xref: /openbmc/qemu/hw/i386/pc.c (revision 9e9b10c6)
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/pc.h"
28 #include "hw/char/serial.h"
29 #include "hw/char/parallel.h"
30 #include "hw/i386/apic.h"
31 #include "hw/i386/topology.h"
32 #include "hw/i386/fw_cfg.h"
33 #include "sysemu/cpus.h"
34 #include "hw/block/fdc.h"
35 #include "hw/ide.h"
36 #include "hw/pci/pci.h"
37 #include "hw/pci/pci_bus.h"
38 #include "hw/nvram/fw_cfg.h"
39 #include "hw/timer/hpet.h"
40 #include "hw/firmware/smbios.h"
41 #include "hw/loader.h"
42 #include "elf.h"
43 #include "migration/vmstate.h"
44 #include "multiboot.h"
45 #include "hw/timer/mc146818rtc.h"
46 #include "hw/dma/i8257.h"
47 #include "hw/timer/i8254.h"
48 #include "hw/input/i8042.h"
49 #include "hw/irq.h"
50 #include "hw/audio/pcspk.h"
51 #include "hw/pci/msi.h"
52 #include "hw/sysbus.h"
53 #include "sysemu/sysemu.h"
54 #include "sysemu/tcg.h"
55 #include "sysemu/numa.h"
56 #include "sysemu/kvm.h"
57 #include "sysemu/qtest.h"
58 #include "sysemu/reset.h"
59 #include "sysemu/runstate.h"
60 #include "kvm_i386.h"
61 #include "hw/xen/xen.h"
62 #include "hw/xen/start_info.h"
63 #include "ui/qemu-spice.h"
64 #include "exec/memory.h"
65 #include "exec/address-spaces.h"
66 #include "sysemu/arch_init.h"
67 #include "qemu/bitmap.h"
68 #include "qemu/config-file.h"
69 #include "qemu/error-report.h"
70 #include "qemu/option.h"
71 #include "hw/acpi/acpi.h"
72 #include "hw/acpi/cpu_hotplug.h"
73 #include "hw/boards.h"
74 #include "acpi-build.h"
75 #include "hw/mem/pc-dimm.h"
76 #include "qapi/error.h"
77 #include "qapi/qapi-visit-common.h"
78 #include "qapi/visitor.h"
79 #include "qom/cpu.h"
80 #include "hw/nmi.h"
81 #include "hw/usb.h"
82 #include "hw/i386/intel_iommu.h"
83 #include "hw/net/ne2000-isa.h"
84 #include "standard-headers/asm-x86/bootparam.h"
85 #include "hw/virtio/virtio-pmem-pci.h"
86 #include "hw/mem/memory-device.h"
87 #include "sysemu/replay.h"
88 #include "qapi/qmp/qerror.h"
89 #include "config-devices.h"
90 
91 /* debug PC/ISA interrupts */
92 //#define DEBUG_IRQ
93 
94 #ifdef DEBUG_IRQ
95 #define DPRINTF(fmt, ...)                                       \
96     do { printf("CPUIRQ: " fmt , ## __VA_ARGS__); } while (0)
97 #else
98 #define DPRINTF(fmt, ...)
99 #endif
100 
101 #define E820_NR_ENTRIES		16
102 
103 struct e820_entry {
104     uint64_t address;
105     uint64_t length;
106     uint32_t type;
107 } QEMU_PACKED __attribute((__aligned__(4)));
108 
109 struct e820_table {
110     uint32_t count;
111     struct e820_entry entry[E820_NR_ENTRIES];
112 } QEMU_PACKED __attribute((__aligned__(4)));
113 
114 static struct e820_table e820_reserve;
115 static struct e820_entry *e820_table;
116 static unsigned e820_entries;
117 struct hpet_fw_config hpet_cfg = {.count = UINT8_MAX};
118 
119 /* Physical Address of PVH entry point read from kernel ELF NOTE */
120 static size_t pvh_start_addr;
121 
122 GlobalProperty pc_compat_4_0[] = {};
123 const size_t pc_compat_4_0_len = G_N_ELEMENTS(pc_compat_4_0);
124 
125 GlobalProperty pc_compat_3_1[] = {
126     { "intel-iommu", "dma-drain", "off" },
127     { "Opteron_G3" "-" TYPE_X86_CPU, "rdtscp", "off" },
128     { "Opteron_G4" "-" TYPE_X86_CPU, "rdtscp", "off" },
129     { "Opteron_G4" "-" TYPE_X86_CPU, "npt", "off" },
130     { "Opteron_G4" "-" TYPE_X86_CPU, "nrip-save", "off" },
131     { "Opteron_G5" "-" TYPE_X86_CPU, "rdtscp", "off" },
132     { "Opteron_G5" "-" TYPE_X86_CPU, "npt", "off" },
133     { "Opteron_G5" "-" TYPE_X86_CPU, "nrip-save", "off" },
134     { "EPYC" "-" TYPE_X86_CPU, "npt", "off" },
135     { "EPYC" "-" TYPE_X86_CPU, "nrip-save", "off" },
136     { "EPYC-IBPB" "-" TYPE_X86_CPU, "npt", "off" },
137     { "EPYC-IBPB" "-" TYPE_X86_CPU, "nrip-save", "off" },
138     { "Skylake-Client" "-" TYPE_X86_CPU,      "mpx", "on" },
139     { "Skylake-Client-IBRS" "-" TYPE_X86_CPU, "mpx", "on" },
140     { "Skylake-Server" "-" TYPE_X86_CPU,      "mpx", "on" },
141     { "Skylake-Server-IBRS" "-" TYPE_X86_CPU, "mpx", "on" },
142     { "Cascadelake-Server" "-" TYPE_X86_CPU,  "mpx", "on" },
143     { "Icelake-Client" "-" TYPE_X86_CPU,      "mpx", "on" },
144     { "Icelake-Server" "-" TYPE_X86_CPU,      "mpx", "on" },
145     { "Cascadelake-Server" "-" TYPE_X86_CPU, "stepping", "5" },
146     { TYPE_X86_CPU, "x-intel-pt-auto-level", "off" },
147 };
148 const size_t pc_compat_3_1_len = G_N_ELEMENTS(pc_compat_3_1);
149 
150 GlobalProperty pc_compat_3_0[] = {
151     { TYPE_X86_CPU, "x-hv-synic-kvm-only", "on" },
152     { "Skylake-Server" "-" TYPE_X86_CPU, "pku", "off" },
153     { "Skylake-Server-IBRS" "-" TYPE_X86_CPU, "pku", "off" },
154 };
155 const size_t pc_compat_3_0_len = G_N_ELEMENTS(pc_compat_3_0);
156 
157 GlobalProperty pc_compat_2_12[] = {
158     { TYPE_X86_CPU, "legacy-cache", "on" },
159     { TYPE_X86_CPU, "topoext", "off" },
160     { "EPYC-" TYPE_X86_CPU, "xlevel", "0x8000000a" },
161     { "EPYC-IBPB-" TYPE_X86_CPU, "xlevel", "0x8000000a" },
162 };
163 const size_t pc_compat_2_12_len = G_N_ELEMENTS(pc_compat_2_12);
164 
165 GlobalProperty pc_compat_2_11[] = {
166     { TYPE_X86_CPU, "x-migrate-smi-count", "off" },
167     { "Skylake-Server" "-" TYPE_X86_CPU, "clflushopt", "off" },
168 };
169 const size_t pc_compat_2_11_len = G_N_ELEMENTS(pc_compat_2_11);
170 
171 GlobalProperty pc_compat_2_10[] = {
172     { TYPE_X86_CPU, "x-hv-max-vps", "0x40" },
173     { "i440FX-pcihost", "x-pci-hole64-fix", "off" },
174     { "q35-pcihost", "x-pci-hole64-fix", "off" },
175 };
176 const size_t pc_compat_2_10_len = G_N_ELEMENTS(pc_compat_2_10);
177 
178 GlobalProperty pc_compat_2_9[] = {
179     { "mch", "extended-tseg-mbytes", "0" },
180 };
181 const size_t pc_compat_2_9_len = G_N_ELEMENTS(pc_compat_2_9);
182 
183 GlobalProperty pc_compat_2_8[] = {
184     { TYPE_X86_CPU, "tcg-cpuid", "off" },
185     { "kvmclock", "x-mach-use-reliable-get-clock", "off" },
186     { "ICH9-LPC", "x-smi-broadcast", "off" },
187     { TYPE_X86_CPU, "vmware-cpuid-freq", "off" },
188     { "Haswell-" TYPE_X86_CPU, "stepping", "1" },
189 };
190 const size_t pc_compat_2_8_len = G_N_ELEMENTS(pc_compat_2_8);
191 
192 GlobalProperty pc_compat_2_7[] = {
193     { TYPE_X86_CPU, "l3-cache", "off" },
194     { TYPE_X86_CPU, "full-cpuid-auto-level", "off" },
195     { "Opteron_G3" "-" TYPE_X86_CPU, "family", "15" },
196     { "Opteron_G3" "-" TYPE_X86_CPU, "model", "6" },
197     { "Opteron_G3" "-" TYPE_X86_CPU, "stepping", "1" },
198     { "isa-pcspk", "migrate", "off" },
199 };
200 const size_t pc_compat_2_7_len = G_N_ELEMENTS(pc_compat_2_7);
201 
202 GlobalProperty pc_compat_2_6[] = {
203     { TYPE_X86_CPU, "cpuid-0xb", "off" },
204     { "vmxnet3", "romfile", "" },
205     { TYPE_X86_CPU, "fill-mtrr-mask", "off" },
206     { "apic-common", "legacy-instance-id", "on", }
207 };
208 const size_t pc_compat_2_6_len = G_N_ELEMENTS(pc_compat_2_6);
209 
210 GlobalProperty pc_compat_2_5[] = {};
211 const size_t pc_compat_2_5_len = G_N_ELEMENTS(pc_compat_2_5);
212 
213 GlobalProperty pc_compat_2_4[] = {
214     PC_CPU_MODEL_IDS("2.4.0")
215     { "Haswell-" TYPE_X86_CPU, "abm", "off" },
216     { "Haswell-noTSX-" TYPE_X86_CPU, "abm", "off" },
217     { "Broadwell-" TYPE_X86_CPU, "abm", "off" },
218     { "Broadwell-noTSX-" TYPE_X86_CPU, "abm", "off" },
219     { "host" "-" TYPE_X86_CPU, "host-cache-info", "on" },
220     { TYPE_X86_CPU, "check", "off" },
221     { "qemu64" "-" TYPE_X86_CPU, "sse4a", "on" },
222     { "qemu64" "-" TYPE_X86_CPU, "abm", "on" },
223     { "qemu64" "-" TYPE_X86_CPU, "popcnt", "on" },
224     { "qemu32" "-" TYPE_X86_CPU, "popcnt", "on" },
225     { "Opteron_G2" "-" TYPE_X86_CPU, "rdtscp", "on" },
226     { "Opteron_G3" "-" TYPE_X86_CPU, "rdtscp", "on" },
227     { "Opteron_G4" "-" TYPE_X86_CPU, "rdtscp", "on" },
228     { "Opteron_G5" "-" TYPE_X86_CPU, "rdtscp", "on", }
229 };
230 const size_t pc_compat_2_4_len = G_N_ELEMENTS(pc_compat_2_4);
231 
232 GlobalProperty pc_compat_2_3[] = {
233     PC_CPU_MODEL_IDS("2.3.0")
234     { TYPE_X86_CPU, "arat", "off" },
235     { "qemu64" "-" TYPE_X86_CPU, "min-level", "4" },
236     { "kvm64" "-" TYPE_X86_CPU, "min-level", "5" },
237     { "pentium3" "-" TYPE_X86_CPU, "min-level", "2" },
238     { "n270" "-" TYPE_X86_CPU, "min-level", "5" },
239     { "Conroe" "-" TYPE_X86_CPU, "min-level", "4" },
240     { "Penryn" "-" TYPE_X86_CPU, "min-level", "4" },
241     { "Nehalem" "-" TYPE_X86_CPU, "min-level", "4" },
242     { "n270" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
243     { "Penryn" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
244     { "Conroe" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
245     { "Nehalem" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
246     { "Westmere" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
247     { "SandyBridge" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
248     { "IvyBridge" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
249     { "Haswell" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
250     { "Haswell-noTSX" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
251     { "Broadwell" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
252     { "Broadwell-noTSX" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
253     { TYPE_X86_CPU, "kvm-no-smi-migration", "on" },
254 };
255 const size_t pc_compat_2_3_len = G_N_ELEMENTS(pc_compat_2_3);
256 
257 GlobalProperty pc_compat_2_2[] = {
258     PC_CPU_MODEL_IDS("2.2.0")
259     { "kvm64" "-" TYPE_X86_CPU, "vme", "off" },
260     { "kvm32" "-" TYPE_X86_CPU, "vme", "off" },
261     { "Conroe" "-" TYPE_X86_CPU, "vme", "off" },
262     { "Penryn" "-" TYPE_X86_CPU, "vme", "off" },
263     { "Nehalem" "-" TYPE_X86_CPU, "vme", "off" },
264     { "Westmere" "-" TYPE_X86_CPU, "vme", "off" },
265     { "SandyBridge" "-" TYPE_X86_CPU, "vme", "off" },
266     { "Haswell" "-" TYPE_X86_CPU, "vme", "off" },
267     { "Broadwell" "-" TYPE_X86_CPU, "vme", "off" },
268     { "Opteron_G1" "-" TYPE_X86_CPU, "vme", "off" },
269     { "Opteron_G2" "-" TYPE_X86_CPU, "vme", "off" },
270     { "Opteron_G3" "-" TYPE_X86_CPU, "vme", "off" },
271     { "Opteron_G4" "-" TYPE_X86_CPU, "vme", "off" },
272     { "Opteron_G5" "-" TYPE_X86_CPU, "vme", "off" },
273     { "Haswell" "-" TYPE_X86_CPU, "f16c", "off" },
274     { "Haswell" "-" TYPE_X86_CPU, "rdrand", "off" },
275     { "Broadwell" "-" TYPE_X86_CPU, "f16c", "off" },
276     { "Broadwell" "-" TYPE_X86_CPU, "rdrand", "off" },
277 };
278 const size_t pc_compat_2_2_len = G_N_ELEMENTS(pc_compat_2_2);
279 
280 GlobalProperty pc_compat_2_1[] = {
281     PC_CPU_MODEL_IDS("2.1.0")
282     { "coreduo" "-" TYPE_X86_CPU, "vmx", "on" },
283     { "core2duo" "-" TYPE_X86_CPU, "vmx", "on" },
284 };
285 const size_t pc_compat_2_1_len = G_N_ELEMENTS(pc_compat_2_1);
286 
287 GlobalProperty pc_compat_2_0[] = {
288     PC_CPU_MODEL_IDS("2.0.0")
289     { "virtio-scsi-pci", "any_layout", "off" },
290     { "PIIX4_PM", "memory-hotplug-support", "off" },
291     { "apic", "version", "0x11" },
292     { "nec-usb-xhci", "superspeed-ports-first", "off" },
293     { "nec-usb-xhci", "force-pcie-endcap", "on" },
294     { "pci-serial", "prog_if", "0" },
295     { "pci-serial-2x", "prog_if", "0" },
296     { "pci-serial-4x", "prog_if", "0" },
297     { "virtio-net-pci", "guest_announce", "off" },
298     { "ICH9-LPC", "memory-hotplug-support", "off" },
299     { "xio3130-downstream", COMPAT_PROP_PCP, "off" },
300     { "ioh3420", COMPAT_PROP_PCP, "off" },
301 };
302 const size_t pc_compat_2_0_len = G_N_ELEMENTS(pc_compat_2_0);
303 
304 GlobalProperty pc_compat_1_7[] = {
305     PC_CPU_MODEL_IDS("1.7.0")
306     { TYPE_USB_DEVICE, "msos-desc", "no" },
307     { "PIIX4_PM", "acpi-pci-hotplug-with-bridge-support", "off" },
308     { "hpet", HPET_INTCAP, "4" },
309 };
310 const size_t pc_compat_1_7_len = G_N_ELEMENTS(pc_compat_1_7);
311 
312 GlobalProperty pc_compat_1_6[] = {
313     PC_CPU_MODEL_IDS("1.6.0")
314     { "e1000", "mitigation", "off" },
315     { "qemu64-" TYPE_X86_CPU, "model", "2" },
316     { "qemu32-" TYPE_X86_CPU, "model", "3" },
317     { "i440FX-pcihost", "short_root_bus", "1" },
318     { "q35-pcihost", "short_root_bus", "1" },
319 };
320 const size_t pc_compat_1_6_len = G_N_ELEMENTS(pc_compat_1_6);
321 
322 GlobalProperty pc_compat_1_5[] = {
323     PC_CPU_MODEL_IDS("1.5.0")
324     { "Conroe-" TYPE_X86_CPU, "model", "2" },
325     { "Conroe-" TYPE_X86_CPU, "min-level", "2" },
326     { "Penryn-" TYPE_X86_CPU, "model", "2" },
327     { "Penryn-" TYPE_X86_CPU, "min-level", "2" },
328     { "Nehalem-" TYPE_X86_CPU, "model", "2" },
329     { "Nehalem-" TYPE_X86_CPU, "min-level", "2" },
330     { "virtio-net-pci", "any_layout", "off" },
331     { TYPE_X86_CPU, "pmu", "on" },
332     { "i440FX-pcihost", "short_root_bus", "0" },
333     { "q35-pcihost", "short_root_bus", "0" },
334 };
335 const size_t pc_compat_1_5_len = G_N_ELEMENTS(pc_compat_1_5);
336 
337 GlobalProperty pc_compat_1_4[] = {
338     PC_CPU_MODEL_IDS("1.4.0")
339     { "scsi-hd", "discard_granularity", "0" },
340     { "scsi-cd", "discard_granularity", "0" },
341     { "scsi-disk", "discard_granularity", "0" },
342     { "ide-hd", "discard_granularity", "0" },
343     { "ide-cd", "discard_granularity", "0" },
344     { "ide-drive", "discard_granularity", "0" },
345     { "virtio-blk-pci", "discard_granularity", "0" },
346     /* DEV_NVECTORS_UNSPECIFIED as a uint32_t string: */
347     { "virtio-serial-pci", "vectors", "0xFFFFFFFF" },
348     { "virtio-net-pci", "ctrl_guest_offloads", "off" },
349     { "e1000", "romfile", "pxe-e1000.rom" },
350     { "ne2k_pci", "romfile", "pxe-ne2k_pci.rom" },
351     { "pcnet", "romfile", "pxe-pcnet.rom" },
352     { "rtl8139", "romfile", "pxe-rtl8139.rom" },
353     { "virtio-net-pci", "romfile", "pxe-virtio.rom" },
354     { "486-" TYPE_X86_CPU, "model", "0" },
355     { "n270" "-" TYPE_X86_CPU, "movbe", "off" },
356     { "Westmere" "-" TYPE_X86_CPU, "pclmulqdq", "off" },
357 };
358 const size_t pc_compat_1_4_len = G_N_ELEMENTS(pc_compat_1_4);
359 
360 void gsi_handler(void *opaque, int n, int level)
361 {
362     GSIState *s = opaque;
363 
364     DPRINTF("pc: %s GSI %d\n", level ? "raising" : "lowering", n);
365     if (n < ISA_NUM_IRQS) {
366         qemu_set_irq(s->i8259_irq[n], level);
367     }
368     qemu_set_irq(s->ioapic_irq[n], level);
369 }
370 
371 static void ioport80_write(void *opaque, hwaddr addr, uint64_t data,
372                            unsigned size)
373 {
374 }
375 
376 static uint64_t ioport80_read(void *opaque, hwaddr addr, unsigned size)
377 {
378     return 0xffffffffffffffffULL;
379 }
380 
381 /* MSDOS compatibility mode FPU exception support */
382 static qemu_irq ferr_irq;
383 
384 void pc_register_ferr_irq(qemu_irq irq)
385 {
386     ferr_irq = irq;
387 }
388 
389 /* XXX: add IGNNE support */
390 void cpu_set_ferr(CPUX86State *s)
391 {
392     qemu_irq_raise(ferr_irq);
393 }
394 
395 static void ioportF0_write(void *opaque, hwaddr addr, uint64_t data,
396                            unsigned size)
397 {
398     qemu_irq_lower(ferr_irq);
399 }
400 
401 static uint64_t ioportF0_read(void *opaque, hwaddr addr, unsigned size)
402 {
403     return 0xffffffffffffffffULL;
404 }
405 
406 /* TSC handling */
407 uint64_t cpu_get_tsc(CPUX86State *env)
408 {
409     return cpu_get_ticks();
410 }
411 
412 /* IRQ handling */
413 int cpu_get_pic_interrupt(CPUX86State *env)
414 {
415     X86CPU *cpu = env_archcpu(env);
416     int intno;
417 
418     if (!kvm_irqchip_in_kernel()) {
419         intno = apic_get_interrupt(cpu->apic_state);
420         if (intno >= 0) {
421             return intno;
422         }
423         /* read the irq from the PIC */
424         if (!apic_accept_pic_intr(cpu->apic_state)) {
425             return -1;
426         }
427     }
428 
429     intno = pic_read_irq(isa_pic);
430     return intno;
431 }
432 
433 static void pic_irq_request(void *opaque, int irq, int level)
434 {
435     CPUState *cs = first_cpu;
436     X86CPU *cpu = X86_CPU(cs);
437 
438     DPRINTF("pic_irqs: %s irq %d\n", level? "raise" : "lower", irq);
439     if (cpu->apic_state && !kvm_irqchip_in_kernel()) {
440         CPU_FOREACH(cs) {
441             cpu = X86_CPU(cs);
442             if (apic_accept_pic_intr(cpu->apic_state)) {
443                 apic_deliver_pic_intr(cpu->apic_state, level);
444             }
445         }
446     } else {
447         if (level) {
448             cpu_interrupt(cs, CPU_INTERRUPT_HARD);
449         } else {
450             cpu_reset_interrupt(cs, CPU_INTERRUPT_HARD);
451         }
452     }
453 }
454 
455 /* PC cmos mappings */
456 
457 #define REG_EQUIPMENT_BYTE          0x14
458 
459 int cmos_get_fd_drive_type(FloppyDriveType fd0)
460 {
461     int val;
462 
463     switch (fd0) {
464     case FLOPPY_DRIVE_TYPE_144:
465         /* 1.44 Mb 3"5 drive */
466         val = 4;
467         break;
468     case FLOPPY_DRIVE_TYPE_288:
469         /* 2.88 Mb 3"5 drive */
470         val = 5;
471         break;
472     case FLOPPY_DRIVE_TYPE_120:
473         /* 1.2 Mb 5"5 drive */
474         val = 2;
475         break;
476     case FLOPPY_DRIVE_TYPE_NONE:
477     default:
478         val = 0;
479         break;
480     }
481     return val;
482 }
483 
484 static void cmos_init_hd(ISADevice *s, int type_ofs, int info_ofs,
485                          int16_t cylinders, int8_t heads, int8_t sectors)
486 {
487     rtc_set_memory(s, type_ofs, 47);
488     rtc_set_memory(s, info_ofs, cylinders);
489     rtc_set_memory(s, info_ofs + 1, cylinders >> 8);
490     rtc_set_memory(s, info_ofs + 2, heads);
491     rtc_set_memory(s, info_ofs + 3, 0xff);
492     rtc_set_memory(s, info_ofs + 4, 0xff);
493     rtc_set_memory(s, info_ofs + 5, 0xc0 | ((heads > 8) << 3));
494     rtc_set_memory(s, info_ofs + 6, cylinders);
495     rtc_set_memory(s, info_ofs + 7, cylinders >> 8);
496     rtc_set_memory(s, info_ofs + 8, sectors);
497 }
498 
499 /* convert boot_device letter to something recognizable by the bios */
500 static int boot_device2nibble(char boot_device)
501 {
502     switch(boot_device) {
503     case 'a':
504     case 'b':
505         return 0x01; /* floppy boot */
506     case 'c':
507         return 0x02; /* hard drive boot */
508     case 'd':
509         return 0x03; /* CD-ROM boot */
510     case 'n':
511         return 0x04; /* Network boot */
512     }
513     return 0;
514 }
515 
516 static void set_boot_dev(ISADevice *s, const char *boot_device, Error **errp)
517 {
518 #define PC_MAX_BOOT_DEVICES 3
519     int nbds, bds[3] = { 0, };
520     int i;
521 
522     nbds = strlen(boot_device);
523     if (nbds > PC_MAX_BOOT_DEVICES) {
524         error_setg(errp, "Too many boot devices for PC");
525         return;
526     }
527     for (i = 0; i < nbds; i++) {
528         bds[i] = boot_device2nibble(boot_device[i]);
529         if (bds[i] == 0) {
530             error_setg(errp, "Invalid boot device for PC: '%c'",
531                        boot_device[i]);
532             return;
533         }
534     }
535     rtc_set_memory(s, 0x3d, (bds[1] << 4) | bds[0]);
536     rtc_set_memory(s, 0x38, (bds[2] << 4) | (fd_bootchk ? 0x0 : 0x1));
537 }
538 
539 static void pc_boot_set(void *opaque, const char *boot_device, Error **errp)
540 {
541     set_boot_dev(opaque, boot_device, errp);
542 }
543 
544 static void pc_cmos_init_floppy(ISADevice *rtc_state, ISADevice *floppy)
545 {
546     int val, nb, i;
547     FloppyDriveType fd_type[2] = { FLOPPY_DRIVE_TYPE_NONE,
548                                    FLOPPY_DRIVE_TYPE_NONE };
549 
550     /* floppy type */
551     if (floppy) {
552         for (i = 0; i < 2; i++) {
553             fd_type[i] = isa_fdc_get_drive_type(floppy, i);
554         }
555     }
556     val = (cmos_get_fd_drive_type(fd_type[0]) << 4) |
557         cmos_get_fd_drive_type(fd_type[1]);
558     rtc_set_memory(rtc_state, 0x10, val);
559 
560     val = rtc_get_memory(rtc_state, REG_EQUIPMENT_BYTE);
561     nb = 0;
562     if (fd_type[0] != FLOPPY_DRIVE_TYPE_NONE) {
563         nb++;
564     }
565     if (fd_type[1] != FLOPPY_DRIVE_TYPE_NONE) {
566         nb++;
567     }
568     switch (nb) {
569     case 0:
570         break;
571     case 1:
572         val |= 0x01; /* 1 drive, ready for boot */
573         break;
574     case 2:
575         val |= 0x41; /* 2 drives, ready for boot */
576         break;
577     }
578     rtc_set_memory(rtc_state, REG_EQUIPMENT_BYTE, val);
579 }
580 
581 typedef struct pc_cmos_init_late_arg {
582     ISADevice *rtc_state;
583     BusState *idebus[2];
584 } pc_cmos_init_late_arg;
585 
586 typedef struct check_fdc_state {
587     ISADevice *floppy;
588     bool multiple;
589 } CheckFdcState;
590 
591 static int check_fdc(Object *obj, void *opaque)
592 {
593     CheckFdcState *state = opaque;
594     Object *fdc;
595     uint32_t iobase;
596     Error *local_err = NULL;
597 
598     fdc = object_dynamic_cast(obj, TYPE_ISA_FDC);
599     if (!fdc) {
600         return 0;
601     }
602 
603     iobase = object_property_get_uint(obj, "iobase", &local_err);
604     if (local_err || iobase != 0x3f0) {
605         error_free(local_err);
606         return 0;
607     }
608 
609     if (state->floppy) {
610         state->multiple = true;
611     } else {
612         state->floppy = ISA_DEVICE(obj);
613     }
614     return 0;
615 }
616 
617 static const char * const fdc_container_path[] = {
618     "/unattached", "/peripheral", "/peripheral-anon"
619 };
620 
621 /*
622  * Locate the FDC at IO address 0x3f0, in order to configure the CMOS registers
623  * and ACPI objects.
624  */
625 ISADevice *pc_find_fdc0(void)
626 {
627     int i;
628     Object *container;
629     CheckFdcState state = { 0 };
630 
631     for (i = 0; i < ARRAY_SIZE(fdc_container_path); i++) {
632         container = container_get(qdev_get_machine(), fdc_container_path[i]);
633         object_child_foreach(container, check_fdc, &state);
634     }
635 
636     if (state.multiple) {
637         warn_report("multiple floppy disk controllers with "
638                     "iobase=0x3f0 have been found");
639         error_printf("the one being picked for CMOS setup might not reflect "
640                      "your intent");
641     }
642 
643     return state.floppy;
644 }
645 
646 static void pc_cmos_init_late(void *opaque)
647 {
648     pc_cmos_init_late_arg *arg = opaque;
649     ISADevice *s = arg->rtc_state;
650     int16_t cylinders;
651     int8_t heads, sectors;
652     int val;
653     int i, trans;
654 
655     val = 0;
656     if (arg->idebus[0] && ide_get_geometry(arg->idebus[0], 0,
657                                            &cylinders, &heads, &sectors) >= 0) {
658         cmos_init_hd(s, 0x19, 0x1b, cylinders, heads, sectors);
659         val |= 0xf0;
660     }
661     if (arg->idebus[0] && ide_get_geometry(arg->idebus[0], 1,
662                                            &cylinders, &heads, &sectors) >= 0) {
663         cmos_init_hd(s, 0x1a, 0x24, cylinders, heads, sectors);
664         val |= 0x0f;
665     }
666     rtc_set_memory(s, 0x12, val);
667 
668     val = 0;
669     for (i = 0; i < 4; i++) {
670         /* NOTE: ide_get_geometry() returns the physical
671            geometry.  It is always such that: 1 <= sects <= 63, 1
672            <= heads <= 16, 1 <= cylinders <= 16383. The BIOS
673            geometry can be different if a translation is done. */
674         if (arg->idebus[i / 2] &&
675             ide_get_geometry(arg->idebus[i / 2], i % 2,
676                              &cylinders, &heads, &sectors) >= 0) {
677             trans = ide_get_bios_chs_trans(arg->idebus[i / 2], i % 2) - 1;
678             assert((trans & ~3) == 0);
679             val |= trans << (i * 2);
680         }
681     }
682     rtc_set_memory(s, 0x39, val);
683 
684     pc_cmos_init_floppy(s, pc_find_fdc0());
685 
686     qemu_unregister_reset(pc_cmos_init_late, opaque);
687 }
688 
689 void pc_cmos_init(PCMachineState *pcms,
690                   BusState *idebus0, BusState *idebus1,
691                   ISADevice *s)
692 {
693     int val;
694     static pc_cmos_init_late_arg arg;
695 
696     /* various important CMOS locations needed by PC/Bochs bios */
697 
698     /* memory size */
699     /* base memory (first MiB) */
700     val = MIN(pcms->below_4g_mem_size / KiB, 640);
701     rtc_set_memory(s, 0x15, val);
702     rtc_set_memory(s, 0x16, val >> 8);
703     /* extended memory (next 64MiB) */
704     if (pcms->below_4g_mem_size > 1 * MiB) {
705         val = (pcms->below_4g_mem_size - 1 * MiB) / KiB;
706     } else {
707         val = 0;
708     }
709     if (val > 65535)
710         val = 65535;
711     rtc_set_memory(s, 0x17, val);
712     rtc_set_memory(s, 0x18, val >> 8);
713     rtc_set_memory(s, 0x30, val);
714     rtc_set_memory(s, 0x31, val >> 8);
715     /* memory between 16MiB and 4GiB */
716     if (pcms->below_4g_mem_size > 16 * MiB) {
717         val = (pcms->below_4g_mem_size - 16 * MiB) / (64 * KiB);
718     } else {
719         val = 0;
720     }
721     if (val > 65535)
722         val = 65535;
723     rtc_set_memory(s, 0x34, val);
724     rtc_set_memory(s, 0x35, val >> 8);
725     /* memory above 4GiB */
726     val = pcms->above_4g_mem_size / 65536;
727     rtc_set_memory(s, 0x5b, val);
728     rtc_set_memory(s, 0x5c, val >> 8);
729     rtc_set_memory(s, 0x5d, val >> 16);
730 
731     object_property_add_link(OBJECT(pcms), "rtc_state",
732                              TYPE_ISA_DEVICE,
733                              (Object **)&pcms->rtc,
734                              object_property_allow_set_link,
735                              OBJ_PROP_LINK_STRONG, &error_abort);
736     object_property_set_link(OBJECT(pcms), OBJECT(s),
737                              "rtc_state", &error_abort);
738 
739     set_boot_dev(s, MACHINE(pcms)->boot_order, &error_fatal);
740 
741     val = 0;
742     val |= 0x02; /* FPU is there */
743     val |= 0x04; /* PS/2 mouse installed */
744     rtc_set_memory(s, REG_EQUIPMENT_BYTE, val);
745 
746     /* hard drives and FDC */
747     arg.rtc_state = s;
748     arg.idebus[0] = idebus0;
749     arg.idebus[1] = idebus1;
750     qemu_register_reset(pc_cmos_init_late, &arg);
751 }
752 
753 #define TYPE_PORT92 "port92"
754 #define PORT92(obj) OBJECT_CHECK(Port92State, (obj), TYPE_PORT92)
755 
756 /* port 92 stuff: could be split off */
757 typedef struct Port92State {
758     ISADevice parent_obj;
759 
760     MemoryRegion io;
761     uint8_t outport;
762     qemu_irq a20_out;
763 } Port92State;
764 
765 static void port92_write(void *opaque, hwaddr addr, uint64_t val,
766                          unsigned size)
767 {
768     Port92State *s = opaque;
769     int oldval = s->outport;
770 
771     DPRINTF("port92: write 0x%02" PRIx64 "\n", val);
772     s->outport = val;
773     qemu_set_irq(s->a20_out, (val >> 1) & 1);
774     if ((val & 1) && !(oldval & 1)) {
775         qemu_system_reset_request(SHUTDOWN_CAUSE_GUEST_RESET);
776     }
777 }
778 
779 static uint64_t port92_read(void *opaque, hwaddr addr,
780                             unsigned size)
781 {
782     Port92State *s = opaque;
783     uint32_t ret;
784 
785     ret = s->outport;
786     DPRINTF("port92: read 0x%02x\n", ret);
787     return ret;
788 }
789 
790 static void port92_init(ISADevice *dev, qemu_irq a20_out)
791 {
792     qdev_connect_gpio_out_named(DEVICE(dev), PORT92_A20_LINE, 0, a20_out);
793 }
794 
795 static const VMStateDescription vmstate_port92_isa = {
796     .name = "port92",
797     .version_id = 1,
798     .minimum_version_id = 1,
799     .fields = (VMStateField[]) {
800         VMSTATE_UINT8(outport, Port92State),
801         VMSTATE_END_OF_LIST()
802     }
803 };
804 
805 static void port92_reset(DeviceState *d)
806 {
807     Port92State *s = PORT92(d);
808 
809     s->outport &= ~1;
810 }
811 
812 static const MemoryRegionOps port92_ops = {
813     .read = port92_read,
814     .write = port92_write,
815     .impl = {
816         .min_access_size = 1,
817         .max_access_size = 1,
818     },
819     .endianness = DEVICE_LITTLE_ENDIAN,
820 };
821 
822 static void port92_initfn(Object *obj)
823 {
824     Port92State *s = PORT92(obj);
825 
826     memory_region_init_io(&s->io, OBJECT(s), &port92_ops, s, "port92", 1);
827 
828     s->outport = 0;
829 
830     qdev_init_gpio_out_named(DEVICE(obj), &s->a20_out, PORT92_A20_LINE, 1);
831 }
832 
833 static void port92_realizefn(DeviceState *dev, Error **errp)
834 {
835     ISADevice *isadev = ISA_DEVICE(dev);
836     Port92State *s = PORT92(dev);
837 
838     isa_register_ioport(isadev, &s->io, 0x92);
839 }
840 
841 static void port92_class_initfn(ObjectClass *klass, void *data)
842 {
843     DeviceClass *dc = DEVICE_CLASS(klass);
844 
845     dc->realize = port92_realizefn;
846     dc->reset = port92_reset;
847     dc->vmsd = &vmstate_port92_isa;
848     /*
849      * Reason: unlike ordinary ISA devices, this one needs additional
850      * wiring: its A20 output line needs to be wired up by
851      * port92_init().
852      */
853     dc->user_creatable = false;
854 }
855 
856 static const TypeInfo port92_info = {
857     .name          = TYPE_PORT92,
858     .parent        = TYPE_ISA_DEVICE,
859     .instance_size = sizeof(Port92State),
860     .instance_init = port92_initfn,
861     .class_init    = port92_class_initfn,
862 };
863 
864 static void port92_register_types(void)
865 {
866     type_register_static(&port92_info);
867 }
868 
869 type_init(port92_register_types)
870 
871 static void handle_a20_line_change(void *opaque, int irq, int level)
872 {
873     X86CPU *cpu = opaque;
874 
875     /* XXX: send to all CPUs ? */
876     /* XXX: add logic to handle multiple A20 line sources */
877     x86_cpu_set_a20(cpu, level);
878 }
879 
880 int e820_add_entry(uint64_t address, uint64_t length, uint32_t type)
881 {
882     int index = le32_to_cpu(e820_reserve.count);
883     struct e820_entry *entry;
884 
885     if (type != E820_RAM) {
886         /* old FW_CFG_E820_TABLE entry -- reservations only */
887         if (index >= E820_NR_ENTRIES) {
888             return -EBUSY;
889         }
890         entry = &e820_reserve.entry[index++];
891 
892         entry->address = cpu_to_le64(address);
893         entry->length = cpu_to_le64(length);
894         entry->type = cpu_to_le32(type);
895 
896         e820_reserve.count = cpu_to_le32(index);
897     }
898 
899     /* new "etc/e820" file -- include ram too */
900     e820_table = g_renew(struct e820_entry, e820_table, e820_entries + 1);
901     e820_table[e820_entries].address = cpu_to_le64(address);
902     e820_table[e820_entries].length = cpu_to_le64(length);
903     e820_table[e820_entries].type = cpu_to_le32(type);
904     e820_entries++;
905 
906     return e820_entries;
907 }
908 
909 int e820_get_num_entries(void)
910 {
911     return e820_entries;
912 }
913 
914 bool e820_get_entry(int idx, uint32_t type, uint64_t *address, uint64_t *length)
915 {
916     if (idx < e820_entries && e820_table[idx].type == cpu_to_le32(type)) {
917         *address = le64_to_cpu(e820_table[idx].address);
918         *length = le64_to_cpu(e820_table[idx].length);
919         return true;
920     }
921     return false;
922 }
923 
924 /* Calculates initial APIC ID for a specific CPU index
925  *
926  * Currently we need to be able to calculate the APIC ID from the CPU index
927  * alone (without requiring a CPU object), as the QEMU<->Seabios interfaces have
928  * no concept of "CPU index", and the NUMA tables on fw_cfg need the APIC ID of
929  * all CPUs up to max_cpus.
930  */
931 static uint32_t x86_cpu_apic_id_from_index(PCMachineState *pcms,
932                                            unsigned int cpu_index)
933 {
934     MachineState *ms = MACHINE(pcms);
935     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
936     uint32_t correct_id;
937     static bool warned;
938 
939     correct_id = x86_apicid_from_cpu_idx(pcms->smp_dies, ms->smp.cores,
940                                          ms->smp.threads, cpu_index);
941     if (pcmc->compat_apic_id_mode) {
942         if (cpu_index != correct_id && !warned && !qtest_enabled()) {
943             error_report("APIC IDs set in compatibility mode, "
944                          "CPU topology won't match the configuration");
945             warned = true;
946         }
947         return cpu_index;
948     } else {
949         return correct_id;
950     }
951 }
952 
953 static void pc_build_smbios(PCMachineState *pcms)
954 {
955     uint8_t *smbios_tables, *smbios_anchor;
956     size_t smbios_tables_len, smbios_anchor_len;
957     struct smbios_phys_mem_area *mem_array;
958     unsigned i, array_count;
959     MachineState *ms = MACHINE(pcms);
960     X86CPU *cpu = X86_CPU(ms->possible_cpus->cpus[0].cpu);
961 
962     /* tell smbios about cpuid version and features */
963     smbios_set_cpuid(cpu->env.cpuid_version, cpu->env.features[FEAT_1_EDX]);
964 
965     smbios_tables = smbios_get_table_legacy(ms, &smbios_tables_len);
966     if (smbios_tables) {
967         fw_cfg_add_bytes(pcms->fw_cfg, FW_CFG_SMBIOS_ENTRIES,
968                          smbios_tables, smbios_tables_len);
969     }
970 
971     /* build the array of physical mem area from e820 table */
972     mem_array = g_malloc0(sizeof(*mem_array) * e820_get_num_entries());
973     for (i = 0, array_count = 0; i < e820_get_num_entries(); i++) {
974         uint64_t addr, len;
975 
976         if (e820_get_entry(i, E820_RAM, &addr, &len)) {
977             mem_array[array_count].address = addr;
978             mem_array[array_count].length = len;
979             array_count++;
980         }
981     }
982     smbios_get_tables(ms, mem_array, array_count,
983                       &smbios_tables, &smbios_tables_len,
984                       &smbios_anchor, &smbios_anchor_len);
985     g_free(mem_array);
986 
987     if (smbios_anchor) {
988         fw_cfg_add_file(pcms->fw_cfg, "etc/smbios/smbios-tables",
989                         smbios_tables, smbios_tables_len);
990         fw_cfg_add_file(pcms->fw_cfg, "etc/smbios/smbios-anchor",
991                         smbios_anchor, smbios_anchor_len);
992     }
993 }
994 
995 static FWCfgState *bochs_bios_init(AddressSpace *as, PCMachineState *pcms)
996 {
997     FWCfgState *fw_cfg;
998     uint64_t *numa_fw_cfg;
999     int i;
1000     const CPUArchIdList *cpus;
1001     MachineClass *mc = MACHINE_GET_CLASS(pcms);
1002 
1003     fw_cfg = fw_cfg_init_io_dma(FW_CFG_IO_BASE, FW_CFG_IO_BASE + 4, as);
1004     fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
1005 
1006     /* FW_CFG_MAX_CPUS is a bit confusing/problematic on x86:
1007      *
1008      * For machine types prior to 1.8, SeaBIOS needs FW_CFG_MAX_CPUS for
1009      * building MPTable, ACPI MADT, ACPI CPU hotplug and ACPI SRAT table,
1010      * that tables are based on xAPIC ID and QEMU<->SeaBIOS interface
1011      * for CPU hotplug also uses APIC ID and not "CPU index".
1012      * This means that FW_CFG_MAX_CPUS is not the "maximum number of CPUs",
1013      * but the "limit to the APIC ID values SeaBIOS may see".
1014      *
1015      * So for compatibility reasons with old BIOSes we are stuck with
1016      * "etc/max-cpus" actually being apic_id_limit
1017      */
1018     fw_cfg_add_i16(fw_cfg, FW_CFG_MAX_CPUS, (uint16_t)pcms->apic_id_limit);
1019     fw_cfg_add_i64(fw_cfg, FW_CFG_RAM_SIZE, (uint64_t)ram_size);
1020     fw_cfg_add_bytes(fw_cfg, FW_CFG_ACPI_TABLES,
1021                      acpi_tables, acpi_tables_len);
1022     fw_cfg_add_i32(fw_cfg, FW_CFG_IRQ0_OVERRIDE, kvm_allows_irq0_override());
1023 
1024     fw_cfg_add_bytes(fw_cfg, FW_CFG_E820_TABLE,
1025                      &e820_reserve, sizeof(e820_reserve));
1026     fw_cfg_add_file(fw_cfg, "etc/e820", e820_table,
1027                     sizeof(struct e820_entry) * e820_entries);
1028 
1029     fw_cfg_add_bytes(fw_cfg, FW_CFG_HPET, &hpet_cfg, sizeof(hpet_cfg));
1030     /* allocate memory for the NUMA channel: one (64bit) word for the number
1031      * of nodes, one word for each VCPU->node and one word for each node to
1032      * hold the amount of memory.
1033      */
1034     numa_fw_cfg = g_new0(uint64_t, 1 + pcms->apic_id_limit + nb_numa_nodes);
1035     numa_fw_cfg[0] = cpu_to_le64(nb_numa_nodes);
1036     cpus = mc->possible_cpu_arch_ids(MACHINE(pcms));
1037     for (i = 0; i < cpus->len; i++) {
1038         unsigned int apic_id = cpus->cpus[i].arch_id;
1039         assert(apic_id < pcms->apic_id_limit);
1040         numa_fw_cfg[apic_id + 1] = cpu_to_le64(cpus->cpus[i].props.node_id);
1041     }
1042     for (i = 0; i < nb_numa_nodes; i++) {
1043         numa_fw_cfg[pcms->apic_id_limit + 1 + i] =
1044             cpu_to_le64(numa_info[i].node_mem);
1045     }
1046     fw_cfg_add_bytes(fw_cfg, FW_CFG_NUMA, numa_fw_cfg,
1047                      (1 + pcms->apic_id_limit + nb_numa_nodes) *
1048                      sizeof(*numa_fw_cfg));
1049 
1050     return fw_cfg;
1051 }
1052 
1053 static long get_file_size(FILE *f)
1054 {
1055     long where, size;
1056 
1057     /* XXX: on Unix systems, using fstat() probably makes more sense */
1058 
1059     where = ftell(f);
1060     fseek(f, 0, SEEK_END);
1061     size = ftell(f);
1062     fseek(f, where, SEEK_SET);
1063 
1064     return size;
1065 }
1066 
1067 struct setup_data {
1068     uint64_t next;
1069     uint32_t type;
1070     uint32_t len;
1071     uint8_t data[0];
1072 } __attribute__((packed));
1073 
1074 
1075 /*
1076  * The entry point into the kernel for PVH boot is different from
1077  * the native entry point.  The PVH entry is defined by the x86/HVM
1078  * direct boot ABI and is available in an ELFNOTE in the kernel binary.
1079  *
1080  * This function is passed to load_elf() when it is called from
1081  * load_elfboot() which then additionally checks for an ELF Note of
1082  * type XEN_ELFNOTE_PHYS32_ENTRY and passes it to this function to
1083  * parse the PVH entry address from the ELF Note.
1084  *
1085  * Due to trickery in elf_opts.h, load_elf() is actually available as
1086  * load_elf32() or load_elf64() and this routine needs to be able
1087  * to deal with being called as 32 or 64 bit.
1088  *
1089  * The address of the PVH entry point is saved to the 'pvh_start_addr'
1090  * global variable.  (although the entry point is 32-bit, the kernel
1091  * binary can be either 32-bit or 64-bit).
1092  */
1093 static uint64_t read_pvh_start_addr(void *arg1, void *arg2, bool is64)
1094 {
1095     size_t *elf_note_data_addr;
1096 
1097     /* Check if ELF Note header passed in is valid */
1098     if (arg1 == NULL) {
1099         return 0;
1100     }
1101 
1102     if (is64) {
1103         struct elf64_note *nhdr64 = (struct elf64_note *)arg1;
1104         uint64_t nhdr_size64 = sizeof(struct elf64_note);
1105         uint64_t phdr_align = *(uint64_t *)arg2;
1106         uint64_t nhdr_namesz = nhdr64->n_namesz;
1107 
1108         elf_note_data_addr =
1109             ((void *)nhdr64) + nhdr_size64 +
1110             QEMU_ALIGN_UP(nhdr_namesz, phdr_align);
1111     } else {
1112         struct elf32_note *nhdr32 = (struct elf32_note *)arg1;
1113         uint32_t nhdr_size32 = sizeof(struct elf32_note);
1114         uint32_t phdr_align = *(uint32_t *)arg2;
1115         uint32_t nhdr_namesz = nhdr32->n_namesz;
1116 
1117         elf_note_data_addr =
1118             ((void *)nhdr32) + nhdr_size32 +
1119             QEMU_ALIGN_UP(nhdr_namesz, phdr_align);
1120     }
1121 
1122     pvh_start_addr = *elf_note_data_addr;
1123 
1124     return pvh_start_addr;
1125 }
1126 
1127 static bool load_elfboot(const char *kernel_filename,
1128                    int kernel_file_size,
1129                    uint8_t *header,
1130                    size_t pvh_xen_start_addr,
1131                    FWCfgState *fw_cfg)
1132 {
1133     uint32_t flags = 0;
1134     uint32_t mh_load_addr = 0;
1135     uint32_t elf_kernel_size = 0;
1136     uint64_t elf_entry;
1137     uint64_t elf_low, elf_high;
1138     int kernel_size;
1139 
1140     if (ldl_p(header) != 0x464c457f) {
1141         return false; /* no elfboot */
1142     }
1143 
1144     bool elf_is64 = header[EI_CLASS] == ELFCLASS64;
1145     flags = elf_is64 ?
1146         ((Elf64_Ehdr *)header)->e_flags : ((Elf32_Ehdr *)header)->e_flags;
1147 
1148     if (flags & 0x00010004) { /* LOAD_ELF_HEADER_HAS_ADDR */
1149         error_report("elfboot unsupported flags = %x", flags);
1150         exit(1);
1151     }
1152 
1153     uint64_t elf_note_type = XEN_ELFNOTE_PHYS32_ENTRY;
1154     kernel_size = load_elf(kernel_filename, read_pvh_start_addr,
1155                            NULL, &elf_note_type, &elf_entry,
1156                            &elf_low, &elf_high, 0, I386_ELF_MACHINE,
1157                            0, 0);
1158 
1159     if (kernel_size < 0) {
1160         error_report("Error while loading elf kernel");
1161         exit(1);
1162     }
1163     mh_load_addr = elf_low;
1164     elf_kernel_size = elf_high - elf_low;
1165 
1166     if (pvh_start_addr == 0) {
1167         error_report("Error loading uncompressed kernel without PVH ELF Note");
1168         exit(1);
1169     }
1170     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ENTRY, pvh_start_addr);
1171     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ADDR, mh_load_addr);
1172     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_SIZE, elf_kernel_size);
1173 
1174     return true;
1175 }
1176 
1177 static void load_linux(PCMachineState *pcms,
1178                        FWCfgState *fw_cfg)
1179 {
1180     uint16_t protocol;
1181     int setup_size, kernel_size, cmdline_size;
1182     int dtb_size, setup_data_offset;
1183     uint32_t initrd_max;
1184     uint8_t header[8192], *setup, *kernel;
1185     hwaddr real_addr, prot_addr, cmdline_addr, initrd_addr = 0;
1186     FILE *f;
1187     char *vmode;
1188     MachineState *machine = MACHINE(pcms);
1189     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1190     struct setup_data *setup_data;
1191     const char *kernel_filename = machine->kernel_filename;
1192     const char *initrd_filename = machine->initrd_filename;
1193     const char *dtb_filename = machine->dtb;
1194     const char *kernel_cmdline = machine->kernel_cmdline;
1195 
1196     /* Align to 16 bytes as a paranoia measure */
1197     cmdline_size = (strlen(kernel_cmdline)+16) & ~15;
1198 
1199     /* load the kernel header */
1200     f = fopen(kernel_filename, "rb");
1201     if (!f || !(kernel_size = get_file_size(f)) ||
1202         fread(header, 1, MIN(ARRAY_SIZE(header), kernel_size), f) !=
1203         MIN(ARRAY_SIZE(header), kernel_size)) {
1204         fprintf(stderr, "qemu: could not load kernel '%s': %s\n",
1205                 kernel_filename, strerror(errno));
1206         exit(1);
1207     }
1208 
1209     /* kernel protocol version */
1210 #if 0
1211     fprintf(stderr, "header magic: %#x\n", ldl_p(header+0x202));
1212 #endif
1213     if (ldl_p(header+0x202) == 0x53726448) {
1214         protocol = lduw_p(header+0x206);
1215     } else {
1216         /*
1217          * This could be a multiboot kernel. If it is, let's stop treating it
1218          * like a Linux kernel.
1219          * Note: some multiboot images could be in the ELF format (the same of
1220          * PVH), so we try multiboot first since we check the multiboot magic
1221          * header before to load it.
1222          */
1223         if (load_multiboot(fw_cfg, f, kernel_filename, initrd_filename,
1224                            kernel_cmdline, kernel_size, header)) {
1225             return;
1226         }
1227         /*
1228          * Check if the file is an uncompressed kernel file (ELF) and load it,
1229          * saving the PVH entry point used by the x86/HVM direct boot ABI.
1230          * If load_elfboot() is successful, populate the fw_cfg info.
1231          */
1232         if (pcmc->pvh_enabled &&
1233             load_elfboot(kernel_filename, kernel_size,
1234                          header, pvh_start_addr, fw_cfg)) {
1235             fclose(f);
1236 
1237             fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_SIZE,
1238                 strlen(kernel_cmdline) + 1);
1239             fw_cfg_add_string(fw_cfg, FW_CFG_CMDLINE_DATA, kernel_cmdline);
1240 
1241             fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_SIZE, sizeof(header));
1242             fw_cfg_add_bytes(fw_cfg, FW_CFG_SETUP_DATA,
1243                              header, sizeof(header));
1244 
1245             /* load initrd */
1246             if (initrd_filename) {
1247                 GMappedFile *mapped_file;
1248                 gsize initrd_size;
1249                 gchar *initrd_data;
1250                 GError *gerr = NULL;
1251 
1252                 mapped_file = g_mapped_file_new(initrd_filename, false, &gerr);
1253                 if (!mapped_file) {
1254                     fprintf(stderr, "qemu: error reading initrd %s: %s\n",
1255                             initrd_filename, gerr->message);
1256                     exit(1);
1257                 }
1258                 pcms->initrd_mapped_file = mapped_file;
1259 
1260                 initrd_data = g_mapped_file_get_contents(mapped_file);
1261                 initrd_size = g_mapped_file_get_length(mapped_file);
1262                 initrd_max = pcms->below_4g_mem_size - pcmc->acpi_data_size - 1;
1263                 if (initrd_size >= initrd_max) {
1264                     fprintf(stderr, "qemu: initrd is too large, cannot support."
1265                             "(max: %"PRIu32", need %"PRId64")\n",
1266                             initrd_max, (uint64_t)initrd_size);
1267                     exit(1);
1268                 }
1269 
1270                 initrd_addr = (initrd_max - initrd_size) & ~4095;
1271 
1272                 fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_ADDR, initrd_addr);
1273                 fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_SIZE, initrd_size);
1274                 fw_cfg_add_bytes(fw_cfg, FW_CFG_INITRD_DATA, initrd_data,
1275                                  initrd_size);
1276             }
1277 
1278             option_rom[nb_option_roms].bootindex = 0;
1279             option_rom[nb_option_roms].name = "pvh.bin";
1280             nb_option_roms++;
1281 
1282             return;
1283         }
1284         protocol = 0;
1285     }
1286 
1287     if (protocol < 0x200 || !(header[0x211] & 0x01)) {
1288         /* Low kernel */
1289         real_addr    = 0x90000;
1290         cmdline_addr = 0x9a000 - cmdline_size;
1291         prot_addr    = 0x10000;
1292     } else if (protocol < 0x202) {
1293         /* High but ancient kernel */
1294         real_addr    = 0x90000;
1295         cmdline_addr = 0x9a000 - cmdline_size;
1296         prot_addr    = 0x100000;
1297     } else {
1298         /* High and recent kernel */
1299         real_addr    = 0x10000;
1300         cmdline_addr = 0x20000;
1301         prot_addr    = 0x100000;
1302     }
1303 
1304 #if 0
1305     fprintf(stderr,
1306             "qemu: real_addr     = 0x" TARGET_FMT_plx "\n"
1307             "qemu: cmdline_addr  = 0x" TARGET_FMT_plx "\n"
1308             "qemu: prot_addr     = 0x" TARGET_FMT_plx "\n",
1309             real_addr,
1310             cmdline_addr,
1311             prot_addr);
1312 #endif
1313 
1314     /* highest address for loading the initrd */
1315     if (protocol >= 0x20c &&
1316         lduw_p(header+0x236) & XLF_CAN_BE_LOADED_ABOVE_4G) {
1317         /*
1318          * Linux has supported initrd up to 4 GB for a very long time (2007,
1319          * long before XLF_CAN_BE_LOADED_ABOVE_4G which was added in 2013),
1320          * though it only sets initrd_max to 2 GB to "work around bootloader
1321          * bugs". Luckily, QEMU firmware(which does something like bootloader)
1322          * has supported this.
1323          *
1324          * It's believed that if XLF_CAN_BE_LOADED_ABOVE_4G is set, initrd can
1325          * be loaded into any address.
1326          *
1327          * In addition, initrd_max is uint32_t simply because QEMU doesn't
1328          * support the 64-bit boot protocol (specifically the ext_ramdisk_image
1329          * field).
1330          *
1331          * Therefore here just limit initrd_max to UINT32_MAX simply as well.
1332          */
1333         initrd_max = UINT32_MAX;
1334     } else if (protocol >= 0x203) {
1335         initrd_max = ldl_p(header+0x22c);
1336     } else {
1337         initrd_max = 0x37ffffff;
1338     }
1339 
1340     if (initrd_max >= pcms->below_4g_mem_size - pcmc->acpi_data_size) {
1341         initrd_max = pcms->below_4g_mem_size - pcmc->acpi_data_size - 1;
1342     }
1343 
1344     fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_ADDR, cmdline_addr);
1345     fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_SIZE, strlen(kernel_cmdline)+1);
1346     fw_cfg_add_string(fw_cfg, FW_CFG_CMDLINE_DATA, kernel_cmdline);
1347 
1348     if (protocol >= 0x202) {
1349         stl_p(header+0x228, cmdline_addr);
1350     } else {
1351         stw_p(header+0x20, 0xA33F);
1352         stw_p(header+0x22, cmdline_addr-real_addr);
1353     }
1354 
1355     /* handle vga= parameter */
1356     vmode = strstr(kernel_cmdline, "vga=");
1357     if (vmode) {
1358         unsigned int video_mode;
1359         /* skip "vga=" */
1360         vmode += 4;
1361         if (!strncmp(vmode, "normal", 6)) {
1362             video_mode = 0xffff;
1363         } else if (!strncmp(vmode, "ext", 3)) {
1364             video_mode = 0xfffe;
1365         } else if (!strncmp(vmode, "ask", 3)) {
1366             video_mode = 0xfffd;
1367         } else {
1368             video_mode = strtol(vmode, NULL, 0);
1369         }
1370         stw_p(header+0x1fa, video_mode);
1371     }
1372 
1373     /* loader type */
1374     /* High nybble = B reserved for QEMU; low nybble is revision number.
1375        If this code is substantially changed, you may want to consider
1376        incrementing the revision. */
1377     if (protocol >= 0x200) {
1378         header[0x210] = 0xB0;
1379     }
1380     /* heap */
1381     if (protocol >= 0x201) {
1382         header[0x211] |= 0x80;	/* CAN_USE_HEAP */
1383         stw_p(header+0x224, cmdline_addr-real_addr-0x200);
1384     }
1385 
1386     /* load initrd */
1387     if (initrd_filename) {
1388         GMappedFile *mapped_file;
1389         gsize initrd_size;
1390         gchar *initrd_data;
1391         GError *gerr = NULL;
1392 
1393         if (protocol < 0x200) {
1394             fprintf(stderr, "qemu: linux kernel too old to load a ram disk\n");
1395             exit(1);
1396         }
1397 
1398         mapped_file = g_mapped_file_new(initrd_filename, false, &gerr);
1399         if (!mapped_file) {
1400             fprintf(stderr, "qemu: error reading initrd %s: %s\n",
1401                     initrd_filename, gerr->message);
1402             exit(1);
1403         }
1404         pcms->initrd_mapped_file = mapped_file;
1405 
1406         initrd_data = g_mapped_file_get_contents(mapped_file);
1407         initrd_size = g_mapped_file_get_length(mapped_file);
1408         if (initrd_size >= initrd_max) {
1409             fprintf(stderr, "qemu: initrd is too large, cannot support."
1410                     "(max: %"PRIu32", need %"PRId64")\n",
1411                     initrd_max, (uint64_t)initrd_size);
1412             exit(1);
1413         }
1414 
1415         initrd_addr = (initrd_max-initrd_size) & ~4095;
1416 
1417         fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_ADDR, initrd_addr);
1418         fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_SIZE, initrd_size);
1419         fw_cfg_add_bytes(fw_cfg, FW_CFG_INITRD_DATA, initrd_data, initrd_size);
1420 
1421         stl_p(header+0x218, initrd_addr);
1422         stl_p(header+0x21c, initrd_size);
1423     }
1424 
1425     /* load kernel and setup */
1426     setup_size = header[0x1f1];
1427     if (setup_size == 0) {
1428         setup_size = 4;
1429     }
1430     setup_size = (setup_size+1)*512;
1431     if (setup_size > kernel_size) {
1432         fprintf(stderr, "qemu: invalid kernel header\n");
1433         exit(1);
1434     }
1435     kernel_size -= setup_size;
1436 
1437     setup  = g_malloc(setup_size);
1438     kernel = g_malloc(kernel_size);
1439     fseek(f, 0, SEEK_SET);
1440     if (fread(setup, 1, setup_size, f) != setup_size) {
1441         fprintf(stderr, "fread() failed\n");
1442         exit(1);
1443     }
1444     if (fread(kernel, 1, kernel_size, f) != kernel_size) {
1445         fprintf(stderr, "fread() failed\n");
1446         exit(1);
1447     }
1448     fclose(f);
1449 
1450     /* append dtb to kernel */
1451     if (dtb_filename) {
1452         if (protocol < 0x209) {
1453             fprintf(stderr, "qemu: Linux kernel too old to load a dtb\n");
1454             exit(1);
1455         }
1456 
1457         dtb_size = get_image_size(dtb_filename);
1458         if (dtb_size <= 0) {
1459             fprintf(stderr, "qemu: error reading dtb %s: %s\n",
1460                     dtb_filename, strerror(errno));
1461             exit(1);
1462         }
1463 
1464         setup_data_offset = QEMU_ALIGN_UP(kernel_size, 16);
1465         kernel_size = setup_data_offset + sizeof(struct setup_data) + dtb_size;
1466         kernel = g_realloc(kernel, kernel_size);
1467 
1468         stq_p(header+0x250, prot_addr + setup_data_offset);
1469 
1470         setup_data = (struct setup_data *)(kernel + setup_data_offset);
1471         setup_data->next = 0;
1472         setup_data->type = cpu_to_le32(SETUP_DTB);
1473         setup_data->len = cpu_to_le32(dtb_size);
1474 
1475         load_image_size(dtb_filename, setup_data->data, dtb_size);
1476     }
1477 
1478     memcpy(setup, header, MIN(sizeof(header), setup_size));
1479 
1480     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ADDR, prot_addr);
1481     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_SIZE, kernel_size);
1482     fw_cfg_add_bytes(fw_cfg, FW_CFG_KERNEL_DATA, kernel, kernel_size);
1483 
1484     fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_ADDR, real_addr);
1485     fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_SIZE, setup_size);
1486     fw_cfg_add_bytes(fw_cfg, FW_CFG_SETUP_DATA, setup, setup_size);
1487 
1488     option_rom[nb_option_roms].bootindex = 0;
1489     option_rom[nb_option_roms].name = "linuxboot.bin";
1490     if (pcmc->linuxboot_dma_enabled && fw_cfg_dma_enabled(fw_cfg)) {
1491         option_rom[nb_option_roms].name = "linuxboot_dma.bin";
1492     }
1493     nb_option_roms++;
1494 }
1495 
1496 #define NE2000_NB_MAX 6
1497 
1498 static const int ne2000_io[NE2000_NB_MAX] = { 0x300, 0x320, 0x340, 0x360,
1499                                               0x280, 0x380 };
1500 static const int ne2000_irq[NE2000_NB_MAX] = { 9, 10, 11, 3, 4, 5 };
1501 
1502 void pc_init_ne2k_isa(ISABus *bus, NICInfo *nd)
1503 {
1504     static int nb_ne2k = 0;
1505 
1506     if (nb_ne2k == NE2000_NB_MAX)
1507         return;
1508     isa_ne2000_init(bus, ne2000_io[nb_ne2k],
1509                     ne2000_irq[nb_ne2k], nd);
1510     nb_ne2k++;
1511 }
1512 
1513 DeviceState *cpu_get_current_apic(void)
1514 {
1515     if (current_cpu) {
1516         X86CPU *cpu = X86_CPU(current_cpu);
1517         return cpu->apic_state;
1518     } else {
1519         return NULL;
1520     }
1521 }
1522 
1523 void pc_acpi_smi_interrupt(void *opaque, int irq, int level)
1524 {
1525     X86CPU *cpu = opaque;
1526 
1527     if (level) {
1528         cpu_interrupt(CPU(cpu), CPU_INTERRUPT_SMI);
1529     }
1530 }
1531 
1532 static void pc_new_cpu(PCMachineState *pcms, int64_t apic_id, Error **errp)
1533 {
1534     Object *cpu = NULL;
1535     Error *local_err = NULL;
1536     CPUX86State *env = NULL;
1537 
1538     cpu = object_new(MACHINE(pcms)->cpu_type);
1539 
1540     env = &X86_CPU(cpu)->env;
1541     env->nr_dies = pcms->smp_dies;
1542 
1543     object_property_set_uint(cpu, apic_id, "apic-id", &local_err);
1544     object_property_set_bool(cpu, true, "realized", &local_err);
1545 
1546     object_unref(cpu);
1547     error_propagate(errp, local_err);
1548 }
1549 
1550 /*
1551  * This function is very similar to smp_parse()
1552  * in hw/core/machine.c but includes CPU die support.
1553  */
1554 void pc_smp_parse(MachineState *ms, QemuOpts *opts)
1555 {
1556     PCMachineState *pcms = PC_MACHINE(ms);
1557 
1558     if (opts) {
1559         unsigned cpus    = qemu_opt_get_number(opts, "cpus", 0);
1560         unsigned sockets = qemu_opt_get_number(opts, "sockets", 0);
1561         unsigned dies = qemu_opt_get_number(opts, "dies", 1);
1562         unsigned cores   = qemu_opt_get_number(opts, "cores", 0);
1563         unsigned threads = qemu_opt_get_number(opts, "threads", 0);
1564 
1565         /* compute missing values, prefer sockets over cores over threads */
1566         if (cpus == 0 || sockets == 0) {
1567             cores = cores > 0 ? cores : 1;
1568             threads = threads > 0 ? threads : 1;
1569             if (cpus == 0) {
1570                 sockets = sockets > 0 ? sockets : 1;
1571                 cpus = cores * threads * dies * sockets;
1572             } else {
1573                 ms->smp.max_cpus =
1574                         qemu_opt_get_number(opts, "maxcpus", cpus);
1575                 sockets = ms->smp.max_cpus / (cores * threads * dies);
1576             }
1577         } else if (cores == 0) {
1578             threads = threads > 0 ? threads : 1;
1579             cores = cpus / (sockets * dies * threads);
1580             cores = cores > 0 ? cores : 1;
1581         } else if (threads == 0) {
1582             threads = cpus / (cores * dies * sockets);
1583             threads = threads > 0 ? threads : 1;
1584         } else if (sockets * dies * cores * threads < cpus) {
1585             error_report("cpu topology: "
1586                          "sockets (%u) * dies (%u) * cores (%u) * threads (%u) < "
1587                          "smp_cpus (%u)",
1588                          sockets, dies, cores, threads, cpus);
1589             exit(1);
1590         }
1591 
1592         ms->smp.max_cpus =
1593                 qemu_opt_get_number(opts, "maxcpus", cpus);
1594 
1595         if (ms->smp.max_cpus < cpus) {
1596             error_report("maxcpus must be equal to or greater than smp");
1597             exit(1);
1598         }
1599 
1600         if (sockets * dies * cores * threads > ms->smp.max_cpus) {
1601             error_report("cpu topology: "
1602                          "sockets (%u) * dies (%u) * cores (%u) * threads (%u) > "
1603                          "maxcpus (%u)",
1604                          sockets, dies, cores, threads,
1605                          ms->smp.max_cpus);
1606             exit(1);
1607         }
1608 
1609         if (sockets * dies * cores * threads != ms->smp.max_cpus) {
1610             warn_report("Invalid CPU topology deprecated: "
1611                         "sockets (%u) * dies (%u) * cores (%u) * threads (%u) "
1612                         "!= maxcpus (%u)",
1613                         sockets, dies, cores, threads,
1614                         ms->smp.max_cpus);
1615         }
1616 
1617         ms->smp.cpus = cpus;
1618         ms->smp.cores = cores;
1619         ms->smp.threads = threads;
1620         pcms->smp_dies = dies;
1621     }
1622 
1623     if (ms->smp.cpus > 1) {
1624         Error *blocker = NULL;
1625         error_setg(&blocker, QERR_REPLAY_NOT_SUPPORTED, "smp");
1626         replay_add_blocker(blocker);
1627     }
1628 }
1629 
1630 void pc_hot_add_cpu(MachineState *ms, const int64_t id, Error **errp)
1631 {
1632     PCMachineState *pcms = PC_MACHINE(ms);
1633     int64_t apic_id = x86_cpu_apic_id_from_index(pcms, id);
1634     Error *local_err = NULL;
1635 
1636     if (id < 0) {
1637         error_setg(errp, "Invalid CPU id: %" PRIi64, id);
1638         return;
1639     }
1640 
1641     if (apic_id >= ACPI_CPU_HOTPLUG_ID_LIMIT) {
1642         error_setg(errp, "Unable to add CPU: %" PRIi64
1643                    ", resulting APIC ID (%" PRIi64 ") is too large",
1644                    id, apic_id);
1645         return;
1646     }
1647 
1648     pc_new_cpu(PC_MACHINE(ms), apic_id, &local_err);
1649     if (local_err) {
1650         error_propagate(errp, local_err);
1651         return;
1652     }
1653 }
1654 
1655 void pc_cpus_init(PCMachineState *pcms)
1656 {
1657     int i;
1658     const CPUArchIdList *possible_cpus;
1659     MachineState *ms = MACHINE(pcms);
1660     MachineClass *mc = MACHINE_GET_CLASS(pcms);
1661     PCMachineClass *pcmc = PC_MACHINE_CLASS(mc);
1662 
1663     x86_cpu_set_default_version(pcmc->default_cpu_version);
1664 
1665     /* Calculates the limit to CPU APIC ID values
1666      *
1667      * Limit for the APIC ID value, so that all
1668      * CPU APIC IDs are < pcms->apic_id_limit.
1669      *
1670      * This is used for FW_CFG_MAX_CPUS. See comments on bochs_bios_init().
1671      */
1672     pcms->apic_id_limit = x86_cpu_apic_id_from_index(pcms,
1673                                                      ms->smp.max_cpus - 1) + 1;
1674     possible_cpus = mc->possible_cpu_arch_ids(ms);
1675     for (i = 0; i < ms->smp.cpus; i++) {
1676         pc_new_cpu(pcms, possible_cpus->cpus[i].arch_id, &error_fatal);
1677     }
1678 }
1679 
1680 static void pc_build_feature_control_file(PCMachineState *pcms)
1681 {
1682     MachineState *ms = MACHINE(pcms);
1683     X86CPU *cpu = X86_CPU(ms->possible_cpus->cpus[0].cpu);
1684     CPUX86State *env = &cpu->env;
1685     uint32_t unused, ecx, edx;
1686     uint64_t feature_control_bits = 0;
1687     uint64_t *val;
1688 
1689     cpu_x86_cpuid(env, 1, 0, &unused, &unused, &ecx, &edx);
1690     if (ecx & CPUID_EXT_VMX) {
1691         feature_control_bits |= FEATURE_CONTROL_VMXON_ENABLED_OUTSIDE_SMX;
1692     }
1693 
1694     if ((edx & (CPUID_EXT2_MCE | CPUID_EXT2_MCA)) ==
1695         (CPUID_EXT2_MCE | CPUID_EXT2_MCA) &&
1696         (env->mcg_cap & MCG_LMCE_P)) {
1697         feature_control_bits |= FEATURE_CONTROL_LMCE;
1698     }
1699 
1700     if (!feature_control_bits) {
1701         return;
1702     }
1703 
1704     val = g_malloc(sizeof(*val));
1705     *val = cpu_to_le64(feature_control_bits | FEATURE_CONTROL_LOCKED);
1706     fw_cfg_add_file(pcms->fw_cfg, "etc/msr_feature_control", val, sizeof(*val));
1707 }
1708 
1709 static void rtc_set_cpus_count(ISADevice *rtc, uint16_t cpus_count)
1710 {
1711     if (cpus_count > 0xff) {
1712         /* If the number of CPUs can't be represented in 8 bits, the
1713          * BIOS must use "FW_CFG_NB_CPUS". Set RTC field to 0 just
1714          * to make old BIOSes fail more predictably.
1715          */
1716         rtc_set_memory(rtc, 0x5f, 0);
1717     } else {
1718         rtc_set_memory(rtc, 0x5f, cpus_count - 1);
1719     }
1720 }
1721 
1722 static
1723 void pc_machine_done(Notifier *notifier, void *data)
1724 {
1725     PCMachineState *pcms = container_of(notifier,
1726                                         PCMachineState, machine_done);
1727     PCIBus *bus = pcms->bus;
1728 
1729     /* set the number of CPUs */
1730     rtc_set_cpus_count(pcms->rtc, pcms->boot_cpus);
1731 
1732     if (bus) {
1733         int extra_hosts = 0;
1734 
1735         QLIST_FOREACH(bus, &bus->child, sibling) {
1736             /* look for expander root buses */
1737             if (pci_bus_is_root(bus)) {
1738                 extra_hosts++;
1739             }
1740         }
1741         if (extra_hosts && pcms->fw_cfg) {
1742             uint64_t *val = g_malloc(sizeof(*val));
1743             *val = cpu_to_le64(extra_hosts);
1744             fw_cfg_add_file(pcms->fw_cfg,
1745                     "etc/extra-pci-roots", val, sizeof(*val));
1746         }
1747     }
1748 
1749     acpi_setup();
1750     if (pcms->fw_cfg) {
1751         pc_build_smbios(pcms);
1752         pc_build_feature_control_file(pcms);
1753         /* update FW_CFG_NB_CPUS to account for -device added CPUs */
1754         fw_cfg_modify_i16(pcms->fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
1755     }
1756 
1757     if (pcms->apic_id_limit > 255 && !xen_enabled()) {
1758         IntelIOMMUState *iommu = INTEL_IOMMU_DEVICE(x86_iommu_get_default());
1759 
1760         if (!iommu || !x86_iommu_ir_supported(X86_IOMMU_DEVICE(iommu)) ||
1761             iommu->intr_eim != ON_OFF_AUTO_ON) {
1762             error_report("current -smp configuration requires "
1763                          "Extended Interrupt Mode enabled. "
1764                          "You can add an IOMMU using: "
1765                          "-device intel-iommu,intremap=on,eim=on");
1766             exit(EXIT_FAILURE);
1767         }
1768     }
1769 }
1770 
1771 void pc_guest_info_init(PCMachineState *pcms)
1772 {
1773     int i;
1774 
1775     pcms->apic_xrupt_override = kvm_allows_irq0_override();
1776     pcms->numa_nodes = nb_numa_nodes;
1777     pcms->node_mem = g_malloc0(pcms->numa_nodes *
1778                                     sizeof *pcms->node_mem);
1779     for (i = 0; i < nb_numa_nodes; i++) {
1780         pcms->node_mem[i] = numa_info[i].node_mem;
1781     }
1782 
1783     pcms->machine_done.notify = pc_machine_done;
1784     qemu_add_machine_init_done_notifier(&pcms->machine_done);
1785 }
1786 
1787 /* setup pci memory address space mapping into system address space */
1788 void pc_pci_as_mapping_init(Object *owner, MemoryRegion *system_memory,
1789                             MemoryRegion *pci_address_space)
1790 {
1791     /* Set to lower priority than RAM */
1792     memory_region_add_subregion_overlap(system_memory, 0x0,
1793                                         pci_address_space, -1);
1794 }
1795 
1796 void xen_load_linux(PCMachineState *pcms)
1797 {
1798     int i;
1799     FWCfgState *fw_cfg;
1800 
1801     assert(MACHINE(pcms)->kernel_filename != NULL);
1802 
1803     fw_cfg = fw_cfg_init_io(FW_CFG_IO_BASE);
1804     fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
1805     rom_set_fw(fw_cfg);
1806 
1807     load_linux(pcms, fw_cfg);
1808     for (i = 0; i < nb_option_roms; i++) {
1809         assert(!strcmp(option_rom[i].name, "linuxboot.bin") ||
1810                !strcmp(option_rom[i].name, "linuxboot_dma.bin") ||
1811                !strcmp(option_rom[i].name, "pvh.bin") ||
1812                !strcmp(option_rom[i].name, "multiboot.bin"));
1813         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1814     }
1815     pcms->fw_cfg = fw_cfg;
1816 }
1817 
1818 void pc_memory_init(PCMachineState *pcms,
1819                     MemoryRegion *system_memory,
1820                     MemoryRegion *rom_memory,
1821                     MemoryRegion **ram_memory)
1822 {
1823     int linux_boot, i;
1824     MemoryRegion *ram, *option_rom_mr;
1825     MemoryRegion *ram_below_4g, *ram_above_4g;
1826     FWCfgState *fw_cfg;
1827     MachineState *machine = MACHINE(pcms);
1828     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1829 
1830     assert(machine->ram_size == pcms->below_4g_mem_size +
1831                                 pcms->above_4g_mem_size);
1832 
1833     linux_boot = (machine->kernel_filename != NULL);
1834 
1835     /* Allocate RAM.  We allocate it as a single memory region and use
1836      * aliases to address portions of it, mostly for backwards compatibility
1837      * with older qemus that used qemu_ram_alloc().
1838      */
1839     ram = g_malloc(sizeof(*ram));
1840     memory_region_allocate_system_memory(ram, NULL, "pc.ram",
1841                                          machine->ram_size);
1842     *ram_memory = ram;
1843     ram_below_4g = g_malloc(sizeof(*ram_below_4g));
1844     memory_region_init_alias(ram_below_4g, NULL, "ram-below-4g", ram,
1845                              0, pcms->below_4g_mem_size);
1846     memory_region_add_subregion(system_memory, 0, ram_below_4g);
1847     e820_add_entry(0, pcms->below_4g_mem_size, E820_RAM);
1848     if (pcms->above_4g_mem_size > 0) {
1849         ram_above_4g = g_malloc(sizeof(*ram_above_4g));
1850         memory_region_init_alias(ram_above_4g, NULL, "ram-above-4g", ram,
1851                                  pcms->below_4g_mem_size,
1852                                  pcms->above_4g_mem_size);
1853         memory_region_add_subregion(system_memory, 0x100000000ULL,
1854                                     ram_above_4g);
1855         e820_add_entry(0x100000000ULL, pcms->above_4g_mem_size, E820_RAM);
1856     }
1857 
1858     if (!pcmc->has_reserved_memory &&
1859         (machine->ram_slots ||
1860          (machine->maxram_size > machine->ram_size))) {
1861         MachineClass *mc = MACHINE_GET_CLASS(machine);
1862 
1863         error_report("\"-memory 'slots|maxmem'\" is not supported by: %s",
1864                      mc->name);
1865         exit(EXIT_FAILURE);
1866     }
1867 
1868     /* always allocate the device memory information */
1869     machine->device_memory = g_malloc0(sizeof(*machine->device_memory));
1870 
1871     /* initialize device memory address space */
1872     if (pcmc->has_reserved_memory &&
1873         (machine->ram_size < machine->maxram_size)) {
1874         ram_addr_t device_mem_size = machine->maxram_size - machine->ram_size;
1875 
1876         if (machine->ram_slots > ACPI_MAX_RAM_SLOTS) {
1877             error_report("unsupported amount of memory slots: %"PRIu64,
1878                          machine->ram_slots);
1879             exit(EXIT_FAILURE);
1880         }
1881 
1882         if (QEMU_ALIGN_UP(machine->maxram_size,
1883                           TARGET_PAGE_SIZE) != machine->maxram_size) {
1884             error_report("maximum memory size must by aligned to multiple of "
1885                          "%d bytes", TARGET_PAGE_SIZE);
1886             exit(EXIT_FAILURE);
1887         }
1888 
1889         machine->device_memory->base =
1890             ROUND_UP(0x100000000ULL + pcms->above_4g_mem_size, 1 * GiB);
1891 
1892         if (pcmc->enforce_aligned_dimm) {
1893             /* size device region assuming 1G page max alignment per slot */
1894             device_mem_size += (1 * GiB) * machine->ram_slots;
1895         }
1896 
1897         if ((machine->device_memory->base + device_mem_size) <
1898             device_mem_size) {
1899             error_report("unsupported amount of maximum memory: " RAM_ADDR_FMT,
1900                          machine->maxram_size);
1901             exit(EXIT_FAILURE);
1902         }
1903 
1904         memory_region_init(&machine->device_memory->mr, OBJECT(pcms),
1905                            "device-memory", device_mem_size);
1906         memory_region_add_subregion(system_memory, machine->device_memory->base,
1907                                     &machine->device_memory->mr);
1908     }
1909 
1910     /* Initialize PC system firmware */
1911     pc_system_firmware_init(pcms, rom_memory);
1912 
1913     option_rom_mr = g_malloc(sizeof(*option_rom_mr));
1914     memory_region_init_ram(option_rom_mr, NULL, "pc.rom", PC_ROM_SIZE,
1915                            &error_fatal);
1916     if (pcmc->pci_enabled) {
1917         memory_region_set_readonly(option_rom_mr, true);
1918     }
1919     memory_region_add_subregion_overlap(rom_memory,
1920                                         PC_ROM_MIN_VGA,
1921                                         option_rom_mr,
1922                                         1);
1923 
1924     fw_cfg = bochs_bios_init(&address_space_memory, pcms);
1925 
1926     rom_set_fw(fw_cfg);
1927 
1928     if (pcmc->has_reserved_memory && machine->device_memory->base) {
1929         uint64_t *val = g_malloc(sizeof(*val));
1930         PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1931         uint64_t res_mem_end = machine->device_memory->base;
1932 
1933         if (!pcmc->broken_reserved_end) {
1934             res_mem_end += memory_region_size(&machine->device_memory->mr);
1935         }
1936         *val = cpu_to_le64(ROUND_UP(res_mem_end, 1 * GiB));
1937         fw_cfg_add_file(fw_cfg, "etc/reserved-memory-end", val, sizeof(*val));
1938     }
1939 
1940     if (linux_boot) {
1941         load_linux(pcms, fw_cfg);
1942     }
1943 
1944     for (i = 0; i < nb_option_roms; i++) {
1945         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1946     }
1947     pcms->fw_cfg = fw_cfg;
1948 
1949     /* Init default IOAPIC address space */
1950     pcms->ioapic_as = &address_space_memory;
1951 }
1952 
1953 /*
1954  * The 64bit pci hole starts after "above 4G RAM" and
1955  * potentially the space reserved for memory hotplug.
1956  */
1957 uint64_t pc_pci_hole64_start(void)
1958 {
1959     PCMachineState *pcms = PC_MACHINE(qdev_get_machine());
1960     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1961     MachineState *ms = MACHINE(pcms);
1962     uint64_t hole64_start = 0;
1963 
1964     if (pcmc->has_reserved_memory && ms->device_memory->base) {
1965         hole64_start = ms->device_memory->base;
1966         if (!pcmc->broken_reserved_end) {
1967             hole64_start += memory_region_size(&ms->device_memory->mr);
1968         }
1969     } else {
1970         hole64_start = 0x100000000ULL + pcms->above_4g_mem_size;
1971     }
1972 
1973     return ROUND_UP(hole64_start, 1 * GiB);
1974 }
1975 
1976 qemu_irq pc_allocate_cpu_irq(void)
1977 {
1978     return qemu_allocate_irq(pic_irq_request, NULL, 0);
1979 }
1980 
1981 DeviceState *pc_vga_init(ISABus *isa_bus, PCIBus *pci_bus)
1982 {
1983     DeviceState *dev = NULL;
1984 
1985     rom_set_order_override(FW_CFG_ORDER_OVERRIDE_VGA);
1986     if (pci_bus) {
1987         PCIDevice *pcidev = pci_vga_init(pci_bus);
1988         dev = pcidev ? &pcidev->qdev : NULL;
1989     } else if (isa_bus) {
1990         ISADevice *isadev = isa_vga_init(isa_bus);
1991         dev = isadev ? DEVICE(isadev) : NULL;
1992     }
1993     rom_reset_order_override();
1994     return dev;
1995 }
1996 
1997 static const MemoryRegionOps ioport80_io_ops = {
1998     .write = ioport80_write,
1999     .read = ioport80_read,
2000     .endianness = DEVICE_NATIVE_ENDIAN,
2001     .impl = {
2002         .min_access_size = 1,
2003         .max_access_size = 1,
2004     },
2005 };
2006 
2007 static const MemoryRegionOps ioportF0_io_ops = {
2008     .write = ioportF0_write,
2009     .read = ioportF0_read,
2010     .endianness = DEVICE_NATIVE_ENDIAN,
2011     .impl = {
2012         .min_access_size = 1,
2013         .max_access_size = 1,
2014     },
2015 };
2016 
2017 static void pc_superio_init(ISABus *isa_bus, bool create_fdctrl, bool no_vmport)
2018 {
2019     int i;
2020     DriveInfo *fd[MAX_FD];
2021     qemu_irq *a20_line;
2022     ISADevice *i8042, *port92, *vmmouse;
2023 
2024     serial_hds_isa_init(isa_bus, 0, MAX_ISA_SERIAL_PORTS);
2025     parallel_hds_isa_init(isa_bus, MAX_PARALLEL_PORTS);
2026 
2027     for (i = 0; i < MAX_FD; i++) {
2028         fd[i] = drive_get(IF_FLOPPY, 0, i);
2029         create_fdctrl |= !!fd[i];
2030     }
2031     if (create_fdctrl) {
2032         fdctrl_init_isa(isa_bus, fd);
2033     }
2034 
2035     i8042 = isa_create_simple(isa_bus, "i8042");
2036     if (!no_vmport) {
2037         vmport_init(isa_bus);
2038         vmmouse = isa_try_create(isa_bus, "vmmouse");
2039     } else {
2040         vmmouse = NULL;
2041     }
2042     if (vmmouse) {
2043         DeviceState *dev = DEVICE(vmmouse);
2044         qdev_prop_set_ptr(dev, "ps2_mouse", i8042);
2045         qdev_init_nofail(dev);
2046     }
2047     port92 = isa_create_simple(isa_bus, "port92");
2048 
2049     a20_line = qemu_allocate_irqs(handle_a20_line_change, first_cpu, 2);
2050     i8042_setup_a20_line(i8042, a20_line[0]);
2051     port92_init(port92, a20_line[1]);
2052     g_free(a20_line);
2053 }
2054 
2055 void pc_basic_device_init(ISABus *isa_bus, qemu_irq *gsi,
2056                           ISADevice **rtc_state,
2057                           bool create_fdctrl,
2058                           bool no_vmport,
2059                           bool has_pit,
2060                           uint32_t hpet_irqs)
2061 {
2062     int i;
2063     DeviceState *hpet = NULL;
2064     int pit_isa_irq = 0;
2065     qemu_irq pit_alt_irq = NULL;
2066     qemu_irq rtc_irq = NULL;
2067     ISADevice *pit = NULL;
2068     MemoryRegion *ioport80_io = g_new(MemoryRegion, 1);
2069     MemoryRegion *ioportF0_io = g_new(MemoryRegion, 1);
2070 
2071     memory_region_init_io(ioport80_io, NULL, &ioport80_io_ops, NULL, "ioport80", 1);
2072     memory_region_add_subregion(isa_bus->address_space_io, 0x80, ioport80_io);
2073 
2074     memory_region_init_io(ioportF0_io, NULL, &ioportF0_io_ops, NULL, "ioportF0", 1);
2075     memory_region_add_subregion(isa_bus->address_space_io, 0xf0, ioportF0_io);
2076 
2077     /*
2078      * Check if an HPET shall be created.
2079      *
2080      * Without KVM_CAP_PIT_STATE2, we cannot switch off the in-kernel PIT
2081      * when the HPET wants to take over. Thus we have to disable the latter.
2082      */
2083     if (!no_hpet && (!kvm_irqchip_in_kernel() || kvm_has_pit_state2())) {
2084         /* In order to set property, here not using sysbus_try_create_simple */
2085         hpet = qdev_try_create(NULL, TYPE_HPET);
2086         if (hpet) {
2087             /* For pc-piix-*, hpet's intcap is always IRQ2. For pc-q35-1.7
2088              * and earlier, use IRQ2 for compat. Otherwise, use IRQ16~23,
2089              * IRQ8 and IRQ2.
2090              */
2091             uint8_t compat = object_property_get_uint(OBJECT(hpet),
2092                     HPET_INTCAP, NULL);
2093             if (!compat) {
2094                 qdev_prop_set_uint32(hpet, HPET_INTCAP, hpet_irqs);
2095             }
2096             qdev_init_nofail(hpet);
2097             sysbus_mmio_map(SYS_BUS_DEVICE(hpet), 0, HPET_BASE);
2098 
2099             for (i = 0; i < GSI_NUM_PINS; i++) {
2100                 sysbus_connect_irq(SYS_BUS_DEVICE(hpet), i, gsi[i]);
2101             }
2102             pit_isa_irq = -1;
2103             pit_alt_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_PIT_INT);
2104             rtc_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_RTC_INT);
2105         }
2106     }
2107     *rtc_state = mc146818_rtc_init(isa_bus, 2000, rtc_irq);
2108 
2109     qemu_register_boot_set(pc_boot_set, *rtc_state);
2110 
2111     if (!xen_enabled() && has_pit) {
2112         if (kvm_pit_in_kernel()) {
2113             pit = kvm_pit_init(isa_bus, 0x40);
2114         } else {
2115             pit = i8254_pit_init(isa_bus, 0x40, pit_isa_irq, pit_alt_irq);
2116         }
2117         if (hpet) {
2118             /* connect PIT to output control line of the HPET */
2119             qdev_connect_gpio_out(hpet, 0, qdev_get_gpio_in(DEVICE(pit), 0));
2120         }
2121         pcspk_init(isa_bus, pit);
2122     }
2123 
2124     i8257_dma_init(isa_bus, 0);
2125 
2126     /* Super I/O */
2127     pc_superio_init(isa_bus, create_fdctrl, no_vmport);
2128 }
2129 
2130 void pc_nic_init(PCMachineClass *pcmc, ISABus *isa_bus, PCIBus *pci_bus)
2131 {
2132     int i;
2133 
2134     rom_set_order_override(FW_CFG_ORDER_OVERRIDE_NIC);
2135     for (i = 0; i < nb_nics; i++) {
2136         NICInfo *nd = &nd_table[i];
2137         const char *model = nd->model ? nd->model : pcmc->default_nic_model;
2138 
2139         if (g_str_equal(model, "ne2k_isa")) {
2140             pc_init_ne2k_isa(isa_bus, nd);
2141         } else {
2142             pci_nic_init_nofail(nd, pci_bus, model, NULL);
2143         }
2144     }
2145     rom_reset_order_override();
2146 }
2147 
2148 void ioapic_init_gsi(GSIState *gsi_state, const char *parent_name)
2149 {
2150     DeviceState *dev;
2151     SysBusDevice *d;
2152     unsigned int i;
2153 
2154     if (kvm_ioapic_in_kernel()) {
2155         dev = qdev_create(NULL, TYPE_KVM_IOAPIC);
2156     } else {
2157         dev = qdev_create(NULL, TYPE_IOAPIC);
2158     }
2159     if (parent_name) {
2160         object_property_add_child(object_resolve_path(parent_name, NULL),
2161                                   "ioapic", OBJECT(dev), NULL);
2162     }
2163     qdev_init_nofail(dev);
2164     d = SYS_BUS_DEVICE(dev);
2165     sysbus_mmio_map(d, 0, IO_APIC_DEFAULT_ADDRESS);
2166 
2167     for (i = 0; i < IOAPIC_NUM_PINS; i++) {
2168         gsi_state->ioapic_irq[i] = qdev_get_gpio_in(dev, i);
2169     }
2170 }
2171 
2172 static void pc_memory_pre_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
2173                                Error **errp)
2174 {
2175     const PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2176     const PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
2177     const MachineState *ms = MACHINE(hotplug_dev);
2178     const bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2179     const uint64_t legacy_align = TARGET_PAGE_SIZE;
2180     Error *local_err = NULL;
2181 
2182     /*
2183      * When -no-acpi is used with Q35 machine type, no ACPI is built,
2184      * but pcms->acpi_dev is still created. Check !acpi_enabled in
2185      * addition to cover this case.
2186      */
2187     if (!pcms->acpi_dev || !acpi_enabled) {
2188         error_setg(errp,
2189                    "memory hotplug is not enabled: missing acpi device or acpi disabled");
2190         return;
2191     }
2192 
2193     if (is_nvdimm && !ms->nvdimms_state->is_enabled) {
2194         error_setg(errp, "nvdimm is not enabled: missing 'nvdimm' in '-M'");
2195         return;
2196     }
2197 
2198     hotplug_handler_pre_plug(pcms->acpi_dev, dev, &local_err);
2199     if (local_err) {
2200         error_propagate(errp, local_err);
2201         return;
2202     }
2203 
2204     pc_dimm_pre_plug(PC_DIMM(dev), MACHINE(hotplug_dev),
2205                      pcmc->enforce_aligned_dimm ? NULL : &legacy_align, errp);
2206 }
2207 
2208 static void pc_memory_plug(HotplugHandler *hotplug_dev,
2209                            DeviceState *dev, Error **errp)
2210 {
2211     Error *local_err = NULL;
2212     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2213     MachineState *ms = MACHINE(hotplug_dev);
2214     bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2215 
2216     pc_dimm_plug(PC_DIMM(dev), MACHINE(pcms), &local_err);
2217     if (local_err) {
2218         goto out;
2219     }
2220 
2221     if (is_nvdimm) {
2222         nvdimm_plug(ms->nvdimms_state);
2223     }
2224 
2225     hotplug_handler_plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &error_abort);
2226 out:
2227     error_propagate(errp, local_err);
2228 }
2229 
2230 static void pc_memory_unplug_request(HotplugHandler *hotplug_dev,
2231                                      DeviceState *dev, Error **errp)
2232 {
2233     Error *local_err = NULL;
2234     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2235 
2236     /*
2237      * When -no-acpi is used with Q35 machine type, no ACPI is built,
2238      * but pcms->acpi_dev is still created. Check !acpi_enabled in
2239      * addition to cover this case.
2240      */
2241     if (!pcms->acpi_dev || !acpi_enabled) {
2242         error_setg(&local_err,
2243                    "memory hotplug is not enabled: missing acpi device or acpi disabled");
2244         goto out;
2245     }
2246 
2247     if (object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM)) {
2248         error_setg(&local_err,
2249                    "nvdimm device hot unplug is not supported yet.");
2250         goto out;
2251     }
2252 
2253     hotplug_handler_unplug_request(HOTPLUG_HANDLER(pcms->acpi_dev), dev,
2254                                    &local_err);
2255 out:
2256     error_propagate(errp, local_err);
2257 }
2258 
2259 static void pc_memory_unplug(HotplugHandler *hotplug_dev,
2260                              DeviceState *dev, Error **errp)
2261 {
2262     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2263     Error *local_err = NULL;
2264 
2265     hotplug_handler_unplug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
2266     if (local_err) {
2267         goto out;
2268     }
2269 
2270     pc_dimm_unplug(PC_DIMM(dev), MACHINE(pcms));
2271     object_property_set_bool(OBJECT(dev), false, "realized", NULL);
2272  out:
2273     error_propagate(errp, local_err);
2274 }
2275 
2276 static int pc_apic_cmp(const void *a, const void *b)
2277 {
2278    CPUArchId *apic_a = (CPUArchId *)a;
2279    CPUArchId *apic_b = (CPUArchId *)b;
2280 
2281    return apic_a->arch_id - apic_b->arch_id;
2282 }
2283 
2284 /* returns pointer to CPUArchId descriptor that matches CPU's apic_id
2285  * in ms->possible_cpus->cpus, if ms->possible_cpus->cpus has no
2286  * entry corresponding to CPU's apic_id returns NULL.
2287  */
2288 static CPUArchId *pc_find_cpu_slot(MachineState *ms, uint32_t id, int *idx)
2289 {
2290     CPUArchId apic_id, *found_cpu;
2291 
2292     apic_id.arch_id = id;
2293     found_cpu = bsearch(&apic_id, ms->possible_cpus->cpus,
2294         ms->possible_cpus->len, sizeof(*ms->possible_cpus->cpus),
2295         pc_apic_cmp);
2296     if (found_cpu && idx) {
2297         *idx = found_cpu - ms->possible_cpus->cpus;
2298     }
2299     return found_cpu;
2300 }
2301 
2302 static void pc_cpu_plug(HotplugHandler *hotplug_dev,
2303                         DeviceState *dev, Error **errp)
2304 {
2305     CPUArchId *found_cpu;
2306     Error *local_err = NULL;
2307     X86CPU *cpu = X86_CPU(dev);
2308     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2309 
2310     if (pcms->acpi_dev) {
2311         hotplug_handler_plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
2312         if (local_err) {
2313             goto out;
2314         }
2315     }
2316 
2317     /* increment the number of CPUs */
2318     pcms->boot_cpus++;
2319     if (pcms->rtc) {
2320         rtc_set_cpus_count(pcms->rtc, pcms->boot_cpus);
2321     }
2322     if (pcms->fw_cfg) {
2323         fw_cfg_modify_i16(pcms->fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
2324     }
2325 
2326     found_cpu = pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, NULL);
2327     found_cpu->cpu = OBJECT(dev);
2328 out:
2329     error_propagate(errp, local_err);
2330 }
2331 static void pc_cpu_unplug_request_cb(HotplugHandler *hotplug_dev,
2332                                      DeviceState *dev, Error **errp)
2333 {
2334     int idx = -1;
2335     Error *local_err = NULL;
2336     X86CPU *cpu = X86_CPU(dev);
2337     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2338 
2339     if (!pcms->acpi_dev) {
2340         error_setg(&local_err, "CPU hot unplug not supported without ACPI");
2341         goto out;
2342     }
2343 
2344     pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, &idx);
2345     assert(idx != -1);
2346     if (idx == 0) {
2347         error_setg(&local_err, "Boot CPU is unpluggable");
2348         goto out;
2349     }
2350 
2351     hotplug_handler_unplug_request(HOTPLUG_HANDLER(pcms->acpi_dev), dev,
2352                                    &local_err);
2353     if (local_err) {
2354         goto out;
2355     }
2356 
2357  out:
2358     error_propagate(errp, local_err);
2359 
2360 }
2361 
2362 static void pc_cpu_unplug_cb(HotplugHandler *hotplug_dev,
2363                              DeviceState *dev, Error **errp)
2364 {
2365     CPUArchId *found_cpu;
2366     Error *local_err = NULL;
2367     X86CPU *cpu = X86_CPU(dev);
2368     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2369 
2370     hotplug_handler_unplug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
2371     if (local_err) {
2372         goto out;
2373     }
2374 
2375     found_cpu = pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, NULL);
2376     found_cpu->cpu = NULL;
2377     object_property_set_bool(OBJECT(dev), false, "realized", NULL);
2378 
2379     /* decrement the number of CPUs */
2380     pcms->boot_cpus--;
2381     /* Update the number of CPUs in CMOS */
2382     rtc_set_cpus_count(pcms->rtc, pcms->boot_cpus);
2383     fw_cfg_modify_i16(pcms->fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
2384  out:
2385     error_propagate(errp, local_err);
2386 }
2387 
2388 static void pc_cpu_pre_plug(HotplugHandler *hotplug_dev,
2389                             DeviceState *dev, Error **errp)
2390 {
2391     int idx;
2392     CPUState *cs;
2393     CPUArchId *cpu_slot;
2394     X86CPUTopoInfo topo;
2395     X86CPU *cpu = X86_CPU(dev);
2396     CPUX86State *env = &cpu->env;
2397     MachineState *ms = MACHINE(hotplug_dev);
2398     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2399     unsigned int smp_cores = ms->smp.cores;
2400     unsigned int smp_threads = ms->smp.threads;
2401 
2402     if(!object_dynamic_cast(OBJECT(cpu), ms->cpu_type)) {
2403         error_setg(errp, "Invalid CPU type, expected cpu type: '%s'",
2404                    ms->cpu_type);
2405         return;
2406     }
2407 
2408     env->nr_dies = pcms->smp_dies;
2409 
2410     /*
2411      * If APIC ID is not set,
2412      * set it based on socket/die/core/thread properties.
2413      */
2414     if (cpu->apic_id == UNASSIGNED_APIC_ID) {
2415         int max_socket = (ms->smp.max_cpus - 1) /
2416                                 smp_threads / smp_cores / pcms->smp_dies;
2417 
2418         if (cpu->socket_id < 0) {
2419             error_setg(errp, "CPU socket-id is not set");
2420             return;
2421         } else if (cpu->socket_id > max_socket) {
2422             error_setg(errp, "Invalid CPU socket-id: %u must be in range 0:%u",
2423                        cpu->socket_id, max_socket);
2424             return;
2425         } else if (cpu->die_id > pcms->smp_dies - 1) {
2426             error_setg(errp, "Invalid CPU die-id: %u must be in range 0:%u",
2427                        cpu->die_id, max_socket);
2428             return;
2429         }
2430         if (cpu->core_id < 0) {
2431             error_setg(errp, "CPU core-id is not set");
2432             return;
2433         } else if (cpu->core_id > (smp_cores - 1)) {
2434             error_setg(errp, "Invalid CPU core-id: %u must be in range 0:%u",
2435                        cpu->core_id, smp_cores - 1);
2436             return;
2437         }
2438         if (cpu->thread_id < 0) {
2439             error_setg(errp, "CPU thread-id is not set");
2440             return;
2441         } else if (cpu->thread_id > (smp_threads - 1)) {
2442             error_setg(errp, "Invalid CPU thread-id: %u must be in range 0:%u",
2443                        cpu->thread_id, smp_threads - 1);
2444             return;
2445         }
2446 
2447         topo.pkg_id = cpu->socket_id;
2448         topo.die_id = cpu->die_id;
2449         topo.core_id = cpu->core_id;
2450         topo.smt_id = cpu->thread_id;
2451         cpu->apic_id = apicid_from_topo_ids(pcms->smp_dies, smp_cores,
2452                                             smp_threads, &topo);
2453     }
2454 
2455     cpu_slot = pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, &idx);
2456     if (!cpu_slot) {
2457         MachineState *ms = MACHINE(pcms);
2458 
2459         x86_topo_ids_from_apicid(cpu->apic_id, pcms->smp_dies,
2460                                  smp_cores, smp_threads, &topo);
2461         error_setg(errp,
2462             "Invalid CPU [socket: %u, die: %u, core: %u, thread: %u] with"
2463             " APIC ID %" PRIu32 ", valid index range 0:%d",
2464             topo.pkg_id, topo.die_id, topo.core_id, topo.smt_id,
2465             cpu->apic_id, ms->possible_cpus->len - 1);
2466         return;
2467     }
2468 
2469     if (cpu_slot->cpu) {
2470         error_setg(errp, "CPU[%d] with APIC ID %" PRIu32 " exists",
2471                    idx, cpu->apic_id);
2472         return;
2473     }
2474 
2475     /* if 'address' properties socket-id/core-id/thread-id are not set, set them
2476      * so that machine_query_hotpluggable_cpus would show correct values
2477      */
2478     /* TODO: move socket_id/core_id/thread_id checks into x86_cpu_realizefn()
2479      * once -smp refactoring is complete and there will be CPU private
2480      * CPUState::nr_cores and CPUState::nr_threads fields instead of globals */
2481     x86_topo_ids_from_apicid(cpu->apic_id, pcms->smp_dies,
2482                              smp_cores, smp_threads, &topo);
2483     if (cpu->socket_id != -1 && cpu->socket_id != topo.pkg_id) {
2484         error_setg(errp, "property socket-id: %u doesn't match set apic-id:"
2485             " 0x%x (socket-id: %u)", cpu->socket_id, cpu->apic_id, topo.pkg_id);
2486         return;
2487     }
2488     cpu->socket_id = topo.pkg_id;
2489 
2490     if (cpu->die_id != -1 && cpu->die_id != topo.die_id) {
2491         error_setg(errp, "property die-id: %u doesn't match set apic-id:"
2492             " 0x%x (die-id: %u)", cpu->die_id, cpu->apic_id, topo.die_id);
2493         return;
2494     }
2495     cpu->die_id = topo.die_id;
2496 
2497     if (cpu->core_id != -1 && cpu->core_id != topo.core_id) {
2498         error_setg(errp, "property core-id: %u doesn't match set apic-id:"
2499             " 0x%x (core-id: %u)", cpu->core_id, cpu->apic_id, topo.core_id);
2500         return;
2501     }
2502     cpu->core_id = topo.core_id;
2503 
2504     if (cpu->thread_id != -1 && cpu->thread_id != topo.smt_id) {
2505         error_setg(errp, "property thread-id: %u doesn't match set apic-id:"
2506             " 0x%x (thread-id: %u)", cpu->thread_id, cpu->apic_id, topo.smt_id);
2507         return;
2508     }
2509     cpu->thread_id = topo.smt_id;
2510 
2511     if (hyperv_feat_enabled(cpu, HYPERV_FEAT_VPINDEX) &&
2512         !kvm_hv_vpindex_settable()) {
2513         error_setg(errp, "kernel doesn't allow setting HyperV VP_INDEX");
2514         return;
2515     }
2516 
2517     cs = CPU(cpu);
2518     cs->cpu_index = idx;
2519 
2520     numa_cpu_pre_plug(cpu_slot, dev, errp);
2521 }
2522 
2523 static void pc_virtio_pmem_pci_pre_plug(HotplugHandler *hotplug_dev,
2524                                         DeviceState *dev, Error **errp)
2525 {
2526     HotplugHandler *hotplug_dev2 = qdev_get_bus_hotplug_handler(dev);
2527     Error *local_err = NULL;
2528 
2529     if (!hotplug_dev2) {
2530         /*
2531          * Without a bus hotplug handler, we cannot control the plug/unplug
2532          * order. This should never be the case on x86, however better add
2533          * a safety net.
2534          */
2535         error_setg(errp, "virtio-pmem-pci not supported on this bus.");
2536         return;
2537     }
2538     /*
2539      * First, see if we can plug this memory device at all. If that
2540      * succeeds, branch of to the actual hotplug handler.
2541      */
2542     memory_device_pre_plug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev), NULL,
2543                            &local_err);
2544     if (!local_err) {
2545         hotplug_handler_pre_plug(hotplug_dev2, dev, &local_err);
2546     }
2547     error_propagate(errp, local_err);
2548 }
2549 
2550 static void pc_virtio_pmem_pci_plug(HotplugHandler *hotplug_dev,
2551                                     DeviceState *dev, Error **errp)
2552 {
2553     HotplugHandler *hotplug_dev2 = qdev_get_bus_hotplug_handler(dev);
2554     Error *local_err = NULL;
2555 
2556     /*
2557      * Plug the memory device first and then branch off to the actual
2558      * hotplug handler. If that one fails, we can easily undo the memory
2559      * device bits.
2560      */
2561     memory_device_plug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev));
2562     hotplug_handler_plug(hotplug_dev2, dev, &local_err);
2563     if (local_err) {
2564         memory_device_unplug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev));
2565     }
2566     error_propagate(errp, local_err);
2567 }
2568 
2569 static void pc_virtio_pmem_pci_unplug_request(HotplugHandler *hotplug_dev,
2570                                               DeviceState *dev, Error **errp)
2571 {
2572     /* We don't support virtio pmem hot unplug */
2573     error_setg(errp, "virtio pmem device unplug not supported.");
2574 }
2575 
2576 static void pc_virtio_pmem_pci_unplug(HotplugHandler *hotplug_dev,
2577                                       DeviceState *dev, Error **errp)
2578 {
2579     /* We don't support virtio pmem hot unplug */
2580 }
2581 
2582 static void pc_machine_device_pre_plug_cb(HotplugHandler *hotplug_dev,
2583                                           DeviceState *dev, Error **errp)
2584 {
2585     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2586         pc_memory_pre_plug(hotplug_dev, dev, errp);
2587     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2588         pc_cpu_pre_plug(hotplug_dev, dev, errp);
2589     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI)) {
2590         pc_virtio_pmem_pci_pre_plug(hotplug_dev, dev, errp);
2591     }
2592 }
2593 
2594 static void pc_machine_device_plug_cb(HotplugHandler *hotplug_dev,
2595                                       DeviceState *dev, Error **errp)
2596 {
2597     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2598         pc_memory_plug(hotplug_dev, dev, errp);
2599     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2600         pc_cpu_plug(hotplug_dev, dev, errp);
2601     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI)) {
2602         pc_virtio_pmem_pci_plug(hotplug_dev, dev, errp);
2603     }
2604 }
2605 
2606 static void pc_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
2607                                                 DeviceState *dev, Error **errp)
2608 {
2609     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2610         pc_memory_unplug_request(hotplug_dev, dev, errp);
2611     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2612         pc_cpu_unplug_request_cb(hotplug_dev, dev, errp);
2613     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI)) {
2614         pc_virtio_pmem_pci_unplug_request(hotplug_dev, dev, errp);
2615     } else {
2616         error_setg(errp, "acpi: device unplug request for not supported device"
2617                    " type: %s", object_get_typename(OBJECT(dev)));
2618     }
2619 }
2620 
2621 static void pc_machine_device_unplug_cb(HotplugHandler *hotplug_dev,
2622                                         DeviceState *dev, Error **errp)
2623 {
2624     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2625         pc_memory_unplug(hotplug_dev, dev, errp);
2626     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2627         pc_cpu_unplug_cb(hotplug_dev, dev, errp);
2628     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI)) {
2629         pc_virtio_pmem_pci_unplug(hotplug_dev, dev, errp);
2630     } else {
2631         error_setg(errp, "acpi: device unplug for not supported device"
2632                    " type: %s", object_get_typename(OBJECT(dev)));
2633     }
2634 }
2635 
2636 static HotplugHandler *pc_get_hotplug_handler(MachineState *machine,
2637                                              DeviceState *dev)
2638 {
2639     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) ||
2640         object_dynamic_cast(OBJECT(dev), TYPE_CPU) ||
2641         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI)) {
2642         return HOTPLUG_HANDLER(machine);
2643     }
2644 
2645     return NULL;
2646 }
2647 
2648 static void
2649 pc_machine_get_device_memory_region_size(Object *obj, Visitor *v,
2650                                          const char *name, void *opaque,
2651                                          Error **errp)
2652 {
2653     MachineState *ms = MACHINE(obj);
2654     int64_t value = 0;
2655 
2656     if (ms->device_memory) {
2657         value = memory_region_size(&ms->device_memory->mr);
2658     }
2659 
2660     visit_type_int(v, name, &value, errp);
2661 }
2662 
2663 static void pc_machine_get_max_ram_below_4g(Object *obj, Visitor *v,
2664                                             const char *name, void *opaque,
2665                                             Error **errp)
2666 {
2667     PCMachineState *pcms = PC_MACHINE(obj);
2668     uint64_t value = pcms->max_ram_below_4g;
2669 
2670     visit_type_size(v, name, &value, errp);
2671 }
2672 
2673 static void pc_machine_set_max_ram_below_4g(Object *obj, Visitor *v,
2674                                             const char *name, void *opaque,
2675                                             Error **errp)
2676 {
2677     PCMachineState *pcms = PC_MACHINE(obj);
2678     Error *error = NULL;
2679     uint64_t value;
2680 
2681     visit_type_size(v, name, &value, &error);
2682     if (error) {
2683         error_propagate(errp, error);
2684         return;
2685     }
2686     if (value > 4 * GiB) {
2687         error_setg(&error,
2688                    "Machine option 'max-ram-below-4g=%"PRIu64
2689                    "' expects size less than or equal to 4G", value);
2690         error_propagate(errp, error);
2691         return;
2692     }
2693 
2694     if (value < 1 * MiB) {
2695         warn_report("Only %" PRIu64 " bytes of RAM below the 4GiB boundary,"
2696                     "BIOS may not work with less than 1MiB", value);
2697     }
2698 
2699     pcms->max_ram_below_4g = value;
2700 }
2701 
2702 static void pc_machine_get_vmport(Object *obj, Visitor *v, const char *name,
2703                                   void *opaque, Error **errp)
2704 {
2705     PCMachineState *pcms = PC_MACHINE(obj);
2706     OnOffAuto vmport = pcms->vmport;
2707 
2708     visit_type_OnOffAuto(v, name, &vmport, errp);
2709 }
2710 
2711 static void pc_machine_set_vmport(Object *obj, Visitor *v, const char *name,
2712                                   void *opaque, Error **errp)
2713 {
2714     PCMachineState *pcms = PC_MACHINE(obj);
2715 
2716     visit_type_OnOffAuto(v, name, &pcms->vmport, errp);
2717 }
2718 
2719 bool pc_machine_is_smm_enabled(PCMachineState *pcms)
2720 {
2721     bool smm_available = false;
2722 
2723     if (pcms->smm == ON_OFF_AUTO_OFF) {
2724         return false;
2725     }
2726 
2727     if (tcg_enabled() || qtest_enabled()) {
2728         smm_available = true;
2729     } else if (kvm_enabled()) {
2730         smm_available = kvm_has_smm();
2731     }
2732 
2733     if (smm_available) {
2734         return true;
2735     }
2736 
2737     if (pcms->smm == ON_OFF_AUTO_ON) {
2738         error_report("System Management Mode not supported by this hypervisor.");
2739         exit(1);
2740     }
2741     return false;
2742 }
2743 
2744 static void pc_machine_get_smm(Object *obj, Visitor *v, const char *name,
2745                                void *opaque, Error **errp)
2746 {
2747     PCMachineState *pcms = PC_MACHINE(obj);
2748     OnOffAuto smm = pcms->smm;
2749 
2750     visit_type_OnOffAuto(v, name, &smm, errp);
2751 }
2752 
2753 static void pc_machine_set_smm(Object *obj, Visitor *v, const char *name,
2754                                void *opaque, Error **errp)
2755 {
2756     PCMachineState *pcms = PC_MACHINE(obj);
2757 
2758     visit_type_OnOffAuto(v, name, &pcms->smm, errp);
2759 }
2760 
2761 static bool pc_machine_get_smbus(Object *obj, Error **errp)
2762 {
2763     PCMachineState *pcms = PC_MACHINE(obj);
2764 
2765     return pcms->smbus_enabled;
2766 }
2767 
2768 static void pc_machine_set_smbus(Object *obj, bool value, Error **errp)
2769 {
2770     PCMachineState *pcms = PC_MACHINE(obj);
2771 
2772     pcms->smbus_enabled = value;
2773 }
2774 
2775 static bool pc_machine_get_sata(Object *obj, Error **errp)
2776 {
2777     PCMachineState *pcms = PC_MACHINE(obj);
2778 
2779     return pcms->sata_enabled;
2780 }
2781 
2782 static void pc_machine_set_sata(Object *obj, bool value, Error **errp)
2783 {
2784     PCMachineState *pcms = PC_MACHINE(obj);
2785 
2786     pcms->sata_enabled = value;
2787 }
2788 
2789 static bool pc_machine_get_pit(Object *obj, Error **errp)
2790 {
2791     PCMachineState *pcms = PC_MACHINE(obj);
2792 
2793     return pcms->pit_enabled;
2794 }
2795 
2796 static void pc_machine_set_pit(Object *obj, bool value, Error **errp)
2797 {
2798     PCMachineState *pcms = PC_MACHINE(obj);
2799 
2800     pcms->pit_enabled = value;
2801 }
2802 
2803 static void pc_machine_initfn(Object *obj)
2804 {
2805     PCMachineState *pcms = PC_MACHINE(obj);
2806 
2807     pcms->max_ram_below_4g = 0; /* use default */
2808     pcms->smm = ON_OFF_AUTO_AUTO;
2809 #ifdef CONFIG_VMPORT
2810     pcms->vmport = ON_OFF_AUTO_AUTO;
2811 #else
2812     pcms->vmport = ON_OFF_AUTO_OFF;
2813 #endif /* CONFIG_VMPORT */
2814     /* acpi build is enabled by default if machine supports it */
2815     pcms->acpi_build_enabled = PC_MACHINE_GET_CLASS(pcms)->has_acpi_build;
2816     pcms->smbus_enabled = true;
2817     pcms->sata_enabled = true;
2818     pcms->pit_enabled = true;
2819     pcms->smp_dies = 1;
2820 
2821     pc_system_flash_create(pcms);
2822 }
2823 
2824 static void pc_machine_reset(MachineState *machine)
2825 {
2826     CPUState *cs;
2827     X86CPU *cpu;
2828 
2829     qemu_devices_reset();
2830 
2831     /* Reset APIC after devices have been reset to cancel
2832      * any changes that qemu_devices_reset() might have done.
2833      */
2834     CPU_FOREACH(cs) {
2835         cpu = X86_CPU(cs);
2836 
2837         if (cpu->apic_state) {
2838             device_reset(cpu->apic_state);
2839         }
2840     }
2841 }
2842 
2843 static CpuInstanceProperties
2844 pc_cpu_index_to_props(MachineState *ms, unsigned cpu_index)
2845 {
2846     MachineClass *mc = MACHINE_GET_CLASS(ms);
2847     const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(ms);
2848 
2849     assert(cpu_index < possible_cpus->len);
2850     return possible_cpus->cpus[cpu_index].props;
2851 }
2852 
2853 static int64_t pc_get_default_cpu_node_id(const MachineState *ms, int idx)
2854 {
2855    X86CPUTopoInfo topo;
2856    PCMachineState *pcms = PC_MACHINE(ms);
2857 
2858    assert(idx < ms->possible_cpus->len);
2859    x86_topo_ids_from_apicid(ms->possible_cpus->cpus[idx].arch_id,
2860                             pcms->smp_dies, ms->smp.cores,
2861                             ms->smp.threads, &topo);
2862    return topo.pkg_id % nb_numa_nodes;
2863 }
2864 
2865 static const CPUArchIdList *pc_possible_cpu_arch_ids(MachineState *ms)
2866 {
2867     PCMachineState *pcms = PC_MACHINE(ms);
2868     int i;
2869     unsigned int max_cpus = ms->smp.max_cpus;
2870 
2871     if (ms->possible_cpus) {
2872         /*
2873          * make sure that max_cpus hasn't changed since the first use, i.e.
2874          * -smp hasn't been parsed after it
2875         */
2876         assert(ms->possible_cpus->len == max_cpus);
2877         return ms->possible_cpus;
2878     }
2879 
2880     ms->possible_cpus = g_malloc0(sizeof(CPUArchIdList) +
2881                                   sizeof(CPUArchId) * max_cpus);
2882     ms->possible_cpus->len = max_cpus;
2883     for (i = 0; i < ms->possible_cpus->len; i++) {
2884         X86CPUTopoInfo topo;
2885 
2886         ms->possible_cpus->cpus[i].type = ms->cpu_type;
2887         ms->possible_cpus->cpus[i].vcpus_count = 1;
2888         ms->possible_cpus->cpus[i].arch_id = x86_cpu_apic_id_from_index(pcms, i);
2889         x86_topo_ids_from_apicid(ms->possible_cpus->cpus[i].arch_id,
2890                                  pcms->smp_dies, ms->smp.cores,
2891                                  ms->smp.threads, &topo);
2892         ms->possible_cpus->cpus[i].props.has_socket_id = true;
2893         ms->possible_cpus->cpus[i].props.socket_id = topo.pkg_id;
2894         ms->possible_cpus->cpus[i].props.has_die_id = true;
2895         ms->possible_cpus->cpus[i].props.die_id = topo.die_id;
2896         ms->possible_cpus->cpus[i].props.has_core_id = true;
2897         ms->possible_cpus->cpus[i].props.core_id = topo.core_id;
2898         ms->possible_cpus->cpus[i].props.has_thread_id = true;
2899         ms->possible_cpus->cpus[i].props.thread_id = topo.smt_id;
2900     }
2901     return ms->possible_cpus;
2902 }
2903 
2904 static void x86_nmi(NMIState *n, int cpu_index, Error **errp)
2905 {
2906     /* cpu index isn't used */
2907     CPUState *cs;
2908 
2909     CPU_FOREACH(cs) {
2910         X86CPU *cpu = X86_CPU(cs);
2911 
2912         if (!cpu->apic_state) {
2913             cpu_interrupt(cs, CPU_INTERRUPT_NMI);
2914         } else {
2915             apic_deliver_nmi(cpu->apic_state);
2916         }
2917     }
2918 }
2919 
2920 static void pc_machine_class_init(ObjectClass *oc, void *data)
2921 {
2922     MachineClass *mc = MACHINE_CLASS(oc);
2923     PCMachineClass *pcmc = PC_MACHINE_CLASS(oc);
2924     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
2925     NMIClass *nc = NMI_CLASS(oc);
2926 
2927     pcmc->pci_enabled = true;
2928     pcmc->has_acpi_build = true;
2929     pcmc->rsdp_in_ram = true;
2930     pcmc->smbios_defaults = true;
2931     pcmc->smbios_uuid_encoded = true;
2932     pcmc->gigabyte_align = true;
2933     pcmc->has_reserved_memory = true;
2934     pcmc->kvmclock_enabled = true;
2935     pcmc->enforce_aligned_dimm = true;
2936     /* BIOS ACPI tables: 128K. Other BIOS datastructures: less than 4K reported
2937      * to be used at the moment, 32K should be enough for a while.  */
2938     pcmc->acpi_data_size = 0x20000 + 0x8000;
2939     pcmc->save_tsc_khz = true;
2940     pcmc->linuxboot_dma_enabled = true;
2941     pcmc->pvh_enabled = true;
2942     assert(!mc->get_hotplug_handler);
2943     mc->get_hotplug_handler = pc_get_hotplug_handler;
2944     mc->cpu_index_to_instance_props = pc_cpu_index_to_props;
2945     mc->get_default_cpu_node_id = pc_get_default_cpu_node_id;
2946     mc->possible_cpu_arch_ids = pc_possible_cpu_arch_ids;
2947     mc->auto_enable_numa_with_memhp = true;
2948     mc->has_hotpluggable_cpus = true;
2949     mc->default_boot_order = "cad";
2950     mc->hot_add_cpu = pc_hot_add_cpu;
2951     mc->smp_parse = pc_smp_parse;
2952     mc->block_default_type = IF_IDE;
2953     mc->max_cpus = 255;
2954     mc->reset = pc_machine_reset;
2955     hc->pre_plug = pc_machine_device_pre_plug_cb;
2956     hc->plug = pc_machine_device_plug_cb;
2957     hc->unplug_request = pc_machine_device_unplug_request_cb;
2958     hc->unplug = pc_machine_device_unplug_cb;
2959     nc->nmi_monitor_handler = x86_nmi;
2960     mc->default_cpu_type = TARGET_DEFAULT_CPU_TYPE;
2961     mc->nvdimm_supported = true;
2962     mc->numa_mem_supported = true;
2963 
2964     object_class_property_add(oc, PC_MACHINE_DEVMEM_REGION_SIZE, "int",
2965         pc_machine_get_device_memory_region_size, NULL,
2966         NULL, NULL, &error_abort);
2967 
2968     object_class_property_add(oc, PC_MACHINE_MAX_RAM_BELOW_4G, "size",
2969         pc_machine_get_max_ram_below_4g, pc_machine_set_max_ram_below_4g,
2970         NULL, NULL, &error_abort);
2971 
2972     object_class_property_set_description(oc, PC_MACHINE_MAX_RAM_BELOW_4G,
2973         "Maximum ram below the 4G boundary (32bit boundary)", &error_abort);
2974 
2975     object_class_property_add(oc, PC_MACHINE_SMM, "OnOffAuto",
2976         pc_machine_get_smm, pc_machine_set_smm,
2977         NULL, NULL, &error_abort);
2978     object_class_property_set_description(oc, PC_MACHINE_SMM,
2979         "Enable SMM (pc & q35)", &error_abort);
2980 
2981     object_class_property_add(oc, PC_MACHINE_VMPORT, "OnOffAuto",
2982         pc_machine_get_vmport, pc_machine_set_vmport,
2983         NULL, NULL, &error_abort);
2984     object_class_property_set_description(oc, PC_MACHINE_VMPORT,
2985         "Enable vmport (pc & q35)", &error_abort);
2986 
2987     object_class_property_add_bool(oc, PC_MACHINE_SMBUS,
2988         pc_machine_get_smbus, pc_machine_set_smbus, &error_abort);
2989 
2990     object_class_property_add_bool(oc, PC_MACHINE_SATA,
2991         pc_machine_get_sata, pc_machine_set_sata, &error_abort);
2992 
2993     object_class_property_add_bool(oc, PC_MACHINE_PIT,
2994         pc_machine_get_pit, pc_machine_set_pit, &error_abort);
2995 }
2996 
2997 static const TypeInfo pc_machine_info = {
2998     .name = TYPE_PC_MACHINE,
2999     .parent = TYPE_MACHINE,
3000     .abstract = true,
3001     .instance_size = sizeof(PCMachineState),
3002     .instance_init = pc_machine_initfn,
3003     .class_size = sizeof(PCMachineClass),
3004     .class_init = pc_machine_class_init,
3005     .interfaces = (InterfaceInfo[]) {
3006          { TYPE_HOTPLUG_HANDLER },
3007          { TYPE_NMI },
3008          { }
3009     },
3010 };
3011 
3012 static void pc_machine_register_types(void)
3013 {
3014     type_register_static(&pc_machine_info);
3015 }
3016 
3017 type_init(pc_machine_register_types)
3018