xref: /openbmc/qemu/hw/i386/pc.c (revision 679cb8e1)
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 "sysemu/cpus.h"
34 #include "hw/block/fdc.h"
35 #include "hw/ide.h"
36 #include "hw/pci/pci.h"
37 #include "hw/pci/pci_bus.h"
38 #include "hw/nvram/fw_cfg.h"
39 #include "hw/timer/hpet.h"
40 #include "hw/firmware/smbios.h"
41 #include "hw/loader.h"
42 #include "elf.h"
43 #include "multiboot.h"
44 #include "hw/timer/mc146818rtc.h"
45 #include "hw/dma/i8257.h"
46 #include "hw/timer/i8254.h"
47 #include "hw/input/i8042.h"
48 #include "hw/audio/pcspk.h"
49 #include "hw/pci/msi.h"
50 #include "hw/sysbus.h"
51 #include "sysemu/sysemu.h"
52 #include "sysemu/numa.h"
53 #include "sysemu/kvm.h"
54 #include "sysemu/qtest.h"
55 #include "kvm_i386.h"
56 #include "hw/xen/xen.h"
57 #include "hw/xen/start_info.h"
58 #include "ui/qemu-spice.h"
59 #include "exec/memory.h"
60 #include "exec/address-spaces.h"
61 #include "sysemu/arch_init.h"
62 #include "qemu/bitmap.h"
63 #include "qemu/config-file.h"
64 #include "qemu/error-report.h"
65 #include "qemu/option.h"
66 #include "hw/acpi/acpi.h"
67 #include "hw/acpi/cpu_hotplug.h"
68 #include "hw/boards.h"
69 #include "acpi-build.h"
70 #include "hw/mem/pc-dimm.h"
71 #include "qapi/error.h"
72 #include "qapi/qapi-visit-common.h"
73 #include "qapi/visitor.h"
74 #include "qom/cpu.h"
75 #include "hw/nmi.h"
76 #include "hw/usb.h"
77 #include "hw/i386/intel_iommu.h"
78 #include "hw/net/ne2000-isa.h"
79 #include "standard-headers/asm-x86/bootparam.h"
80 
81 /* debug PC/ISA interrupts */
82 //#define DEBUG_IRQ
83 
84 #ifdef DEBUG_IRQ
85 #define DPRINTF(fmt, ...)                                       \
86     do { printf("CPUIRQ: " fmt , ## __VA_ARGS__); } while (0)
87 #else
88 #define DPRINTF(fmt, ...)
89 #endif
90 
91 #define FW_CFG_ACPI_TABLES (FW_CFG_ARCH_LOCAL + 0)
92 #define FW_CFG_SMBIOS_ENTRIES (FW_CFG_ARCH_LOCAL + 1)
93 #define FW_CFG_IRQ0_OVERRIDE (FW_CFG_ARCH_LOCAL + 2)
94 #define FW_CFG_E820_TABLE (FW_CFG_ARCH_LOCAL + 3)
95 #define FW_CFG_HPET (FW_CFG_ARCH_LOCAL + 4)
96 
97 #define E820_NR_ENTRIES		16
98 
99 struct e820_entry {
100     uint64_t address;
101     uint64_t length;
102     uint32_t type;
103 } QEMU_PACKED __attribute((__aligned__(4)));
104 
105 struct e820_table {
106     uint32_t count;
107     struct e820_entry entry[E820_NR_ENTRIES];
108 } QEMU_PACKED __attribute((__aligned__(4)));
109 
110 static struct e820_table e820_reserve;
111 static struct e820_entry *e820_table;
112 static unsigned e820_entries;
113 struct hpet_fw_config hpet_cfg = {.count = UINT8_MAX};
114 
115 /* Physical Address of PVH entry point read from kernel ELF NOTE */
116 static size_t pvh_start_addr;
117 
118 GlobalProperty pc_compat_3_1[] = {
119     { "intel-iommu", "dma-drain", "off" },
120     { "Opteron_G3" "-" TYPE_X86_CPU, "rdtscp", "off" },
121     { "Opteron_G4" "-" TYPE_X86_CPU, "rdtscp", "off" },
122     { "Opteron_G4" "-" TYPE_X86_CPU, "npt", "off" },
123     { "Opteron_G4" "-" TYPE_X86_CPU, "nrip-save", "off" },
124     { "Opteron_G5" "-" TYPE_X86_CPU, "rdtscp", "off" },
125     { "Opteron_G5" "-" TYPE_X86_CPU, "npt", "off" },
126     { "Opteron_G5" "-" TYPE_X86_CPU, "nrip-save", "off" },
127     { "EPYC" "-" TYPE_X86_CPU, "npt", "off" },
128     { "EPYC" "-" TYPE_X86_CPU, "nrip-save", "off" },
129     { "EPYC-IBPB" "-" TYPE_X86_CPU, "npt", "off" },
130     { "EPYC-IBPB" "-" TYPE_X86_CPU, "nrip-save", "off" },
131     { "Skylake-Client" "-" TYPE_X86_CPU,      "mpx", "on" },
132     { "Skylake-Client-IBRS" "-" TYPE_X86_CPU, "mpx", "on" },
133     { "Skylake-Server" "-" TYPE_X86_CPU,      "mpx", "on" },
134     { "Skylake-Server-IBRS" "-" TYPE_X86_CPU, "mpx", "on" },
135     { "Cascadelake-Server" "-" TYPE_X86_CPU,  "mpx", "on" },
136     { "Icelake-Client" "-" TYPE_X86_CPU,      "mpx", "on" },
137     { "Icelake-Server" "-" TYPE_X86_CPU,      "mpx", "on" },
138     { "Cascadelake-Server" "-" TYPE_X86_CPU, "stepping", "5" },
139     { TYPE_X86_CPU, "x-intel-pt-auto-level", "off" },
140 };
141 const size_t pc_compat_3_1_len = G_N_ELEMENTS(pc_compat_3_1);
142 
143 GlobalProperty pc_compat_3_0[] = {
144     { TYPE_X86_CPU, "x-hv-synic-kvm-only", "on" },
145     { "Skylake-Server" "-" TYPE_X86_CPU, "pku", "off" },
146     { "Skylake-Server-IBRS" "-" TYPE_X86_CPU, "pku", "off" },
147 };
148 const size_t pc_compat_3_0_len = G_N_ELEMENTS(pc_compat_3_0);
149 
150 GlobalProperty pc_compat_2_12[] = {
151     { TYPE_X86_CPU, "legacy-cache", "on" },
152     { TYPE_X86_CPU, "topoext", "off" },
153     { "EPYC-" TYPE_X86_CPU, "xlevel", "0x8000000a" },
154     { "EPYC-IBPB-" TYPE_X86_CPU, "xlevel", "0x8000000a" },
155 };
156 const size_t pc_compat_2_12_len = G_N_ELEMENTS(pc_compat_2_12);
157 
158 GlobalProperty pc_compat_2_11[] = {
159     { TYPE_X86_CPU, "x-migrate-smi-count", "off" },
160     { "Skylake-Server" "-" TYPE_X86_CPU, "clflushopt", "off" },
161 };
162 const size_t pc_compat_2_11_len = G_N_ELEMENTS(pc_compat_2_11);
163 
164 GlobalProperty pc_compat_2_10[] = {
165     { TYPE_X86_CPU, "x-hv-max-vps", "0x40" },
166     { "i440FX-pcihost", "x-pci-hole64-fix", "off" },
167     { "q35-pcihost", "x-pci-hole64-fix", "off" },
168 };
169 const size_t pc_compat_2_10_len = G_N_ELEMENTS(pc_compat_2_10);
170 
171 GlobalProperty pc_compat_2_9[] = {
172     { "mch", "extended-tseg-mbytes", "0" },
173 };
174 const size_t pc_compat_2_9_len = G_N_ELEMENTS(pc_compat_2_9);
175 
176 GlobalProperty pc_compat_2_8[] = {
177     { TYPE_X86_CPU, "tcg-cpuid", "off" },
178     { "kvmclock", "x-mach-use-reliable-get-clock", "off" },
179     { "ICH9-LPC", "x-smi-broadcast", "off" },
180     { TYPE_X86_CPU, "vmware-cpuid-freq", "off" },
181     { "Haswell-" TYPE_X86_CPU, "stepping", "1" },
182 };
183 const size_t pc_compat_2_8_len = G_N_ELEMENTS(pc_compat_2_8);
184 
185 GlobalProperty pc_compat_2_7[] = {
186     { TYPE_X86_CPU, "l3-cache", "off" },
187     { TYPE_X86_CPU, "full-cpuid-auto-level", "off" },
188     { "Opteron_G3" "-" TYPE_X86_CPU, "family", "15" },
189     { "Opteron_G3" "-" TYPE_X86_CPU, "model", "6" },
190     { "Opteron_G3" "-" TYPE_X86_CPU, "stepping", "1" },
191     { "isa-pcspk", "migrate", "off" },
192 };
193 const size_t pc_compat_2_7_len = G_N_ELEMENTS(pc_compat_2_7);
194 
195 GlobalProperty pc_compat_2_6[] = {
196     { TYPE_X86_CPU, "cpuid-0xb", "off" },
197     { "vmxnet3", "romfile", "" },
198     { TYPE_X86_CPU, "fill-mtrr-mask", "off" },
199     { "apic-common", "legacy-instance-id", "on", }
200 };
201 const size_t pc_compat_2_6_len = G_N_ELEMENTS(pc_compat_2_6);
202 
203 GlobalProperty pc_compat_2_5[] = {};
204 const size_t pc_compat_2_5_len = G_N_ELEMENTS(pc_compat_2_5);
205 
206 GlobalProperty pc_compat_2_4[] = {
207     PC_CPU_MODEL_IDS("2.4.0")
208     { "Haswell-" TYPE_X86_CPU, "abm", "off" },
209     { "Haswell-noTSX-" TYPE_X86_CPU, "abm", "off" },
210     { "Broadwell-" TYPE_X86_CPU, "abm", "off" },
211     { "Broadwell-noTSX-" TYPE_X86_CPU, "abm", "off" },
212     { "host" "-" TYPE_X86_CPU, "host-cache-info", "on" },
213     { TYPE_X86_CPU, "check", "off" },
214     { "qemu64" "-" TYPE_X86_CPU, "sse4a", "on" },
215     { "qemu64" "-" TYPE_X86_CPU, "abm", "on" },
216     { "qemu64" "-" TYPE_X86_CPU, "popcnt", "on" },
217     { "qemu32" "-" TYPE_X86_CPU, "popcnt", "on" },
218     { "Opteron_G2" "-" TYPE_X86_CPU, "rdtscp", "on" },
219     { "Opteron_G3" "-" TYPE_X86_CPU, "rdtscp", "on" },
220     { "Opteron_G4" "-" TYPE_X86_CPU, "rdtscp", "on" },
221     { "Opteron_G5" "-" TYPE_X86_CPU, "rdtscp", "on", }
222 };
223 const size_t pc_compat_2_4_len = G_N_ELEMENTS(pc_compat_2_4);
224 
225 GlobalProperty pc_compat_2_3[] = {
226     PC_CPU_MODEL_IDS("2.3.0")
227     { TYPE_X86_CPU, "arat", "off" },
228     { "qemu64" "-" TYPE_X86_CPU, "min-level", "4" },
229     { "kvm64" "-" TYPE_X86_CPU, "min-level", "5" },
230     { "pentium3" "-" TYPE_X86_CPU, "min-level", "2" },
231     { "n270" "-" TYPE_X86_CPU, "min-level", "5" },
232     { "Conroe" "-" TYPE_X86_CPU, "min-level", "4" },
233     { "Penryn" "-" TYPE_X86_CPU, "min-level", "4" },
234     { "Nehalem" "-" TYPE_X86_CPU, "min-level", "4" },
235     { "n270" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
236     { "Penryn" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
237     { "Conroe" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
238     { "Nehalem" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
239     { "Westmere" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
240     { "SandyBridge" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
241     { "IvyBridge" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
242     { "Haswell" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
243     { "Haswell-noTSX" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
244     { "Broadwell" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
245     { "Broadwell-noTSX" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
246     { TYPE_X86_CPU, "kvm-no-smi-migration", "on" },
247 };
248 const size_t pc_compat_2_3_len = G_N_ELEMENTS(pc_compat_2_3);
249 
250 GlobalProperty pc_compat_2_2[] = {
251     PC_CPU_MODEL_IDS("2.2.0")
252     { "kvm64" "-" TYPE_X86_CPU, "vme", "off" },
253     { "kvm32" "-" TYPE_X86_CPU, "vme", "off" },
254     { "Conroe" "-" TYPE_X86_CPU, "vme", "off" },
255     { "Penryn" "-" TYPE_X86_CPU, "vme", "off" },
256     { "Nehalem" "-" TYPE_X86_CPU, "vme", "off" },
257     { "Westmere" "-" TYPE_X86_CPU, "vme", "off" },
258     { "SandyBridge" "-" TYPE_X86_CPU, "vme", "off" },
259     { "Haswell" "-" TYPE_X86_CPU, "vme", "off" },
260     { "Broadwell" "-" TYPE_X86_CPU, "vme", "off" },
261     { "Opteron_G1" "-" TYPE_X86_CPU, "vme", "off" },
262     { "Opteron_G2" "-" TYPE_X86_CPU, "vme", "off" },
263     { "Opteron_G3" "-" TYPE_X86_CPU, "vme", "off" },
264     { "Opteron_G4" "-" TYPE_X86_CPU, "vme", "off" },
265     { "Opteron_G5" "-" TYPE_X86_CPU, "vme", "off" },
266     { "Haswell" "-" TYPE_X86_CPU, "f16c", "off" },
267     { "Haswell" "-" TYPE_X86_CPU, "rdrand", "off" },
268     { "Broadwell" "-" TYPE_X86_CPU, "f16c", "off" },
269     { "Broadwell" "-" TYPE_X86_CPU, "rdrand", "off" },
270 };
271 const size_t pc_compat_2_2_len = G_N_ELEMENTS(pc_compat_2_2);
272 
273 GlobalProperty pc_compat_2_1[] = {
274     PC_CPU_MODEL_IDS("2.1.0")
275     { "coreduo" "-" TYPE_X86_CPU, "vmx", "on" },
276     { "core2duo" "-" TYPE_X86_CPU, "vmx", "on" },
277 };
278 const size_t pc_compat_2_1_len = G_N_ELEMENTS(pc_compat_2_1);
279 
280 GlobalProperty pc_compat_2_0[] = {
281     PC_CPU_MODEL_IDS("2.0.0")
282     { "virtio-scsi-pci", "any_layout", "off" },
283     { "PIIX4_PM", "memory-hotplug-support", "off" },
284     { "apic", "version", "0x11" },
285     { "nec-usb-xhci", "superspeed-ports-first", "off" },
286     { "nec-usb-xhci", "force-pcie-endcap", "on" },
287     { "pci-serial", "prog_if", "0" },
288     { "pci-serial-2x", "prog_if", "0" },
289     { "pci-serial-4x", "prog_if", "0" },
290     { "virtio-net-pci", "guest_announce", "off" },
291     { "ICH9-LPC", "memory-hotplug-support", "off" },
292     { "xio3130-downstream", COMPAT_PROP_PCP, "off" },
293     { "ioh3420", COMPAT_PROP_PCP, "off" },
294 };
295 const size_t pc_compat_2_0_len = G_N_ELEMENTS(pc_compat_2_0);
296 
297 GlobalProperty pc_compat_1_7[] = {
298     PC_CPU_MODEL_IDS("1.7.0")
299     { TYPE_USB_DEVICE, "msos-desc", "no" },
300     { "PIIX4_PM", "acpi-pci-hotplug-with-bridge-support", "off" },
301     { "hpet", HPET_INTCAP, "4" },
302 };
303 const size_t pc_compat_1_7_len = G_N_ELEMENTS(pc_compat_1_7);
304 
305 GlobalProperty pc_compat_1_6[] = {
306     PC_CPU_MODEL_IDS("1.6.0")
307     { "e1000", "mitigation", "off" },
308     { "qemu64-" TYPE_X86_CPU, "model", "2" },
309     { "qemu32-" TYPE_X86_CPU, "model", "3" },
310     { "i440FX-pcihost", "short_root_bus", "1" },
311     { "q35-pcihost", "short_root_bus", "1" },
312 };
313 const size_t pc_compat_1_6_len = G_N_ELEMENTS(pc_compat_1_6);
314 
315 GlobalProperty pc_compat_1_5[] = {
316     PC_CPU_MODEL_IDS("1.5.0")
317     { "Conroe-" TYPE_X86_CPU, "model", "2" },
318     { "Conroe-" TYPE_X86_CPU, "min-level", "2" },
319     { "Penryn-" TYPE_X86_CPU, "model", "2" },
320     { "Penryn-" TYPE_X86_CPU, "min-level", "2" },
321     { "Nehalem-" TYPE_X86_CPU, "model", "2" },
322     { "Nehalem-" TYPE_X86_CPU, "min-level", "2" },
323     { "virtio-net-pci", "any_layout", "off" },
324     { TYPE_X86_CPU, "pmu", "on" },
325     { "i440FX-pcihost", "short_root_bus", "0" },
326     { "q35-pcihost", "short_root_bus", "0" },
327 };
328 const size_t pc_compat_1_5_len = G_N_ELEMENTS(pc_compat_1_5);
329 
330 GlobalProperty pc_compat_1_4[] = {
331     PC_CPU_MODEL_IDS("1.4.0")
332     { "scsi-hd", "discard_granularity", "0" },
333     { "scsi-cd", "discard_granularity", "0" },
334     { "scsi-disk", "discard_granularity", "0" },
335     { "ide-hd", "discard_granularity", "0" },
336     { "ide-cd", "discard_granularity", "0" },
337     { "ide-drive", "discard_granularity", "0" },
338     { "virtio-blk-pci", "discard_granularity", "0" },
339     /* DEV_NVECTORS_UNSPECIFIED as a uint32_t string: */
340     { "virtio-serial-pci", "vectors", "0xFFFFFFFF" },
341     { "virtio-net-pci", "ctrl_guest_offloads", "off" },
342     { "e1000", "romfile", "pxe-e1000.rom" },
343     { "ne2k_pci", "romfile", "pxe-ne2k_pci.rom" },
344     { "pcnet", "romfile", "pxe-pcnet.rom" },
345     { "rtl8139", "romfile", "pxe-rtl8139.rom" },
346     { "virtio-net-pci", "romfile", "pxe-virtio.rom" },
347     { "486-" TYPE_X86_CPU, "model", "0" },
348     { "n270" "-" TYPE_X86_CPU, "movbe", "off" },
349     { "Westmere" "-" TYPE_X86_CPU, "pclmulqdq", "off" },
350 };
351 const size_t pc_compat_1_4_len = G_N_ELEMENTS(pc_compat_1_4);
352 
353 void gsi_handler(void *opaque, int n, int level)
354 {
355     GSIState *s = opaque;
356 
357     DPRINTF("pc: %s GSI %d\n", level ? "raising" : "lowering", n);
358     if (n < ISA_NUM_IRQS) {
359         qemu_set_irq(s->i8259_irq[n], level);
360     }
361     qemu_set_irq(s->ioapic_irq[n], level);
362 }
363 
364 static void ioport80_write(void *opaque, hwaddr addr, uint64_t data,
365                            unsigned size)
366 {
367 }
368 
369 static uint64_t ioport80_read(void *opaque, hwaddr addr, unsigned size)
370 {
371     return 0xffffffffffffffffULL;
372 }
373 
374 /* MSDOS compatibility mode FPU exception support */
375 static qemu_irq ferr_irq;
376 
377 void pc_register_ferr_irq(qemu_irq irq)
378 {
379     ferr_irq = irq;
380 }
381 
382 /* XXX: add IGNNE support */
383 void cpu_set_ferr(CPUX86State *s)
384 {
385     qemu_irq_raise(ferr_irq);
386 }
387 
388 static void ioportF0_write(void *opaque, hwaddr addr, uint64_t data,
389                            unsigned size)
390 {
391     qemu_irq_lower(ferr_irq);
392 }
393 
394 static uint64_t ioportF0_read(void *opaque, hwaddr addr, unsigned size)
395 {
396     return 0xffffffffffffffffULL;
397 }
398 
399 /* TSC handling */
400 uint64_t cpu_get_tsc(CPUX86State *env)
401 {
402     return cpu_get_ticks();
403 }
404 
405 /* IRQ handling */
406 int cpu_get_pic_interrupt(CPUX86State *env)
407 {
408     X86CPU *cpu = x86_env_get_cpu(env);
409     int intno;
410 
411     if (!kvm_irqchip_in_kernel()) {
412         intno = apic_get_interrupt(cpu->apic_state);
413         if (intno >= 0) {
414             return intno;
415         }
416         /* read the irq from the PIC */
417         if (!apic_accept_pic_intr(cpu->apic_state)) {
418             return -1;
419         }
420     }
421 
422     intno = pic_read_irq(isa_pic);
423     return intno;
424 }
425 
426 static void pic_irq_request(void *opaque, int irq, int level)
427 {
428     CPUState *cs = first_cpu;
429     X86CPU *cpu = X86_CPU(cs);
430 
431     DPRINTF("pic_irqs: %s irq %d\n", level? "raise" : "lower", irq);
432     if (cpu->apic_state && !kvm_irqchip_in_kernel()) {
433         CPU_FOREACH(cs) {
434             cpu = X86_CPU(cs);
435             if (apic_accept_pic_intr(cpu->apic_state)) {
436                 apic_deliver_pic_intr(cpu->apic_state, level);
437             }
438         }
439     } else {
440         if (level) {
441             cpu_interrupt(cs, CPU_INTERRUPT_HARD);
442         } else {
443             cpu_reset_interrupt(cs, CPU_INTERRUPT_HARD);
444         }
445     }
446 }
447 
448 /* PC cmos mappings */
449 
450 #define REG_EQUIPMENT_BYTE          0x14
451 
452 int cmos_get_fd_drive_type(FloppyDriveType fd0)
453 {
454     int val;
455 
456     switch (fd0) {
457     case FLOPPY_DRIVE_TYPE_144:
458         /* 1.44 Mb 3"5 drive */
459         val = 4;
460         break;
461     case FLOPPY_DRIVE_TYPE_288:
462         /* 2.88 Mb 3"5 drive */
463         val = 5;
464         break;
465     case FLOPPY_DRIVE_TYPE_120:
466         /* 1.2 Mb 5"5 drive */
467         val = 2;
468         break;
469     case FLOPPY_DRIVE_TYPE_NONE:
470     default:
471         val = 0;
472         break;
473     }
474     return val;
475 }
476 
477 static void cmos_init_hd(ISADevice *s, int type_ofs, int info_ofs,
478                          int16_t cylinders, int8_t heads, int8_t sectors)
479 {
480     rtc_set_memory(s, type_ofs, 47);
481     rtc_set_memory(s, info_ofs, cylinders);
482     rtc_set_memory(s, info_ofs + 1, cylinders >> 8);
483     rtc_set_memory(s, info_ofs + 2, heads);
484     rtc_set_memory(s, info_ofs + 3, 0xff);
485     rtc_set_memory(s, info_ofs + 4, 0xff);
486     rtc_set_memory(s, info_ofs + 5, 0xc0 | ((heads > 8) << 3));
487     rtc_set_memory(s, info_ofs + 6, cylinders);
488     rtc_set_memory(s, info_ofs + 7, cylinders >> 8);
489     rtc_set_memory(s, info_ofs + 8, sectors);
490 }
491 
492 /* convert boot_device letter to something recognizable by the bios */
493 static int boot_device2nibble(char boot_device)
494 {
495     switch(boot_device) {
496     case 'a':
497     case 'b':
498         return 0x01; /* floppy boot */
499     case 'c':
500         return 0x02; /* hard drive boot */
501     case 'd':
502         return 0x03; /* CD-ROM boot */
503     case 'n':
504         return 0x04; /* Network boot */
505     }
506     return 0;
507 }
508 
509 static void set_boot_dev(ISADevice *s, const char *boot_device, Error **errp)
510 {
511 #define PC_MAX_BOOT_DEVICES 3
512     int nbds, bds[3] = { 0, };
513     int i;
514 
515     nbds = strlen(boot_device);
516     if (nbds > PC_MAX_BOOT_DEVICES) {
517         error_setg(errp, "Too many boot devices for PC");
518         return;
519     }
520     for (i = 0; i < nbds; i++) {
521         bds[i] = boot_device2nibble(boot_device[i]);
522         if (bds[i] == 0) {
523             error_setg(errp, "Invalid boot device for PC: '%c'",
524                        boot_device[i]);
525             return;
526         }
527     }
528     rtc_set_memory(s, 0x3d, (bds[1] << 4) | bds[0]);
529     rtc_set_memory(s, 0x38, (bds[2] << 4) | (fd_bootchk ? 0x0 : 0x1));
530 }
531 
532 static void pc_boot_set(void *opaque, const char *boot_device, Error **errp)
533 {
534     set_boot_dev(opaque, boot_device, errp);
535 }
536 
537 static void pc_cmos_init_floppy(ISADevice *rtc_state, ISADevice *floppy)
538 {
539     int val, nb, i;
540     FloppyDriveType fd_type[2] = { FLOPPY_DRIVE_TYPE_NONE,
541                                    FLOPPY_DRIVE_TYPE_NONE };
542 
543     /* floppy type */
544     if (floppy) {
545         for (i = 0; i < 2; i++) {
546             fd_type[i] = isa_fdc_get_drive_type(floppy, i);
547         }
548     }
549     val = (cmos_get_fd_drive_type(fd_type[0]) << 4) |
550         cmos_get_fd_drive_type(fd_type[1]);
551     rtc_set_memory(rtc_state, 0x10, val);
552 
553     val = rtc_get_memory(rtc_state, REG_EQUIPMENT_BYTE);
554     nb = 0;
555     if (fd_type[0] != FLOPPY_DRIVE_TYPE_NONE) {
556         nb++;
557     }
558     if (fd_type[1] != FLOPPY_DRIVE_TYPE_NONE) {
559         nb++;
560     }
561     switch (nb) {
562     case 0:
563         break;
564     case 1:
565         val |= 0x01; /* 1 drive, ready for boot */
566         break;
567     case 2:
568         val |= 0x41; /* 2 drives, ready for boot */
569         break;
570     }
571     rtc_set_memory(rtc_state, REG_EQUIPMENT_BYTE, val);
572 }
573 
574 typedef struct pc_cmos_init_late_arg {
575     ISADevice *rtc_state;
576     BusState *idebus[2];
577 } pc_cmos_init_late_arg;
578 
579 typedef struct check_fdc_state {
580     ISADevice *floppy;
581     bool multiple;
582 } CheckFdcState;
583 
584 static int check_fdc(Object *obj, void *opaque)
585 {
586     CheckFdcState *state = opaque;
587     Object *fdc;
588     uint32_t iobase;
589     Error *local_err = NULL;
590 
591     fdc = object_dynamic_cast(obj, TYPE_ISA_FDC);
592     if (!fdc) {
593         return 0;
594     }
595 
596     iobase = object_property_get_uint(obj, "iobase", &local_err);
597     if (local_err || iobase != 0x3f0) {
598         error_free(local_err);
599         return 0;
600     }
601 
602     if (state->floppy) {
603         state->multiple = true;
604     } else {
605         state->floppy = ISA_DEVICE(obj);
606     }
607     return 0;
608 }
609 
610 static const char * const fdc_container_path[] = {
611     "/unattached", "/peripheral", "/peripheral-anon"
612 };
613 
614 /*
615  * Locate the FDC at IO address 0x3f0, in order to configure the CMOS registers
616  * and ACPI objects.
617  */
618 ISADevice *pc_find_fdc0(void)
619 {
620     int i;
621     Object *container;
622     CheckFdcState state = { 0 };
623 
624     for (i = 0; i < ARRAY_SIZE(fdc_container_path); i++) {
625         container = container_get(qdev_get_machine(), fdc_container_path[i]);
626         object_child_foreach(container, check_fdc, &state);
627     }
628 
629     if (state.multiple) {
630         warn_report("multiple floppy disk controllers with "
631                     "iobase=0x3f0 have been found");
632         error_printf("the one being picked for CMOS setup might not reflect "
633                      "your intent");
634     }
635 
636     return state.floppy;
637 }
638 
639 static void pc_cmos_init_late(void *opaque)
640 {
641     pc_cmos_init_late_arg *arg = opaque;
642     ISADevice *s = arg->rtc_state;
643     int16_t cylinders;
644     int8_t heads, sectors;
645     int val;
646     int i, trans;
647 
648     val = 0;
649     if (arg->idebus[0] && ide_get_geometry(arg->idebus[0], 0,
650                                            &cylinders, &heads, &sectors) >= 0) {
651         cmos_init_hd(s, 0x19, 0x1b, cylinders, heads, sectors);
652         val |= 0xf0;
653     }
654     if (arg->idebus[0] && ide_get_geometry(arg->idebus[0], 1,
655                                            &cylinders, &heads, &sectors) >= 0) {
656         cmos_init_hd(s, 0x1a, 0x24, cylinders, heads, sectors);
657         val |= 0x0f;
658     }
659     rtc_set_memory(s, 0x12, val);
660 
661     val = 0;
662     for (i = 0; i < 4; i++) {
663         /* NOTE: ide_get_geometry() returns the physical
664            geometry.  It is always such that: 1 <= sects <= 63, 1
665            <= heads <= 16, 1 <= cylinders <= 16383. The BIOS
666            geometry can be different if a translation is done. */
667         if (arg->idebus[i / 2] &&
668             ide_get_geometry(arg->idebus[i / 2], i % 2,
669                              &cylinders, &heads, &sectors) >= 0) {
670             trans = ide_get_bios_chs_trans(arg->idebus[i / 2], i % 2) - 1;
671             assert((trans & ~3) == 0);
672             val |= trans << (i * 2);
673         }
674     }
675     rtc_set_memory(s, 0x39, val);
676 
677     pc_cmos_init_floppy(s, pc_find_fdc0());
678 
679     qemu_unregister_reset(pc_cmos_init_late, opaque);
680 }
681 
682 void pc_cmos_init(PCMachineState *pcms,
683                   BusState *idebus0, BusState *idebus1,
684                   ISADevice *s)
685 {
686     int val;
687     static pc_cmos_init_late_arg arg;
688 
689     /* various important CMOS locations needed by PC/Bochs bios */
690 
691     /* memory size */
692     /* base memory (first MiB) */
693     val = MIN(pcms->below_4g_mem_size / KiB, 640);
694     rtc_set_memory(s, 0x15, val);
695     rtc_set_memory(s, 0x16, val >> 8);
696     /* extended memory (next 64MiB) */
697     if (pcms->below_4g_mem_size > 1 * MiB) {
698         val = (pcms->below_4g_mem_size - 1 * MiB) / KiB;
699     } else {
700         val = 0;
701     }
702     if (val > 65535)
703         val = 65535;
704     rtc_set_memory(s, 0x17, val);
705     rtc_set_memory(s, 0x18, val >> 8);
706     rtc_set_memory(s, 0x30, val);
707     rtc_set_memory(s, 0x31, val >> 8);
708     /* memory between 16MiB and 4GiB */
709     if (pcms->below_4g_mem_size > 16 * MiB) {
710         val = (pcms->below_4g_mem_size - 16 * MiB) / (64 * KiB);
711     } else {
712         val = 0;
713     }
714     if (val > 65535)
715         val = 65535;
716     rtc_set_memory(s, 0x34, val);
717     rtc_set_memory(s, 0x35, val >> 8);
718     /* memory above 4GiB */
719     val = pcms->above_4g_mem_size / 65536;
720     rtc_set_memory(s, 0x5b, val);
721     rtc_set_memory(s, 0x5c, val >> 8);
722     rtc_set_memory(s, 0x5d, val >> 16);
723 
724     object_property_add_link(OBJECT(pcms), "rtc_state",
725                              TYPE_ISA_DEVICE,
726                              (Object **)&pcms->rtc,
727                              object_property_allow_set_link,
728                              OBJ_PROP_LINK_STRONG, &error_abort);
729     object_property_set_link(OBJECT(pcms), OBJECT(s),
730                              "rtc_state", &error_abort);
731 
732     set_boot_dev(s, MACHINE(pcms)->boot_order, &error_fatal);
733 
734     val = 0;
735     val |= 0x02; /* FPU is there */
736     val |= 0x04; /* PS/2 mouse installed */
737     rtc_set_memory(s, REG_EQUIPMENT_BYTE, val);
738 
739     /* hard drives and FDC */
740     arg.rtc_state = s;
741     arg.idebus[0] = idebus0;
742     arg.idebus[1] = idebus1;
743     qemu_register_reset(pc_cmos_init_late, &arg);
744 }
745 
746 #define TYPE_PORT92 "port92"
747 #define PORT92(obj) OBJECT_CHECK(Port92State, (obj), TYPE_PORT92)
748 
749 /* port 92 stuff: could be split off */
750 typedef struct Port92State {
751     ISADevice parent_obj;
752 
753     MemoryRegion io;
754     uint8_t outport;
755     qemu_irq a20_out;
756 } Port92State;
757 
758 static void port92_write(void *opaque, hwaddr addr, uint64_t val,
759                          unsigned size)
760 {
761     Port92State *s = opaque;
762     int oldval = s->outport;
763 
764     DPRINTF("port92: write 0x%02" PRIx64 "\n", val);
765     s->outport = val;
766     qemu_set_irq(s->a20_out, (val >> 1) & 1);
767     if ((val & 1) && !(oldval & 1)) {
768         qemu_system_reset_request(SHUTDOWN_CAUSE_GUEST_RESET);
769     }
770 }
771 
772 static uint64_t port92_read(void *opaque, hwaddr addr,
773                             unsigned size)
774 {
775     Port92State *s = opaque;
776     uint32_t ret;
777 
778     ret = s->outport;
779     DPRINTF("port92: read 0x%02x\n", ret);
780     return ret;
781 }
782 
783 static void port92_init(ISADevice *dev, qemu_irq a20_out)
784 {
785     qdev_connect_gpio_out_named(DEVICE(dev), PORT92_A20_LINE, 0, a20_out);
786 }
787 
788 static const VMStateDescription vmstate_port92_isa = {
789     .name = "port92",
790     .version_id = 1,
791     .minimum_version_id = 1,
792     .fields = (VMStateField[]) {
793         VMSTATE_UINT8(outport, Port92State),
794         VMSTATE_END_OF_LIST()
795     }
796 };
797 
798 static void port92_reset(DeviceState *d)
799 {
800     Port92State *s = PORT92(d);
801 
802     s->outport &= ~1;
803 }
804 
805 static const MemoryRegionOps port92_ops = {
806     .read = port92_read,
807     .write = port92_write,
808     .impl = {
809         .min_access_size = 1,
810         .max_access_size = 1,
811     },
812     .endianness = DEVICE_LITTLE_ENDIAN,
813 };
814 
815 static void port92_initfn(Object *obj)
816 {
817     Port92State *s = PORT92(obj);
818 
819     memory_region_init_io(&s->io, OBJECT(s), &port92_ops, s, "port92", 1);
820 
821     s->outport = 0;
822 
823     qdev_init_gpio_out_named(DEVICE(obj), &s->a20_out, PORT92_A20_LINE, 1);
824 }
825 
826 static void port92_realizefn(DeviceState *dev, Error **errp)
827 {
828     ISADevice *isadev = ISA_DEVICE(dev);
829     Port92State *s = PORT92(dev);
830 
831     isa_register_ioport(isadev, &s->io, 0x92);
832 }
833 
834 static void port92_class_initfn(ObjectClass *klass, void *data)
835 {
836     DeviceClass *dc = DEVICE_CLASS(klass);
837 
838     dc->realize = port92_realizefn;
839     dc->reset = port92_reset;
840     dc->vmsd = &vmstate_port92_isa;
841     /*
842      * Reason: unlike ordinary ISA devices, this one needs additional
843      * wiring: its A20 output line needs to be wired up by
844      * port92_init().
845      */
846     dc->user_creatable = false;
847 }
848 
849 static const TypeInfo port92_info = {
850     .name          = TYPE_PORT92,
851     .parent        = TYPE_ISA_DEVICE,
852     .instance_size = sizeof(Port92State),
853     .instance_init = port92_initfn,
854     .class_init    = port92_class_initfn,
855 };
856 
857 static void port92_register_types(void)
858 {
859     type_register_static(&port92_info);
860 }
861 
862 type_init(port92_register_types)
863 
864 static void handle_a20_line_change(void *opaque, int irq, int level)
865 {
866     X86CPU *cpu = opaque;
867 
868     /* XXX: send to all CPUs ? */
869     /* XXX: add logic to handle multiple A20 line sources */
870     x86_cpu_set_a20(cpu, level);
871 }
872 
873 int e820_add_entry(uint64_t address, uint64_t length, uint32_t type)
874 {
875     int index = le32_to_cpu(e820_reserve.count);
876     struct e820_entry *entry;
877 
878     if (type != E820_RAM) {
879         /* old FW_CFG_E820_TABLE entry -- reservations only */
880         if (index >= E820_NR_ENTRIES) {
881             return -EBUSY;
882         }
883         entry = &e820_reserve.entry[index++];
884 
885         entry->address = cpu_to_le64(address);
886         entry->length = cpu_to_le64(length);
887         entry->type = cpu_to_le32(type);
888 
889         e820_reserve.count = cpu_to_le32(index);
890     }
891 
892     /* new "etc/e820" file -- include ram too */
893     e820_table = g_renew(struct e820_entry, e820_table, e820_entries + 1);
894     e820_table[e820_entries].address = cpu_to_le64(address);
895     e820_table[e820_entries].length = cpu_to_le64(length);
896     e820_table[e820_entries].type = cpu_to_le32(type);
897     e820_entries++;
898 
899     return e820_entries;
900 }
901 
902 int e820_get_num_entries(void)
903 {
904     return e820_entries;
905 }
906 
907 bool e820_get_entry(int idx, uint32_t type, uint64_t *address, uint64_t *length)
908 {
909     if (idx < e820_entries && e820_table[idx].type == cpu_to_le32(type)) {
910         *address = le64_to_cpu(e820_table[idx].address);
911         *length = le64_to_cpu(e820_table[idx].length);
912         return true;
913     }
914     return false;
915 }
916 
917 /* Enables contiguous-apic-ID mode, for compatibility */
918 static bool compat_apic_id_mode;
919 
920 void enable_compat_apic_id_mode(void)
921 {
922     compat_apic_id_mode = true;
923 }
924 
925 /* Calculates initial APIC ID for a specific CPU index
926  *
927  * Currently we need to be able to calculate the APIC ID from the CPU index
928  * alone (without requiring a CPU object), as the QEMU<->Seabios interfaces have
929  * no concept of "CPU index", and the NUMA tables on fw_cfg need the APIC ID of
930  * all CPUs up to max_cpus.
931  */
932 static uint32_t x86_cpu_apic_id_from_index(unsigned int cpu_index)
933 {
934     uint32_t correct_id;
935     static bool warned;
936 
937     correct_id = x86_apicid_from_cpu_idx(smp_cores, smp_threads, cpu_index);
938     if (compat_apic_id_mode) {
939         if (cpu_index != correct_id && !warned && !qtest_enabled()) {
940             error_report("APIC IDs set in compatibility mode, "
941                          "CPU topology won't match the configuration");
942             warned = true;
943         }
944         return cpu_index;
945     } else {
946         return correct_id;
947     }
948 }
949 
950 static void pc_build_smbios(PCMachineState *pcms)
951 {
952     uint8_t *smbios_tables, *smbios_anchor;
953     size_t smbios_tables_len, smbios_anchor_len;
954     struct smbios_phys_mem_area *mem_array;
955     unsigned i, array_count;
956     MachineState *ms = MACHINE(pcms);
957     X86CPU *cpu = X86_CPU(ms->possible_cpus->cpus[0].cpu);
958 
959     /* tell smbios about cpuid version and features */
960     smbios_set_cpuid(cpu->env.cpuid_version, cpu->env.features[FEAT_1_EDX]);
961 
962     smbios_tables = smbios_get_table_legacy(&smbios_tables_len);
963     if (smbios_tables) {
964         fw_cfg_add_bytes(pcms->fw_cfg, FW_CFG_SMBIOS_ENTRIES,
965                          smbios_tables, smbios_tables_len);
966     }
967 
968     /* build the array of physical mem area from e820 table */
969     mem_array = g_malloc0(sizeof(*mem_array) * e820_get_num_entries());
970     for (i = 0, array_count = 0; i < e820_get_num_entries(); i++) {
971         uint64_t addr, len;
972 
973         if (e820_get_entry(i, E820_RAM, &addr, &len)) {
974             mem_array[array_count].address = addr;
975             mem_array[array_count].length = len;
976             array_count++;
977         }
978     }
979     smbios_get_tables(mem_array, array_count,
980                       &smbios_tables, &smbios_tables_len,
981                       &smbios_anchor, &smbios_anchor_len);
982     g_free(mem_array);
983 
984     if (smbios_anchor) {
985         fw_cfg_add_file(pcms->fw_cfg, "etc/smbios/smbios-tables",
986                         smbios_tables, smbios_tables_len);
987         fw_cfg_add_file(pcms->fw_cfg, "etc/smbios/smbios-anchor",
988                         smbios_anchor, smbios_anchor_len);
989     }
990 }
991 
992 static FWCfgState *bochs_bios_init(AddressSpace *as, PCMachineState *pcms)
993 {
994     FWCfgState *fw_cfg;
995     uint64_t *numa_fw_cfg;
996     int i;
997     const CPUArchIdList *cpus;
998     MachineClass *mc = MACHINE_GET_CLASS(pcms);
999 
1000     fw_cfg = fw_cfg_init_io_dma(FW_CFG_IO_BASE, FW_CFG_IO_BASE + 4, as);
1001     fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
1002 
1003     /* FW_CFG_MAX_CPUS is a bit confusing/problematic on x86:
1004      *
1005      * For machine types prior to 1.8, SeaBIOS needs FW_CFG_MAX_CPUS for
1006      * building MPTable, ACPI MADT, ACPI CPU hotplug and ACPI SRAT table,
1007      * that tables are based on xAPIC ID and QEMU<->SeaBIOS interface
1008      * for CPU hotplug also uses APIC ID and not "CPU index".
1009      * This means that FW_CFG_MAX_CPUS is not the "maximum number of CPUs",
1010      * but the "limit to the APIC ID values SeaBIOS may see".
1011      *
1012      * So for compatibility reasons with old BIOSes we are stuck with
1013      * "etc/max-cpus" actually being apic_id_limit
1014      */
1015     fw_cfg_add_i16(fw_cfg, FW_CFG_MAX_CPUS, (uint16_t)pcms->apic_id_limit);
1016     fw_cfg_add_i64(fw_cfg, FW_CFG_RAM_SIZE, (uint64_t)ram_size);
1017     fw_cfg_add_bytes(fw_cfg, FW_CFG_ACPI_TABLES,
1018                      acpi_tables, acpi_tables_len);
1019     fw_cfg_add_i32(fw_cfg, FW_CFG_IRQ0_OVERRIDE, kvm_allows_irq0_override());
1020 
1021     fw_cfg_add_bytes(fw_cfg, FW_CFG_E820_TABLE,
1022                      &e820_reserve, sizeof(e820_reserve));
1023     fw_cfg_add_file(fw_cfg, "etc/e820", e820_table,
1024                     sizeof(struct e820_entry) * e820_entries);
1025 
1026     fw_cfg_add_bytes(fw_cfg, FW_CFG_HPET, &hpet_cfg, sizeof(hpet_cfg));
1027     /* allocate memory for the NUMA channel: one (64bit) word for the number
1028      * of nodes, one word for each VCPU->node and one word for each node to
1029      * hold the amount of memory.
1030      */
1031     numa_fw_cfg = g_new0(uint64_t, 1 + pcms->apic_id_limit + nb_numa_nodes);
1032     numa_fw_cfg[0] = cpu_to_le64(nb_numa_nodes);
1033     cpus = mc->possible_cpu_arch_ids(MACHINE(pcms));
1034     for (i = 0; i < cpus->len; i++) {
1035         unsigned int apic_id = cpus->cpus[i].arch_id;
1036         assert(apic_id < pcms->apic_id_limit);
1037         numa_fw_cfg[apic_id + 1] = cpu_to_le64(cpus->cpus[i].props.node_id);
1038     }
1039     for (i = 0; i < nb_numa_nodes; i++) {
1040         numa_fw_cfg[pcms->apic_id_limit + 1 + i] =
1041             cpu_to_le64(numa_info[i].node_mem);
1042     }
1043     fw_cfg_add_bytes(fw_cfg, FW_CFG_NUMA, numa_fw_cfg,
1044                      (1 + pcms->apic_id_limit + nb_numa_nodes) *
1045                      sizeof(*numa_fw_cfg));
1046 
1047     return fw_cfg;
1048 }
1049 
1050 static long get_file_size(FILE *f)
1051 {
1052     long where, size;
1053 
1054     /* XXX: on Unix systems, using fstat() probably makes more sense */
1055 
1056     where = ftell(f);
1057     fseek(f, 0, SEEK_END);
1058     size = ftell(f);
1059     fseek(f, where, SEEK_SET);
1060 
1061     return size;
1062 }
1063 
1064 struct setup_data {
1065     uint64_t next;
1066     uint32_t type;
1067     uint32_t len;
1068     uint8_t data[0];
1069 } __attribute__((packed));
1070 
1071 
1072 /*
1073  * The entry point into the kernel for PVH boot is different from
1074  * the native entry point.  The PVH entry is defined by the x86/HVM
1075  * direct boot ABI and is available in an ELFNOTE in the kernel binary.
1076  *
1077  * This function is passed to load_elf() when it is called from
1078  * load_elfboot() which then additionally checks for an ELF Note of
1079  * type XEN_ELFNOTE_PHYS32_ENTRY and passes it to this function to
1080  * parse the PVH entry address from the ELF Note.
1081  *
1082  * Due to trickery in elf_opts.h, load_elf() is actually available as
1083  * load_elf32() or load_elf64() and this routine needs to be able
1084  * to deal with being called as 32 or 64 bit.
1085  *
1086  * The address of the PVH entry point is saved to the 'pvh_start_addr'
1087  * global variable.  (although the entry point is 32-bit, the kernel
1088  * binary can be either 32-bit or 64-bit).
1089  */
1090 static uint64_t read_pvh_start_addr(void *arg1, void *arg2, bool is64)
1091 {
1092     size_t *elf_note_data_addr;
1093 
1094     /* Check if ELF Note header passed in is valid */
1095     if (arg1 == NULL) {
1096         return 0;
1097     }
1098 
1099     if (is64) {
1100         struct elf64_note *nhdr64 = (struct elf64_note *)arg1;
1101         uint64_t nhdr_size64 = sizeof(struct elf64_note);
1102         uint64_t phdr_align = *(uint64_t *)arg2;
1103         uint64_t nhdr_namesz = nhdr64->n_namesz;
1104 
1105         elf_note_data_addr =
1106             ((void *)nhdr64) + nhdr_size64 +
1107             QEMU_ALIGN_UP(nhdr_namesz, phdr_align);
1108     } else {
1109         struct elf32_note *nhdr32 = (struct elf32_note *)arg1;
1110         uint32_t nhdr_size32 = sizeof(struct elf32_note);
1111         uint32_t phdr_align = *(uint32_t *)arg2;
1112         uint32_t nhdr_namesz = nhdr32->n_namesz;
1113 
1114         elf_note_data_addr =
1115             ((void *)nhdr32) + nhdr_size32 +
1116             QEMU_ALIGN_UP(nhdr_namesz, phdr_align);
1117     }
1118 
1119     pvh_start_addr = *elf_note_data_addr;
1120 
1121     return pvh_start_addr;
1122 }
1123 
1124 static bool load_elfboot(const char *kernel_filename,
1125                    int kernel_file_size,
1126                    uint8_t *header,
1127                    size_t pvh_xen_start_addr,
1128                    FWCfgState *fw_cfg)
1129 {
1130     uint32_t flags = 0;
1131     uint32_t mh_load_addr = 0;
1132     uint32_t elf_kernel_size = 0;
1133     uint64_t elf_entry;
1134     uint64_t elf_low, elf_high;
1135     int kernel_size;
1136 
1137     if (ldl_p(header) != 0x464c457f) {
1138         return false; /* no elfboot */
1139     }
1140 
1141     bool elf_is64 = header[EI_CLASS] == ELFCLASS64;
1142     flags = elf_is64 ?
1143         ((Elf64_Ehdr *)header)->e_flags : ((Elf32_Ehdr *)header)->e_flags;
1144 
1145     if (flags & 0x00010004) { /* LOAD_ELF_HEADER_HAS_ADDR */
1146         error_report("elfboot unsupported flags = %x", flags);
1147         exit(1);
1148     }
1149 
1150     uint64_t elf_note_type = XEN_ELFNOTE_PHYS32_ENTRY;
1151     kernel_size = load_elf(kernel_filename, read_pvh_start_addr,
1152                            NULL, &elf_note_type, &elf_entry,
1153                            &elf_low, &elf_high, 0, I386_ELF_MACHINE,
1154                            0, 0);
1155 
1156     if (kernel_size < 0) {
1157         error_report("Error while loading elf kernel");
1158         exit(1);
1159     }
1160     mh_load_addr = elf_low;
1161     elf_kernel_size = elf_high - elf_low;
1162 
1163     if (pvh_start_addr == 0) {
1164         error_report("Error loading uncompressed kernel without PVH ELF Note");
1165         exit(1);
1166     }
1167     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ENTRY, pvh_start_addr);
1168     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ADDR, mh_load_addr);
1169     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_SIZE, elf_kernel_size);
1170 
1171     return true;
1172 }
1173 
1174 static void load_linux(PCMachineState *pcms,
1175                        FWCfgState *fw_cfg)
1176 {
1177     uint16_t protocol;
1178     int setup_size, kernel_size, cmdline_size;
1179     int dtb_size, setup_data_offset;
1180     uint32_t initrd_max;
1181     uint8_t header[8192], *setup, *kernel;
1182     hwaddr real_addr, prot_addr, cmdline_addr, initrd_addr = 0;
1183     FILE *f;
1184     char *vmode;
1185     MachineState *machine = MACHINE(pcms);
1186     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1187     struct setup_data *setup_data;
1188     const char *kernel_filename = machine->kernel_filename;
1189     const char *initrd_filename = machine->initrd_filename;
1190     const char *dtb_filename = machine->dtb;
1191     const char *kernel_cmdline = machine->kernel_cmdline;
1192 
1193     /* Align to 16 bytes as a paranoia measure */
1194     cmdline_size = (strlen(kernel_cmdline)+16) & ~15;
1195 
1196     /* load the kernel header */
1197     f = fopen(kernel_filename, "rb");
1198     if (!f || !(kernel_size = get_file_size(f)) ||
1199         fread(header, 1, MIN(ARRAY_SIZE(header), kernel_size), f) !=
1200         MIN(ARRAY_SIZE(header), kernel_size)) {
1201         fprintf(stderr, "qemu: could not load kernel '%s': %s\n",
1202                 kernel_filename, strerror(errno));
1203         exit(1);
1204     }
1205 
1206     /* kernel protocol version */
1207 #if 0
1208     fprintf(stderr, "header magic: %#x\n", ldl_p(header+0x202));
1209 #endif
1210     if (ldl_p(header+0x202) == 0x53726448) {
1211         protocol = lduw_p(header+0x206);
1212     } else {
1213         /*
1214          * This could be a multiboot kernel. If it is, let's stop treating it
1215          * like a Linux kernel.
1216          * Note: some multiboot images could be in the ELF format (the same of
1217          * PVH), so we try multiboot first since we check the multiboot magic
1218          * header before to load it.
1219          */
1220         if (load_multiboot(fw_cfg, f, kernel_filename, initrd_filename,
1221                            kernel_cmdline, kernel_size, header)) {
1222             return;
1223         }
1224         /*
1225          * Check if the file is an uncompressed kernel file (ELF) and load it,
1226          * saving the PVH entry point used by the x86/HVM direct boot ABI.
1227          * If load_elfboot() is successful, populate the fw_cfg info.
1228          */
1229         if (pcmc->pvh_enabled &&
1230             load_elfboot(kernel_filename, kernel_size,
1231                          header, pvh_start_addr, fw_cfg)) {
1232             fclose(f);
1233 
1234             fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_SIZE,
1235                 strlen(kernel_cmdline) + 1);
1236             fw_cfg_add_string(fw_cfg, FW_CFG_CMDLINE_DATA, kernel_cmdline);
1237 
1238             fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_SIZE, sizeof(header));
1239             fw_cfg_add_bytes(fw_cfg, FW_CFG_SETUP_DATA,
1240                              header, sizeof(header));
1241 
1242             /* load initrd */
1243             if (initrd_filename) {
1244                 gsize initrd_size;
1245                 gchar *initrd_data;
1246                 GError *gerr = NULL;
1247 
1248                 if (!g_file_get_contents(initrd_filename, &initrd_data,
1249                             &initrd_size, &gerr)) {
1250                     fprintf(stderr, "qemu: error reading initrd %s: %s\n",
1251                             initrd_filename, gerr->message);
1252                     exit(1);
1253                 }
1254 
1255                 initrd_max = pcms->below_4g_mem_size - pcmc->acpi_data_size - 1;
1256                 if (initrd_size >= initrd_max) {
1257                     fprintf(stderr, "qemu: initrd is too large, cannot support."
1258                             "(max: %"PRIu32", need %"PRId64")\n",
1259                             initrd_max, (uint64_t)initrd_size);
1260                     exit(1);
1261                 }
1262 
1263                 initrd_addr = (initrd_max - initrd_size) & ~4095;
1264 
1265                 fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_ADDR, initrd_addr);
1266                 fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_SIZE, initrd_size);
1267                 fw_cfg_add_bytes(fw_cfg, FW_CFG_INITRD_DATA, initrd_data,
1268                                  initrd_size);
1269             }
1270 
1271             option_rom[nb_option_roms].bootindex = 0;
1272             option_rom[nb_option_roms].name = "pvh.bin";
1273             nb_option_roms++;
1274 
1275             return;
1276         }
1277         protocol = 0;
1278     }
1279 
1280     if (protocol < 0x200 || !(header[0x211] & 0x01)) {
1281         /* Low kernel */
1282         real_addr    = 0x90000;
1283         cmdline_addr = 0x9a000 - cmdline_size;
1284         prot_addr    = 0x10000;
1285     } else if (protocol < 0x202) {
1286         /* High but ancient kernel */
1287         real_addr    = 0x90000;
1288         cmdline_addr = 0x9a000 - cmdline_size;
1289         prot_addr    = 0x100000;
1290     } else {
1291         /* High and recent kernel */
1292         real_addr    = 0x10000;
1293         cmdline_addr = 0x20000;
1294         prot_addr    = 0x100000;
1295     }
1296 
1297 #if 0
1298     fprintf(stderr,
1299             "qemu: real_addr     = 0x" TARGET_FMT_plx "\n"
1300             "qemu: cmdline_addr  = 0x" TARGET_FMT_plx "\n"
1301             "qemu: prot_addr     = 0x" TARGET_FMT_plx "\n",
1302             real_addr,
1303             cmdline_addr,
1304             prot_addr);
1305 #endif
1306 
1307     /* highest address for loading the initrd */
1308     if (protocol >= 0x20c &&
1309         lduw_p(header+0x236) & XLF_CAN_BE_LOADED_ABOVE_4G) {
1310         /*
1311          * Linux has supported initrd up to 4 GB for a very long time (2007,
1312          * long before XLF_CAN_BE_LOADED_ABOVE_4G which was added in 2013),
1313          * though it only sets initrd_max to 2 GB to "work around bootloader
1314          * bugs". Luckily, QEMU firmware(which does something like bootloader)
1315          * has supported this.
1316          *
1317          * It's believed that if XLF_CAN_BE_LOADED_ABOVE_4G is set, initrd can
1318          * be loaded into any address.
1319          *
1320          * In addition, initrd_max is uint32_t simply because QEMU doesn't
1321          * support the 64-bit boot protocol (specifically the ext_ramdisk_image
1322          * field).
1323          *
1324          * Therefore here just limit initrd_max to UINT32_MAX simply as well.
1325          */
1326         initrd_max = UINT32_MAX;
1327     } else if (protocol >= 0x203) {
1328         initrd_max = ldl_p(header+0x22c);
1329     } else {
1330         initrd_max = 0x37ffffff;
1331     }
1332 
1333     if (initrd_max >= pcms->below_4g_mem_size - pcmc->acpi_data_size) {
1334         initrd_max = pcms->below_4g_mem_size - pcmc->acpi_data_size - 1;
1335     }
1336 
1337     fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_ADDR, cmdline_addr);
1338     fw_cfg_add_i32(fw_cfg, FW_CFG_CMDLINE_SIZE, strlen(kernel_cmdline)+1);
1339     fw_cfg_add_string(fw_cfg, FW_CFG_CMDLINE_DATA, kernel_cmdline);
1340 
1341     if (protocol >= 0x202) {
1342         stl_p(header+0x228, cmdline_addr);
1343     } else {
1344         stw_p(header+0x20, 0xA33F);
1345         stw_p(header+0x22, cmdline_addr-real_addr);
1346     }
1347 
1348     /* handle vga= parameter */
1349     vmode = strstr(kernel_cmdline, "vga=");
1350     if (vmode) {
1351         unsigned int video_mode;
1352         /* skip "vga=" */
1353         vmode += 4;
1354         if (!strncmp(vmode, "normal", 6)) {
1355             video_mode = 0xffff;
1356         } else if (!strncmp(vmode, "ext", 3)) {
1357             video_mode = 0xfffe;
1358         } else if (!strncmp(vmode, "ask", 3)) {
1359             video_mode = 0xfffd;
1360         } else {
1361             video_mode = strtol(vmode, NULL, 0);
1362         }
1363         stw_p(header+0x1fa, video_mode);
1364     }
1365 
1366     /* loader type */
1367     /* High nybble = B reserved for QEMU; low nybble is revision number.
1368        If this code is substantially changed, you may want to consider
1369        incrementing the revision. */
1370     if (protocol >= 0x200) {
1371         header[0x210] = 0xB0;
1372     }
1373     /* heap */
1374     if (protocol >= 0x201) {
1375         header[0x211] |= 0x80;	/* CAN_USE_HEAP */
1376         stw_p(header+0x224, cmdline_addr-real_addr-0x200);
1377     }
1378 
1379     /* load initrd */
1380     if (initrd_filename) {
1381         gsize initrd_size;
1382         gchar *initrd_data;
1383         GError *gerr = NULL;
1384 
1385         if (protocol < 0x200) {
1386             fprintf(stderr, "qemu: linux kernel too old to load a ram disk\n");
1387             exit(1);
1388         }
1389 
1390         if (!g_file_get_contents(initrd_filename, &initrd_data,
1391                                  &initrd_size, &gerr)) {
1392             fprintf(stderr, "qemu: error reading initrd %s: %s\n",
1393                     initrd_filename, gerr->message);
1394             exit(1);
1395         }
1396         if (initrd_size >= initrd_max) {
1397             fprintf(stderr, "qemu: initrd is too large, cannot support."
1398                     "(max: %"PRIu32", need %"PRId64")\n",
1399                     initrd_max, (uint64_t)initrd_size);
1400             exit(1);
1401         }
1402 
1403         initrd_addr = (initrd_max-initrd_size) & ~4095;
1404 
1405         fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_ADDR, initrd_addr);
1406         fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_SIZE, initrd_size);
1407         fw_cfg_add_bytes(fw_cfg, FW_CFG_INITRD_DATA, initrd_data, initrd_size);
1408 
1409         stl_p(header+0x218, initrd_addr);
1410         stl_p(header+0x21c, initrd_size);
1411     }
1412 
1413     /* load kernel and setup */
1414     setup_size = header[0x1f1];
1415     if (setup_size == 0) {
1416         setup_size = 4;
1417     }
1418     setup_size = (setup_size+1)*512;
1419     if (setup_size > kernel_size) {
1420         fprintf(stderr, "qemu: invalid kernel header\n");
1421         exit(1);
1422     }
1423     kernel_size -= setup_size;
1424 
1425     setup  = g_malloc(setup_size);
1426     kernel = g_malloc(kernel_size);
1427     fseek(f, 0, SEEK_SET);
1428     if (fread(setup, 1, setup_size, f) != setup_size) {
1429         fprintf(stderr, "fread() failed\n");
1430         exit(1);
1431     }
1432     if (fread(kernel, 1, kernel_size, f) != kernel_size) {
1433         fprintf(stderr, "fread() failed\n");
1434         exit(1);
1435     }
1436     fclose(f);
1437 
1438     /* append dtb to kernel */
1439     if (dtb_filename) {
1440         if (protocol < 0x209) {
1441             fprintf(stderr, "qemu: Linux kernel too old to load a dtb\n");
1442             exit(1);
1443         }
1444 
1445         dtb_size = get_image_size(dtb_filename);
1446         if (dtb_size <= 0) {
1447             fprintf(stderr, "qemu: error reading dtb %s: %s\n",
1448                     dtb_filename, strerror(errno));
1449             exit(1);
1450         }
1451 
1452         setup_data_offset = QEMU_ALIGN_UP(kernel_size, 16);
1453         kernel_size = setup_data_offset + sizeof(struct setup_data) + dtb_size;
1454         kernel = g_realloc(kernel, kernel_size);
1455 
1456         stq_p(header+0x250, prot_addr + setup_data_offset);
1457 
1458         setup_data = (struct setup_data *)(kernel + setup_data_offset);
1459         setup_data->next = 0;
1460         setup_data->type = cpu_to_le32(SETUP_DTB);
1461         setup_data->len = cpu_to_le32(dtb_size);
1462 
1463         load_image_size(dtb_filename, setup_data->data, dtb_size);
1464     }
1465 
1466     memcpy(setup, header, MIN(sizeof(header), setup_size));
1467 
1468     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ADDR, prot_addr);
1469     fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_SIZE, kernel_size);
1470     fw_cfg_add_bytes(fw_cfg, FW_CFG_KERNEL_DATA, kernel, kernel_size);
1471 
1472     fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_ADDR, real_addr);
1473     fw_cfg_add_i32(fw_cfg, FW_CFG_SETUP_SIZE, setup_size);
1474     fw_cfg_add_bytes(fw_cfg, FW_CFG_SETUP_DATA, setup, setup_size);
1475 
1476     option_rom[nb_option_roms].bootindex = 0;
1477     option_rom[nb_option_roms].name = "linuxboot.bin";
1478     if (pcmc->linuxboot_dma_enabled && fw_cfg_dma_enabled(fw_cfg)) {
1479         option_rom[nb_option_roms].name = "linuxboot_dma.bin";
1480     }
1481     nb_option_roms++;
1482 }
1483 
1484 #define NE2000_NB_MAX 6
1485 
1486 static const int ne2000_io[NE2000_NB_MAX] = { 0x300, 0x320, 0x340, 0x360,
1487                                               0x280, 0x380 };
1488 static const int ne2000_irq[NE2000_NB_MAX] = { 9, 10, 11, 3, 4, 5 };
1489 
1490 void pc_init_ne2k_isa(ISABus *bus, NICInfo *nd)
1491 {
1492     static int nb_ne2k = 0;
1493 
1494     if (nb_ne2k == NE2000_NB_MAX)
1495         return;
1496     isa_ne2000_init(bus, ne2000_io[nb_ne2k],
1497                     ne2000_irq[nb_ne2k], nd);
1498     nb_ne2k++;
1499 }
1500 
1501 DeviceState *cpu_get_current_apic(void)
1502 {
1503     if (current_cpu) {
1504         X86CPU *cpu = X86_CPU(current_cpu);
1505         return cpu->apic_state;
1506     } else {
1507         return NULL;
1508     }
1509 }
1510 
1511 void pc_acpi_smi_interrupt(void *opaque, int irq, int level)
1512 {
1513     X86CPU *cpu = opaque;
1514 
1515     if (level) {
1516         cpu_interrupt(CPU(cpu), CPU_INTERRUPT_SMI);
1517     }
1518 }
1519 
1520 static void pc_new_cpu(const char *typename, int64_t apic_id, Error **errp)
1521 {
1522     Object *cpu = NULL;
1523     Error *local_err = NULL;
1524 
1525     cpu = object_new(typename);
1526 
1527     object_property_set_uint(cpu, apic_id, "apic-id", &local_err);
1528     object_property_set_bool(cpu, true, "realized", &local_err);
1529 
1530     object_unref(cpu);
1531     error_propagate(errp, local_err);
1532 }
1533 
1534 void pc_hot_add_cpu(const int64_t id, Error **errp)
1535 {
1536     MachineState *ms = MACHINE(qdev_get_machine());
1537     int64_t apic_id = x86_cpu_apic_id_from_index(id);
1538     Error *local_err = NULL;
1539 
1540     if (id < 0) {
1541         error_setg(errp, "Invalid CPU id: %" PRIi64, id);
1542         return;
1543     }
1544 
1545     if (apic_id >= ACPI_CPU_HOTPLUG_ID_LIMIT) {
1546         error_setg(errp, "Unable to add CPU: %" PRIi64
1547                    ", resulting APIC ID (%" PRIi64 ") is too large",
1548                    id, apic_id);
1549         return;
1550     }
1551 
1552     pc_new_cpu(ms->cpu_type, apic_id, &local_err);
1553     if (local_err) {
1554         error_propagate(errp, local_err);
1555         return;
1556     }
1557 }
1558 
1559 void pc_cpus_init(PCMachineState *pcms)
1560 {
1561     int i;
1562     const CPUArchIdList *possible_cpus;
1563     MachineState *ms = MACHINE(pcms);
1564     MachineClass *mc = MACHINE_GET_CLASS(pcms);
1565 
1566     /* Calculates the limit to CPU APIC ID values
1567      *
1568      * Limit for the APIC ID value, so that all
1569      * CPU APIC IDs are < pcms->apic_id_limit.
1570      *
1571      * This is used for FW_CFG_MAX_CPUS. See comments on bochs_bios_init().
1572      */
1573     pcms->apic_id_limit = x86_cpu_apic_id_from_index(max_cpus - 1) + 1;
1574     possible_cpus = mc->possible_cpu_arch_ids(ms);
1575     for (i = 0; i < smp_cpus; i++) {
1576         pc_new_cpu(possible_cpus->cpus[i].type, possible_cpus->cpus[i].arch_id,
1577                    &error_fatal);
1578     }
1579 }
1580 
1581 static void pc_build_feature_control_file(PCMachineState *pcms)
1582 {
1583     MachineState *ms = MACHINE(pcms);
1584     X86CPU *cpu = X86_CPU(ms->possible_cpus->cpus[0].cpu);
1585     CPUX86State *env = &cpu->env;
1586     uint32_t unused, ecx, edx;
1587     uint64_t feature_control_bits = 0;
1588     uint64_t *val;
1589 
1590     cpu_x86_cpuid(env, 1, 0, &unused, &unused, &ecx, &edx);
1591     if (ecx & CPUID_EXT_VMX) {
1592         feature_control_bits |= FEATURE_CONTROL_VMXON_ENABLED_OUTSIDE_SMX;
1593     }
1594 
1595     if ((edx & (CPUID_EXT2_MCE | CPUID_EXT2_MCA)) ==
1596         (CPUID_EXT2_MCE | CPUID_EXT2_MCA) &&
1597         (env->mcg_cap & MCG_LMCE_P)) {
1598         feature_control_bits |= FEATURE_CONTROL_LMCE;
1599     }
1600 
1601     if (!feature_control_bits) {
1602         return;
1603     }
1604 
1605     val = g_malloc(sizeof(*val));
1606     *val = cpu_to_le64(feature_control_bits | FEATURE_CONTROL_LOCKED);
1607     fw_cfg_add_file(pcms->fw_cfg, "etc/msr_feature_control", val, sizeof(*val));
1608 }
1609 
1610 static void rtc_set_cpus_count(ISADevice *rtc, uint16_t cpus_count)
1611 {
1612     if (cpus_count > 0xff) {
1613         /* If the number of CPUs can't be represented in 8 bits, the
1614          * BIOS must use "FW_CFG_NB_CPUS". Set RTC field to 0 just
1615          * to make old BIOSes fail more predictably.
1616          */
1617         rtc_set_memory(rtc, 0x5f, 0);
1618     } else {
1619         rtc_set_memory(rtc, 0x5f, cpus_count - 1);
1620     }
1621 }
1622 
1623 static
1624 void pc_machine_done(Notifier *notifier, void *data)
1625 {
1626     PCMachineState *pcms = container_of(notifier,
1627                                         PCMachineState, machine_done);
1628     PCIBus *bus = pcms->bus;
1629 
1630     /* set the number of CPUs */
1631     rtc_set_cpus_count(pcms->rtc, pcms->boot_cpus);
1632 
1633     if (bus) {
1634         int extra_hosts = 0;
1635 
1636         QLIST_FOREACH(bus, &bus->child, sibling) {
1637             /* look for expander root buses */
1638             if (pci_bus_is_root(bus)) {
1639                 extra_hosts++;
1640             }
1641         }
1642         if (extra_hosts && pcms->fw_cfg) {
1643             uint64_t *val = g_malloc(sizeof(*val));
1644             *val = cpu_to_le64(extra_hosts);
1645             fw_cfg_add_file(pcms->fw_cfg,
1646                     "etc/extra-pci-roots", val, sizeof(*val));
1647         }
1648     }
1649 
1650     acpi_setup();
1651     if (pcms->fw_cfg) {
1652         pc_build_smbios(pcms);
1653         pc_build_feature_control_file(pcms);
1654         /* update FW_CFG_NB_CPUS to account for -device added CPUs */
1655         fw_cfg_modify_i16(pcms->fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
1656     }
1657 
1658     if (pcms->apic_id_limit > 255 && !xen_enabled()) {
1659         IntelIOMMUState *iommu = INTEL_IOMMU_DEVICE(x86_iommu_get_default());
1660 
1661         if (!iommu || !x86_iommu_ir_supported(X86_IOMMU_DEVICE(iommu)) ||
1662             iommu->intr_eim != ON_OFF_AUTO_ON) {
1663             error_report("current -smp configuration requires "
1664                          "Extended Interrupt Mode enabled. "
1665                          "You can add an IOMMU using: "
1666                          "-device intel-iommu,intremap=on,eim=on");
1667             exit(EXIT_FAILURE);
1668         }
1669     }
1670 }
1671 
1672 void pc_guest_info_init(PCMachineState *pcms)
1673 {
1674     int i;
1675 
1676     pcms->apic_xrupt_override = kvm_allows_irq0_override();
1677     pcms->numa_nodes = nb_numa_nodes;
1678     pcms->node_mem = g_malloc0(pcms->numa_nodes *
1679                                     sizeof *pcms->node_mem);
1680     for (i = 0; i < nb_numa_nodes; i++) {
1681         pcms->node_mem[i] = numa_info[i].node_mem;
1682     }
1683 
1684     pcms->machine_done.notify = pc_machine_done;
1685     qemu_add_machine_init_done_notifier(&pcms->machine_done);
1686 }
1687 
1688 /* setup pci memory address space mapping into system address space */
1689 void pc_pci_as_mapping_init(Object *owner, MemoryRegion *system_memory,
1690                             MemoryRegion *pci_address_space)
1691 {
1692     /* Set to lower priority than RAM */
1693     memory_region_add_subregion_overlap(system_memory, 0x0,
1694                                         pci_address_space, -1);
1695 }
1696 
1697 void xen_load_linux(PCMachineState *pcms)
1698 {
1699     int i;
1700     FWCfgState *fw_cfg;
1701 
1702     assert(MACHINE(pcms)->kernel_filename != NULL);
1703 
1704     fw_cfg = fw_cfg_init_io(FW_CFG_IO_BASE);
1705     fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
1706     rom_set_fw(fw_cfg);
1707 
1708     load_linux(pcms, fw_cfg);
1709     for (i = 0; i < nb_option_roms; i++) {
1710         assert(!strcmp(option_rom[i].name, "linuxboot.bin") ||
1711                !strcmp(option_rom[i].name, "linuxboot_dma.bin") ||
1712                !strcmp(option_rom[i].name, "pvh.bin") ||
1713                !strcmp(option_rom[i].name, "multiboot.bin"));
1714         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1715     }
1716     pcms->fw_cfg = fw_cfg;
1717 }
1718 
1719 void pc_memory_init(PCMachineState *pcms,
1720                     MemoryRegion *system_memory,
1721                     MemoryRegion *rom_memory,
1722                     MemoryRegion **ram_memory)
1723 {
1724     int linux_boot, i;
1725     MemoryRegion *ram, *option_rom_mr;
1726     MemoryRegion *ram_below_4g, *ram_above_4g;
1727     FWCfgState *fw_cfg;
1728     MachineState *machine = MACHINE(pcms);
1729     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1730 
1731     assert(machine->ram_size == pcms->below_4g_mem_size +
1732                                 pcms->above_4g_mem_size);
1733 
1734     linux_boot = (machine->kernel_filename != NULL);
1735 
1736     /* Allocate RAM.  We allocate it as a single memory region and use
1737      * aliases to address portions of it, mostly for backwards compatibility
1738      * with older qemus that used qemu_ram_alloc().
1739      */
1740     ram = g_malloc(sizeof(*ram));
1741     memory_region_allocate_system_memory(ram, NULL, "pc.ram",
1742                                          machine->ram_size);
1743     *ram_memory = ram;
1744     ram_below_4g = g_malloc(sizeof(*ram_below_4g));
1745     memory_region_init_alias(ram_below_4g, NULL, "ram-below-4g", ram,
1746                              0, pcms->below_4g_mem_size);
1747     memory_region_add_subregion(system_memory, 0, ram_below_4g);
1748     e820_add_entry(0, pcms->below_4g_mem_size, E820_RAM);
1749     if (pcms->above_4g_mem_size > 0) {
1750         ram_above_4g = g_malloc(sizeof(*ram_above_4g));
1751         memory_region_init_alias(ram_above_4g, NULL, "ram-above-4g", ram,
1752                                  pcms->below_4g_mem_size,
1753                                  pcms->above_4g_mem_size);
1754         memory_region_add_subregion(system_memory, 0x100000000ULL,
1755                                     ram_above_4g);
1756         e820_add_entry(0x100000000ULL, pcms->above_4g_mem_size, E820_RAM);
1757     }
1758 
1759     if (!pcmc->has_reserved_memory &&
1760         (machine->ram_slots ||
1761          (machine->maxram_size > machine->ram_size))) {
1762         MachineClass *mc = MACHINE_GET_CLASS(machine);
1763 
1764         error_report("\"-memory 'slots|maxmem'\" is not supported by: %s",
1765                      mc->name);
1766         exit(EXIT_FAILURE);
1767     }
1768 
1769     /* always allocate the device memory information */
1770     machine->device_memory = g_malloc0(sizeof(*machine->device_memory));
1771 
1772     /* initialize device memory address space */
1773     if (pcmc->has_reserved_memory &&
1774         (machine->ram_size < machine->maxram_size)) {
1775         ram_addr_t device_mem_size = machine->maxram_size - machine->ram_size;
1776 
1777         if (machine->ram_slots > ACPI_MAX_RAM_SLOTS) {
1778             error_report("unsupported amount of memory slots: %"PRIu64,
1779                          machine->ram_slots);
1780             exit(EXIT_FAILURE);
1781         }
1782 
1783         if (QEMU_ALIGN_UP(machine->maxram_size,
1784                           TARGET_PAGE_SIZE) != machine->maxram_size) {
1785             error_report("maximum memory size must by aligned to multiple of "
1786                          "%d bytes", TARGET_PAGE_SIZE);
1787             exit(EXIT_FAILURE);
1788         }
1789 
1790         machine->device_memory->base =
1791             ROUND_UP(0x100000000ULL + pcms->above_4g_mem_size, 1 * GiB);
1792 
1793         if (pcmc->enforce_aligned_dimm) {
1794             /* size device region assuming 1G page max alignment per slot */
1795             device_mem_size += (1 * GiB) * machine->ram_slots;
1796         }
1797 
1798         if ((machine->device_memory->base + device_mem_size) <
1799             device_mem_size) {
1800             error_report("unsupported amount of maximum memory: " RAM_ADDR_FMT,
1801                          machine->maxram_size);
1802             exit(EXIT_FAILURE);
1803         }
1804 
1805         memory_region_init(&machine->device_memory->mr, OBJECT(pcms),
1806                            "device-memory", device_mem_size);
1807         memory_region_add_subregion(system_memory, machine->device_memory->base,
1808                                     &machine->device_memory->mr);
1809     }
1810 
1811     /* Initialize PC system firmware */
1812     pc_system_firmware_init(pcms, rom_memory);
1813 
1814     option_rom_mr = g_malloc(sizeof(*option_rom_mr));
1815     memory_region_init_ram(option_rom_mr, NULL, "pc.rom", PC_ROM_SIZE,
1816                            &error_fatal);
1817     if (pcmc->pci_enabled) {
1818         memory_region_set_readonly(option_rom_mr, true);
1819     }
1820     memory_region_add_subregion_overlap(rom_memory,
1821                                         PC_ROM_MIN_VGA,
1822                                         option_rom_mr,
1823                                         1);
1824 
1825     fw_cfg = bochs_bios_init(&address_space_memory, pcms);
1826 
1827     rom_set_fw(fw_cfg);
1828 
1829     if (pcmc->has_reserved_memory && machine->device_memory->base) {
1830         uint64_t *val = g_malloc(sizeof(*val));
1831         PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1832         uint64_t res_mem_end = machine->device_memory->base;
1833 
1834         if (!pcmc->broken_reserved_end) {
1835             res_mem_end += memory_region_size(&machine->device_memory->mr);
1836         }
1837         *val = cpu_to_le64(ROUND_UP(res_mem_end, 1 * GiB));
1838         fw_cfg_add_file(fw_cfg, "etc/reserved-memory-end", val, sizeof(*val));
1839     }
1840 
1841     if (linux_boot) {
1842         load_linux(pcms, fw_cfg);
1843     }
1844 
1845     for (i = 0; i < nb_option_roms; i++) {
1846         rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1847     }
1848     pcms->fw_cfg = fw_cfg;
1849 
1850     /* Init default IOAPIC address space */
1851     pcms->ioapic_as = &address_space_memory;
1852 }
1853 
1854 /*
1855  * The 64bit pci hole starts after "above 4G RAM" and
1856  * potentially the space reserved for memory hotplug.
1857  */
1858 uint64_t pc_pci_hole64_start(void)
1859 {
1860     PCMachineState *pcms = PC_MACHINE(qdev_get_machine());
1861     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1862     MachineState *ms = MACHINE(pcms);
1863     uint64_t hole64_start = 0;
1864 
1865     if (pcmc->has_reserved_memory && ms->device_memory->base) {
1866         hole64_start = ms->device_memory->base;
1867         if (!pcmc->broken_reserved_end) {
1868             hole64_start += memory_region_size(&ms->device_memory->mr);
1869         }
1870     } else {
1871         hole64_start = 0x100000000ULL + pcms->above_4g_mem_size;
1872     }
1873 
1874     return ROUND_UP(hole64_start, 1 * GiB);
1875 }
1876 
1877 qemu_irq pc_allocate_cpu_irq(void)
1878 {
1879     return qemu_allocate_irq(pic_irq_request, NULL, 0);
1880 }
1881 
1882 DeviceState *pc_vga_init(ISABus *isa_bus, PCIBus *pci_bus)
1883 {
1884     DeviceState *dev = NULL;
1885 
1886     rom_set_order_override(FW_CFG_ORDER_OVERRIDE_VGA);
1887     if (pci_bus) {
1888         PCIDevice *pcidev = pci_vga_init(pci_bus);
1889         dev = pcidev ? &pcidev->qdev : NULL;
1890     } else if (isa_bus) {
1891         ISADevice *isadev = isa_vga_init(isa_bus);
1892         dev = isadev ? DEVICE(isadev) : NULL;
1893     }
1894     rom_reset_order_override();
1895     return dev;
1896 }
1897 
1898 static const MemoryRegionOps ioport80_io_ops = {
1899     .write = ioport80_write,
1900     .read = ioport80_read,
1901     .endianness = DEVICE_NATIVE_ENDIAN,
1902     .impl = {
1903         .min_access_size = 1,
1904         .max_access_size = 1,
1905     },
1906 };
1907 
1908 static const MemoryRegionOps ioportF0_io_ops = {
1909     .write = ioportF0_write,
1910     .read = ioportF0_read,
1911     .endianness = DEVICE_NATIVE_ENDIAN,
1912     .impl = {
1913         .min_access_size = 1,
1914         .max_access_size = 1,
1915     },
1916 };
1917 
1918 static void pc_superio_init(ISABus *isa_bus, bool create_fdctrl, bool no_vmport)
1919 {
1920     int i;
1921     DriveInfo *fd[MAX_FD];
1922     qemu_irq *a20_line;
1923     ISADevice *i8042, *port92, *vmmouse;
1924 
1925     serial_hds_isa_init(isa_bus, 0, MAX_ISA_SERIAL_PORTS);
1926     parallel_hds_isa_init(isa_bus, MAX_PARALLEL_PORTS);
1927 
1928     for (i = 0; i < MAX_FD; i++) {
1929         fd[i] = drive_get(IF_FLOPPY, 0, i);
1930         create_fdctrl |= !!fd[i];
1931     }
1932     if (create_fdctrl) {
1933         fdctrl_init_isa(isa_bus, fd);
1934     }
1935 
1936     i8042 = isa_create_simple(isa_bus, "i8042");
1937     if (!no_vmport) {
1938         vmport_init(isa_bus);
1939         vmmouse = isa_try_create(isa_bus, "vmmouse");
1940     } else {
1941         vmmouse = NULL;
1942     }
1943     if (vmmouse) {
1944         DeviceState *dev = DEVICE(vmmouse);
1945         qdev_prop_set_ptr(dev, "ps2_mouse", i8042);
1946         qdev_init_nofail(dev);
1947     }
1948     port92 = isa_create_simple(isa_bus, "port92");
1949 
1950     a20_line = qemu_allocate_irqs(handle_a20_line_change, first_cpu, 2);
1951     i8042_setup_a20_line(i8042, a20_line[0]);
1952     port92_init(port92, a20_line[1]);
1953     g_free(a20_line);
1954 }
1955 
1956 void pc_basic_device_init(ISABus *isa_bus, qemu_irq *gsi,
1957                           ISADevice **rtc_state,
1958                           bool create_fdctrl,
1959                           bool no_vmport,
1960                           bool has_pit,
1961                           uint32_t hpet_irqs)
1962 {
1963     int i;
1964     DeviceState *hpet = NULL;
1965     int pit_isa_irq = 0;
1966     qemu_irq pit_alt_irq = NULL;
1967     qemu_irq rtc_irq = NULL;
1968     ISADevice *pit = NULL;
1969     MemoryRegion *ioport80_io = g_new(MemoryRegion, 1);
1970     MemoryRegion *ioportF0_io = g_new(MemoryRegion, 1);
1971 
1972     memory_region_init_io(ioport80_io, NULL, &ioport80_io_ops, NULL, "ioport80", 1);
1973     memory_region_add_subregion(isa_bus->address_space_io, 0x80, ioport80_io);
1974 
1975     memory_region_init_io(ioportF0_io, NULL, &ioportF0_io_ops, NULL, "ioportF0", 1);
1976     memory_region_add_subregion(isa_bus->address_space_io, 0xf0, ioportF0_io);
1977 
1978     /*
1979      * Check if an HPET shall be created.
1980      *
1981      * Without KVM_CAP_PIT_STATE2, we cannot switch off the in-kernel PIT
1982      * when the HPET wants to take over. Thus we have to disable the latter.
1983      */
1984     if (!no_hpet && (!kvm_irqchip_in_kernel() || kvm_has_pit_state2())) {
1985         /* In order to set property, here not using sysbus_try_create_simple */
1986         hpet = qdev_try_create(NULL, TYPE_HPET);
1987         if (hpet) {
1988             /* For pc-piix-*, hpet's intcap is always IRQ2. For pc-q35-1.7
1989              * and earlier, use IRQ2 for compat. Otherwise, use IRQ16~23,
1990              * IRQ8 and IRQ2.
1991              */
1992             uint8_t compat = object_property_get_uint(OBJECT(hpet),
1993                     HPET_INTCAP, NULL);
1994             if (!compat) {
1995                 qdev_prop_set_uint32(hpet, HPET_INTCAP, hpet_irqs);
1996             }
1997             qdev_init_nofail(hpet);
1998             sysbus_mmio_map(SYS_BUS_DEVICE(hpet), 0, HPET_BASE);
1999 
2000             for (i = 0; i < GSI_NUM_PINS; i++) {
2001                 sysbus_connect_irq(SYS_BUS_DEVICE(hpet), i, gsi[i]);
2002             }
2003             pit_isa_irq = -1;
2004             pit_alt_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_PIT_INT);
2005             rtc_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_RTC_INT);
2006         }
2007     }
2008     *rtc_state = mc146818_rtc_init(isa_bus, 2000, rtc_irq);
2009 
2010     qemu_register_boot_set(pc_boot_set, *rtc_state);
2011 
2012     if (!xen_enabled() && has_pit) {
2013         if (kvm_pit_in_kernel()) {
2014             pit = kvm_pit_init(isa_bus, 0x40);
2015         } else {
2016             pit = i8254_pit_init(isa_bus, 0x40, pit_isa_irq, pit_alt_irq);
2017         }
2018         if (hpet) {
2019             /* connect PIT to output control line of the HPET */
2020             qdev_connect_gpio_out(hpet, 0, qdev_get_gpio_in(DEVICE(pit), 0));
2021         }
2022         pcspk_init(isa_bus, pit);
2023     }
2024 
2025     i8257_dma_init(isa_bus, 0);
2026 
2027     /* Super I/O */
2028     pc_superio_init(isa_bus, create_fdctrl, no_vmport);
2029 }
2030 
2031 void pc_nic_init(PCMachineClass *pcmc, ISABus *isa_bus, PCIBus *pci_bus)
2032 {
2033     int i;
2034 
2035     rom_set_order_override(FW_CFG_ORDER_OVERRIDE_NIC);
2036     for (i = 0; i < nb_nics; i++) {
2037         NICInfo *nd = &nd_table[i];
2038         const char *model = nd->model ? nd->model : pcmc->default_nic_model;
2039 
2040         if (g_str_equal(model, "ne2k_isa")) {
2041             pc_init_ne2k_isa(isa_bus, nd);
2042         } else {
2043             pci_nic_init_nofail(nd, pci_bus, model, NULL);
2044         }
2045     }
2046     rom_reset_order_override();
2047 }
2048 
2049 void ioapic_init_gsi(GSIState *gsi_state, const char *parent_name)
2050 {
2051     DeviceState *dev;
2052     SysBusDevice *d;
2053     unsigned int i;
2054 
2055     if (kvm_ioapic_in_kernel()) {
2056         dev = qdev_create(NULL, TYPE_KVM_IOAPIC);
2057     } else {
2058         dev = qdev_create(NULL, TYPE_IOAPIC);
2059     }
2060     if (parent_name) {
2061         object_property_add_child(object_resolve_path(parent_name, NULL),
2062                                   "ioapic", OBJECT(dev), NULL);
2063     }
2064     qdev_init_nofail(dev);
2065     d = SYS_BUS_DEVICE(dev);
2066     sysbus_mmio_map(d, 0, IO_APIC_DEFAULT_ADDRESS);
2067 
2068     for (i = 0; i < IOAPIC_NUM_PINS; i++) {
2069         gsi_state->ioapic_irq[i] = qdev_get_gpio_in(dev, i);
2070     }
2071 }
2072 
2073 static void pc_memory_pre_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
2074                                Error **errp)
2075 {
2076     const PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2077     const PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
2078     const MachineState *ms = MACHINE(hotplug_dev);
2079     const bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2080     const uint64_t legacy_align = TARGET_PAGE_SIZE;
2081     Error *local_err = NULL;
2082 
2083     /*
2084      * When -no-acpi is used with Q35 machine type, no ACPI is built,
2085      * but pcms->acpi_dev is still created. Check !acpi_enabled in
2086      * addition to cover this case.
2087      */
2088     if (!pcms->acpi_dev || !acpi_enabled) {
2089         error_setg(errp,
2090                    "memory hotplug is not enabled: missing acpi device or acpi disabled");
2091         return;
2092     }
2093 
2094     if (is_nvdimm && !ms->nvdimms_state->is_enabled) {
2095         error_setg(errp, "nvdimm is not enabled: missing 'nvdimm' in '-M'");
2096         return;
2097     }
2098 
2099     hotplug_handler_pre_plug(pcms->acpi_dev, dev, &local_err);
2100     if (local_err) {
2101         error_propagate(errp, local_err);
2102         return;
2103     }
2104 
2105     pc_dimm_pre_plug(PC_DIMM(dev), MACHINE(hotplug_dev),
2106                      pcmc->enforce_aligned_dimm ? NULL : &legacy_align, errp);
2107 }
2108 
2109 static void pc_memory_plug(HotplugHandler *hotplug_dev,
2110                            DeviceState *dev, Error **errp)
2111 {
2112     Error *local_err = NULL;
2113     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2114     MachineState *ms = MACHINE(hotplug_dev);
2115     bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2116 
2117     pc_dimm_plug(PC_DIMM(dev), MACHINE(pcms), &local_err);
2118     if (local_err) {
2119         goto out;
2120     }
2121 
2122     if (is_nvdimm) {
2123         nvdimm_plug(ms->nvdimms_state);
2124     }
2125 
2126     hotplug_handler_plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &error_abort);
2127 out:
2128     error_propagate(errp, local_err);
2129 }
2130 
2131 static void pc_memory_unplug_request(HotplugHandler *hotplug_dev,
2132                                      DeviceState *dev, Error **errp)
2133 {
2134     Error *local_err = NULL;
2135     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2136 
2137     /*
2138      * When -no-acpi is used with Q35 machine type, no ACPI is built,
2139      * but pcms->acpi_dev is still created. Check !acpi_enabled in
2140      * addition to cover this case.
2141      */
2142     if (!pcms->acpi_dev || !acpi_enabled) {
2143         error_setg(&local_err,
2144                    "memory hotplug is not enabled: missing acpi device or acpi disabled");
2145         goto out;
2146     }
2147 
2148     if (object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM)) {
2149         error_setg(&local_err,
2150                    "nvdimm device hot unplug is not supported yet.");
2151         goto out;
2152     }
2153 
2154     hotplug_handler_unplug_request(HOTPLUG_HANDLER(pcms->acpi_dev), dev,
2155                                    &local_err);
2156 out:
2157     error_propagate(errp, local_err);
2158 }
2159 
2160 static void pc_memory_unplug(HotplugHandler *hotplug_dev,
2161                              DeviceState *dev, Error **errp)
2162 {
2163     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2164     Error *local_err = NULL;
2165 
2166     hotplug_handler_unplug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
2167     if (local_err) {
2168         goto out;
2169     }
2170 
2171     pc_dimm_unplug(PC_DIMM(dev), MACHINE(pcms));
2172     object_property_set_bool(OBJECT(dev), false, "realized", NULL);
2173  out:
2174     error_propagate(errp, local_err);
2175 }
2176 
2177 static int pc_apic_cmp(const void *a, const void *b)
2178 {
2179    CPUArchId *apic_a = (CPUArchId *)a;
2180    CPUArchId *apic_b = (CPUArchId *)b;
2181 
2182    return apic_a->arch_id - apic_b->arch_id;
2183 }
2184 
2185 /* returns pointer to CPUArchId descriptor that matches CPU's apic_id
2186  * in ms->possible_cpus->cpus, if ms->possible_cpus->cpus has no
2187  * entry corresponding to CPU's apic_id returns NULL.
2188  */
2189 static CPUArchId *pc_find_cpu_slot(MachineState *ms, uint32_t id, int *idx)
2190 {
2191     CPUArchId apic_id, *found_cpu;
2192 
2193     apic_id.arch_id = id;
2194     found_cpu = bsearch(&apic_id, ms->possible_cpus->cpus,
2195         ms->possible_cpus->len, sizeof(*ms->possible_cpus->cpus),
2196         pc_apic_cmp);
2197     if (found_cpu && idx) {
2198         *idx = found_cpu - ms->possible_cpus->cpus;
2199     }
2200     return found_cpu;
2201 }
2202 
2203 static void pc_cpu_plug(HotplugHandler *hotplug_dev,
2204                         DeviceState *dev, Error **errp)
2205 {
2206     CPUArchId *found_cpu;
2207     Error *local_err = NULL;
2208     X86CPU *cpu = X86_CPU(dev);
2209     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2210 
2211     if (pcms->acpi_dev) {
2212         hotplug_handler_plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
2213         if (local_err) {
2214             goto out;
2215         }
2216     }
2217 
2218     /* increment the number of CPUs */
2219     pcms->boot_cpus++;
2220     if (pcms->rtc) {
2221         rtc_set_cpus_count(pcms->rtc, pcms->boot_cpus);
2222     }
2223     if (pcms->fw_cfg) {
2224         fw_cfg_modify_i16(pcms->fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
2225     }
2226 
2227     found_cpu = pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, NULL);
2228     found_cpu->cpu = OBJECT(dev);
2229 out:
2230     error_propagate(errp, local_err);
2231 }
2232 static void pc_cpu_unplug_request_cb(HotplugHandler *hotplug_dev,
2233                                      DeviceState *dev, Error **errp)
2234 {
2235     int idx = -1;
2236     Error *local_err = NULL;
2237     X86CPU *cpu = X86_CPU(dev);
2238     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2239 
2240     if (!pcms->acpi_dev) {
2241         error_setg(&local_err, "CPU hot unplug not supported without ACPI");
2242         goto out;
2243     }
2244 
2245     pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, &idx);
2246     assert(idx != -1);
2247     if (idx == 0) {
2248         error_setg(&local_err, "Boot CPU is unpluggable");
2249         goto out;
2250     }
2251 
2252     hotplug_handler_unplug_request(HOTPLUG_HANDLER(pcms->acpi_dev), dev,
2253                                    &local_err);
2254     if (local_err) {
2255         goto out;
2256     }
2257 
2258  out:
2259     error_propagate(errp, local_err);
2260 
2261 }
2262 
2263 static void pc_cpu_unplug_cb(HotplugHandler *hotplug_dev,
2264                              DeviceState *dev, Error **errp)
2265 {
2266     CPUArchId *found_cpu;
2267     Error *local_err = NULL;
2268     X86CPU *cpu = X86_CPU(dev);
2269     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2270 
2271     hotplug_handler_unplug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
2272     if (local_err) {
2273         goto out;
2274     }
2275 
2276     found_cpu = pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, NULL);
2277     found_cpu->cpu = NULL;
2278     object_property_set_bool(OBJECT(dev), false, "realized", NULL);
2279 
2280     /* decrement the number of CPUs */
2281     pcms->boot_cpus--;
2282     /* Update the number of CPUs in CMOS */
2283     rtc_set_cpus_count(pcms->rtc, pcms->boot_cpus);
2284     fw_cfg_modify_i16(pcms->fw_cfg, FW_CFG_NB_CPUS, pcms->boot_cpus);
2285  out:
2286     error_propagate(errp, local_err);
2287 }
2288 
2289 static void pc_cpu_pre_plug(HotplugHandler *hotplug_dev,
2290                             DeviceState *dev, Error **errp)
2291 {
2292     int idx;
2293     CPUState *cs;
2294     CPUArchId *cpu_slot;
2295     X86CPUTopoInfo topo;
2296     X86CPU *cpu = X86_CPU(dev);
2297     MachineState *ms = MACHINE(hotplug_dev);
2298     PCMachineState *pcms = PC_MACHINE(hotplug_dev);
2299 
2300     if(!object_dynamic_cast(OBJECT(cpu), ms->cpu_type)) {
2301         error_setg(errp, "Invalid CPU type, expected cpu type: '%s'",
2302                    ms->cpu_type);
2303         return;
2304     }
2305 
2306     /* if APIC ID is not set, set it based on socket/core/thread properties */
2307     if (cpu->apic_id == UNASSIGNED_APIC_ID) {
2308         int max_socket = (max_cpus - 1) / smp_threads / smp_cores;
2309 
2310         if (cpu->socket_id < 0) {
2311             error_setg(errp, "CPU socket-id is not set");
2312             return;
2313         } else if (cpu->socket_id > max_socket) {
2314             error_setg(errp, "Invalid CPU socket-id: %u must be in range 0:%u",
2315                        cpu->socket_id, max_socket);
2316             return;
2317         }
2318         if (cpu->core_id < 0) {
2319             error_setg(errp, "CPU core-id is not set");
2320             return;
2321         } else if (cpu->core_id > (smp_cores - 1)) {
2322             error_setg(errp, "Invalid CPU core-id: %u must be in range 0:%u",
2323                        cpu->core_id, smp_cores - 1);
2324             return;
2325         }
2326         if (cpu->thread_id < 0) {
2327             error_setg(errp, "CPU thread-id is not set");
2328             return;
2329         } else if (cpu->thread_id > (smp_threads - 1)) {
2330             error_setg(errp, "Invalid CPU thread-id: %u must be in range 0:%u",
2331                        cpu->thread_id, smp_threads - 1);
2332             return;
2333         }
2334 
2335         topo.pkg_id = cpu->socket_id;
2336         topo.core_id = cpu->core_id;
2337         topo.smt_id = cpu->thread_id;
2338         cpu->apic_id = apicid_from_topo_ids(smp_cores, smp_threads, &topo);
2339     }
2340 
2341     cpu_slot = pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, &idx);
2342     if (!cpu_slot) {
2343         MachineState *ms = MACHINE(pcms);
2344 
2345         x86_topo_ids_from_apicid(cpu->apic_id, smp_cores, smp_threads, &topo);
2346         error_setg(errp, "Invalid CPU [socket: %u, core: %u, thread: %u] with"
2347                   " APIC ID %" PRIu32 ", valid index range 0:%d",
2348                    topo.pkg_id, topo.core_id, topo.smt_id, cpu->apic_id,
2349                    ms->possible_cpus->len - 1);
2350         return;
2351     }
2352 
2353     if (cpu_slot->cpu) {
2354         error_setg(errp, "CPU[%d] with APIC ID %" PRIu32 " exists",
2355                    idx, cpu->apic_id);
2356         return;
2357     }
2358 
2359     /* if 'address' properties socket-id/core-id/thread-id are not set, set them
2360      * so that machine_query_hotpluggable_cpus would show correct values
2361      */
2362     /* TODO: move socket_id/core_id/thread_id checks into x86_cpu_realizefn()
2363      * once -smp refactoring is complete and there will be CPU private
2364      * CPUState::nr_cores and CPUState::nr_threads fields instead of globals */
2365     x86_topo_ids_from_apicid(cpu->apic_id, smp_cores, smp_threads, &topo);
2366     if (cpu->socket_id != -1 && cpu->socket_id != topo.pkg_id) {
2367         error_setg(errp, "property socket-id: %u doesn't match set apic-id:"
2368             " 0x%x (socket-id: %u)", cpu->socket_id, cpu->apic_id, topo.pkg_id);
2369         return;
2370     }
2371     cpu->socket_id = topo.pkg_id;
2372 
2373     if (cpu->core_id != -1 && cpu->core_id != topo.core_id) {
2374         error_setg(errp, "property core-id: %u doesn't match set apic-id:"
2375             " 0x%x (core-id: %u)", cpu->core_id, cpu->apic_id, topo.core_id);
2376         return;
2377     }
2378     cpu->core_id = topo.core_id;
2379 
2380     if (cpu->thread_id != -1 && cpu->thread_id != topo.smt_id) {
2381         error_setg(errp, "property thread-id: %u doesn't match set apic-id:"
2382             " 0x%x (thread-id: %u)", cpu->thread_id, cpu->apic_id, topo.smt_id);
2383         return;
2384     }
2385     cpu->thread_id = topo.smt_id;
2386 
2387     if (cpu->hyperv_vpindex && !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