xref: /openbmc/qemu/hw/i386/pc.c (revision 08528271)
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/hw.h"
28 #include "hw/i386/pc.h"
29 #include "hw/char/serial.h"
30 #include "hw/char/parallel.h"
31 #include "hw/i386/apic.h"
32 #include "hw/i386/topology.h"
33 #include "hw/i386/fw_cfg.h"
34 #include "sysemu/cpus.h"
35 #include "hw/block/fdc.h"
36 #include "hw/ide.h"
37 #include "hw/pci/pci.h"
38 #include "hw/pci/pci_bus.h"
39 #include "hw/nvram/fw_cfg.h"
40 #include "hw/timer/hpet.h"
41 #include "hw/firmware/smbios.h"
42 #include "hw/loader.h"
43 #include "elf.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/audio/pcspk.h"
50 #include "hw/pci/msi.h"
51 #include "hw/sysbus.h"
52 #include "sysemu/sysemu.h"
53 #include "sysemu/tcg.h"
54 #include "sysemu/numa.h"
55 #include "sysemu/kvm.h"
56 #include "sysemu/qtest.h"
57 #include "kvm_i386.h"
58 #include "hw/xen/xen.h"
59 #include "hw/xen/start_info.h"
60 #include "ui/qemu-spice.h"
61 #include "exec/memory.h"
62 #include "exec/address-spaces.h"
63 #include "sysemu/arch_init.h"
64 #include "qemu/bitmap.h"
65 #include "qemu/config-file.h"
66 #include "qemu/error-report.h"
67 #include "qemu/option.h"
68 #include "hw/acpi/acpi.h"
69 #include "hw/acpi/cpu_hotplug.h"
70 #include "hw/boards.h"
71 #include "acpi-build.h"
72 #include "hw/mem/pc-dimm.h"
73 #include "qapi/error.h"
74 #include "qapi/qapi-visit-common.h"
75 #include "qapi/visitor.h"
76 #include "qom/cpu.h"
77 #include "hw/nmi.h"
78 #include "hw/usb.h"
79 #include "hw/i386/intel_iommu.h"
80 #include "hw/net/ne2000-isa.h"
81 #include "standard-headers/asm-x86/bootparam.h"
82 
83 /* debug PC/ISA interrupts */
84 //#define DEBUG_IRQ
85 
86 #ifdef DEBUG_IRQ
87 #define DPRINTF(fmt, ...)                                       \
88     do { printf("CPUIRQ: " fmt , ## __VA_ARGS__); } while (0)
89 #else
90 #define DPRINTF(fmt, ...)
91 #endif
92 
93 #define E820_NR_ENTRIES		16
94 
95 struct e820_entry {
96     uint64_t address;
97     uint64_t length;
98     uint32_t type;
99 } QEMU_PACKED __attribute((__aligned__(4)));
100 
101 struct e820_table {
102     uint32_t count;
103     struct e820_entry entry[E820_NR_ENTRIES];
104 } QEMU_PACKED __attribute((__aligned__(4)));
105 
106 static struct e820_table e820_reserve;
107 static struct e820_entry *e820_table;
108 static unsigned e820_entries;
109 struct hpet_fw_config hpet_cfg = {.count = UINT8_MAX};
110 
111 /* Physical Address of PVH entry point read from kernel ELF NOTE */
112 static size_t pvh_start_addr;
113 
114 GlobalProperty pc_compat_4_0[] = {};
115 const size_t pc_compat_4_0_len = G_N_ELEMENTS(pc_compat_4_0);
116 
117 GlobalProperty pc_compat_3_1[] = {
118     { "intel-iommu", "dma-drain", "off" },
119     { "Opteron_G3" "-" TYPE_X86_CPU, "rdtscp", "off" },
120     { "Opteron_G4" "-" TYPE_X86_CPU, "rdtscp", "off" },
121     { "Opteron_G4" "-" TYPE_X86_CPU, "npt", "off" },
122     { "Opteron_G4" "-" TYPE_X86_CPU, "nrip-save", "off" },
123     { "Opteron_G5" "-" TYPE_X86_CPU, "rdtscp", "off" },
124     { "Opteron_G5" "-" TYPE_X86_CPU, "npt", "off" },
125     { "Opteron_G5" "-" TYPE_X86_CPU, "nrip-save", "off" },
126     { "EPYC" "-" TYPE_X86_CPU, "npt", "off" },
127     { "EPYC" "-" TYPE_X86_CPU, "nrip-save", "off" },
128     { "EPYC-IBPB" "-" TYPE_X86_CPU, "npt", "off" },
129     { "EPYC-IBPB" "-" TYPE_X86_CPU, "nrip-save", "off" },
130     { "Skylake-Client" "-" TYPE_X86_CPU,      "mpx", "on" },
131     { "Skylake-Client-IBRS" "-" TYPE_X86_CPU, "mpx", "on" },
132     { "Skylake-Server" "-" TYPE_X86_CPU,      "mpx", "on" },
133     { "Skylake-Server-IBRS" "-" TYPE_X86_CPU, "mpx", "on" },
134     { "Cascadelake-Server" "-" TYPE_X86_CPU,  "mpx", "on" },
135     { "Icelake-Client" "-" TYPE_X86_CPU,      "mpx", "on" },
136     { "Icelake-Server" "-" TYPE_X86_CPU,      "mpx", "on" },
137     { "Cascadelake-Server" "-" TYPE_X86_CPU, "stepping", "5" },
138     { TYPE_X86_CPU, "x-intel-pt-auto-level", "off" },
139 };
140 const size_t pc_compat_3_1_len = G_N_ELEMENTS(pc_compat_3_1);
141 
142 GlobalProperty pc_compat_3_0[] = {
143     { TYPE_X86_CPU, "x-hv-synic-kvm-only", "on" },
144     { "Skylake-Server" "-" TYPE_X86_CPU, "pku", "off" },
145     { "Skylake-Server-IBRS" "-" TYPE_X86_CPU, "pku", "off" },
146 };
147 const size_t pc_compat_3_0_len = G_N_ELEMENTS(pc_compat_3_0);
148 
149 GlobalProperty pc_compat_2_12[] = {
150     { TYPE_X86_CPU, "legacy-cache", "on" },
151     { TYPE_X86_CPU, "topoext", "off" },
152     { "EPYC-" TYPE_X86_CPU, "xlevel", "0x8000000a" },
153     { "EPYC-IBPB-" TYPE_X86_CPU, "xlevel", "0x8000000a" },
154 };
155 const size_t pc_compat_2_12_len = G_N_ELEMENTS(pc_compat_2_12);
156 
157 GlobalProperty pc_compat_2_11[] = {
158     { TYPE_X86_CPU, "x-migrate-smi-count", "off" },
159     { "Skylake-Server" "-" TYPE_X86_CPU, "clflushopt", "off" },
160 };
161 const size_t pc_compat_2_11_len = G_N_ELEMENTS(pc_compat_2_11);
162 
163 GlobalProperty pc_compat_2_10[] = {
164     { TYPE_X86_CPU, "x-hv-max-vps", "0x40" },
165     { "i440FX-pcihost", "x-pci-hole64-fix", "off" },
166     { "q35-pcihost", "x-pci-hole64-fix", "off" },
167 };
168 const size_t pc_compat_2_10_len = G_N_ELEMENTS(pc_compat_2_10);
169 
170 GlobalProperty pc_compat_2_9[] = {
171     { "mch", "extended-tseg-mbytes", "0" },
172 };
173 const size_t pc_compat_2_9_len = G_N_ELEMENTS(pc_compat_2_9);
174 
175 GlobalProperty pc_compat_2_8[] = {
176     { TYPE_X86_CPU, "tcg-cpuid", "off" },
177     { "kvmclock", "x-mach-use-reliable-get-clock", "off" },
178     { "ICH9-LPC", "x-smi-broadcast", "off" },
179     { TYPE_X86_CPU, "vmware-cpuid-freq", "off" },
180     { "Haswell-" TYPE_X86_CPU, "stepping", "1" },
181 };
182 const size_t pc_compat_2_8_len = G_N_ELEMENTS(pc_compat_2_8);
183 
184 GlobalProperty pc_compat_2_7[] = {
185     { TYPE_X86_CPU, "l3-cache", "off" },
186     { TYPE_X86_CPU, "full-cpuid-auto-level", "off" },
187     { "Opteron_G3" "-" TYPE_X86_CPU, "family", "15" },
188     { "Opteron_G3" "-" TYPE_X86_CPU, "model", "6" },
189     { "Opteron_G3" "-" TYPE_X86_CPU, "stepping", "1" },
190     { "isa-pcspk", "migrate", "off" },
191 };
192 const size_t pc_compat_2_7_len = G_N_ELEMENTS(pc_compat_2_7);
193 
194 GlobalProperty pc_compat_2_6[] = {
195     { TYPE_X86_CPU, "cpuid-0xb", "off" },
196     { "vmxnet3", "romfile", "" },
197     { TYPE_X86_CPU, "fill-mtrr-mask", "off" },
198     { "apic-common", "legacy-instance-id", "on", }
199 };
200 const size_t pc_compat_2_6_len = G_N_ELEMENTS(pc_compat_2_6);
201 
202 GlobalProperty pc_compat_2_5[] = {};
203 const size_t pc_compat_2_5_len = G_N_ELEMENTS(pc_compat_2_5);
204 
205 GlobalProperty pc_compat_2_4[] = {
206     PC_CPU_MODEL_IDS("2.4.0")
207     { "Haswell-" TYPE_X86_CPU, "abm", "off" },
208     { "Haswell-noTSX-" TYPE_X86_CPU, "abm", "off" },
209     { "Broadwell-" TYPE_X86_CPU, "abm", "off" },
210     { "Broadwell-noTSX-" TYPE_X86_CPU, "abm", "off" },
211     { "host" "-" TYPE_X86_CPU, "host-cache-info", "on" },
212     { TYPE_X86_CPU, "check", "off" },
213     { "qemu64" "-" TYPE_X86_CPU, "sse4a", "on" },
214     { "qemu64" "-" TYPE_X86_CPU, "abm", "on" },
215     { "qemu64" "-" TYPE_X86_CPU, "popcnt", "on" },
216     { "qemu32" "-" TYPE_X86_CPU, "popcnt", "on" },
217     { "Opteron_G2" "-" TYPE_X86_CPU, "rdtscp", "on" },
218     { "Opteron_G3" "-" TYPE_X86_CPU, "rdtscp", "on" },
219     { "Opteron_G4" "-" TYPE_X86_CPU, "rdtscp", "on" },
220     { "Opteron_G5" "-" TYPE_X86_CPU, "rdtscp", "on", }
221 };
222 const size_t pc_compat_2_4_len = G_N_ELEMENTS(pc_compat_2_4);
223 
224 GlobalProperty pc_compat_2_3[] = {
225     PC_CPU_MODEL_IDS("2.3.0")
226     { TYPE_X86_CPU, "arat", "off" },
227     { "qemu64" "-" TYPE_X86_CPU, "min-level", "4" },
228     { "kvm64" "-" TYPE_X86_CPU, "min-level", "5" },
229     { "pentium3" "-" TYPE_X86_CPU, "min-level", "2" },
230     { "n270" "-" TYPE_X86_CPU, "min-level", "5" },
231     { "Conroe" "-" TYPE_X86_CPU, "min-level", "4" },
232     { "Penryn" "-" TYPE_X86_CPU, "min-level", "4" },
233     { "Nehalem" "-" TYPE_X86_CPU, "min-level", "4" },
234     { "n270" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
235     { "Penryn" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
236     { "Conroe" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
237     { "Nehalem" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
238     { "Westmere" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
239     { "SandyBridge" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
240     { "IvyBridge" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
241     { "Haswell" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
242     { "Haswell-noTSX" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
243     { "Broadwell" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
244     { "Broadwell-noTSX" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
245     { TYPE_X86_CPU, "kvm-no-smi-migration", "on" },
246 };
247 const size_t pc_compat_2_3_len = G_N_ELEMENTS(pc_compat_2_3);
248 
249 GlobalProperty pc_compat_2_2[] = {
250     PC_CPU_MODEL_IDS("2.2.0")
251     { "kvm64" "-" TYPE_X86_CPU, "vme", "off" },
252     { "kvm32" "-" TYPE_X86_CPU, "vme", "off" },
253     { "Conroe" "-" TYPE_X86_CPU, "vme", "off" },
254     { "Penryn" "-" TYPE_X86_CPU, "vme", "off" },
255     { "Nehalem" "-" TYPE_X86_CPU, "vme", "off" },
256     { "Westmere" "-" TYPE_X86_CPU, "vme", "off" },
257     { "SandyBridge" "-" TYPE_X86_CPU, "vme", "off" },
258     { "Haswell" "-" TYPE_X86_CPU, "vme", "off" },
259     { "Broadwell" "-" TYPE_X86_CPU, "vme", "off" },
260     { "Opteron_G1" "-" TYPE_X86_CPU, "vme", "off" },
261     { "Opteron_G2" "-" TYPE_X86_CPU, "vme", "off" },
262     { "Opteron_G3" "-" TYPE_X86_CPU, "vme", "off" },
263     { "Opteron_G4" "-" TYPE_X86_CPU, "vme", "off" },
264     { "Opteron_G5" "-" TYPE_X86_CPU, "vme", "off" },
265     { "Haswell" "-" TYPE_X86_CPU, "f16c", "off" },
266     { "Haswell" "-" TYPE_X86_CPU, "rdrand", "off" },
267     { "Broadwell" "-" TYPE_X86_CPU, "f16c", "off" },
268     { "Broadwell" "-" TYPE_X86_CPU, "rdrand", "off" },
269 };
270 const size_t pc_compat_2_2_len = G_N_ELEMENTS(pc_compat_2_2);
271 
272 GlobalProperty pc_compat_2_1[] = {
273     PC_CPU_MODEL_IDS("2.1.0")
274     { "coreduo" "-" TYPE_X86_CPU, "vmx", "on" },
275     { "core2duo" "-" TYPE_X86_CPU, "vmx", "on" },
276 };
277 const size_t pc_compat_2_1_len = G_N_ELEMENTS(pc_compat_2_1);
278 
279 GlobalProperty pc_compat_2_0[] = {
280     PC_CPU_MODEL_IDS("2.0.0")
281     { "virtio-scsi-pci", "any_layout", "off" },
282     { "PIIX4_PM", "memory-hotplug-support", "off" },
283     { "apic", "version", "0x11" },
284     { "nec-usb-xhci", "superspeed-ports-first", "off" },
285     { "nec-usb-xhci", "force-pcie-endcap", "on" },
286     { "pci-serial", "prog_if", "0" },
287     { "pci-serial-2x", "prog_if", "0" },
288     { "pci-serial-4x", "prog_if", "0" },
289     { "virtio-net-pci", "guest_announce", "off" },
290     { "ICH9-LPC", "memory-hotplug-support", "off" },
291     { "xio3130-downstream", COMPAT_PROP_PCP, "off" },
292     { "ioh3420", COMPAT_PROP_PCP, "off" },
293 };
294 const size_t pc_compat_2_0_len = G_N_ELEMENTS(pc_compat_2_0);
295 
296 GlobalProperty pc_compat_1_7[] = {
297     PC_CPU_MODEL_IDS("1.7.0")
298     { TYPE_USB_DEVICE, "msos-desc", "no" },
299     { "PIIX4_PM", "acpi-pci-hotplug-with-bridge-support", "off" },
300     { "hpet", HPET_INTCAP, "4" },
301 };
302 const size_t pc_compat_1_7_len = G_N_ELEMENTS(pc_compat_1_7);
303 
304 GlobalProperty pc_compat_1_6[] = {
305     PC_CPU_MODEL_IDS("1.6.0")
306     { "e1000", "mitigation", "off" },
307     { "qemu64-" TYPE_X86_CPU, "model", "2" },
308     { "qemu32-" TYPE_X86_CPU, "model", "3" },
309     { "i440FX-pcihost", "short_root_bus", "1" },
310     { "q35-pcihost", "short_root_bus", "1" },
311 };
312 const size_t pc_compat_1_6_len = G_N_ELEMENTS(pc_compat_1_6);
313 
314 GlobalProperty pc_compat_1_5[] = {
315     PC_CPU_MODEL_IDS("1.5.0")
316     { "Conroe-" TYPE_X86_CPU, "model", "2" },
317     { "Conroe-" TYPE_X86_CPU, "min-level", "2" },
318     { "Penryn-" TYPE_X86_CPU, "model", "2" },
319     { "Penryn-" TYPE_X86_CPU, "min-level", "2" },
320     { "Nehalem-" TYPE_X86_CPU, "model", "2" },
321     { "Nehalem-" TYPE_X86_CPU, "min-level", "2" },
322     { "virtio-net-pci", "any_layout", "off" },
323     { TYPE_X86_CPU, "pmu", "on" },
324     { "i440FX-pcihost", "short_root_bus", "0" },
325     { "q35-pcihost", "short_root_bus", "0" },
326 };
327 const size_t pc_compat_1_5_len = G_N_ELEMENTS(pc_compat_1_5);
328 
329 GlobalProperty pc_compat_1_4[] = {
330     PC_CPU_MODEL_IDS("1.4.0")
331     { "scsi-hd", "discard_granularity", "0" },
332     { "scsi-cd", "discard_granularity", "0" },
333     { "scsi-disk", "discard_granularity", "0" },
334     { "ide-hd", "discard_granularity", "0" },
335     { "ide-cd", "discard_granularity", "0" },
336     { "ide-drive", "discard_granularity", "0" },
337     { "virtio-blk-pci", "discard_granularity", "0" },
338     /* DEV_NVECTORS_UNSPECIFIED as a uint32_t string: */
339     { "virtio-serial-pci", "vectors", "0xFFFFFFFF" },
340     { "virtio-net-pci", "ctrl_guest_offloads", "off" },
341     { "e1000", "romfile", "pxe-e1000.rom" },
342     { "ne2k_pci", "romfile", "pxe-ne2k_pci.rom" },
343     { "pcnet", "romfile", "pxe-pcnet.rom" },
344     { "rtl8139", "romfile", "pxe-rtl8139.rom" },
345     { "virtio-net-pci", "romfile", "pxe-virtio.rom" },
346     { "486-" TYPE_X86_CPU, "model", "0" },
347     { "n270" "-" TYPE_X86_CPU, "movbe", "off" },
348     { "Westmere" "-" TYPE_X86_CPU, "pclmulqdq", "off" },
349 };
350 const size_t pc_compat_1_4_len = G_N_ELEMENTS(pc_compat_1_4);
351 
352 void gsi_handler(void *opaque, int n, int level)
353 {
354     GSIState *s = opaque;
355 
356     DPRINTF("pc: %s GSI %d\n", level ? "raising" : "lowering", n);
357     if (n < ISA_NUM_IRQS) {
358         qemu_set_irq(s->i8259_irq[n], level);
359     }
360     qemu_set_irq(s->ioapic_irq[n], level);
361 }
362 
363 static void ioport80_write(void *opaque, hwaddr addr, uint64_t data,
364                            unsigned size)
365 {
366 }
367 
368 static uint64_t ioport80_read(void *opaque, hwaddr addr, unsigned size)
369 {
370     return 0xffffffffffffffffULL;
371 }
372 
373 /* MSDOS compatibility mode FPU exception support */
374 static qemu_irq ferr_irq;
375 
376 void pc_register_ferr_irq(qemu_irq irq)
377 {
378     ferr_irq = irq;
379 }
380 
381 /* XXX: add IGNNE support */
382 void cpu_set_ferr(CPUX86State *s)
383 {
384     qemu_irq_raise(ferr_irq);
385 }
386 
387 static void ioportF0_write(void *opaque, hwaddr addr, uint64_t data,
388                            unsigned size)
389 {
390     qemu_irq_lower(ferr_irq);
391 }
392 
393 static uint64_t ioportF0_read(void *opaque, hwaddr addr, unsigned size)
394 {
395     return 0xffffffffffffffffULL;
396 }
397 
398 /* TSC handling */
399 uint64_t cpu_get_tsc(CPUX86State *env)
400 {
401     return cpu_get_ticks();
402 }
403 
404 /* IRQ handling */
405 int cpu_get_pic_interrupt(CPUX86State *env)
406 {
407     X86CPU *cpu = env_archcpu(env);
408     int intno;
409 
410     if (!kvm_irqchip_in_kernel()) {
411         intno = apic_get_interrupt(cpu->apic_state);
412         if (intno >= 0) {
413             return intno;
414         }
415         /* read the irq from the PIC */
416         if (!apic_accept_pic_intr(cpu->apic_state)) {
417             return -1;
418         }
419     }
420 
421     intno = pic_read_irq(isa_pic);
422     return intno;
423 }
424 
425 static void pic_irq_request(void *opaque, int irq, int level)
426 {
427     CPUState *cs = first_cpu;
428     X86CPU *cpu = X86_CPU(cs);
429 
430     DPRINTF("pic_irqs: %s irq %d\n", level? "raise" : "lower", irq);
431     if (cpu->apic_state && !kvm_irqchip_in_kernel()) {
432         CPU_FOREACH(cs) {
433             cpu = X86_CPU(cs);
434             if (apic_accept_pic_intr(cpu->apic_state)) {
435                 apic_deliver_pic_intr(cpu->apic_state, level);
436             }
437         }
438     } else {
439         if (level) {
440             cpu_interrupt(cs, CPU_INTERRUPT_HARD);
441         } else {
442             cpu_reset_interrupt(cs, CPU_INTERRUPT_HARD);
443         }
444     }
445 }
446 
447 /* PC cmos mappings */
448 
449 #define REG_EQUIPMENT_BYTE          0x14
450 
451 int cmos_get_fd_drive_type(FloppyDriveType fd0)
452 {
453     int val;
454 
455     switch (fd0) {
456     case FLOPPY_DRIVE_TYPE_144:
457         /* 1.44 Mb 3"5 drive */
458         val = 4;
459         break;
460     case FLOPPY_DRIVE_TYPE_288:
461         /* 2.88 Mb 3"5 drive */
462         val = 5;
463         break;
464     case FLOPPY_DRIVE_TYPE_120:
465         /* 1.2 Mb 5"5 drive */
466         val = 2;
467         break;
468     case FLOPPY_DRIVE_TYPE_NONE:
469     default:
470         val = 0;
471         break;
472     }
473     return val;
474 }
475 
476 static void cmos_init_hd(ISADevice *s, int type_ofs, int info_ofs,
477                          int16_t cylinders, int8_t heads, int8_t sectors)
478 {
479     rtc_set_memory(s, type_ofs, 47);
480     rtc_set_memory(s, info_ofs, cylinders);
481     rtc_set_memory(s, info_ofs + 1, cylinders >> 8);
482     rtc_set_memory(s, info_ofs + 2, heads);
483     rtc_set_memory(s, info_ofs + 3, 0xff);
484     rtc_set_memory(s, info_ofs + 4, 0xff);
485     rtc_set_memory(s, info_ofs + 5, 0xc0 | ((heads > 8) << 3));
486     rtc_set_memory(s, info_ofs + 6, cylinders);
487     rtc_set_memory(s, info_ofs + 7, cylinders >> 8);
488     rtc_set_memory(s, info_ofs + 8, sectors);
489 }
490 
491 /* convert boot_device letter to something recognizable by the bios */
492 static int boot_device2nibble(char boot_device)
493 {
494     switch(boot_device) {
495     case 'a':
496     case 'b':
497         return 0x01; /* floppy boot */
498     case 'c':
499         return 0x02; /* hard drive boot */
500     case 'd':
501         return 0x03; /* CD-ROM boot */
502     case 'n':
503         return 0x04; /* Network boot */
504     }
505     return 0;
506 }
507 
508 static void set_boot_dev(ISADevice *s, const char *boot_device, Error **errp)
509 {
510 #define PC_MAX_BOOT_DEVICES 3
511     int nbds, bds[3] = { 0, };
512     int i;
513 
514     nbds = strlen(boot_device);
515     if (nbds > PC_MAX_BOOT_DEVICES) {
516         error_setg(errp, "Too many boot devices for PC");
517         return;
518     }
519     for (i = 0; i < nbds; i++) {
520         bds[i] = boot_device2nibble(boot_device[i]);
521         if (bds[i] == 0) {
522             error_setg(errp, "Invalid boot device for PC: '%c'",
523                        boot_device[i]);
524             return;
525         }
526     }
527     rtc_set_memory(s, 0x3d, (bds[1] << 4) | bds[0]);
528     rtc_set_memory(s, 0x38, (bds[2] << 4) | (fd_bootchk ? 0x0 : 0x1));
529 }
530 
531 static void pc_boot_set(void *opaque, const char *boot_device, Error **errp)
532 {
533     set_boot_dev(opaque, boot_device, errp);
534 }
535 
536 static void pc_cmos_init_floppy(ISADevice *rtc_state, ISADevice *floppy)
537 {
538     int val, nb, i;
539     FloppyDriveType fd_type[2] = { FLOPPY_DRIVE_TYPE_NONE,
540                                    FLOPPY_DRIVE_TYPE_NONE };
541 
542     /* floppy type */
543     if (floppy) {
544         for (i = 0; i < 2; i++) {
545             fd_type[i] = isa_fdc_get_drive_type(floppy, i);
546         }
547     }
548     val = (cmos_get_fd_drive_type(fd_type[0]) << 4) |
549         cmos_get_fd_drive_type(fd_type[1]);
550     rtc_set_memory(rtc_state, 0x10, val);
551 
552     val = rtc_get_memory(rtc_state, REG_EQUIPMENT_BYTE);
553     nb = 0;
554     if (fd_type[0] != FLOPPY_DRIVE_TYPE_NONE) {
555         nb++;
556     }
557     if (fd_type[1] != FLOPPY_DRIVE_TYPE_NONE) {
558         nb++;
559     }
560     switch (nb) {
561     case 0:
562         break;
563     case 1:
564         val |= 0x01; /* 1 drive, ready for boot */
565         break;
566     case 2:
567         val |= 0x41; /* 2 drives, ready for boot */
568         break;
569     }
570     rtc_set_memory(rtc_state, REG_EQUIPMENT_BYTE, val);
571 }
572 
573 typedef struct pc_cmos_init_late_arg {
574     ISADevice *rtc_state;
575     BusState *idebus[2];
576 } pc_cmos_init_late_arg;
577 
578 typedef struct check_fdc_state {
579     ISADevice *floppy;
580     bool multiple;
581 } CheckFdcState;
582 
583 static int check_fdc(Object *obj, void *opaque)
584 {
585     CheckFdcState *state = opaque;
586     Object *fdc;
587     uint32_t iobase;
588     Error *local_err = NULL;
589 
590     fdc = object_dynamic_cast(obj, TYPE_ISA_FDC);
591     if (!fdc) {
592         return 0;
593     }
594 
595     iobase = object_property_get_uint(obj, "iobase", &local_err);
596     if (local_err || iobase != 0x3f0) {
597         error_free(local_err);
598         return 0;
599     }
600 
601     if (state->floppy) {
602         state->multiple = true;
603     } else {
604         state->floppy = ISA_DEVICE(obj);
605     }
606     return 0;
607 }
608 
609 static const char * const fdc_container_path[] = {
610     "/unattached", "/peripheral", "/peripheral-anon"
611 };
612 
613 /*
614  * Locate the FDC at IO address 0x3f0, in order to configure the CMOS registers
615  * and ACPI objects.
616  */
617 ISADevice *pc_find_fdc0(void)
618 {
619     int i;
620     Object *container;
621     CheckFdcState state = { 0 };
622 
623     for (i = 0; i < ARRAY_SIZE(fdc_container_path); i++) {
624         container = container_get(qdev_get_machine(), fdc_container_path[i]);
625         object_child_foreach(container, check_fdc, &state);
626     }
627 
628     if (state.multiple) {
629         warn_report("multiple floppy disk controllers with "
630                     "iobase=0x3f0 have been found");
631         error_printf("the one being picked for CMOS setup might not reflect "
632                      "your intent");
633     }
634 
635     return state.floppy;
636 }
637 
638 static void pc_cmos_init_late(void *opaque)
639 {
640     pc_cmos_init_late_arg *arg = opaque;
641     ISADevice *s = arg->rtc_state;
642     int16_t cylinders;
643     int8_t heads, sectors;
644     int val;
645     int i, trans;
646 
647     val = 0;
648     if (arg->idebus[0] && ide_get_geometry(arg->idebus[0], 0,
649                                            &cylinders, &heads, &sectors) >= 0) {
650         cmos_init_hd(s, 0x19, 0x1b, cylinders, heads, sectors);
651         val |= 0xf0;
652     }
653     if (arg->idebus[0] && ide_get_geometry(arg->idebus[0], 1,
654                                            &cylinders, &heads, &sectors) >= 0) {
655         cmos_init_hd(s, 0x1a, 0x24, cylinders, heads, sectors);
656         val |= 0x0f;
657     }
658     rtc_set_memory(s, 0x12, val);
659 
660     val = 0;
661     for (i = 0; i < 4; i++) {
662         /* NOTE: ide_get_geometry() returns the physical
663            geometry.  It is always such that: 1 <= sects <= 63, 1
664            <= heads <= 16, 1 <= cylinders <= 16383. The BIOS
665            geometry can be different if a translation is done. */
666         if (arg->idebus[i / 2] &&
667             ide_get_geometry(arg->idebus[i / 2], i % 2,
668                              &cylinders, &heads, &sectors) >= 0) {
669             trans = ide_get_bios_chs_trans(arg->idebus[i / 2], i % 2) - 1;
670             assert((trans & ~3) == 0);
671             val |= trans << (i * 2);
672         }
673     }
674     rtc_set_memory(s, 0x39, val);
675 
676     pc_cmos_init_floppy(s, pc_find_fdc0());
677 
678     qemu_unregister_reset(pc_cmos_init_late, opaque);
679 }
680 
681 void pc_cmos_init(PCMachineState *pcms,
682                   BusState *idebus0, BusState *idebus1,
683                   ISADevice *s)
684 {
685     int val;
686     static pc_cmos_init_late_arg arg;
687 
688     /* various important CMOS locations needed by PC/Bochs bios */
689 
690     /* memory size */
691     /* base memory (first MiB) */
692     val = MIN(pcms->below_4g_mem_size / KiB, 640);
693     rtc_set_memory(s, 0x15, val);
694     rtc_set_memory(s, 0x16, val >> 8);
695     /* extended memory (next 64MiB) */
696     if (pcms->below_4g_mem_size > 1 * MiB) {
697         val = (pcms->below_4g_mem_size - 1 * MiB) / KiB;
698     } else {
699         val = 0;
700     }
701     if (val > 65535)
702         val = 65535;
703     rtc_set_memory(s, 0x17, val);
704     rtc_set_memory(s, 0x18, val >> 8);
705     rtc_set_memory(s, 0x30, val);
706     rtc_set_memory(s, 0x31, val >> 8);
707     /* memory between 16MiB and 4GiB */
708     if (pcms->below_4g_mem_size > 16 * MiB) {
709         val = (pcms->below_4g_mem_size - 16 * MiB) / (64 * KiB);
710     } else {
711         val = 0;
712     }
713     if (val > 65535)
714         val = 65535;
715     rtc_set_memory(s, 0x34, val);
716     rtc_set_memory(s, 0x35, val >> 8);
717     /* memory above 4GiB */
718     val = pcms->above_4g_mem_size / 65536;
719     rtc_set_memory(s, 0x5b, val);
720     rtc_set_memory(s, 0x5c, val >> 8);
721     rtc_set_memory(s, 0x5d, val >> 16);
722 
723     object_property_add_link(OBJECT(pcms), "rtc_state",
724                              TYPE_ISA_DEVICE,
725                              (Object **)&pcms->rtc,
726                              object_property_allow_set_link,
727                              OBJ_PROP_LINK_STRONG, &error_abort);
728     object_property_set_link(OBJECT(pcms), OBJECT(s),
729                              "rtc_state", &error_abort);
730 
731     set_boot_dev(s, MACHINE(pcms)->boot_order, &error_fatal);
732 
733     val = 0;
734     val |= 0x02; /* FPU is there */
735     val |= 0x04; /* PS/2 mouse installed */
736     rtc_set_memory(s, REG_EQUIPMENT_BYTE, val);
737 
738     /* hard drives and FDC */
739     arg.rtc_state = s;
740     arg.idebus[0] = idebus0;
741     arg.idebus[1] = idebus1;
742     qemu_register_reset(pc_cmos_init_late, &arg);
743 }
744 
745 #define TYPE_PORT92 "port92"
746 #define PORT92(obj) OBJECT_CHECK(Port92State, (obj), TYPE_PORT92)
747 
748 /* port 92 stuff: could be split off */
749 typedef struct Port92State {
750     ISADevice parent_obj;
751 
752     MemoryRegion io;
753     uint8_t outport;
754     qemu_irq a20_out;
755 } Port92State;
756 
757 static void port92_write(void *opaque, hwaddr addr, uint64_t val,
758                          unsigned size)
759 {
760     Port92State *s = opaque;
761     int oldval = s->outport;
762 
763     DPRINTF("port92: write 0x%02" PRIx64 "\n", val);
764     s->outport = val;
765     qemu_set_irq(s->a20_out, (val >> 1) & 1);
766     if ((val & 1) && !(oldval & 1)) {
767         qemu_system_reset_request(SHUTDOWN_CAUSE_GUEST_RESET);
768     }
769 }
770 
771 static uint64_t port92_read(void *opaque, hwaddr addr,
772                             unsigned size)
773 {
774     Port92State *s = opaque;
775     uint32_t ret;
776 
777     ret = s->outport;
778     DPRINTF("port92: read 0x%02x\n", ret);
779     return ret;
780 }
781 
782 static void port92_init(ISADevice *dev, qemu_irq a20_out)
783 {
784     qdev_connect_gpio_out_named(DEVICE(dev), PORT92_A20_LINE, 0, a20_out);
785 }
786 
787 static const VMStateDescription vmstate_port92_isa = {
788     .name = "port92",
789     .version_id = 1,
790     .minimum_version_id = 1,
791     .fields = (VMStateField[]) {
792         VMSTATE_UINT8(outport, Port92State),
793         VMSTATE_END_OF_LIST()
794     }
795 };
796 
797 static void port92_reset(DeviceState *d)
798 {
799     Port92State *s = PORT92(d);
800 
801     s->outport &= ~1;
802 }
803 
804 static const MemoryRegionOps port92_ops = {
805     .read = port92_read,
806     .write = port92_write,
807     .impl = {
808         .min_access_size = 1,
809         .max_access_size = 1,
810     },
811     .endianness = DEVICE_LITTLE_ENDIAN,
812 };
813 
814 static void port92_initfn(Object *obj)
815 {
816     Port92State *s = PORT92(obj);
817 
818     memory_region_init_io(&s->io, OBJECT(s), &port92_ops, s, "port92", 1);
819 
820     s->outport = 0;
821 
822     qdev_init_gpio_out_named(DEVICE(obj), &s->a20_out, PORT92_A20_LINE, 1);
823 }
824 
825 static void port92_realizefn(DeviceState *dev, Error **errp)
826 {
827     ISADevice *isadev = ISA_DEVICE(dev);
828     Port92State *s = PORT92(dev);
829 
830     isa_register_ioport(isadev, &s->io, 0x92);
831 }
832 
833 static void port92_class_initfn(ObjectClass *klass, void *data)
834 {
835     DeviceClass *dc = DEVICE_CLASS(klass);
836 
837     dc->realize = port92_realizefn;
838     dc->reset = port92_reset;
839     dc->vmsd = &vmstate_port92_isa;
840     /*
841      * Reason: unlike ordinary ISA devices, this one needs additional
842      * wiring: its A20 output line needs to be wired up by
843      * port92_init().
844      */
845     dc->user_creatable = false;
846 }
847 
848 static const TypeInfo port92_info = {
849     .name          = TYPE_PORT92,
850     .parent        = TYPE_ISA_DEVICE,
851     .instance_size = sizeof(Port92State),
852     .instance_init = port92_initfn,
853     .class_init    = port92_class_initfn,
854 };
855 
856 static void port92_register_types(void)
857 {
858     type_register_static(&port92_info);
859 }
860 
861 type_init(port92_register_types)
862 
863 static void handle_a20_line_change(void *opaque, int irq, int level)
864 {
865     X86CPU *cpu = opaque;
866 
867     /* XXX: send to all CPUs ? */
868     /* XXX: add logic to handle multiple A20 line sources */
869     x86_cpu_set_a20(cpu, level);
870 }
871 
872 int e820_add_entry(uint64_t address, uint64_t length, uint32_t type)
873 {
874     int index = le32_to_cpu(e820_reserve.count);
875     struct e820_entry *entry;
876 
877     if (type != E820_RAM) {
878         /* old FW_CFG_E820_TABLE entry -- reservations only */
879         if (index >= E820_NR_ENTRIES) {
880             return -EBUSY;
881         }
882         entry = &e820_reserve.entry[index++];
883 
884         entry->address = cpu_to_le64(address);
885         entry->length = cpu_to_le64(length);
886         entry->type = cpu_to_le32(type);
887 
888         e820_reserve.count = cpu_to_le32(index);
889     }
890 
891     /* new "etc/e820" file -- include ram too */
892     e820_table = g_renew(struct e820_entry, e820_table, e820_entries + 1);
893     e820_table[e820_entries].address = cpu_to_le64(address);
894     e820_table[e820_entries].length = cpu_to_le64(length);
895     e820_table[e820_entries].type = cpu_to_le32(type);
896     e820_entries++;
897 
898     return e820_entries;
899 }
900 
901 int e820_get_num_entries(void)
902 {
903     return e820_entries;
904 }
905 
906 bool e820_get_entry(int idx, uint32_t type, uint64_t *address, uint64_t *length)
907 {
908     if (idx < e820_entries && e820_table[idx].type == cpu_to_le32(type)) {
909         *address = le64_to_cpu(e820_table[idx].address);
910         *length = le64_to_cpu(e820_table[idx].length);
911         return true;
912     }
913     return false;
914 }
915 
916 /* Enables contiguous-apic-ID mode, for compatibility */
917 static bool compat_apic_id_mode;
918 
919 void enable_compat_apic_id_mode(void)
920 {
921     compat_apic_id_mode = true;
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(unsigned int cpu_index)
932 {
933     uint32_t correct_id;
934     static bool warned;
935 
936     correct_id = x86_apicid_from_cpu_idx(smp_cores, smp_threads, cpu_index);
937     if (compat_apic_id_mode) {
938         if (cpu_index != correct_id && !warned && !qtest_enabled()) {
939             error_report("APIC IDs set in compatibility mode, "
940                          "CPU topology won't match the configuration");
941             warned = true;
942         }
943         return cpu_index;
944     } else {
945         return correct_id;
946     }
947 }
948 
949 static void pc_build_smbios(PCMachineState *pcms)
950 {
951     uint8_t *smbios_tables, *smbios_anchor;
952     size_t smbios_tables_len, smbios_anchor_len;
953     struct smbios_phys_mem_area *mem_array;
954     unsigned i, array_count;
955     MachineState *ms = MACHINE(pcms);
956     X86CPU *cpu = X86_CPU(ms->possible_cpus->cpus[0].cpu);
957 
958     /* tell smbios about cpuid version and features */
959     smbios_set_cpuid(cpu->env.cpuid_version, cpu->env.features[FEAT_1_EDX]);
960 
961     smbios_tables = smbios_get_table_legacy(&smbios_tables_len);
962     if (smbios_tables) {
963         fw_cfg_add_bytes(pcms->fw_cfg, FW_CFG_SMBIOS_ENTRIES,
964                          smbios_tables, smbios_tables_len);
965     }
966 
967     /* build the array of physical mem area from e820 table */
968     mem_array = g_malloc0(sizeof(*mem_array) * e820_get_num_entries());
969     for (i = 0, array_count = 0; i < e820_get_num_entries(); i++) {
970         uint64_t addr, len;
971 
972         if (e820_get_entry(i, E820_RAM, &addr, &len)) {
973             mem_array[array_count].address = addr;
974             mem_array[array_count].length = len;
975             array_count++;
976         }
977     }
978     smbios_get_tables(mem_array, array_count,
979                       &smbios_tables, &smbios_tables_len,
980                       &smbios_anchor, &smbios_anchor_len);
981     g_free(mem_array);
982 
983     if (smbios_anchor) {
984         fw_cfg_add_file(pcms->fw_cfg, "etc/smbios/smbios-tables",
985                         smbios_tables, smbios_tables_len);
986         fw_cfg_add_file(pcms->fw_cfg, "etc/smbios/smbios-anchor",
987                         smbios_anchor, smbios_anchor_len);
988     }
989 }
990 
991 static FWCfgState *bochs_bios_init(AddressSpace *as, PCMachineState *pcms)
992 {
993     FWCfgState *fw_cfg;
994     uint64_t *numa_fw_cfg;
995     int i;
996     const CPUArchIdList *cpus;
997     MachineClass *mc = MACHINE_GET_CLASS(pcms);
998 
999     fw_cfg = fw_cfg_init_io_dma(FW_CFG_IO_BASE, FW_CFG_IO_BASE + 4, as);
1000     fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
1001 
1002     /* FW_CFG_MAX_CPUS is a bit confusing/problematic on x86:
1003      *
1004      * For machine types prior to 1.8, SeaBIOS needs FW_CFG_MAX_CPUS for
1005      * building MPTable, ACPI MADT, ACPI CPU hotplug and ACPI SRAT table,
1006      * that tables are based on xAPIC ID and QEMU<->SeaBIOS interface
1007      * for CPU hotplug also uses APIC ID and not "CPU index".
1008      * This means that FW_CFG_MAX_CPUS is not the "maximum number of CPUs",
1009      * but the "limit to the APIC ID values SeaBIOS may see".
1010      *
1011      * So for compatibility reasons with old BIOSes we are stuck with
1012      * "etc/max-cpus" actually being apic_id_limit
1013      */
1014     fw_cfg_add_i16(fw_cfg, FW_CFG_MAX_CPUS, (uint16_t)pcms->apic_id_limit);
1015     fw_cfg_add_i64(fw_cfg, FW_CFG_RAM_SIZE, (uint64_t)ram_size);
1016     fw_cfg_add_bytes(fw_cfg, FW_CFG_ACPI_TABLES,
1017                      acpi_tables, acpi_tables_len);
1018     fw_cfg_add_i32(fw_cfg, FW_CFG_IRQ0_OVERRIDE, kvm_allows_irq0_override());
1019 
1020     fw_cfg_add_bytes(fw_cfg, FW_CFG_E820_TABLE,
1021                      &e820_reserve, sizeof(e820_reserve));
1022     fw_cfg_add_file(fw_cfg, "etc/e820", e820_table,
1023                     sizeof(struct e820_entry) * e820_entries);
1024 
1025     fw_cfg_add_bytes(fw_cfg, FW_CFG_HPET, &hpet_cfg, sizeof(hpet_cfg));
1026     /* allocate memory for the NUMA channel: one (64bit) word for the number
1027      * of nodes, one word for each VCPU->node and one word for each node to
1028      * hold the amount of memory.
1029      */
1030     numa_fw_cfg = g_new0(uint64_t, 1 + pcms->apic_id_limit + nb_numa_nodes);
1031     numa_fw_cfg[0] = cpu_to_le64(nb_numa_nodes);
1032     cpus = mc->possible_cpu_arch_ids(MACHINE(pcms));
1033     for (i = 0; i < cpus->len; i++) {
1034         unsigned int apic_id = cpus->cpus[i].arch_id;
1035         assert(apic_id < pcms->apic_id_limit);
1036         numa_fw_cfg[apic_id + 1] = cpu_to_le64(cpus->cpus[i].props.node_id);
1037     }
1038     for (i = 0; i < nb_numa_nodes; i++) {
1039         numa_fw_cfg[pcms->apic_id_limit + 1 + i] =
1040             cpu_to_le64(numa_info[i].node_mem);
1041     }
1042     fw_cfg_add_bytes(fw_cfg, FW_CFG_NUMA, numa_fw_cfg,
1043                      (1 + pcms->apic_id_limit + nb_numa_nodes) *
1044                      sizeof(*numa_fw_cfg));
1045 
1046     return fw_cfg;
1047 }
1048 
1049 static long get_file_size(FILE *f)
1050 {
1051     long where, size;
1052 
1053     /* XXX: on Unix systems, using fstat() probably makes more sense */
1054 
1055     where = ftell(f);
1056     fseek(f, 0, SEEK_END);
1057     size = ftell(f);
1058     fseek(f, where, SEEK_SET);
1059 
1060     return size;
1061 }
1062 
1063 struct setup_data {
1064     uint64_t next;
1065     uint32_t type;
1066     uint32_t len;
1067     uint8_t data[0];
1068 } __attribute__((packed));
1069 
1070 
1071 /*
1072  * The entry point into the kernel for PVH boot is different from
1073  * the native entry point.  The PVH entry is defined by the x86/HVM
1074  * direct boot ABI and is available in an ELFNOTE in the kernel binary.
1075  *
1076  * This function is passed to load_elf() when it is called from
1077  * load_elfboot() which then additionally checks for an ELF Note of
1078  * type XEN_ELFNOTE_PHYS32_ENTRY and passes it to this function to
1079  * parse the PVH entry address from the ELF Note.
1080  *
1081  * Due to trickery in elf_opts.h, load_elf() is actually available as
1082  * load_elf32() or load_elf64() and this routine needs to be able
1083  * to deal with being called as 32 or 64 bit.
1084  *
1085  * The address of the PVH entry point is saved to the 'pvh_start_addr'
1086  * global variable.  (although the entry point is 32-bit, the kernel
1087  * binary can be either 32-bit or 64-bit).
1088  */
1089 static uint64_t read_pvh_start_addr(void *arg1, void *arg2, bool is64)
1090 {
1091     size_t *elf_note_data_addr;
1092 
1093     /* Check if ELF Note header passed in is valid */
1094     if (arg1 == NULL) {
1095         return 0;
1096     }
1097 
1098     if (is64) {
1099         struct elf64_note *nhdr64 = (struct elf64_note *)arg1;
1100         uint64_t nhdr_size64 = sizeof(struct elf64_note);
1101         uint64_t phdr_align = *(uint64_t *)arg2;
1102         uint64_t nhdr_namesz = nhdr64->n_namesz;
1103 
1104         elf_note_data_addr =
1105             ((void *)nhdr64) + nhdr_size64 +
1106             QEMU_ALIGN_UP(nhdr_namesz, phdr_align);
1107     } else {
1108         struct elf32_note *nhdr32 = (struct elf32_note *)arg1;
1109         uint32_t nhdr_size32 = sizeof(struct elf32_note);
1110         uint32_t phdr_align = *(uint32_t *)arg2;
1111         uint32_t nhdr_namesz = nhdr32->n_namesz;
1112 
1113         elf_note_data_addr =
1114             ((void *)nhdr32) + nhdr_size32 +
1115             QEMU_ALIGN_UP(nhdr_namesz, phdr_align);
1116     }
1117 
1118     pvh_start_addr = *elf_note_data_addr;
1119 
1120     return pvh_start_addr;
1121 }
1122 
1123 static bool load_elfboot(const char *kernel_filename,
1124                    int kernel_file_size,
1125                    uint8_t *header,
1126                    size_t pvh_xen_start_addr,
1127                    FWCfgState *fw_cfg)
1128 {
1129     uint32_t flags = 0;
1130     uint32_t mh_load_addr = 0;
1131     uint32_t elf_kernel_size = 0;
1132     uint64_t elf_entry;
1133     uint64_t elf_low, elf_high;
1134     int kernel_size;
1135 
1136     if (ldl_p(header) != 0x464c457f) {
1137         return false; /* no elfboot */
1138     }
1139 
1140     bool elf_is64 = header[EI_CLASS] == ELFCLASS64;
1141     flags = elf_is64 ?
1142         ((Elf64_Ehdr *)header)->e_flags : ((Elf32_Ehdr *)header)->e_flags;
1143 
1144     if (flags & 0x00010004) { /* LOAD_ELF_HEADER_HAS_ADDR */
1145         error_report("elfboot unsupported flags = %x", flags);
1146         exit(1);
1147     }
1148 
1149     uint64_t elf_note_type = XEN_ELFNOTE_PHYS32_ENTRY;
1150     kernel_size = load_elf(kernel_filename, read_pvh_start_addr,
1151                            NULL, &elf_note_type, &elf_entry,
1152                            &elf_low, &elf_high, 0, I386_ELF_MACHINE,
1153                            0, 0);
1154 
1155     if (kernel_size < 0) {
1156         error_report("Error while loading elf kernel");
1157         exit(1);
1158     }
1159     mh_load_addr = elf_low;
1160     elf_kernel_size = elf_high - elf_low;
1161 
1162     if (pvh_start_addr == 0) {
1163         error_report("Error loading uncompressed kernel without PVH ELF Note");
1164         exit(1);
1165     }
1166     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ENTRY, pvh_start_addr);
1167     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ADDR, mh_load_addr);
1168     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_SIZE, elf_kernel_size);
1169 
1170     return true;
1171 }
1172 
1173 static void load_linux(PCMachineState *pcms,
1174                        FWCfgState *fw_cfg)
1175 {
1176     uint16_t protocol;
1177     int setup_size, kernel_size, cmdline_size;
1178     int dtb_size, setup_data_offset;
1179     uint32_t initrd_max;
1180     uint8_t header[8192], *setup, *kernel;
1181     hwaddr real_addr, prot_addr, cmdline_addr, initrd_addr = 0;
1182     FILE *f;
1183     char *vmode;
1184     MachineState *machine = MACHINE(pcms);
1185     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1186     struct setup_data *setup_data;
1187     const char *kernel_filename = machine->kernel_filename;
1188     const char *initrd_filename = machine->initrd_filename;
1189     const char *dtb_filename = machine->dtb;
1190     const char *kernel_cmdline = machine->kernel_cmdline;
1191 
1192     /* Align to 16 bytes as a paranoia measure */
1193     cmdline_size = (strlen(kernel_cmdline)+16) & ~15;
1194 
1195     /* load the kernel header */
1196     f = fopen(kernel_filename, "rb");
1197     if (!f || !(kernel_size = get_file_size(f)) ||
1198         fread(header, 1, MIN(ARRAY_SIZE(header), kernel_size), f) !=
1199         MIN(ARRAY_SIZE(header), kernel_size)) {
1200         fprintf(stderr, "qemu: could not load kernel '%s': %s\n",
1201                 kernel_filename, strerror(errno));
1202         exit(1);
1203     }
1204 
1205     /* kernel protocol version */
1206 #if 0
1207     fprintf(stderr, "header magic: %#x\n", ldl_p(header+0x202));
1208 #endif
1209     if (ldl_p(header+0x202) == 0x53726448) {
1210         protocol = lduw_p(header+0x206);
1211     } else {
1212         /*
1213          * This could be a multiboot kernel. If it is, let's stop treating it
1214          * like a Linux kernel.
1215          * Note: some multiboot images could be in the ELF format (the same of
1216          * PVH), so we try multiboot first since we check the multiboot magic
1217          * header before to load it.
1218          */
1219         if (load_multiboot(fw_cfg, f, kernel_filename, initrd_filename,
1220                            kernel_cmdline, kernel_size, header)) {
1221             return;
1222         }
1223         /*
1224          * Check if the file is an uncompressed kernel file (ELF) and load it,
1225          * saving the PVH entry point used by the x86/HVM direct boot ABI.
1226          * If load_elfboot() is successful, populate the fw_cfg info.
1227          */
1228         if (pcmc->pvh_enabled &&
1229             load_elfboot(kernel_filename, kernel_size,
1230                          header, pvh_start_addr, fw_cfg)) {
1231             fclose(f);
1232 
1233             fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_SIZE,
1234                 strlen(kernel_cmdline) + 1);
1235             fw_cfg_add_string(fw_cfg, FW_CFG_CMDLINE_DATA, kernel_cmdline);
1236 
1237             fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_SIZE, sizeof(header));
1238             fw_cfg_add_bytes(fw_cfg, FW_CFG_SETUP_DATA,
1239                              header, sizeof(header));
1240 
1241             /* load initrd */
1242             if (initrd_filename) {
1243                 gsize initrd_size;
1244                 gchar *initrd_data;
1245                 GError *gerr = NULL;
1246 
1247                 if (!g_file_get_contents(initrd_filename, &initrd_data,
1248                             &initrd_size, &gerr)) {
1249                     fprintf(stderr, "qemu: error reading initrd %s: %s\n",
1250                             initrd_filename, gerr->message);
1251                     exit(1);
1252                 }
1253 
1254                 initrd_max = pcms->below_4g_mem_size - pcmc->acpi_data_size - 1;
1255                 if (initrd_size >= initrd_max) {
1256                     fprintf(stderr, "qemu: initrd is too large, cannot support."
1257                             "(max: %"PRIu32", need %"PRId64")\n",
1258                             initrd_max, (uint64_t)initrd_size);
1259                     exit(1);
1260                 }
1261 
1262                 initrd_addr = (initrd_max - initrd_size) & ~4095;
1263 
1264                 fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_ADDR, initrd_addr);
1265                 fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_SIZE, initrd_size);
1266                 fw_cfg_add_bytes(fw_cfg, FW_CFG_INITRD_DATA, initrd_data,
1267                                  initrd_size);
1268             }
1269 
1270             option_rom[nb_option_roms].bootindex = 0;
1271             option_rom[nb_option_roms].name = "pvh.bin";
1272             nb_option_roms++;
1273 
1274             return;
1275         }
1276         protocol = 0;
1277     }
1278 
1279     if (protocol < 0x200 || !(header[0x211] & 0x01)) {
1280         /* Low kernel */
1281         real_addr    = 0x90000;
1282         cmdline_addr = 0x9a000 - cmdline_size;
1283         prot_addr    = 0x10000;
1284     } else if (protocol < 0x202) {
1285         /* High but ancient kernel */
1286         real_addr    = 0x90000;
1287         cmdline_addr = 0x9a000 - cmdline_size;
1288         prot_addr    = 0x100000;
1289     } else {
1290         /* High and recent kernel */
1291         real_addr    = 0x10000;
1292         cmdline_addr = 0x20000;
1293         prot_addr    = 0x100000;
1294     }
1295 
1296 #if 0
1297     fprintf(stderr,
1298             "qemu: real_addr     = 0x" TARGET_FMT_plx "\n"
1299             "qemu: cmdline_addr  = 0x" TARGET_FMT_plx "\n"
1300             "qemu: prot_addr     = 0x" TARGET_FMT_plx "\n",
1301             real_addr,
1302             cmdline_addr,
1303             prot_addr);
1304 #endif
1305 
1306     /* highest address for loading the initrd */
1307     if (protocol >= 0x20c &&
1308         lduw_p(header+0x236) & XLF_CAN_BE_LOADED_ABOVE_4G) {
1309         /*
1310          * Linux has supported initrd up to 4 GB for a very long time (2007,
1311          * long before XLF_CAN_BE_LOADED_ABOVE_4G which was added in 2013),
1312          * though it only sets initrd_max to 2 GB to "work around bootloader
1313          * bugs". Luckily, QEMU firmware(which does something like bootloader)
1314          * has supported this.
1315          *
1316          * It's believed that if XLF_CAN_BE_LOADED_ABOVE_4G is set, initrd can
1317          * be loaded into any address.
1318          *
1319          * In addition, initrd_max is uint32_t simply because QEMU doesn't
1320          * support the 64-bit boot protocol (specifically the ext_ramdisk_image
1321          * field).
1322          *
1323          * Therefore here just limit initrd_max to UINT32_MAX simply as well.
1324          */
1325         initrd_max = UINT32_MAX;
1326     } else if (protocol >= 0x203) {
1327         initrd_max = ldl_p(header+0x22c);
1328     } else {
1329         initrd_max = 0x37ffffff;
1330     }
1331 
1332     if (initrd_max >= pcms->below_4g_mem_size - pcmc->acpi_data_size) {
1333         initrd_max = pcms->below_4g_mem_size - pcmc->acpi_data_size - 1;
1334     }
1335 
1336     fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_ADDR, cmdline_addr);
1337     fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_SIZE, strlen(kernel_cmdline)+1);
1338     fw_cfg_add_string(fw_cfg, FW_CFG_CMDLINE_DATA, kernel_cmdline);
1339 
1340     if (protocol >= 0x202) {
1341         stl_p(header+0x228, cmdline_addr);
1342     } else {
1343         stw_p(header+0x20, 0xA33F);
1344         stw_p(header+0x22, cmdline_addr-real_addr);
1345     }
1346 
1347     /* handle vga= parameter */
1348     vmode = strstr(kernel_cmdline, "vga=");
1349     if (vmode) {
1350         unsigned int video_mode;
1351         /* skip "vga=" */
1352         vmode += 4;
1353         if (!strncmp(vmode, "normal", 6)) {
1354             video_mode = 0xffff;
1355         } else if (!strncmp(vmode, "ext", 3)) {
1356             video_mode = 0xfffe;
1357         } else if (!strncmp(vmode, "ask", 3)) {
1358             video_mode = 0xfffd;
1359         } else {
1360             video_mode = strtol(vmode, NULL, 0);
1361         }
1362         stw_p(header+0x1fa, video_mode);
1363     }
1364 
1365     /* loader type */
1366     /* High nybble = B reserved for QEMU; low nybble is revision number.
1367        If this code is substantially changed, you may want to consider
1368        incrementing the revision. */
1369     if (protocol >= 0x200) {
1370         header[0x210] = 0xB0;
1371     }
1372     /* heap */
1373     if (protocol >= 0x201) {
1374         header[0x211] |= 0x80;	/* CAN_USE_HEAP */
1375         stw_p(header+0x224, cmdline_addr-real_addr-0x200);
1376     }
1377 
1378     /* load initrd */
1379     if (initrd_filename) {
1380         gsize initrd_size;
1381         gchar *initrd_data;
1382         GError *gerr = NULL;
1383 
1384         if (protocol < 0x200) {
1385             fprintf(stderr, "qemu: linux kernel too old to load a ram disk\n");
1386             exit(1);
1387         }
1388 
1389         if (!g_file_get_contents(initrd_filename, &initrd_data,
1390                                  &initrd_size, &gerr)) {
1391             fprintf(stderr, "qemu: error reading initrd %s: %s\n",
1392                     initrd_filename, gerr->message);
1393             exit(1);
1394         }
1395         if (initrd_size >= initrd_max) {
1396             fprintf(stderr, "qemu: initrd is too large, cannot support."
1397                     "(max: %"PRIu32", need %"PRId64")\n",
1398                     initrd_max, (uint64_t)initrd_size);
1399             exit(1);
1400         }
1401 
1402         initrd_addr = (initrd_max-initrd_size) & ~4095;
1403 
1404         fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_ADDR, initrd_addr);
1405         fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_SIZE, initrd_size);
1406         fw_cfg_add_bytes(fw_cfg, FW_CFG_INITRD_DATA, initrd_data, initrd_size);
1407 
1408         stl_p(header+0x218, initrd_addr);
1409         stl_p(header+0x21c, initrd_size);
1410     }
1411 
1412     /* load kernel and setup */
1413     setup_size = header[0x1f1];
1414     if (setup_size == 0) {
1415         setup_size = 4;
1416     }
1417     setup_size = (setup_size+1)*512;
1418     if (setup_size > kernel_size) {
1419         fprintf(stderr, "qemu: invalid kernel header\n");
1420         exit(1);
1421     }
1422     kernel_size -= setup_size;
1423 
1424     setup  = g_malloc(setup_size);
1425     kernel = g_malloc(kernel_size);
1426     fseek(f, 0, SEEK_SET);
1427     if (fread(setup, 1, setup_size, f) != setup_size) {
1428         fprintf(stderr, "fread() failed\n");
1429         exit(1);
1430     }
1431     if (fread(kernel, 1, kernel_size, f) != kernel_size) {
1432         fprintf(stderr, "fread() failed\n");
1433         exit(1);
1434     }
1435     fclose(f);
1436 
1437     /* append dtb to kernel */
1438     if (dtb_filename) {
1439         if (protocol < 0x209) {
1440             fprintf(stderr, "qemu: Linux kernel too old to load a dtb\n");
1441             exit(1);
1442         }
1443 
1444         dtb_size = get_image_size(dtb_filename);
1445         if (dtb_size <= 0) {
1446             fprintf(stderr, "qemu: error reading dtb %s: %s\n",
1447                     dtb_filename, strerror(errno));
1448             exit(1);
1449         }
1450 
1451         setup_data_offset = QEMU_ALIGN_UP(kernel_size, 16);
1452         kernel_size = setup_data_offset + sizeof(struct setup_data) + dtb_size;
1453         kernel = g_realloc(kernel, kernel_size);
1454 
1455         stq_p(header+0x250, prot_addr + setup_data_offset);
1456 
1457         setup_data = (struct setup_data *)(kernel + setup_data_offset);
1458         setup_data->next = 0;
1459         setup_data->type = cpu_to_le32(SETUP_DTB);
1460         setup_data->len = cpu_to_le32(dtb_size);
1461 
1462         load_image_size(dtb_filename, setup_data->data, dtb_size);
1463     }
1464 
1465     memcpy(setup, header, MIN(sizeof(header), setup_size));
1466 
1467     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ADDR, prot_addr);
1468     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_SIZE, kernel_size);
1469     fw_cfg_add_bytes(fw_cfg, FW_CFG_KERNEL_DATA, kernel, kernel_size);
1470 
1471     fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_ADDR, real_addr);
1472     fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_SIZE, setup_size);
1473     fw_cfg_add_bytes(fw_cfg, FW_CFG_SETUP_DATA, setup, setup_size);
1474 
1475     option_rom[nb_option_roms].bootindex = 0;
1476     option_rom[nb_option_roms].name = "linuxboot.bin";
1477     if (pcmc->linuxboot_dma_enabled && fw_cfg_dma_enabled(fw_cfg)) {
1478         option_rom[nb_option_roms].name = "linuxboot_dma.bin";
1479     }
1480     nb_option_roms++;
1481 }
1482 
1483 #define NE2000_NB_MAX 6
1484 
1485 static const int ne2000_io[NE2000_NB_MAX] = { 0x300, 0x320, 0x340, 0x360,
1486                                               0x280, 0x380 };
1487 static const int ne2000_irq[NE2000_NB_MAX] = { 9, 10, 11, 3, 4, 5 };
1488 
1489 void pc_init_ne2k_isa(ISABus *bus, NICInfo *nd)
1490 {
1491     static int nb_ne2k = 0;
1492 
1493     if (nb_ne2k == NE2000_NB_MAX)
1494         return;
1495     isa_ne2000_init(bus, ne2000_io[nb_ne2k],
1496                     ne2000_irq[nb_ne2k], nd);
1497     nb_ne2k++;
1498 }
1499 
1500 DeviceState *cpu_get_current_apic(void)
1501 {
1502     if (current_cpu) {
1503         X86CPU *cpu = X86_CPU(current_cpu);
1504         return cpu->apic_state;
1505     } else {
1506         return NULL;
1507     }
1508 }
1509 
1510 void pc_acpi_smi_interrupt(void *opaque, int irq, int level)
1511 {
1512     X86CPU *cpu = opaque;
1513 
1514     if (level) {
1515         cpu_interrupt(CPU(cpu), CPU_INTERRUPT_SMI);
1516     }
1517 }
1518 
1519 static void pc_new_cpu(const char *typename, int64_t apic_id, Error **errp)
1520 {
1521     Object *cpu = NULL;
1522     Error *local_err = NULL;
1523 
1524     cpu = object_new(typename);
1525 
1526     object_property_set_uint(cpu, apic_id, "apic-id", &local_err);
1527     object_property_set_bool(cpu, true, "realized", &local_err);
1528 
1529     object_unref(cpu);
1530     error_propagate(errp, local_err);
1531 }
1532 
1533 void pc_hot_add_cpu(const int64_t id, Error **errp)
1534 {
1535     MachineState *ms = MACHINE(qdev_get_machine());
1536     int64_t apic_id = x86_cpu_apic_id_from_index(id);
1537     Error *local_err = NULL;
1538 
1539     if (id < 0) {
1540         error_setg(errp, "Invalid CPU id: %" PRIi64, id);
1541         return;
1542     }
1543 
1544     if (apic_id >= ACPI_CPU_HOTPLUG_ID_LIMIT) {
1545         error_setg(errp, "Unable to add CPU: %" PRIi64
1546                    ", resulting APIC ID (%" PRIi64 ") is too large",
1547                    id, apic_id);
1548         return;
1549     }
1550 
1551     pc_new_cpu(ms->cpu_type, apic_id, &local_err);
1552     if (local_err) {
1553         error_propagate(errp, local_err);
1554         return;
1555     }
1556 }
1557 
1558 void pc_cpus_init(PCMachineState *pcms)
1559 {
1560     int i;
1561     const CPUArchIdList *possible_cpus;
1562     MachineState *ms = MACHINE(pcms);
1563     MachineClass *mc = MACHINE_GET_CLASS(pcms);
1564 
1565     /* Calculates the limit to CPU APIC ID values
1566      *
1567      * Limit for the APIC ID value, so that all
1568      * CPU APIC IDs are < pcms->apic_id_limit.
1569      *
1570      * This is used for FW_CFG_MAX_CPUS. See comments on bochs_bios_init().
1571      */
1572     pcms->apic_id_limit = x86_cpu_apic_id_from_index(max_cpus - 1) + 1;
1573     possible_cpus = mc->possible_cpu_arch_ids(ms);
1574     for (i = 0; i < smp_cpus; i++) {
1575         pc_new_cpu(possible_cpus->cpus[i].type, possible_cpus->cpus[i].arch_id,
1576                    &error_fatal);
1577     }
1578 }
1579 
1580 static void pc_build_feature_control_file(PCMachineState *pcms)
1581 {
1582     MachineState *ms = MACHINE(pcms);
1583     X86CPU *cpu = X86_CPU(ms->possible_cpus->cpus[0].cpu);
1584     CPUX86State *env = &cpu->env;
1585     uint32_t unused, ecx, edx;
1586     uint64_t feature_control_bits = 0;
1587     uint64_t *val;
1588 
1589     cpu_x86_cpuid(env, 1, 0, &unused, &unused, &ecx, &edx);
1590     if (ecx & CPUID_EXT_VMX) {
1591         feature_control_bits |= FEATURE_CONTROL_VMXON_ENABLED_OUTSIDE_SMX;
1592     }
1593 
1594     if ((edx & (CPUID_EXT2_MCE | CPUID_EXT2_MCA)) ==
1595         (CPUID_EXT2_MCE | CPUID_EXT2_MCA) &&
1596         (env->mcg_cap & MCG_LMCE_P)) {
1597         feature_control_bits |= FEATURE_CONTROL_LMCE;
1598     }
1599 
1600     if (!feature_control_bits) {
1601         return;
1602     }
1603 
1604     val = g_malloc(sizeof(*val));
1605     *val = cpu_to_le64(feature_control_bits | FEATURE_CONTROL_LOCKED);
1606     fw_cfg_add_file(pcms->fw_cfg, "etc/msr_feature_control", val, sizeof(*val));
1607 }
1608 
1609 static void rtc_set_cpus_count(ISADevice *rtc, uint16_t cpus_count)
1610 {
1611     if (cpus_count > 0xff) {
1612         /* If the number of CPUs can't be represented in 8 bits, the
1613          * BIOS must use "FW_CFG_NB_CPUS". Set RTC field to 0 just
1614          * to make old BIOSes fail more predictably.
1615          */
1616         rtc_set_memory(rtc, 0x5f, 0);
1617     } else {
1618         rtc_set_memory(rtc, 0x5f, cpus_count - 1);
1619     }
1620 }
1621 
1622 static
1623 void pc_machine_done(Notifier *notifier, void *data)
1624 {
1625     PCMachineState *pcms = container_of(notifier,
1626                                         PCMachineState, machine_done);
1627     PCIBus *bus = pcms->bus;
1628 
1629     /* set the number of CPUs */
1630     rtc_set_cpus_count(pcms->rtc, pcms->boot_cpus);
1631 
1632     if (bus) {
1633         int extra_hosts = 0;
1634 
1635         QLIST_FOREACH(bus, &bus->child, sibling) {
1636             /* look for expander root buses */
1637             if (pci_bus_is_root(bus)) {
1638                 extra_hosts++;
1639             }
1640         }
1641         if (extra_hosts && pcms->fw_cfg) {
1642             uint64_t *val = g_malloc(sizeof(*val));
1643             *val = cpu_to_le64(extra_hosts);
1644             fw_cfg_add_file(pcms->fw_cfg,
1645                     "etc/extra-pci-roots", val, sizeof(*val));
1646         }
1647     }
1648 
1649     acpi_setup();
1650     if (pcms->fw_cfg) {
1651         pc_build_smbios(pcms);
1652         pc_build_feature_control_file(pcms);
1653         /* update FW_CFG_NB_CPUS to account for -device added CPUs */
1654         fw_cfg_modify_i16(pcms->fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
1655     }
1656 
1657     if (pcms->apic_id_limit > 255 && !xen_enabled()) {
1658         IntelIOMMUState *iommu = INTEL_IOMMU_DEVICE(x86_iommu_get_default());
1659 
1660         if (!iommu || !x86_iommu_ir_supported(X86_IOMMU_DEVICE(iommu)) ||
1661             iommu->intr_eim != ON_OFF_AUTO_ON) {
1662             error_report("current -smp configuration requires "
1663                          "Extended Interrupt Mode enabled. "
1664                          "You can add an IOMMU using: "
1665                          "-device intel-iommu,intremap=on,eim=on");
1666             exit(EXIT_FAILURE);
1667         }
1668     }
1669 }
1670 
1671 void pc_guest_info_init(PCMachineState *pcms)
1672 {
1673     int i;
1674 
1675     pcms->apic_xrupt_override = kvm_allows_irq0_override();
1676     pcms->numa_nodes = nb_numa_nodes;
1677     pcms->node_mem = g_malloc0(pcms->numa_nodes *
1678                                     sizeof *pcms->node_mem);
1679     for (i = 0; i < nb_numa_nodes; i++) {
1680         pcms->node_mem[i] = numa_info[i].node_mem;
1681     }
1682 
1683     pcms->machine_done.notify = pc_machine_done;
1684     qemu_add_machine_init_done_notifier(&pcms->machine_done);
1685 }
1686 
1687 /* setup pci memory address space mapping into system address space */
1688 void pc_pci_as_mapping_init(Object *owner, MemoryRegion *system_memory,
1689                             MemoryRegion *pci_address_space)
1690 {
1691     /* Set to lower priority than RAM */
1692     memory_region_add_subregion_overlap(system_memory, 0x0,
1693                                         pci_address_space, -1);
1694 }
1695 
1696 void xen_load_linux(PCMachineState *pcms)
1697 {
1698     int i;
1699     FWCfgState *fw_cfg;
1700 
1701     assert(MACHINE(pcms)->kernel_filename != NULL);
1702 
1703     fw_cfg = fw_cfg_init_io(FW_CFG_IO_BASE);
1704     fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
1705     rom_set_fw(fw_cfg);
1706 
1707     load_linux(pcms, fw_cfg);
1708     for (i = 0; i < nb_option_roms; i++) {
1709         assert(!strcmp(option_rom[i].name, "linuxboot.bin") ||
1710                !strcmp(option_rom[i].name, "linuxboot_dma.bin") ||
1711                !strcmp(option_rom[i].name, "pvh.bin") ||
1712                !strcmp(option_rom[i].name, "multiboot.bin"));
1713         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1714     }
1715     pcms->fw_cfg = fw_cfg;
1716 }
1717 
1718 void pc_memory_init(PCMachineState *pcms,
1719                     MemoryRegion *system_memory,
1720                     MemoryRegion *rom_memory,
1721                     MemoryRegion **ram_memory)
1722 {
1723     int linux_boot, i;
1724     MemoryRegion *ram, *option_rom_mr;
1725     MemoryRegion *ram_below_4g, *ram_above_4g;
1726     FWCfgState *fw_cfg;
1727     MachineState *machine = MACHINE(pcms);
1728     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1729 
1730     assert(machine->ram_size == pcms->below_4g_mem_size +
1731                                 pcms->above_4g_mem_size);
1732 
1733     linux_boot = (machine->kernel_filename != NULL);
1734 
1735     /* Allocate RAM.  We allocate it as a single memory region and use
1736      * aliases to address portions of it, mostly for backwards compatibility
1737      * with older qemus that used qemu_ram_alloc().
1738      */
1739     ram = g_malloc(sizeof(*ram));
1740     memory_region_allocate_system_memory(ram, NULL, "pc.ram",
1741                                          machine->ram_size);
1742     *ram_memory = ram;
1743     ram_below_4g = g_malloc(sizeof(*ram_below_4g));
1744     memory_region_init_alias(ram_below_4g, NULL, "ram-below-4g", ram,
1745                              0, pcms->below_4g_mem_size);
1746     memory_region_add_subregion(system_memory, 0, ram_below_4g);
1747     e820_add_entry(0, pcms->below_4g_mem_size, E820_RAM);
1748     if (pcms->above_4g_mem_size > 0) {
1749         ram_above_4g = g_malloc(sizeof(*ram_above_4g));
1750         memory_region_init_alias(ram_above_4g, NULL, "ram-above-4g", ram,
1751                                  pcms->below_4g_mem_size,
1752                                  pcms->above_4g_mem_size);
1753         memory_region_add_subregion(system_memory, 0x100000000ULL,
1754                                     ram_above_4g);
1755         e820_add_entry(0x100000000ULL, pcms->above_4g_mem_size, E820_RAM);
1756     }
1757 
1758     if (!pcmc->has_reserved_memory &&
1759         (machine->ram_slots ||
1760          (machine->maxram_size > machine->ram_size))) {
1761         MachineClass *mc = MACHINE_GET_CLASS(machine);
1762 
1763         error_report("\"-memory 'slots|maxmem'\" is not supported by: %s",
1764                      mc->name);
1765         exit(EXIT_FAILURE);
1766     }
1767 
1768     /* always allocate the device memory information */
1769     machine->device_memory = g_malloc0(sizeof(*machine->device_memory));
1770 
1771     /* initialize device memory address space */
1772     if (pcmc->has_reserved_memory &&
1773         (machine->ram_size < machine->maxram_size)) {
1774         ram_addr_t device_mem_size = machine->maxram_size - machine->ram_size;
1775 
1776         if (machine->ram_slots > ACPI_MAX_RAM_SLOTS) {
1777             error_report("unsupported amount of memory slots: %"PRIu64,
1778                          machine->ram_slots);
1779             exit(EXIT_FAILURE);
1780         }
1781 
1782         if (QEMU_ALIGN_UP(machine->maxram_size,
1783                           TARGET_PAGE_SIZE) != machine->maxram_size) {
1784             error_report("maximum memory size must by aligned to multiple of "
1785                          "%d bytes", TARGET_PAGE_SIZE);
1786             exit(EXIT_FAILURE);
1787         }
1788 
1789         machine->device_memory->base =
1790             ROUND_UP(0x100000000ULL + pcms->above_4g_mem_size, 1 * GiB);
1791 
1792         if (pcmc->enforce_aligned_dimm) {
1793             /* size device region assuming 1G page max alignment per slot */
1794             device_mem_size += (1 * GiB) * machine->ram_slots;
1795         }
1796 
1797         if ((machine->device_memory->base + device_mem_size) <
1798             device_mem_size) {
1799             error_report("unsupported amount of maximum memory: " RAM_ADDR_FMT,
1800                          machine->maxram_size);
1801             exit(EXIT_FAILURE);
1802         }
1803 
1804         memory_region_init(&machine->device_memory->mr, OBJECT(pcms),
1805                            "device-memory", device_mem_size);
1806         memory_region_add_subregion(system_memory, machine->device_memory->base,
1807                                     &machine->device_memory->mr);
1808     }
1809 
1810     /* Initialize PC system firmware */
1811     pc_system_firmware_init(pcms, rom_memory);
1812 
1813     option_rom_mr = g_malloc(sizeof(*option_rom_mr));
1814     memory_region_init_ram(option_rom_mr, NULL, "pc.rom", PC_ROM_SIZE,
1815                            &error_fatal);
1816     if (pcmc->pci_enabled) {
1817         memory_region_set_readonly(option_rom_mr, true);
1818     }
1819     memory_region_add_subregion_overlap(rom_memory,
1820                                         PC_ROM_MIN_VGA,
1821                                         option_rom_mr,
1822                                         1);
1823 
1824     fw_cfg = bochs_bios_init(&address_space_memory, pcms);
1825 
1826     rom_set_fw(fw_cfg);
1827 
1828     if (pcmc->has_reserved_memory && machine->device_memory->base) {
1829         uint64_t *val = g_malloc(sizeof(*val));
1830         PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1831         uint64_t res_mem_end = machine->device_memory->base;
1832 
1833         if (!pcmc->broken_reserved_end) {
1834             res_mem_end += memory_region_size(&machine->device_memory->mr);
1835         }
1836         *val = cpu_to_le64(ROUND_UP(res_mem_end, 1 * GiB));
1837         fw_cfg_add_file(fw_cfg, "etc/reserved-memory-end", val, sizeof(*val));
1838     }
1839 
1840     if (linux_boot) {
1841         load_linux(pcms, fw_cfg);
1842     }
1843 
1844     for (i = 0; i < nb_option_roms; i++) {
1845         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1846     }
1847     pcms->fw_cfg = fw_cfg;
1848 
1849     /* Init default IOAPIC address space */
1850     pcms->ioapic_as = &address_space_memory;
1851 }
1852 
1853 /*
1854  * The 64bit pci hole starts after "above 4G RAM" and
1855  * potentially the space reserved for memory hotplug.
1856  */
1857 uint64_t pc_pci_hole64_start(void)
1858 {
1859     PCMachineState *pcms = PC_MACHINE(qdev_get_machine());
1860     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1861     MachineState *ms = MACHINE(pcms);
1862     uint64_t hole64_start = 0;
1863 
1864     if (pcmc->has_reserved_memory && ms->device_memory->base) {
1865         hole64_start = ms->device_memory->base;
1866         if (!pcmc->broken_reserved_end) {
1867             hole64_start += memory_region_size(&ms->device_memory->mr);
1868         }
1869     } else {
1870         hole64_start = 0x100000000ULL + pcms->above_4g_mem_size;
1871     }
1872 
1873     return ROUND_UP(hole64_start, 1 * GiB);
1874 }
1875 
1876 qemu_irq pc_allocate_cpu_irq(void)
1877 {
1878     return qemu_allocate_irq(pic_irq_request, NULL, 0);
1879 }
1880 
1881 DeviceState *pc_vga_init(ISABus *isa_bus, PCIBus *pci_bus)
1882 {
1883     DeviceState *dev = NULL;
1884 
1885     rom_set_order_override(FW_CFG_ORDER_OVERRIDE_VGA);
1886     if (pci_bus) {
1887         PCIDevice *pcidev = pci_vga_init(pci_bus);
1888         dev = pcidev ? &pcidev->qdev : NULL;
1889     } else if (isa_bus) {
1890         ISADevice *isadev = isa_vga_init(isa_bus);
1891         dev = isadev ? DEVICE(isadev) : NULL;
1892     }
1893     rom_reset_order_override();
1894     return dev;
1895 }
1896 
1897 static const MemoryRegionOps ioport80_io_ops = {
1898     .write = ioport80_write,
1899     .read = ioport80_read,
1900     .endianness = DEVICE_NATIVE_ENDIAN,
1901     .impl = {
1902         .min_access_size = 1,
1903         .max_access_size = 1,
1904     },
1905 };
1906 
1907 static const MemoryRegionOps ioportF0_io_ops = {
1908     .write = ioportF0_write,
1909     .read = ioportF0_read,
1910     .endianness = DEVICE_NATIVE_ENDIAN,
1911     .impl = {
1912         .min_access_size = 1,
1913         .max_access_size = 1,
1914     },
1915 };
1916 
1917 static void pc_superio_init(ISABus *isa_bus, bool create_fdctrl, bool no_vmport)
1918 {
1919     int i;
1920     DriveInfo *fd[MAX_FD];
1921     qemu_irq *a20_line;
1922     ISADevice *i8042, *port92, *vmmouse;
1923 
1924     serial_hds_isa_init(isa_bus, 0, MAX_ISA_SERIAL_PORTS);
1925     parallel_hds_isa_init(isa_bus, MAX_PARALLEL_PORTS);
1926 
1927     for (i = 0; i < MAX_FD; i++) {
1928         fd[i] = drive_get(IF_FLOPPY, 0, i);
1929         create_fdctrl |= !!fd[i];
1930     }
1931     if (create_fdctrl) {
1932         fdctrl_init_isa(isa_bus, fd);
1933     }
1934 
1935     i8042 = isa_create_simple(isa_bus, "i8042");
1936     if (!no_vmport) {
1937         vmport_init(isa_bus);
1938         vmmouse = isa_try_create(isa_bus, "vmmouse");
1939     } else {
1940         vmmouse = NULL;
1941     }
1942     if (vmmouse) {
1943         DeviceState *dev = DEVICE(vmmouse);
1944         qdev_prop_set_ptr(dev, "ps2_mouse", i8042);
1945         qdev_init_nofail(dev);
1946     }
1947     port92 = isa_create_simple(isa_bus, "port92");
1948 
1949     a20_line = qemu_allocate_irqs(handle_a20_line_change, first_cpu, 2);
1950     i8042_setup_a20_line(i8042, a20_line[0]);
1951     port92_init(port92, a20_line[1]);
1952     g_free(a20_line);
1953 }
1954 
1955 void pc_basic_device_init(ISABus *isa_bus, qemu_irq *gsi,
1956                           ISADevice **rtc_state,
1957                           bool create_fdctrl,
1958                           bool no_vmport,
1959                           bool has_pit,
1960                           uint32_t hpet_irqs)
1961 {
1962     int i;
1963     DeviceState *hpet = NULL;
1964     int pit_isa_irq = 0;
1965     qemu_irq pit_alt_irq = NULL;
1966     qemu_irq rtc_irq = NULL;
1967     ISADevice *pit = NULL;
1968     MemoryRegion *ioport80_io = g_new(MemoryRegion, 1);
1969     MemoryRegion *ioportF0_io = g_new(MemoryRegion, 1);
1970 
1971     memory_region_init_io(ioport80_io, NULL, &ioport80_io_ops, NULL, "ioport80", 1);
1972     memory_region_add_subregion(isa_bus->address_space_io, 0x80, ioport80_io);
1973 
1974     memory_region_init_io(ioportF0_io, NULL, &ioportF0_io_ops, NULL, "ioportF0", 1);
1975     memory_region_add_subregion(isa_bus->address_space_io, 0xf0, ioportF0_io);
1976 
1977     /*
1978      * Check if an HPET shall be created.
1979      *
1980      * Without KVM_CAP_PIT_STATE2, we cannot switch off the in-kernel PIT
1981      * when the HPET wants to take over. Thus we have to disable the latter.
1982      */
1983     if (!no_hpet && (!kvm_irqchip_in_kernel() || kvm_has_pit_state2())) {
1984         /* In order to set property, here not using sysbus_try_create_simple */
1985         hpet = qdev_try_create(NULL, TYPE_HPET);
1986         if (hpet) {
1987             /* For pc-piix-*, hpet's intcap is always IRQ2. For pc-q35-1.7
1988              * and earlier, use IRQ2 for compat. Otherwise, use IRQ16~23,
1989              * IRQ8 and IRQ2.
1990              */
1991             uint8_t compat = object_property_get_uint(OBJECT(hpet),
1992                     HPET_INTCAP, NULL);
1993             if (!compat) {
1994                 qdev_prop_set_uint32(hpet, HPET_INTCAP, hpet_irqs);
1995             }
1996             qdev_init_nofail(hpet);
1997             sysbus_mmio_map(SYS_BUS_DEVICE(hpet), 0, HPET_BASE);
1998 
1999             for (i = 0; i < GSI_NUM_PINS; i++) {
2000                 sysbus_connect_irq(SYS_BUS_DEVICE(hpet), i, gsi[i]);
2001             }
2002             pit_isa_irq = -1;
2003             pit_alt_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_PIT_INT);
2004             rtc_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_RTC_INT);
2005         }
2006     }
2007     *rtc_state = mc146818_rtc_init(isa_bus, 2000, rtc_irq);
2008 
2009     qemu_register_boot_set(pc_boot_set, *rtc_state);
2010 
2011     if (!xen_enabled() && has_pit) {
2012         if (kvm_pit_in_kernel()) {
2013             pit = kvm_pit_init(isa_bus, 0x40);
2014         } else {
2015             pit = i8254_pit_init(isa_bus, 0x40, pit_isa_irq, pit_alt_irq);
2016         }
2017         if (hpet) {
2018             /* connect PIT to output control line of the HPET */
2019             qdev_connect_gpio_out(hpet, 0, qdev_get_gpio_in(DEVICE(pit), 0));
2020         }
2021         pcspk_init(isa_bus, pit);
2022     }
2023 
2024     i8257_dma_init(isa_bus, 0);
2025 
2026     /* Super I/O */
2027     pc_superio_init(isa_bus, create_fdctrl, no_vmport);
2028 }
2029 
2030 void pc_nic_init(PCMachineClass *pcmc, ISABus *isa_bus, PCIBus *pci_bus)
2031 {
2032     int i;
2033 
2034     rom_set_order_override(FW_CFG_ORDER_OVERRIDE_NIC);
2035     for (i = 0; i < nb_nics; i++) {
2036         NICInfo *nd = &nd_table[i];
2037         const char *model = nd->model ? nd->model : pcmc->default_nic_model;
2038 
2039         if (g_str_equal(model, "ne2k_isa")) {
2040             pc_init_ne2k_isa(isa_bus, nd);
2041         } else {
2042             pci_nic_init_nofail(nd, pci_bus, model, NULL);
2043         }
2044     }
2045     rom_reset_order_override();
2046 }
2047 
2048 void ioapic_init_gsi(GSIState *gsi_state, const char *parent_name)
2049 {
2050     DeviceState *dev;
2051     SysBusDevice *d;
2052     unsigned int i;
2053 
2054     if (kvm_ioapic_in_kernel()) {
2055         dev = qdev_create(NULL, TYPE_KVM_IOAPIC);
2056     } else {
2057         dev = qdev_create(NULL, TYPE_IOAPIC);
2058     }
2059     if (parent_name) {
2060         object_property_add_child(object_resolve_path(parent_name, NULL),
2061                                   "ioapic", OBJECT(dev), NULL);
2062     }
2063     qdev_init_nofail(dev);
2064     d = SYS_BUS_DEVICE(dev);
2065     sysbus_mmio_map(d, 0, IO_APIC_DEFAULT_ADDRESS);
2066 
2067     for (i = 0; i < IOAPIC_NUM_PINS; i++) {
2068         gsi_state->ioapic_irq[i] = qdev_get_gpio_in(dev, i);
2069     }
2070 }
2071 
2072 static void pc_memory_pre_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
2073                                Error **errp)
2074 {
2075     const PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2076     const PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
2077     const MachineState *ms = MACHINE(hotplug_dev);
2078     const bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2079     const uint64_t legacy_align = TARGET_PAGE_SIZE;
2080     Error *local_err = NULL;
2081 
2082     /*
2083      * When -no-acpi is used with Q35 machine type, no ACPI is built,
2084      * but pcms->acpi_dev is still created. Check !acpi_enabled in
2085      * addition to cover this case.
2086      */
2087     if (!pcms->acpi_dev || !acpi_enabled) {
2088         error_setg(errp,
2089                    "memory hotplug is not enabled: missing acpi device or acpi disabled");
2090         return;
2091     }
2092 
2093     if (is_nvdimm && !ms->nvdimms_state->is_enabled) {
2094         error_setg(errp, "nvdimm is not enabled: missing 'nvdimm' in '-M'");
2095         return;
2096     }
2097 
2098     hotplug_handler_pre_plug(pcms->acpi_dev, dev, &local_err);
2099     if (local_err) {
2100         error_propagate(errp, local_err);
2101         return;
2102     }
2103 
2104     pc_dimm_pre_plug(PC_DIMM(dev), MACHINE(hotplug_dev),
2105                      pcmc->enforce_aligned_dimm ? NULL : &legacy_align, errp);
2106 }
2107 
2108 static void pc_memory_plug(HotplugHandler *hotplug_dev,
2109                            DeviceState *dev, Error **errp)
2110 {
2111     Error *local_err = NULL;
2112     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2113     MachineState *ms = MACHINE(hotplug_dev);
2114     bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2115 
2116     pc_dimm_plug(PC_DIMM(dev), MACHINE(pcms), &local_err);
2117     if (local_err) {
2118         goto out;
2119     }
2120 
2121     if (is_nvdimm) {
2122         nvdimm_plug(ms->nvdimms_state);
2123     }
2124 
2125     hotplug_handler_plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &error_abort);
2126 out:
2127     error_propagate(errp, local_err);
2128 }
2129 
2130 static void pc_memory_unplug_request(HotplugHandler *hotplug_dev,
2131                                      DeviceState *dev, Error **errp)
2132 {
2133     Error *local_err = NULL;
2134     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2135 
2136     /*
2137      * When -no-acpi is used with Q35 machine type, no ACPI is built,
2138      * but pcms->acpi_dev is still created. Check !acpi_enabled in
2139      * addition to cover this case.
2140      */
2141     if (!pcms->acpi_dev || !acpi_enabled) {
2142         error_setg(&local_err,
2143                    "memory hotplug is not enabled: missing acpi device or acpi disabled");
2144         goto out;
2145     }
2146 
2147     if (object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM)) {
2148         error_setg(&local_err,
2149                    "nvdimm device hot unplug is not supported yet.");
2150         goto out;
2151     }
2152 
2153     hotplug_handler_unplug_request(HOTPLUG_HANDLER(pcms->acpi_dev), dev,
2154                                    &local_err);
2155 out:
2156     error_propagate(errp, local_err);
2157 }
2158 
2159 static void pc_memory_unplug(HotplugHandler *hotplug_dev,
2160                              DeviceState *dev, Error **errp)
2161 {
2162     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2163     Error *local_err = NULL;
2164 
2165     hotplug_handler_unplug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
2166     if (local_err) {
2167         goto out;
2168     }
2169 
2170     pc_dimm_unplug(PC_DIMM(dev), MACHINE(pcms));
2171     object_property_set_bool(OBJECT(dev), false, "realized", NULL);
2172  out:
2173     error_propagate(errp, local_err);
2174 }
2175 
2176 static int pc_apic_cmp(const void *a, const void *b)
2177 {
2178    CPUArchId *apic_a = (CPUArchId *)a;
2179    CPUArchId *apic_b = (CPUArchId *)b;
2180 
2181    return apic_a->arch_id - apic_b->arch_id;
2182 }
2183 
2184 /* returns pointer to CPUArchId descriptor that matches CPU's apic_id
2185  * in ms->possible_cpus->cpus, if ms->possible_cpus->cpus has no
2186  * entry corresponding to CPU's apic_id returns NULL.
2187  */
2188 static CPUArchId *pc_find_cpu_slot(MachineState *ms, uint32_t id, int *idx)
2189 {
2190     CPUArchId apic_id, *found_cpu;
2191 
2192     apic_id.arch_id = id;
2193     found_cpu = bsearch(&apic_id, ms->possible_cpus->cpus,
2194         ms->possible_cpus->len, sizeof(*ms->possible_cpus->cpus),
2195         pc_apic_cmp);
2196     if (found_cpu && idx) {
2197         *idx = found_cpu - ms->possible_cpus->cpus;
2198     }
2199     return found_cpu;
2200 }
2201 
2202 static void pc_cpu_plug(HotplugHandler *hotplug_dev,
2203                         DeviceState *dev, Error **errp)
2204 {
2205     CPUArchId *found_cpu;
2206     Error *local_err = NULL;
2207     X86CPU *cpu = X86_CPU(dev);
2208     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2209 
2210     if (pcms->acpi_dev) {
2211         hotplug_handler_plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
2212         if (local_err) {
2213             goto out;
2214         }
2215     }
2216 
2217     /* increment the number of CPUs */
2218     pcms->boot_cpus++;
2219     if (pcms->rtc) {
2220         rtc_set_cpus_count(pcms->rtc, pcms->boot_cpus);
2221     }
2222     if (pcms->fw_cfg) {
2223         fw_cfg_modify_i16(pcms->fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
2224     }
2225 
2226     found_cpu = pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, NULL);
2227     found_cpu->cpu = OBJECT(dev);
2228 out:
2229     error_propagate(errp, local_err);
2230 }
2231 static void pc_cpu_unplug_request_cb(HotplugHandler *hotplug_dev,
2232                                      DeviceState *dev, Error **errp)
2233 {
2234     int idx = -1;
2235     Error *local_err = NULL;
2236     X86CPU *cpu = X86_CPU(dev);
2237     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2238 
2239     if (!pcms->acpi_dev) {
2240         error_setg(&local_err, "CPU hot unplug not supported without ACPI");
2241         goto out;
2242     }
2243 
2244     pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, &idx);
2245     assert(idx != -1);
2246     if (idx == 0) {
2247         error_setg(&local_err, "Boot CPU is unpluggable");
2248         goto out;
2249     }
2250 
2251     hotplug_handler_unplug_request(HOTPLUG_HANDLER(pcms->acpi_dev), dev,
2252                                    &local_err);
2253     if (local_err) {
2254         goto out;
2255     }
2256 
2257  out:
2258     error_propagate(errp, local_err);
2259 
2260 }
2261 
2262 static void pc_cpu_unplug_cb(HotplugHandler *hotplug_dev,
2263                              DeviceState *dev, Error **errp)
2264 {
2265     CPUArchId *found_cpu;
2266     Error *local_err = NULL;
2267     X86CPU *cpu = X86_CPU(dev);
2268     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2269 
2270     hotplug_handler_unplug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
2271     if (local_err) {
2272         goto out;
2273     }
2274 
2275     found_cpu = pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, NULL);
2276     found_cpu->cpu = NULL;
2277     object_property_set_bool(OBJECT(dev), false, "realized", NULL);
2278 
2279     /* decrement the number of CPUs */
2280     pcms->boot_cpus--;
2281     /* Update the number of CPUs in CMOS */
2282     rtc_set_cpus_count(pcms->rtc, pcms->boot_cpus);
2283     fw_cfg_modify_i16(pcms->fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
2284  out:
2285     error_propagate(errp, local_err);
2286 }
2287 
2288 static void pc_cpu_pre_plug(HotplugHandler *hotplug_dev,
2289                             DeviceState *dev, Error **errp)
2290 {
2291     int idx;
2292     CPUState *cs;
2293     CPUArchId *cpu_slot;
2294     X86CPUTopoInfo topo;
2295     X86CPU *cpu = X86_CPU(dev);
2296     MachineState *ms = MACHINE(hotplug_dev);
2297     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2298 
2299     if(!object_dynamic_cast(OBJECT(cpu), ms->cpu_type)) {
2300         error_setg(errp, "Invalid CPU type, expected cpu type: '%s'",
2301                    ms->cpu_type);
2302         return;
2303     }
2304 
2305     /* if APIC ID is not set, set it based on socket/core/thread properties */
2306     if (cpu->apic_id == UNASSIGNED_APIC_ID) {
2307         int max_socket = (max_cpus - 1) / smp_threads / smp_cores;
2308 
2309         if (cpu->socket_id < 0) {
2310             error_setg(errp, "CPU socket-id is not set");
2311             return;
2312         } else if (cpu->socket_id > max_socket) {
2313             error_setg(errp, "Invalid CPU socket-id: %u must be in range 0:%u",
2314                        cpu->socket_id, max_socket);
2315             return;
2316         }
2317         if (cpu->core_id < 0) {
2318             error_setg(errp, "CPU core-id is not set");
2319             return;
2320         } else if (cpu->core_id > (smp_cores - 1)) {
2321             error_setg(errp, "Invalid CPU core-id: %u must be in range 0:%u",
2322                        cpu->core_id, smp_cores - 1);
2323             return;
2324         }
2325         if (cpu->thread_id < 0) {
2326             error_setg(errp, "CPU thread-id is not set");
2327             return;
2328         } else if (cpu->thread_id > (smp_threads - 1)) {
2329             error_setg(errp, "Invalid CPU thread-id: %u must be in range 0:%u",
2330                        cpu->thread_id, smp_threads - 1);
2331             return;
2332         }
2333 
2334         topo.pkg_id = cpu->socket_id;
2335         topo.core_id = cpu->core_id;
2336         topo.smt_id = cpu->thread_id;
2337         cpu->apic_id = apicid_from_topo_ids(smp_cores, smp_threads, &topo);
2338     }
2339 
2340     cpu_slot = pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, &idx);
2341     if (!cpu_slot) {
2342         MachineState *ms = MACHINE(pcms);
2343 
2344         x86_topo_ids_from_apicid(cpu->apic_id, smp_cores, smp_threads, &topo);
2345         error_setg(errp, "Invalid CPU [socket: %u, core: %u, thread: %u] with"
2346                   " APIC ID %" PRIu32 ", valid index range 0:%d",
2347                    topo.pkg_id, topo.core_id, topo.smt_id, cpu->apic_id,
2348                    ms->possible_cpus->len - 1);
2349         return;
2350     }
2351 
2352     if (cpu_slot->cpu) {
2353         error_setg(errp, "CPU[%d] with APIC ID %" PRIu32 " exists",
2354                    idx, cpu->apic_id);
2355         return;
2356     }
2357 
2358     /* if 'address' properties socket-id/core-id/thread-id are not set, set them
2359      * so that machine_query_hotpluggable_cpus would show correct values
2360      */
2361     /* TODO: move socket_id/core_id/thread_id checks into x86_cpu_realizefn()
2362      * once -smp refactoring is complete and there will be CPU private
2363      * CPUState::nr_cores and CPUState::nr_threads fields instead of globals */
2364     x86_topo_ids_from_apicid(cpu->apic_id, smp_cores, smp_threads, &topo);
2365     if (cpu->socket_id != -1 && cpu->socket_id != topo.pkg_id) {
2366         error_setg(errp, "property socket-id: %u doesn't match set apic-id:"
2367             " 0x%x (socket-id: %u)", cpu->socket_id, cpu->apic_id, topo.pkg_id);
2368         return;
2369     }
2370     cpu->socket_id = topo.pkg_id;
2371 
2372     if (cpu->core_id != -1 && cpu->core_id != topo.core_id) {
2373         error_setg(errp, "property core-id: %u doesn't match set apic-id:"
2374             " 0x%x (core-id: %u)", cpu->core_id, cpu->apic_id, topo.core_id);
2375         return;
2376     }
2377     cpu->core_id = topo.core_id;
2378 
2379     if (cpu->thread_id != -1 && cpu->thread_id != topo.smt_id) {
2380         error_setg(errp, "property thread-id: %u doesn't match set apic-id:"
2381             " 0x%x (thread-id: %u)", cpu->thread_id, cpu->apic_id, topo.smt_id);
2382         return;
2383     }
2384     cpu->thread_id = topo.smt_id;
2385 
2386     if (hyperv_feat_enabled(cpu, HYPERV_FEAT_VPINDEX) &&
2387         !kvm_hv_vpindex_settable()) {
2388         error_setg(errp, "kernel doesn't allow setting HyperV VP_INDEX");
2389         return;
2390     }
2391 
2392     cs = CPU(cpu);
2393     cs->cpu_index = idx;
2394 
2395     numa_cpu_pre_plug(cpu_slot, dev, errp);
2396 }
2397 
2398 static void pc_machine_device_pre_plug_cb(HotplugHandler *hotplug_dev,
2399                                           DeviceState *dev, Error **errp)
2400 {
2401     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2402         pc_memory_pre_plug(hotplug_dev, dev, errp);
2403     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2404         pc_cpu_pre_plug(hotplug_dev, dev, errp);
2405     }
2406 }
2407 
2408 static void pc_machine_device_plug_cb(HotplugHandler *hotplug_dev,
2409                                       DeviceState *dev, Error **errp)
2410 {
2411     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2412         pc_memory_plug(hotplug_dev, dev, errp);
2413     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2414         pc_cpu_plug(hotplug_dev, dev, errp);
2415     }
2416 }
2417 
2418 static void pc_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
2419                                                 DeviceState *dev, Error **errp)
2420 {
2421     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2422         pc_memory_unplug_request(hotplug_dev, dev, errp);
2423     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2424         pc_cpu_unplug_request_cb(hotplug_dev, dev, errp);
2425     } else {
2426         error_setg(errp, "acpi: device unplug request for not supported device"
2427                    " type: %s", object_get_typename(OBJECT(dev)));
2428     }
2429 }
2430 
2431 static void pc_machine_device_unplug_cb(HotplugHandler *hotplug_dev,
2432                                         DeviceState *dev, Error **errp)
2433 {
2434     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2435         pc_memory_unplug(hotplug_dev, dev, errp);
2436     } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2437         pc_cpu_unplug_cb(hotplug_dev, dev, errp);
2438     } else {
2439         error_setg(errp, "acpi: device unplug for not supported device"
2440                    " type: %s", object_get_typename(OBJECT(dev)));
2441     }
2442 }
2443 
2444 static HotplugHandler *pc_get_hotplug_handler(MachineState *machine,
2445                                              DeviceState *dev)
2446 {
2447     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) ||
2448         object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
2449         return HOTPLUG_HANDLER(machine);
2450     }
2451 
2452     return NULL;
2453 }
2454 
2455 static void
2456 pc_machine_get_device_memory_region_size(Object *obj, Visitor *v,
2457                                          const char *name, void *opaque,
2458                                          Error **errp)
2459 {
2460     MachineState *ms = MACHINE(obj);
2461     int64_t value = memory_region_size(&ms->device_memory->mr);
2462 
2463     visit_type_int(v, name, &value, errp);
2464 }
2465 
2466 static void pc_machine_get_max_ram_below_4g(Object *obj, Visitor *v,
2467                                             const char *name, void *opaque,
2468                                             Error **errp)
2469 {
2470     PCMachineState *pcms = PC_MACHINE(obj);
2471     uint64_t value = pcms->max_ram_below_4g;
2472 
2473     visit_type_size(v, name, &value, errp);
2474 }
2475 
2476 static void pc_machine_set_max_ram_below_4g(Object *obj, Visitor *v,
2477                                             const char *name, void *opaque,
2478                                             Error **errp)
2479 {
2480     PCMachineState *pcms = PC_MACHINE(obj);
2481     Error *error = NULL;
2482     uint64_t value;
2483 
2484     visit_type_size(v, name, &value, &error);
2485     if (error) {
2486         error_propagate(errp, error);
2487         return;
2488     }
2489     if (value > 4 * GiB) {
2490         error_setg(&error,
2491                    "Machine option 'max-ram-below-4g=%"PRIu64
2492                    "' expects size less than or equal to 4G", value);
2493         error_propagate(errp, error);
2494         return;
2495     }
2496 
2497     if (value < 1 * MiB) {
2498         warn_report("Only %" PRIu64 " bytes of RAM below the 4GiB boundary,"
2499                     "BIOS may not work with less than 1MiB", value);
2500     }
2501 
2502     pcms->max_ram_below_4g = value;
2503 }
2504 
2505 static void pc_machine_get_vmport(Object *obj, Visitor *v, const char *name,
2506                                   void *opaque, Error **errp)
2507 {
2508     PCMachineState *pcms = PC_MACHINE(obj);
2509     OnOffAuto vmport = pcms->vmport;
2510 
2511     visit_type_OnOffAuto(v, name, &vmport, errp);
2512 }
2513 
2514 static void pc_machine_set_vmport(Object *obj, Visitor *v, const char *name,
2515                                   void *opaque, Error **errp)
2516 {
2517     PCMachineState *pcms = PC_MACHINE(obj);
2518 
2519     visit_type_OnOffAuto(v, name, &pcms->vmport, errp);
2520 }
2521 
2522 bool pc_machine_is_smm_enabled(PCMachineState *pcms)
2523 {
2524     bool smm_available = false;
2525 
2526     if (pcms->smm == ON_OFF_AUTO_OFF) {
2527         return false;
2528     }
2529 
2530     if (tcg_enabled() || qtest_enabled()) {
2531         smm_available = true;
2532     } else if (kvm_enabled()) {
2533         smm_available = kvm_has_smm();
2534     }
2535 
2536     if (smm_available) {
2537         return true;
2538     }
2539 
2540     if (pcms->smm == ON_OFF_AUTO_ON) {
2541         error_report("System Management Mode not supported by this hypervisor.");
2542         exit(1);
2543     }
2544     return false;
2545 }
2546 
2547 static void pc_machine_get_smm(Object *obj, Visitor *v, const char *name,
2548                                void *opaque, Error **errp)
2549 {
2550     PCMachineState *pcms = PC_MACHINE(obj);
2551     OnOffAuto smm = pcms->smm;
2552 
2553     visit_type_OnOffAuto(v, name, &smm, errp);
2554 }
2555 
2556 static void pc_machine_set_smm(Object *obj, Visitor *v, const char *name,
2557                                void *opaque, Error **errp)
2558 {
2559     PCMachineState *pcms = PC_MACHINE(obj);
2560 
2561     visit_type_OnOffAuto(v, name, &pcms->smm, errp);
2562 }
2563 
2564 static bool pc_machine_get_smbus(Object *obj, Error **errp)
2565 {
2566     PCMachineState *pcms = PC_MACHINE(obj);
2567 
2568     return pcms->smbus_enabled;
2569 }
2570 
2571 static void pc_machine_set_smbus(Object *obj, bool value, Error **errp)
2572 {
2573     PCMachineState *pcms = PC_MACHINE(obj);
2574 
2575     pcms->smbus_enabled = value;
2576 }
2577 
2578 static bool pc_machine_get_sata(Object *obj, Error **errp)
2579 {
2580     PCMachineState *pcms = PC_MACHINE(obj);
2581 
2582     return pcms->sata_enabled;
2583 }
2584 
2585 static void pc_machine_set_sata(Object *obj, bool value, Error **errp)
2586 {
2587     PCMachineState *pcms = PC_MACHINE(obj);
2588 
2589     pcms->sata_enabled = value;
2590 }
2591 
2592 static bool pc_machine_get_pit(Object *obj, Error **errp)
2593 {
2594     PCMachineState *pcms = PC_MACHINE(obj);
2595 
2596     return pcms->pit_enabled;
2597 }
2598 
2599 static void pc_machine_set_pit(Object *obj, bool value, Error **errp)
2600 {
2601     PCMachineState *pcms = PC_MACHINE(obj);
2602 
2603     pcms->pit_enabled = value;
2604 }
2605 
2606 static void pc_machine_initfn(Object *obj)
2607 {
2608     PCMachineState *pcms = PC_MACHINE(obj);
2609 
2610     pcms->max_ram_below_4g = 0; /* use default */
2611     pcms->smm = ON_OFF_AUTO_AUTO;
2612     pcms->vmport = ON_OFF_AUTO_AUTO;
2613     /* acpi build is enabled by default if machine supports it */
2614     pcms->acpi_build_enabled = PC_MACHINE_GET_CLASS(pcms)->has_acpi_build;
2615     pcms->smbus_enabled = true;
2616     pcms->sata_enabled = true;
2617     pcms->pit_enabled = true;
2618 
2619     pc_system_flash_create(pcms);
2620 }
2621 
2622 static void pc_machine_reset(void)
2623 {
2624     CPUState *cs;
2625     X86CPU *cpu;
2626 
2627     qemu_devices_reset();
2628 
2629     /* Reset APIC after devices have been reset to cancel
2630      * any changes that qemu_devices_reset() might have done.
2631      */
2632     CPU_FOREACH(cs) {
2633         cpu = X86_CPU(cs);
2634 
2635         if (cpu->apic_state) {
2636             device_reset(cpu->apic_state);
2637         }
2638     }
2639 }
2640 
2641 static CpuInstanceProperties
2642 pc_cpu_index_to_props(MachineState *ms, unsigned cpu_index)
2643 {
2644     MachineClass *mc = MACHINE_GET_CLASS(ms);
2645     const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(ms);
2646 
2647     assert(cpu_index < possible_cpus->len);
2648     return possible_cpus->cpus[cpu_index].props;
2649 }
2650 
2651 static int64_t pc_get_default_cpu_node_id(const MachineState *ms, int idx)
2652 {
2653    X86CPUTopoInfo topo;
2654 
2655    assert(idx < ms->possible_cpus->len);
2656    x86_topo_ids_from_apicid(ms->possible_cpus->cpus[idx].arch_id,
2657                             smp_cores, smp_threads, &topo);
2658    return topo.pkg_id % nb_numa_nodes;
2659 }
2660 
2661 static const CPUArchIdList *pc_possible_cpu_arch_ids(MachineState *ms)
2662 {
2663     int i;
2664 
2665     if (ms->possible_cpus) {
2666         /*
2667          * make sure that max_cpus hasn't changed since the first use, i.e.
2668          * -smp hasn't been parsed after it
2669         */
2670         assert(ms->possible_cpus->len == max_cpus);
2671         return ms->possible_cpus;
2672     }
2673 
2674     ms->possible_cpus = g_malloc0(sizeof(CPUArchIdList) +
2675                                   sizeof(CPUArchId) * max_cpus);
2676     ms->possible_cpus->len = max_cpus;
2677     for (i = 0; i < ms->possible_cpus->len; i++) {
2678         X86CPUTopoInfo topo;
2679 
2680         ms->possible_cpus->cpus[i].type = ms->cpu_type;
2681         ms->possible_cpus->cpus[i].vcpus_count = 1;
2682         ms->possible_cpus->cpus[i].arch_id = x86_cpu_apic_id_from_index(i);
2683         x86_topo_ids_from_apicid(ms->possible_cpus->cpus[i].arch_id,
2684                                  smp_cores, smp_threads, &topo);
2685         ms->possible_cpus->cpus[i].props.has_socket_id = true;
2686         ms->possible_cpus->cpus[i].props.socket_id = topo.pkg_id;
2687         ms->possible_cpus->cpus[i].props.has_core_id = true;
2688         ms->possible_cpus->cpus[i].props.core_id = topo.core_id;
2689         ms->possible_cpus->cpus[i].props.has_thread_id = true;
2690         ms->possible_cpus->cpus[i].props.thread_id = topo.smt_id;
2691     }
2692     return ms->possible_cpus;
2693 }
2694 
2695 static void x86_nmi(NMIState *n, int cpu_index, Error **errp)
2696 {
2697     /* cpu index isn't used */
2698     CPUState *cs;
2699 
2700     CPU_FOREACH(cs) {
2701         X86CPU *cpu = X86_CPU(cs);
2702 
2703         if (!cpu->apic_state) {
2704             cpu_interrupt(cs, CPU_INTERRUPT_NMI);
2705         } else {
2706             apic_deliver_nmi(cpu->apic_state);
2707         }
2708     }
2709 }
2710 
2711 static void pc_machine_class_init(ObjectClass *oc, void *data)
2712 {
2713     MachineClass *mc = MACHINE_CLASS(oc);
2714     PCMachineClass *pcmc = PC_MACHINE_CLASS(oc);
2715     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
2716     NMIClass *nc = NMI_CLASS(oc);
2717 
2718     pcmc->pci_enabled = true;
2719     pcmc->has_acpi_build = true;
2720     pcmc->rsdp_in_ram = true;
2721     pcmc->smbios_defaults = true;
2722     pcmc->smbios_uuid_encoded = true;
2723     pcmc->gigabyte_align = true;
2724     pcmc->has_reserved_memory = true;
2725     pcmc->kvmclock_enabled = true;
2726     pcmc->enforce_aligned_dimm = true;
2727     /* BIOS ACPI tables: 128K. Other BIOS datastructures: less than 4K reported
2728      * to be used at the moment, 32K should be enough for a while.  */
2729     pcmc->acpi_data_size = 0x20000 + 0x8000;
2730     pcmc->save_tsc_khz = true;
2731     pcmc->linuxboot_dma_enabled = true;
2732     pcmc->pvh_enabled = true;
2733     assert(!mc->get_hotplug_handler);
2734     mc->get_hotplug_handler = pc_get_hotplug_handler;
2735     mc->cpu_index_to_instance_props = pc_cpu_index_to_props;
2736     mc->get_default_cpu_node_id = pc_get_default_cpu_node_id;
2737     mc->possible_cpu_arch_ids = pc_possible_cpu_arch_ids;
2738     mc->auto_enable_numa_with_memhp = true;
2739     mc->has_hotpluggable_cpus = true;
2740     mc->default_boot_order = "cad";
2741     mc->hot_add_cpu = pc_hot_add_cpu;
2742     mc->block_default_type = IF_IDE;
2743     mc->max_cpus = 255;
2744     mc->reset = pc_machine_reset;
2745     hc->pre_plug = pc_machine_device_pre_plug_cb;
2746     hc->plug = pc_machine_device_plug_cb;
2747     hc->unplug_request = pc_machine_device_unplug_request_cb;
2748     hc->unplug = pc_machine_device_unplug_cb;
2749     nc->nmi_monitor_handler = x86_nmi;
2750     mc->default_cpu_type = TARGET_DEFAULT_CPU_TYPE;
2751     mc->nvdimm_supported = true;
2752 
2753     object_class_property_add(oc, PC_MACHINE_DEVMEM_REGION_SIZE, "int",
2754         pc_machine_get_device_memory_region_size, NULL,
2755         NULL, NULL, &error_abort);
2756 
2757     object_class_property_add(oc, PC_MACHINE_MAX_RAM_BELOW_4G, "size",
2758         pc_machine_get_max_ram_below_4g, pc_machine_set_max_ram_below_4g,
2759         NULL, NULL, &error_abort);
2760 
2761     object_class_property_set_description(oc, PC_MACHINE_MAX_RAM_BELOW_4G,
2762         "Maximum ram below the 4G boundary (32bit boundary)", &error_abort);
2763 
2764     object_class_property_add(oc, PC_MACHINE_SMM, "OnOffAuto",
2765         pc_machine_get_smm, pc_machine_set_smm,
2766         NULL, NULL, &error_abort);
2767     object_class_property_set_description(oc, PC_MACHINE_SMM,
2768         "Enable SMM (pc & q35)", &error_abort);
2769 
2770     object_class_property_add(oc, PC_MACHINE_VMPORT, "OnOffAuto",
2771         pc_machine_get_vmport, pc_machine_set_vmport,
2772         NULL, NULL, &error_abort);
2773     object_class_property_set_description(oc, PC_MACHINE_VMPORT,
2774         "Enable vmport (pc & q35)", &error_abort);
2775 
2776     object_class_property_add_bool(oc, PC_MACHINE_SMBUS,
2777         pc_machine_get_smbus, pc_machine_set_smbus, &error_abort);
2778 
2779     object_class_property_add_bool(oc, PC_MACHINE_SATA,
2780         pc_machine_get_sata, pc_machine_set_sata, &error_abort);
2781 
2782     object_class_property_add_bool(oc, PC_MACHINE_PIT,
2783         pc_machine_get_pit, pc_machine_set_pit, &error_abort);
2784 }
2785 
2786 static const TypeInfo pc_machine_info = {
2787     .name = TYPE_PC_MACHINE,
2788     .parent = TYPE_MACHINE,
2789     .abstract = true,
2790     .instance_size = sizeof(PCMachineState),
2791     .instance_init = pc_machine_initfn,
2792     .class_size = sizeof(PCMachineClass),
2793     .class_init = pc_machine_class_init,
2794     .interfaces = (InterfaceInfo[]) {
2795          { TYPE_HOTPLUG_HANDLER },
2796          { TYPE_NMI },
2797          { }
2798     },
2799 };
2800 
2801 static void pc_machine_register_types(void)
2802 {
2803     type_register_static(&pc_machine_info);
2804 }
2805 
2806 type_init(pc_machine_register_types)
2807