_CAP_PIT_STATE2, we cannot switch off the in-kernel PIT
1184      * when the HPET wants to take over. Thus we have to disable the latter.
1185      */
1186     if (!no_hpet && (!kvm_irqchip_in_kernel() || kvm_has_pit_state2())) {
1187         hpet = qdev_try_new(TYPE_HPET);
1188         if (hpet) {
1189             /* For pc-piix-*, hpet's intcap is always IRQ2. For pc-q35-1.7
1190              * and earlier, use IRQ2 for compat. Otherwise, use IRQ16~23,
1191              * IRQ8 and IRQ2.
1192              */
1193             uint8_t compat = object_property_get_uint(OBJECT(hpet),
1194                     HPET_INTCAP, NULL);
1195             if (!compat) {
1196                 qdev_prop_set_uint32(hpet, HPET_INTCAP, hpet_irqs);
1197             }
1198             sysbus_realize_and_unref(SYS_BUS_DEVICE(hpet), &error_fatal);
1199             sysbus_mmio_map(SYS_BUS_DEVICE(hpet), 0, HPET_BASE);
1200 
1201             for (i = 0; i < GSI_NUM_PINS; i++) {
1202                 sysbus_connect_irq(SYS_BUS_DEVICE(hpet), i, gsi[i]);
1203             }
1204             pit_isa_irq = -1;
1205             pit_alt_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_PIT_INT);
1206             rtc_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_RTC_INT);
1207         }
1208     }
1209     *rtc_state = mc146818_rtc_init(isa_bus, 2000, rtc_irq);
1210 
1211     qemu_register_boot_set(pc_boot_set, *rtc_state);
1212 
1213     if (!xen_enabled() && has_pit) {
1214         if (kvm_pit_in_kernel()) {
1215             pit = kvm_pit_init(isa_bus, 0x40);
1216         } else {
1217             pit = i8254_pit_init(isa_bus, 0x40, pit_isa_irq, pit_alt_irq);
1218         }
1219         if (hpet) {
1220             /* connect PIT to output control line of the HPET */
1221             qdev_connect_gpio_out(hpet, 0, qdev_get_gpio_in(DEVICE(pit), 0));
1222         }
1223         pcspk_init(isa_bus, pit);
1224     }
1225 
1226     i8257_dma_init(isa_bus, 0);
1227 
1228     /* Super I/O */
1229     pc_superio_init(isa_bus, create_fdctrl, no_vmport);
1230 }
1231 
1232 void pc_nic_init(PCMachineClass *pcmc, ISABus *isa_bus, PCIBus *pci_bus)
1233 {
1234     int i;
1235 
1236     rom_set_order_override(FW_CFG_ORDER_OVERRIDE_NIC);
1237     for (i = 0; i < nb_nics; i++) {
1238         NICInfo *nd = &nd_table[i];
1239         const char *model = nd->model ? nd->model : pcmc->default_nic_model;
1240 
1241         if (g_str_equal(model, "ne2k_isa")) {
1242             pc_init_ne2k_isa(isa_bus, nd);
1243         } else {
1244             pci_nic_init_nofail(nd, pci_bus, model, NULL);
1245         }
1246     }
1247     rom_reset_order_override();
1248 }
1249 
1250 void pc_i8259_create(ISABus *isa_bus, qemu_irq *i8259_irqs)
1251 {
1252     qemu_irq *i8259;
1253 
1254     if (kvm_pic_in_kernel()) {
1255         i8259 = kvm_i8259_init(isa_bus);
1256     } else if (xen_enabled()) {
1257         i8259 = xen_interrupt_controller_init();
1258     } else {
1259         i8259 = i8259_init(isa_bus, x86_allocate_cpu_irq());
1260     }
1261 
1262     for (size_t i = 0; i < ISA_NUM_IRQS; i++) {
1263         i8259_irqs[i] = i8259[i];
1264     }
1265 
1266     g_free(i8259);
1267 }
1268 
1269 static void pc_memory_pre_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
1270                                Error **errp)
1271 {
1272     const PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1273     const PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1274     const MachineState *ms = MACHINE(hotplug_dev);
1275     const bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
1276     const uint64_t legacy_align = TARGET_PAGE_SIZE;
1277     Error *local_err = NULL;
1278 
1279     /*
1280      * When -no-acpi is used with Q35 machine type, no ACPI is built,
1281      * but pcms->acpi_dev is still created. Check !acpi_enabled in
1282      * addition to cover this case.
1283      */
1284     if (!pcms->acpi_dev || !x86_machine_is_acpi_enabled(X86_MACHINE(pcms))) {
1285         error_setg(errp,
1286                    "memory hotplug is not enabled: missing acpi device or acpi disabled");
1287         return;
1288     }
1289 
1290     if (is_nvdimm && !ms->nvdimms_state->is_enabled) {
1291         error_setg(errp, "nvdimm is not enabled: missing 'nvdimm' in '-M'");
1292         return;
1293     }
1294 
1295     hotplug_handler_pre_plug(pcms->acpi_dev, dev, &local_err);
1296     if (local_err) {
1297         error_propagate(errp, local_err);
1298         return;
1299     }
1300 
1301     pc_dimm_pre_plug(PC_DIMM(dev), MACHINE(hotplug_dev),
1302                      pcmc->enforce_aligned_dimm ? NULL : &legacy_align, errp);
1303 }
1304 
1305 static void pc_memory_plug(HotplugHandler *hotplug_dev,
1306                            DeviceState *dev, Error **errp)
1307 {
1308     Error *local_err = NULL;
1309     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1310     MachineState *ms = MACHINE(hotplug_dev);
1311     bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
1312 
1313     pc_dimm_plug(PC_DIMM(dev), MACHINE(pcms), &local_err);
1314     if (local_err) {
1315         goto out;
1316     }
1317 
1318     if (is_nvdimm) {
1319         nvdimm_plug(ms->nvdimms_state);
1320     }
1321 
1322     hotplug_handler_plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &error_abort);
1323 out:
1324     error_propagate(errp, local_err);
1325 }
1326 
1327 static void pc_memory_unplug_request(HotplugHandler *hotplug_dev,
1328                                      DeviceState *dev, Error **errp)
1329 {
1330     Error *local_err = NULL;
1331     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1332 
1333     /*
1334      * When -no-acpi is used with Q35 machine type, no ACPI is built,
1335      * but pcms->acpi_dev is still created. Check !acpi_enabled in
1336      * addition to cover this case.
1337      */
1338     if (!pcms->acpi_dev || !x86_machine_is_acpi_enabled(X86_MACHINE(pcms))) {
1339         error_setg(&local_err,
1340                    "memory hotplug is not enabled: missing acpi device or acpi disabled");
1341         goto out;
1342     }
1343 
1344     if (object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM)) {
1345         error_setg(&local_err,
1346                    "nvdimm device hot unplug is not supported yet.");
1347         goto out;
1348     }
1349 
1350     hotplug_handler_unplug_request(HOTPLUG_HANDLER(pcms->acpi_dev), dev,
1351                                    &local_err);
1352 out:
1353     error_propagate(errp, local_err);
1354 }
1355 
1356 static void pc_memory_unplug(HotplugHandler *hotplug_dev,
1357                              DeviceState *dev, Error **errp)
1358 {
1359     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1360     Error *local_err = NULL;
1361 
1362     hotplug_handler_unplug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
1363     if (local_err) {
1364         goto out;
1365     }
1366 
1367     pc_dimm_unplug(PC_DIMM(dev), MACHINE(pcms));
1368     qdev_unrealize(dev);
1369  out:
1370     error_propagate(errp, local_err);
1371 }
1372 
1373 static int pc_apic_cmp(const void *a, const void *b)
1374 {
1375    CPUArchId *apic_a = (CPUArchId *)a;
1376    CPUArchId *apic_b = (CPUArchId *)b;
1377 
1378    return apic_a->arch_id - apic_b->arch_id;
1379 }
1380 
1381 /* returns pointer to CPUArchId descriptor that matches CPU's apic_id
1382  * in ms->possible_cpus->cpus, if ms->possible_cpus->cpus has no
1383  * entry corresponding to CPU's apic_id returns NULL.
1384  */
1385 static CPUArchId *pc_find_cpu_slot(MachineState *ms, uint32_t id, int *idx)
1386 {
1387     CPUArchId apic_id, *found_cpu;
1388 
1389     apic_id.arch_id = id;
1390     found_cpu = bsearch(&apic_id, ms->possible_cpus->cpus,
1391         ms->possible_cpus->len, sizeof(*ms->possible_cpus->cpus),
1392         pc_apic_cmp);
1393     if (found_cpu && idx) {
1394         *idx = found_cpu - ms->possible_cpus->cpus;
1395     }
1396     return found_cpu;
1397 }
1398 
1399 static void pc_cpu_plug(HotplugHandler *hotplug_dev,
1400                         DeviceState *dev, Error **errp)
1401 {
1402     CPUArchId *found_cpu;
1403     Error *local_err = NULL;
1404     X86CPU *cpu = X86_CPU(dev);
1405     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1406     X86MachineState *x86ms = X86_MACHINE(pcms);
1407 
1408     if (pcms->acpi_dev) {
1409         hotplug_handler_plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
1410         if (local_err) {
1411             goto out;
1412         }
1413     }
1414 
1415     /* increment the number of CPUs */
1416     x86ms->boot_cpus++;
1417     if (x86ms->rtc) {
1418         rtc_set_cpus_count(x86ms->rtc, x86ms->boot_cpus);
1419     }
1420     if (x86ms->fw_cfg) {
1421         fw_cfg_modify_i16(x86ms->fw_cfg, FW_CFG_NB_CPUS, x86ms->boot_cpus);
1422     }
1423 
1424     found_cpu = pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, NULL);
1425     found_cpu->cpu = OBJECT(dev);
1426 out:
1427     error_propagate(errp, local_err);
1428 }
1429 static void pc_cpu_unplug_request_cb(HotplugHandler *hotplug_dev,
1430                                      DeviceState *dev, Error **errp)
1431 {
1432     int idx = -1;
1433     Error *local_err = NULL;
1434     X86CPU *cpu = X86_CPU(dev);
1435     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1436 
1437     if (!pcms->acpi_dev) {
1438         error_setg(&local_err, "CPU hot unplug not supported without ACPI");
1439         goto out;
1440     }
1441 
1442     pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, &idx);
1443     assert(idx != -1);
1444     if (idx == 0) {
1445         error_setg(&local_err, "Boot CPU is unpluggable");
1446         goto out;
1447     }
1448 
1449     hotplug_handler_unplug_request(HOTPLUG_HANDLER(pcms->acpi_dev), dev,
1450                                    &local_err);
1451     if (local_err) {
1452         goto out;
1453     }
1454 
1455  out:
1456     error_propagate(errp, local_err);
1457 
1458 }
1459 
1460 static void pc_cpu_unplug_cb(HotplugHandler *hotplug_dev,
1461                              DeviceState *dev, Error **errp)
1462 {
1463     CPUArchId *found_cpu;
1464     Error *local_err = NULL;
1465     X86CPU *cpu = X86_CPU(dev);
1466     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1467     X86MachineState *x86ms = X86_MACHINE(pcms);
1468 
1469     hotplug_handler_unplug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
1470     if (local_err) {
1471         goto out;
1472     }
1473 
1474     found_cpu = pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, NULL);
1475     found_cpu->cpu = NULL;
1476     qdev_unrealize(dev);
1477 
1478     /* decrement the number of CPUs */
1479     x86ms->boot_cpus--;
1480     /* Update the number of CPUs in CMOS */
1481     rtc_set_cpus_count(x86ms->rtc, x86ms->boot_cpus);
1482     fw_cfg_modify_i16(x86ms->fw_cfg, FW_CFG_NB_CPUS, x86ms->boot_cpus);
1483  out:
1484     error_propagate(errp, local_err);
1485 }
1486 
1487 static void pc_cpu_pre_plug(HotplugHandler *hotplug_dev,
1488                             DeviceState *dev, Error **errp)
1489 {
1490     int idx;
1491     CPUState *cs;
1492     CPUArchId *cpu_slot;
1493     X86CPUTopoIDs topo_ids;
1494     X86CPU *cpu = X86_CPU(dev);
1495     CPUX86State *env = &cpu->env;
1496     MachineState *ms = MACHINE(hotplug_dev);
1497     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1498     X86MachineState *x86ms = X86_MACHINE(pcms);
1499     unsigned int smp_cores = ms->smp.cores;
1500     unsigned int smp_threads = ms->smp.threads;
1501     X86CPUTopoInfo topo_info;
1502 
1503     if(!object_dynamic_cast(OBJECT(cpu), ms->cpu_type)) {
1504         error_setg(errp, "Invalid CPU type, expected cpu type: '%s'",
1505                    ms->cpu_type);
1506         return;
1507     }
1508 
1509     init_topo_info(&topo_info, x86ms);
1510 
1511     env->nr_dies = x86ms->smp_dies;
1512     env->nr_nodes = topo_info.nodes_per_pkg;
1513     env->pkg_offset = x86ms->apicid_pkg_offset(&topo_info);
1514 
1515     /*
1516      * If APIC ID is not set,
1517      * set it based on socket/die/core/thread properties.
1518      */
1519     if (cpu->apic_id == UNASSIGNED_APIC_ID) {
1520         int max_socket = (ms->smp.max_cpus - 1) /
1521                                 smp_threads / smp_cores / x86ms->smp_dies;
1522 
1523         /*
1524          * die-id was optional in QEMU 4.0 and older, so keep it optional
1525          * if there's only one die per socket.
1526          */
1527         if (cpu->die_id < 0 && x86ms->smp_dies == 1) {
1528             cpu->die_id = 0;
1529         }
1530 
1531         if (cpu->socket_id < 0) {
1532             error_setg(errp, "CPU socket-id is not set");
1533             return;
1534         } else if (cpu->socket_id > max_socket) {
1535             error_setg(errp, "Invalid CPU socket-id: %u must be in range 0:%u",
1536                        cpu->socket_id, max_socket);
1537             return;
1538         }
1539         if (cpu->die_id < 0) {
1540             error_setg(errp, "CPU die-id is not set");
1541             return;
1542         } else if (cpu->die_id > x86ms->smp_dies - 1) {
1543             error_setg(errp, "Invalid CPU die-id: %u must be in range 0:%u",
1544                        cpu->die_id, x86ms->smp_dies - 1);
1545             return;
1546         }
1547         if (cpu->core_id < 0) {
1548             error_setg(errp, "CPU core-id is not set");
1549             return;
1550         } else if (cpu->core_id > (smp_cores - 1)) {
1551             error_setg(errp, "Invalid CPU core-id: %u must be in range 0:%u",
1552                        cpu->core_id, smp_cores - 1);
1553             return;
1554         }
1555         if (cpu->thread_id < 0) {
1556             error_setg(errp, "CPU thread-id is not set");
1557             return;
1558         } else if (cpu->thread_id > (smp_threads - 1)) {
1559             error_setg(errp, "Invalid CPU thread-id: %u must be in range 0:%u",
1560                        cpu->thread_id, smp_threads - 1);
1561             return;
1562         }
1563 
1564         topo_ids.pkg_id = cpu->socket_id;
1565         topo_ids.die_id = cpu->die_id;
1566         topo_ids.core_id = cpu->core_id;
1567         topo_ids.smt_id = cpu->thread_id;
1568         cpu->apic_id = x86ms->apicid_from_topo_ids(&topo_info, &topo_ids);
1569     }
1570 
1571     cpu_slot = pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, &idx);
1572     if (!cpu_slot) {
1573         MachineState *ms = MACHINE(pcms);
1574 
1575         x86ms->topo_ids_from_apicid(cpu->apic_id, &topo_info, &topo_ids);
1576         error_setg(errp,
1577             "Invalid CPU [socket: %u, die: %u, core: %u, thread: %u] with"
1578             " APIC ID %" PRIu32 ", valid index range 0:%d",
1579             topo_ids.pkg_id, topo_ids.die_id, topo_ids.core_id, topo_ids.smt_id,
1580             cpu->apic_id, ms->possible_cpus->len - 1);
1581         return;
1582     }
1583 
1584     if (cpu_slot->cpu) {
1585         error_setg(errp, "CPU[%d] with APIC ID %" PRIu32 " exists",
1586                    idx, cpu->apic_id);
1587         return;
1588     }
1589 
1590     /* if 'address' properties socket-id/core-id/thread-id are not set, set them
1591      * so that machine_query_hotpluggable_cpus would show correct values
1592      */
1593     /* TODO: move socket_id/core_id/thread_id checks into x86_cpu_realizefn()
1594      * once -smp refactoring is complete and there will be CPU private
1595      * CPUState::nr_cores and CPUState::nr_threads fields instead of globals */
1596     x86ms->topo_ids_from_apicid(cpu->apic_id, &topo_info, &topo_ids);
1597     if (cpu->socket_id != -1 && cpu->socket_id != topo_ids.pkg_id) {
1598         error_setg(errp, "property socket-id: %u doesn't match set apic-id:"
1599             " 0x%x (socket-id: %u)", cpu->socket_id, cpu->apic_id,
1600             topo_ids.pkg_id);
1601         return;
1602     }
1603     cpu->socket_id = topo_ids.pkg_id;
1604 
1605     if (cpu->die_id != -1 && cpu->die_id != topo_ids.die_id) {
1606         error_setg(errp, "property die-id: %u doesn't match set apic-id:"
1607             " 0x%x (die-id: %u)", cpu->die_id, cpu->apic_id, topo_ids.die_id);
1608         return;
1609     }
1610     cpu->die_id = topo_ids.die_id;
1611 
1612     if (cpu->core_id != -1 && cpu->core_id != topo_ids.core_id) {
1613         error_setg(errp, "property core-id: %u doesn't match set apic-id:"
1614             " 0x%x (core-id: %u)", cpu->core_id, cpu->apic_id,
1615             topo_ids.core_id);
1616         return;
1617     }
1618     cpu->core_id = topo_ids.core_id;
1619 
1620     if (cpu->thread_id != -1 && cpu->thread_id != topo_ids.smt_id) {
1621         error_setg(errp, "property thread-id: %u doesn't match set apic-id:"
1622             " 0x%x (thread-id: %u)", cpu->thread_id, cpu->apic_id,
1623             topo_ids.smt_id);
1624         return;
1625     }
1626     cpu->thread_id = topo_ids.smt_id;
1627 
1628     if (hyperv_feat_enabled(cpu, HYPERV_FEAT_VPINDEX) &&
1629         !kvm_hv_vpindex_settable()) {
1630         error_setg(errp, "kernel doesn't allow setting HyperV VP_INDEX");
1631         return;
1632     }
1633 
1634     cs = CPU(cpu);
1635     cs->cpu_index = idx;
1636 
1637     numa_cpu_pre_plug(cpu_slot, dev, errp);
1638 }
1639 
1640 static void pc_virtio_pmem_pci_pre_plug(HotplugHandler *hotplug_dev,
1641                                         DeviceState *dev, Error **errp)
1642 {
1643     HotplugHandler *hotplug_dev2 = qdev_get_bus_hotplug_handler(dev);
1644     Error *local_err = NULL;
1645 
1646     if (!hotplug_dev2) {
1647         /*
1648          * Without a bus hotplug handler, we cannot control the plug/unplug
1649          * order. This should never be the case on x86, however better add
1650          * a safety net.
1651          */
1652         error_setg(errp, "virtio-pmem-pci not supported on this bus.");
1653         return;
1654     }
1655     /*
1656      * First, see if we can plug this memory device at all. If that
1657      * succeeds, branch of to the actual hotplug handler.
1658      */
1659     memory_device_pre_plug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev), NULL,
1660                            &local_err);
1661     if (!local_err) {
1662         hotplug_handler_pre_plug(hotplug_dev2, dev, &local_err);
1663     }
1664     error_propagate(errp, local_err);
1665 }
1666 
1667 static void pc_virtio_pmem_pci_plug(HotplugHandler *hotplug_dev,
1668                                     DeviceState *dev, Error **errp)
1669 {
1670     HotplugHandler *hotplug_dev2 = qdev_get_bus_hotplug_handler(dev);
1671     Error *local_err = NULL;
1672 
1673     /*
1674      * Plug the memory device first and then branch off to the actual
1675      * hotplug handler. If that one fails, we can easily undo the memory
1676      * device bits.
1677      */
1678     memory_device_plug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev));
1679     hotplug_handler_plug(hotplug_dev2, dev, &local_err);
1680     if (local_err) {
1681         memory_device_unplug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev));
1682     }
1683     error_propagate(errp, local_err);
1684 }
1685 
1686 static void pc_virtio_pmem_pci_unplug_request(HotplugHandler *hotplug_dev,
1687                                               DeviceState *dev, Error **errp)
1688 {
1689     /* We don't support virtio pmem hot unplug */
1690     error_setg(errp, "virtio pmem device unplug not supported.");
1691 }
1692 
1693 static void pc_virtio_pmem_pci_unplug(HotplugHandler *hotplug_dev,
1694                                       DeviceState *dev, Error **errp)
1695 {
1696     /* We don't support virtio pmem hot unplug */
1697 }
1698 
1699 static void pc_machine_device_pre_plug_cb(HotplugHandler *hotplug_dev,
1700                                           DeviceState *dev, Error **errp)
1701 {
1702     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1703         pc_memory_pre_plug(hotplug_dev, dev, errp);
1704     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1705         pc_cpu_pre_plug(hotplug_dev, dev, errp);
1706     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI)) {
1707         pc_virtio_pmem_pci_pre_plug(hotplug_dev, dev, errp);
1708     }
1709 }
1710 
1711 static void pc_machine_device_plug_cb(HotplugHandler *hotplug_dev,
1712                                       DeviceState *dev, Error **errp)
1713 {
1714     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1715         pc_memory_plug(hotplug_dev, dev, errp);
1716     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1717         pc_cpu_plug(hotplug_dev, dev, errp);
1718     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI)) {
1719         pc_virtio_pmem_pci_plug(hotplug_dev, dev, errp);
1720     }
1721 }
1722 
1723 static void pc_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
1724                                                 DeviceState *dev, Error **errp)
1725 {
1726     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1727         pc_memory_unplug_request(hotplug_dev, dev, errp);
1728     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1729         pc_cpu_unplug_request_cb(hotplug_dev, dev, errp);
1730     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI)) {
1731         pc_virtio_pmem_pci_unplug_request(hotplug_dev, dev, errp);
1732     } else {
1733         error_setg(errp, "acpi: device unplug request for not supported device"
1734                    " type: %s", object_get_typename(OBJECT(dev)));
1735     }
1736 }
1737 
1738 static void pc_machine_device_unplug_cb(HotplugHandler *hotplug_dev,
1739                                         DeviceState *dev, Error **errp)
1740 {
1741     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1742         pc_memory_unplug(hotplug_dev, dev, errp);
1743     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1744         pc_cpu_unplug_cb(hotplug_dev, dev, errp);
1745     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI)) {
1746         pc_virtio_pmem_pci_unplug(hotplug_dev, dev, errp);
1747     } else {
1748         error_setg(errp, "acpi: device unplug for not supported device"
1749                    " type: %s", object_get_typename(OBJECT(dev)));
1750     }
1751 }
1752 
1753 static HotplugHandler *pc_get_hotplug_handler(MachineState *machine,
1754                                              DeviceState *dev)
1755 {
1756     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) ||
1757         object_dynamic_cast(OBJECT(dev), TYPE_CPU) ||
1758         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI)) {
1759         return HOTPLUG_HANDLER(machine);
1760     }
1761 
1762     return NULL;
1763 }
1764 
1765 static void
1766 pc_machine_get_device_memory_region_size(Object *obj, Visitor *v,
1767                                          const char *name, void *opaque,
1768                                          Error **errp)
1769 {
1770     MachineState *ms = MACHINE(obj);
1771     int64_t value = 0;
1772 
1773     if (ms->device_memory) {
1774         value = memory_region_size(&ms->device_memory->mr);
1775     }
1776 
1777     visit_type_int(v, name, &value, errp);
1778 }
1779 
1780 static void pc_machine_get_vmport(Object *obj, Visitor *v, const char *name,
1781                                   void *opaque, Error **errp)
1782 {
1783     PCMachineState *pcms = PC_MACHINE(obj);
1784     OnOffAuto vmport = pcms->vmport;
1785 
1786     visit_type_OnOffAuto(v, name, &vmport, errp);
1787 }
1788 
1789 static void pc_machine_set_vmport(Object *obj, Visitor *v, const char *name,
1790                                   void *opaque, Error **errp)
1791 {
1792     PCMachineState *pcms = PC_MACHINE(obj);
1793 
1794     visit_type_OnOffAuto(v, name, &pcms->vmport, errp);
1795 }
1796 
1797 static bool pc_machine_get_smbus(Object *obj, Error **errp)
1798 {
1799     PCMachineState *pcms = PC_MACHINE(obj);
1800 
1801     return pcms->smbus_enabled;
1802 }
1803 
1804 static void pc_machine_set_smbus(Object *obj, bool value, Error **errp)
1805 {
1806     PCMachineState *pcms = PC_MACHINE(obj);
1807 
1808     pcms->smbus_enabled = value;
1809 }
1810 
1811 static bool pc_machine_get_sata(Object *obj, Error **errp)
1812 {
1813     PCMachineState *pcms = PC_MACHINE(obj);
1814 
1815     return pcms->sata_enabled;
1816 }
1817 
1818 static void pc_machine_set_sata(Object *obj, bool value, Error **errp)
1819 {
1820     PCMachineState *pcms = PC_MACHINE(obj);
1821 
1822     pcms->sata_enabled = value;
1823 }
1824 
1825 static bool pc_machine_get_pit(Object *obj, Error **errp)
1826 {
1827     PCMachineState *pcms = PC_MACHINE(obj);
1828 
1829     return pcms->pit_enabled;
1830 }
1831 
1832 static void pc_machine_set_pit(Object *obj, bool value, Error **errp)
1833 {
1834     PCMachineState *pcms = PC_MACHINE(obj);
1835 
1836     pcms->pit_enabled = value;
1837 }
1838 
1839 static void pc_machine_get_max_ram_below_4g(Object *obj, Visitor *v,
1840                                             const char *name, void *opaque,
1841                                             Error **errp)
1842 {
1843     PCMachineState *pcms = PC_MACHINE(obj);
1844     uint64_t value = pcms->max_ram_below_4g;
1845 
1846     visit_type_size(v, name, &value, errp);
1847 }
1848 
1849 static void pc_machine_set_max_ram_below_4g(Object *obj, Visitor *v,
1850                                             const char *name, void *opaque,
1851                                             Error **errp)
1852 {
1853     PCMachineState *pcms = PC_MACHINE(obj);
1854     Error *error = NULL;
1855     uint64_t value;
1856 
1857     visit_type_size(v, name, &value, &error);
1858     if (error) {
1859         error_propagate(errp, error);
1860         return;
1861     }
1862     if (value > 4 * GiB) {
1863         error_setg(&error,
1864                    "Machine option 'max-ram-below-4g=%"PRIu64
1865                    "' expects size less than or equal to 4G", value);
1866         error_propagate(errp, error);
1867         return;
1868     }
1869 
1870     if (value < 1 * MiB) {
1871         warn_report("Only %" PRIu64 " bytes of RAM below the 4GiB boundary,"
1872                     "BIOS may not work with less than 1MiB", value);
1873     }
1874 
1875     pcms->max_ram_below_4g = value;
1876 }
1877 
1878 static void pc_machine_initfn(Object *obj)
1879 {
1880     PCMachineState *pcms = PC_MACHINE(obj);
1881 
1882 #ifdef CONFIG_VMPORT
1883     pcms->vmport = ON_OFF_AUTO_AUTO;
1884 #else
1885     pcms->vmport = ON_OFF_AUTO_OFF;
1886 #endif /* CONFIG_VMPORT */
1887     pcms->max_ram_below_4g = 0; /* use default */
1888     /* acpi build is enabled by default if machine supports it */
1889     pcms->acpi_build_enabled = PC_MACHINE_GET_CLASS(pcms)->has_acpi_build;
1890     pcms->smbus_enabled = true;
1891     pcms->sata_enabled = true;
1892     pcms->pit_enabled = true;
1893 
1894     pc_system_flash_create(pcms);
1895 }
1896 
1897 static void pc_machine_reset(MachineState *machine)
1898 {
1899     CPUState *cs;
1900     X86CPU *cpu;
1901 
1902     qemu_devices_reset();
1903 
1904     /* Reset APIC after devices have been reset to cancel
1905      * any changes that qemu_devices_reset() might have done.
1906      */
1907     CPU_FOREACH(cs) {
1908         cpu = X86_CPU(cs);
1909 
1910         if (cpu->apic_state) {
1911             device_legacy_reset(cpu->apic_state);
1912         }
1913     }
1914 }
1915 
1916 static void pc_machine_wakeup(MachineState *machine)
1917 {
1918     cpu_synchronize_all_states();
1919     pc_machine_reset(machine);
1920     cpu_synchronize_all_post_reset();
1921 }
1922 
1923 static bool pc_hotplug_allowed(MachineState *ms, DeviceState *dev, Error **errp)
1924 {
1925     X86IOMMUState *iommu = x86_iommu_get_default();
1926     IntelIOMMUState *intel_iommu;
1927 
1928     if (iommu &&
1929         object_dynamic_cast((Object *)iommu, TYPE_INTEL_IOMMU_DEVICE) &&
1930         object_dynamic_cast((Object *)dev, "vfio-pci")) {
1931         intel_iommu = INTEL_IOMMU_DEVICE(iommu);
1932         if (!intel_iommu->caching_mode) {
1933             error_setg(errp, "Device assignment is not allowed without "
1934                        "enabling caching-mode=on for Intel IOMMU.");
1935             return false;
1936         }
1937     }
1938 
1939     return true;
1940 }
1941 
1942 static void pc_machine_class_init(ObjectClass *oc, void *data)
1943 {
1944     MachineClass *mc = MACHINE_CLASS(oc);
1945     PCMachineClass *pcmc = PC_MACHINE_CLASS(oc);
1946     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
1947 
1948     pcmc->pci_enabled = true;
1949     pcmc->has_acpi_build = true;
1950     pcmc->rsdp_in_ram = true;
1951     pcmc->smbios_defaults = true;
1952     pcmc->smbios_uuid_encoded = true;
1953     pcmc->gigabyte_align = true;
1954     pcmc->has_reserved_memory = true;
1955     pcmc->kvmclock_enabled = true;
1956     pcmc->enforce_aligned_dimm = true;
1957     /* BIOS ACPI tables: 128K. Other BIOS datastructures: less than 4K reported
1958      * to be used at the moment, 32K should be enough for a while.  */
1959     pcmc->acpi_data_size = 0x20000 + 0x8000;
1960     pcmc->linuxboot_dma_enabled = true;
1961     pcmc->pvh_enabled = true;
1962     assert(!mc->get_hotplug_handler);
1963     mc->get_hotplug_handler = pc_get_hotplug_handler;
1964     mc->hotplug_allowed = pc_hotplug_allowed;
1965     mc->cpu_index_to_instance_props = x86_cpu_index_to_props;
1966     mc->get_default_cpu_node_id = x86_get_default_cpu_node_id;
1967     mc->possible_cpu_arch_ids = x86_possible_cpu_arch_ids;
1968     mc->auto_enable_numa_with_memhp = true;
1969     mc->has_hotpluggable_cpus = true;
1970     mc->default_boot_order = "cad";
1971     mc->hot_add_cpu = pc_hot_add_cpu;
1972     mc->smp_parse = pc_smp_parse;
1973     mc->block_default_type = IF_IDE;
1974     mc->max_cpus = 255;
1975     mc->reset = pc_machine_reset;
1976     mc->wakeup = pc_machine_wakeup;
1977     hc->pre_plug = pc_machine_device_pre_plug_cb;
1978     hc->plug = pc_machine_device_plug_cb;
1979     hc->unplug_request = pc_machine_device_unplug_request_cb;
1980     hc->unplug = pc_machine_device_unplug_cb;
1981     mc->default_cpu_type = TARGET_DEFAULT_CPU_TYPE;
1982     mc->nvdimm_supported = true;
1983     mc->default_ram_id = "pc.ram";
1984 
1985     object_class_property_add(oc, PC_MACHINE_MAX_RAM_BELOW_4G, "size",
1986         pc_machine_get_max_ram_below_4g, pc_machine_set_max_ram_below_4g,
1987         NULL, NULL);
1988     object_class_property_set_description(oc, PC_MACHINE_MAX_RAM_BELOW_4G,
1989         "Maximum ram below the 4G boundary (32bit boundary)");
1990 
1991     object_class_property_add(oc, PC_MACHINE_DEVMEM_REGION_SIZE, "int",
1992         pc_machine_get_device_memory_region_size, NULL,
1993         NULL, NULL);
1994 
1995     object_class_property_add(oc, PC_MACHINE_VMPORT, "OnOffAuto",
1996         pc_machine_get_vmport, pc_machine_set_vmport,
1997         NULL, NULL);
1998     object_class_property_set_description(oc, PC_MACHINE_VMPORT,
1999         "Enable vmport (pc & q35)");
2000 
2001     object_class_property_add_bool(oc, PC_MACHINE_SMBUS,
2002         pc_machine_get_smbus, pc_machine_set_smbus);
2003 
2004     object_class_property_add_bool(oc, PC_MACHINE_SATA,
2005         pc_machine_get_sata, pc_machine_set_sata);
2006 
2007     object_class_property_add_bool(oc, PC_MACHINE_PIT,
2008         pc_machine_get_pit, pc_machine_set_pit);
2009 }
2010 
2011 static const TypeInfo pc_machine_info = {
2012     .name = TYPE_PC_MACHINE,
2013     .parent = TYPE_X86_MACHINE,
2014     .abstract = true,
2015     .instance_size = sizeof(PCMachineState),
2016     .instance_init = pc_machine_initfn,
2017     .class_size = sizeof(PCMachineClass),
2018     .class_init = pc_machine_class_init,
2019     .interfaces = (InterfaceInfo[]) {
2020          { TYPE_HOTPLUG_HANDLER },
2021          { }
2022     },
2023 };
2024 
2025 static void pc_machine_register_types(void)
2026 {
2027     type_register_static(&pc_machine_info);
2028 }
2029 
2030 type_init(pc_machine_register_types)
2031