]> git.proxmox.com Git - mirror_qemu.git/blob - hw/i386/pc.c
hw/i386/pc: Convert DPRINTF() to trace events
[mirror_qemu.git] / hw / i386 / pc.c
1 /*
2 * QEMU PC System Emulator
3 *
4 * Copyright (c) 2003-2004 Fabrice Bellard
5 *
6 * Permission is hereby granted, free of charge, to any person obtaining a copy
7 * of this software and associated documentation files (the "Software"), to deal
8 * in the Software without restriction, including without limitation the rights
9 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
10 * copies of the Software, and to permit persons to whom the Software is
11 * furnished to do so, subject to the following conditions:
12 *
13 * The above copyright notice and this permission notice shall be included in
14 * all copies or substantial portions of the Software.
15 *
16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
19 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
22 * THE SOFTWARE.
23 */
24
25 #include "qemu/osdep.h"
26 #include "qemu/units.h"
27 #include "hw/i386/x86.h"
28 #include "hw/i386/pc.h"
29 #include "hw/char/serial.h"
30 #include "hw/char/parallel.h"
31 #include "hw/i386/apic.h"
32 #include "hw/i386/topology.h"
33 #include "hw/i386/fw_cfg.h"
34 #include "sysemu/cpus.h"
35 #include "hw/block/fdc.h"
36 #include "hw/ide.h"
37 #include "hw/pci/pci.h"
38 #include "hw/pci/pci_bus.h"
39 #include "hw/nvram/fw_cfg.h"
40 #include "hw/timer/hpet.h"
41 #include "hw/firmware/smbios.h"
42 #include "hw/loader.h"
43 #include "elf.h"
44 #include "migration/vmstate.h"
45 #include "multiboot.h"
46 #include "hw/rtc/mc146818rtc.h"
47 #include "hw/intc/i8259.h"
48 #include "hw/dma/i8257.h"
49 #include "hw/timer/i8254.h"
50 #include "hw/input/i8042.h"
51 #include "hw/irq.h"
52 #include "hw/audio/pcspk.h"
53 #include "hw/pci/msi.h"
54 #include "hw/sysbus.h"
55 #include "sysemu/sysemu.h"
56 #include "sysemu/tcg.h"
57 #include "sysemu/numa.h"
58 #include "sysemu/kvm.h"
59 #include "sysemu/qtest.h"
60 #include "sysemu/reset.h"
61 #include "sysemu/runstate.h"
62 #include "kvm_i386.h"
63 #include "hw/xen/xen.h"
64 #include "hw/xen/start_info.h"
65 #include "ui/qemu-spice.h"
66 #include "exec/memory.h"
67 #include "exec/address-spaces.h"
68 #include "sysemu/arch_init.h"
69 #include "qemu/bitmap.h"
70 #include "qemu/config-file.h"
71 #include "qemu/error-report.h"
72 #include "qemu/option.h"
73 #include "qemu/cutils.h"
74 #include "hw/acpi/acpi.h"
75 #include "hw/acpi/cpu_hotplug.h"
76 #include "hw/boards.h"
77 #include "acpi-build.h"
78 #include "hw/mem/pc-dimm.h"
79 #include "qapi/error.h"
80 #include "qapi/qapi-visit-common.h"
81 #include "qapi/visitor.h"
82 #include "hw/core/cpu.h"
83 #include "hw/usb.h"
84 #include "hw/i386/intel_iommu.h"
85 #include "hw/net/ne2000-isa.h"
86 #include "standard-headers/asm-x86/bootparam.h"
87 #include "hw/virtio/virtio-pmem-pci.h"
88 #include "hw/mem/memory-device.h"
89 #include "sysemu/replay.h"
90 #include "qapi/qmp/qerror.h"
91 #include "config-devices.h"
92 #include "e820_memory_layout.h"
93 #include "fw_cfg.h"
94 #include "trace.h"
95
96 GlobalProperty pc_compat_4_2[] = {};
97 const size_t pc_compat_4_2_len = G_N_ELEMENTS(pc_compat_4_2);
98
99 GlobalProperty pc_compat_4_1[] = {};
100 const size_t pc_compat_4_1_len = G_N_ELEMENTS(pc_compat_4_1);
101
102 GlobalProperty pc_compat_4_0[] = {};
103 const size_t pc_compat_4_0_len = G_N_ELEMENTS(pc_compat_4_0);
104
105 GlobalProperty pc_compat_3_1[] = {
106 { "intel-iommu", "dma-drain", "off" },
107 { "Opteron_G3" "-" TYPE_X86_CPU, "rdtscp", "off" },
108 { "Opteron_G4" "-" TYPE_X86_CPU, "rdtscp", "off" },
109 { "Opteron_G4" "-" TYPE_X86_CPU, "npt", "off" },
110 { "Opteron_G4" "-" TYPE_X86_CPU, "nrip-save", "off" },
111 { "Opteron_G5" "-" TYPE_X86_CPU, "rdtscp", "off" },
112 { "Opteron_G5" "-" TYPE_X86_CPU, "npt", "off" },
113 { "Opteron_G5" "-" TYPE_X86_CPU, "nrip-save", "off" },
114 { "EPYC" "-" TYPE_X86_CPU, "npt", "off" },
115 { "EPYC" "-" TYPE_X86_CPU, "nrip-save", "off" },
116 { "EPYC-IBPB" "-" TYPE_X86_CPU, "npt", "off" },
117 { "EPYC-IBPB" "-" TYPE_X86_CPU, "nrip-save", "off" },
118 { "Skylake-Client" "-" TYPE_X86_CPU, "mpx", "on" },
119 { "Skylake-Client-IBRS" "-" TYPE_X86_CPU, "mpx", "on" },
120 { "Skylake-Server" "-" TYPE_X86_CPU, "mpx", "on" },
121 { "Skylake-Server-IBRS" "-" TYPE_X86_CPU, "mpx", "on" },
122 { "Cascadelake-Server" "-" TYPE_X86_CPU, "mpx", "on" },
123 { "Icelake-Client" "-" TYPE_X86_CPU, "mpx", "on" },
124 { "Icelake-Server" "-" TYPE_X86_CPU, "mpx", "on" },
125 { "Cascadelake-Server" "-" TYPE_X86_CPU, "stepping", "5" },
126 { TYPE_X86_CPU, "x-intel-pt-auto-level", "off" },
127 };
128 const size_t pc_compat_3_1_len = G_N_ELEMENTS(pc_compat_3_1);
129
130 GlobalProperty pc_compat_3_0[] = {
131 { TYPE_X86_CPU, "x-hv-synic-kvm-only", "on" },
132 { "Skylake-Server" "-" TYPE_X86_CPU, "pku", "off" },
133 { "Skylake-Server-IBRS" "-" TYPE_X86_CPU, "pku", "off" },
134 };
135 const size_t pc_compat_3_0_len = G_N_ELEMENTS(pc_compat_3_0);
136
137 GlobalProperty pc_compat_2_12[] = {
138 { TYPE_X86_CPU, "legacy-cache", "on" },
139 { TYPE_X86_CPU, "topoext", "off" },
140 { "EPYC-" TYPE_X86_CPU, "xlevel", "0x8000000a" },
141 { "EPYC-IBPB-" TYPE_X86_CPU, "xlevel", "0x8000000a" },
142 };
143 const size_t pc_compat_2_12_len = G_N_ELEMENTS(pc_compat_2_12);
144
145 GlobalProperty pc_compat_2_11[] = {
146 { TYPE_X86_CPU, "x-migrate-smi-count", "off" },
147 { "Skylake-Server" "-" TYPE_X86_CPU, "clflushopt", "off" },
148 };
149 const size_t pc_compat_2_11_len = G_N_ELEMENTS(pc_compat_2_11);
150
151 GlobalProperty pc_compat_2_10[] = {
152 { TYPE_X86_CPU, "x-hv-max-vps", "0x40" },
153 { "i440FX-pcihost", "x-pci-hole64-fix", "off" },
154 { "q35-pcihost", "x-pci-hole64-fix", "off" },
155 };
156 const size_t pc_compat_2_10_len = G_N_ELEMENTS(pc_compat_2_10);
157
158 GlobalProperty pc_compat_2_9[] = {
159 { "mch", "extended-tseg-mbytes", "0" },
160 };
161 const size_t pc_compat_2_9_len = G_N_ELEMENTS(pc_compat_2_9);
162
163 GlobalProperty pc_compat_2_8[] = {
164 { TYPE_X86_CPU, "tcg-cpuid", "off" },
165 { "kvmclock", "x-mach-use-reliable-get-clock", "off" },
166 { "ICH9-LPC", "x-smi-broadcast", "off" },
167 { TYPE_X86_CPU, "vmware-cpuid-freq", "off" },
168 { "Haswell-" TYPE_X86_CPU, "stepping", "1" },
169 };
170 const size_t pc_compat_2_8_len = G_N_ELEMENTS(pc_compat_2_8);
171
172 GlobalProperty pc_compat_2_7[] = {
173 { TYPE_X86_CPU, "l3-cache", "off" },
174 { TYPE_X86_CPU, "full-cpuid-auto-level", "off" },
175 { "Opteron_G3" "-" TYPE_X86_CPU, "family", "15" },
176 { "Opteron_G3" "-" TYPE_X86_CPU, "model", "6" },
177 { "Opteron_G3" "-" TYPE_X86_CPU, "stepping", "1" },
178 { "isa-pcspk", "migrate", "off" },
179 };
180 const size_t pc_compat_2_7_len = G_N_ELEMENTS(pc_compat_2_7);
181
182 GlobalProperty pc_compat_2_6[] = {
183 { TYPE_X86_CPU, "cpuid-0xb", "off" },
184 { "vmxnet3", "romfile", "" },
185 { TYPE_X86_CPU, "fill-mtrr-mask", "off" },
186 { "apic-common", "legacy-instance-id", "on", }
187 };
188 const size_t pc_compat_2_6_len = G_N_ELEMENTS(pc_compat_2_6);
189
190 GlobalProperty pc_compat_2_5[] = {};
191 const size_t pc_compat_2_5_len = G_N_ELEMENTS(pc_compat_2_5);
192
193 GlobalProperty pc_compat_2_4[] = {
194 PC_CPU_MODEL_IDS("2.4.0")
195 { "Haswell-" TYPE_X86_CPU, "abm", "off" },
196 { "Haswell-noTSX-" TYPE_X86_CPU, "abm", "off" },
197 { "Broadwell-" TYPE_X86_CPU, "abm", "off" },
198 { "Broadwell-noTSX-" TYPE_X86_CPU, "abm", "off" },
199 { "host" "-" TYPE_X86_CPU, "host-cache-info", "on" },
200 { TYPE_X86_CPU, "check", "off" },
201 { "qemu64" "-" TYPE_X86_CPU, "sse4a", "on" },
202 { "qemu64" "-" TYPE_X86_CPU, "abm", "on" },
203 { "qemu64" "-" TYPE_X86_CPU, "popcnt", "on" },
204 { "qemu32" "-" TYPE_X86_CPU, "popcnt", "on" },
205 { "Opteron_G2" "-" TYPE_X86_CPU, "rdtscp", "on" },
206 { "Opteron_G3" "-" TYPE_X86_CPU, "rdtscp", "on" },
207 { "Opteron_G4" "-" TYPE_X86_CPU, "rdtscp", "on" },
208 { "Opteron_G5" "-" TYPE_X86_CPU, "rdtscp", "on", }
209 };
210 const size_t pc_compat_2_4_len = G_N_ELEMENTS(pc_compat_2_4);
211
212 GlobalProperty pc_compat_2_3[] = {
213 PC_CPU_MODEL_IDS("2.3.0")
214 { TYPE_X86_CPU, "arat", "off" },
215 { "qemu64" "-" TYPE_X86_CPU, "min-level", "4" },
216 { "kvm64" "-" TYPE_X86_CPU, "min-level", "5" },
217 { "pentium3" "-" TYPE_X86_CPU, "min-level", "2" },
218 { "n270" "-" TYPE_X86_CPU, "min-level", "5" },
219 { "Conroe" "-" TYPE_X86_CPU, "min-level", "4" },
220 { "Penryn" "-" TYPE_X86_CPU, "min-level", "4" },
221 { "Nehalem" "-" TYPE_X86_CPU, "min-level", "4" },
222 { "n270" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
223 { "Penryn" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
224 { "Conroe" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
225 { "Nehalem" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
226 { "Westmere" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
227 { "SandyBridge" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
228 { "IvyBridge" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
229 { "Haswell" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
230 { "Haswell-noTSX" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
231 { "Broadwell" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
232 { "Broadwell-noTSX" "-" TYPE_X86_CPU, "min-xlevel", "0x8000000a" },
233 { TYPE_X86_CPU, "kvm-no-smi-migration", "on" },
234 };
235 const size_t pc_compat_2_3_len = G_N_ELEMENTS(pc_compat_2_3);
236
237 GlobalProperty pc_compat_2_2[] = {
238 PC_CPU_MODEL_IDS("2.2.0")
239 { "kvm64" "-" TYPE_X86_CPU, "vme", "off" },
240 { "kvm32" "-" TYPE_X86_CPU, "vme", "off" },
241 { "Conroe" "-" TYPE_X86_CPU, "vme", "off" },
242 { "Penryn" "-" TYPE_X86_CPU, "vme", "off" },
243 { "Nehalem" "-" TYPE_X86_CPU, "vme", "off" },
244 { "Westmere" "-" TYPE_X86_CPU, "vme", "off" },
245 { "SandyBridge" "-" TYPE_X86_CPU, "vme", "off" },
246 { "Haswell" "-" TYPE_X86_CPU, "vme", "off" },
247 { "Broadwell" "-" TYPE_X86_CPU, "vme", "off" },
248 { "Opteron_G1" "-" TYPE_X86_CPU, "vme", "off" },
249 { "Opteron_G2" "-" TYPE_X86_CPU, "vme", "off" },
250 { "Opteron_G3" "-" TYPE_X86_CPU, "vme", "off" },
251 { "Opteron_G4" "-" TYPE_X86_CPU, "vme", "off" },
252 { "Opteron_G5" "-" TYPE_X86_CPU, "vme", "off" },
253 { "Haswell" "-" TYPE_X86_CPU, "f16c", "off" },
254 { "Haswell" "-" TYPE_X86_CPU, "rdrand", "off" },
255 { "Broadwell" "-" TYPE_X86_CPU, "f16c", "off" },
256 { "Broadwell" "-" TYPE_X86_CPU, "rdrand", "off" },
257 };
258 const size_t pc_compat_2_2_len = G_N_ELEMENTS(pc_compat_2_2);
259
260 GlobalProperty pc_compat_2_1[] = {
261 PC_CPU_MODEL_IDS("2.1.0")
262 { "coreduo" "-" TYPE_X86_CPU, "vmx", "on" },
263 { "core2duo" "-" TYPE_X86_CPU, "vmx", "on" },
264 };
265 const size_t pc_compat_2_1_len = G_N_ELEMENTS(pc_compat_2_1);
266
267 GlobalProperty pc_compat_2_0[] = {
268 PC_CPU_MODEL_IDS("2.0.0")
269 { "virtio-scsi-pci", "any_layout", "off" },
270 { "PIIX4_PM", "memory-hotplug-support", "off" },
271 { "apic", "version", "0x11" },
272 { "nec-usb-xhci", "superspeed-ports-first", "off" },
273 { "nec-usb-xhci", "force-pcie-endcap", "on" },
274 { "pci-serial", "prog_if", "0" },
275 { "pci-serial-2x", "prog_if", "0" },
276 { "pci-serial-4x", "prog_if", "0" },
277 { "virtio-net-pci", "guest_announce", "off" },
278 { "ICH9-LPC", "memory-hotplug-support", "off" },
279 { "xio3130-downstream", COMPAT_PROP_PCP, "off" },
280 { "ioh3420", COMPAT_PROP_PCP, "off" },
281 };
282 const size_t pc_compat_2_0_len = G_N_ELEMENTS(pc_compat_2_0);
283
284 GlobalProperty pc_compat_1_7[] = {
285 PC_CPU_MODEL_IDS("1.7.0")
286 { TYPE_USB_DEVICE, "msos-desc", "no" },
287 { "PIIX4_PM", "acpi-pci-hotplug-with-bridge-support", "off" },
288 { "hpet", HPET_INTCAP, "4" },
289 };
290 const size_t pc_compat_1_7_len = G_N_ELEMENTS(pc_compat_1_7);
291
292 GlobalProperty pc_compat_1_6[] = {
293 PC_CPU_MODEL_IDS("1.6.0")
294 { "e1000", "mitigation", "off" },
295 { "qemu64-" TYPE_X86_CPU, "model", "2" },
296 { "qemu32-" TYPE_X86_CPU, "model", "3" },
297 { "i440FX-pcihost", "short_root_bus", "1" },
298 { "q35-pcihost", "short_root_bus", "1" },
299 };
300 const size_t pc_compat_1_6_len = G_N_ELEMENTS(pc_compat_1_6);
301
302 GlobalProperty pc_compat_1_5[] = {
303 PC_CPU_MODEL_IDS("1.5.0")
304 { "Conroe-" TYPE_X86_CPU, "model", "2" },
305 { "Conroe-" TYPE_X86_CPU, "min-level", "2" },
306 { "Penryn-" TYPE_X86_CPU, "model", "2" },
307 { "Penryn-" TYPE_X86_CPU, "min-level", "2" },
308 { "Nehalem-" TYPE_X86_CPU, "model", "2" },
309 { "Nehalem-" TYPE_X86_CPU, "min-level", "2" },
310 { "virtio-net-pci", "any_layout", "off" },
311 { TYPE_X86_CPU, "pmu", "on" },
312 { "i440FX-pcihost", "short_root_bus", "0" },
313 { "q35-pcihost", "short_root_bus", "0" },
314 };
315 const size_t pc_compat_1_5_len = G_N_ELEMENTS(pc_compat_1_5);
316
317 GlobalProperty pc_compat_1_4[] = {
318 PC_CPU_MODEL_IDS("1.4.0")
319 { "scsi-hd", "discard_granularity", "0" },
320 { "scsi-cd", "discard_granularity", "0" },
321 { "scsi-disk", "discard_granularity", "0" },
322 { "ide-hd", "discard_granularity", "0" },
323 { "ide-cd", "discard_granularity", "0" },
324 { "ide-drive", "discard_granularity", "0" },
325 { "virtio-blk-pci", "discard_granularity", "0" },
326 /* DEV_NVECTORS_UNSPECIFIED as a uint32_t string: */
327 { "virtio-serial-pci", "vectors", "0xFFFFFFFF" },
328 { "virtio-net-pci", "ctrl_guest_offloads", "off" },
329 { "e1000", "romfile", "pxe-e1000.rom" },
330 { "ne2k_pci", "romfile", "pxe-ne2k_pci.rom" },
331 { "pcnet", "romfile", "pxe-pcnet.rom" },
332 { "rtl8139", "romfile", "pxe-rtl8139.rom" },
333 { "virtio-net-pci", "romfile", "pxe-virtio.rom" },
334 { "486-" TYPE_X86_CPU, "model", "0" },
335 { "n270" "-" TYPE_X86_CPU, "movbe", "off" },
336 { "Westmere" "-" TYPE_X86_CPU, "pclmulqdq", "off" },
337 };
338 const size_t pc_compat_1_4_len = G_N_ELEMENTS(pc_compat_1_4);
339
340 void gsi_handler(void *opaque, int n, int level)
341 {
342 GSIState *s = opaque;
343
344 trace_pc_gsi_interrupt(n, level);
345 if (n < ISA_NUM_IRQS) {
346 qemu_set_irq(s->i8259_irq[n], level);
347 }
348 qemu_set_irq(s->ioapic_irq[n], level);
349 }
350
351 GSIState *pc_gsi_create(qemu_irq **irqs, bool pci_enabled)
352 {
353 GSIState *s;
354
355 s = g_new0(GSIState, 1);
356 if (kvm_ioapic_in_kernel()) {
357 kvm_pc_setup_irq_routing(pci_enabled);
358 *irqs = qemu_allocate_irqs(kvm_pc_gsi_handler, s, GSI_NUM_PINS);
359 } else {
360 *irqs = qemu_allocate_irqs(gsi_handler, s, GSI_NUM_PINS);
361 }
362
363 return s;
364 }
365
366 static void ioport80_write(void *opaque, hwaddr addr, uint64_t data,
367 unsigned size)
368 {
369 }
370
371 static uint64_t ioport80_read(void *opaque, hwaddr addr, unsigned size)
372 {
373 return 0xffffffffffffffffULL;
374 }
375
376 /* MSDOS compatibility mode FPU exception support */
377 static void ioportF0_write(void *opaque, hwaddr addr, uint64_t data,
378 unsigned size)
379 {
380 if (tcg_enabled()) {
381 cpu_set_ignne();
382 }
383 }
384
385 static uint64_t ioportF0_read(void *opaque, hwaddr addr, unsigned size)
386 {
387 return 0xffffffffffffffffULL;
388 }
389
390 /* TSC handling */
391 uint64_t cpu_get_tsc(CPUX86State *env)
392 {
393 return cpu_get_ticks();
394 }
395
396 /* IRQ handling */
397 int cpu_get_pic_interrupt(CPUX86State *env)
398 {
399 X86CPU *cpu = env_archcpu(env);
400 int intno;
401
402 if (!kvm_irqchip_in_kernel()) {
403 intno = apic_get_interrupt(cpu->apic_state);
404 if (intno >= 0) {
405 return intno;
406 }
407 /* read the irq from the PIC */
408 if (!apic_accept_pic_intr(cpu->apic_state)) {
409 return -1;
410 }
411 }
412
413 intno = pic_read_irq(isa_pic);
414 return intno;
415 }
416
417 static void pic_irq_request(void *opaque, int irq, int level)
418 {
419 CPUState *cs = first_cpu;
420 X86CPU *cpu = X86_CPU(cs);
421
422 trace_pc_pic_interrupt(irq, level);
423 if (cpu->apic_state && !kvm_irqchip_in_kernel()) {
424 CPU_FOREACH(cs) {
425 cpu = X86_CPU(cs);
426 if (apic_accept_pic_intr(cpu->apic_state)) {
427 apic_deliver_pic_intr(cpu->apic_state, level);
428 }
429 }
430 } else {
431 if (level) {
432 cpu_interrupt(cs, CPU_INTERRUPT_HARD);
433 } else {
434 cpu_reset_interrupt(cs, CPU_INTERRUPT_HARD);
435 }
436 }
437 }
438
439 /* PC cmos mappings */
440
441 #define REG_EQUIPMENT_BYTE 0x14
442
443 int cmos_get_fd_drive_type(FloppyDriveType fd0)
444 {
445 int val;
446
447 switch (fd0) {
448 case FLOPPY_DRIVE_TYPE_144:
449 /* 1.44 Mb 3"5 drive */
450 val = 4;
451 break;
452 case FLOPPY_DRIVE_TYPE_288:
453 /* 2.88 Mb 3"5 drive */
454 val = 5;
455 break;
456 case FLOPPY_DRIVE_TYPE_120:
457 /* 1.2 Mb 5"5 drive */
458 val = 2;
459 break;
460 case FLOPPY_DRIVE_TYPE_NONE:
461 default:
462 val = 0;
463 break;
464 }
465 return val;
466 }
467
468 static void cmos_init_hd(ISADevice *s, int type_ofs, int info_ofs,
469 int16_t cylinders, int8_t heads, int8_t sectors)
470 {
471 rtc_set_memory(s, type_ofs, 47);
472 rtc_set_memory(s, info_ofs, cylinders);
473 rtc_set_memory(s, info_ofs + 1, cylinders >> 8);
474 rtc_set_memory(s, info_ofs + 2, heads);
475 rtc_set_memory(s, info_ofs + 3, 0xff);
476 rtc_set_memory(s, info_ofs + 4, 0xff);
477 rtc_set_memory(s, info_ofs + 5, 0xc0 | ((heads > 8) << 3));
478 rtc_set_memory(s, info_ofs + 6, cylinders);
479 rtc_set_memory(s, info_ofs + 7, cylinders >> 8);
480 rtc_set_memory(s, info_ofs + 8, sectors);
481 }
482
483 /* convert boot_device letter to something recognizable by the bios */
484 static int boot_device2nibble(char boot_device)
485 {
486 switch(boot_device) {
487 case 'a':
488 case 'b':
489 return 0x01; /* floppy boot */
490 case 'c':
491 return 0x02; /* hard drive boot */
492 case 'd':
493 return 0x03; /* CD-ROM boot */
494 case 'n':
495 return 0x04; /* Network boot */
496 }
497 return 0;
498 }
499
500 static void set_boot_dev(ISADevice *s, const char *boot_device, Error **errp)
501 {
502 #define PC_MAX_BOOT_DEVICES 3
503 int nbds, bds[3] = { 0, };
504 int i;
505
506 nbds = strlen(boot_device);
507 if (nbds > PC_MAX_BOOT_DEVICES) {
508 error_setg(errp, "Too many boot devices for PC");
509 return;
510 }
511 for (i = 0; i < nbds; i++) {
512 bds[i] = boot_device2nibble(boot_device[i]);
513 if (bds[i] == 0) {
514 error_setg(errp, "Invalid boot device for PC: '%c'",
515 boot_device[i]);
516 return;
517 }
518 }
519 rtc_set_memory(s, 0x3d, (bds[1] << 4) | bds[0]);
520 rtc_set_memory(s, 0x38, (bds[2] << 4) | (fd_bootchk ? 0x0 : 0x1));
521 }
522
523 static void pc_boot_set(void *opaque, const char *boot_device, Error **errp)
524 {
525 set_boot_dev(opaque, boot_device, errp);
526 }
527
528 static void pc_cmos_init_floppy(ISADevice *rtc_state, ISADevice *floppy)
529 {
530 int val, nb, i;
531 FloppyDriveType fd_type[2] = { FLOPPY_DRIVE_TYPE_NONE,
532 FLOPPY_DRIVE_TYPE_NONE };
533
534 /* floppy type */
535 if (floppy) {
536 for (i = 0; i < 2; i++) {
537 fd_type[i] = isa_fdc_get_drive_type(floppy, i);
538 }
539 }
540 val = (cmos_get_fd_drive_type(fd_type[0]) << 4) |
541 cmos_get_fd_drive_type(fd_type[1]);
542 rtc_set_memory(rtc_state, 0x10, val);
543
544 val = rtc_get_memory(rtc_state, REG_EQUIPMENT_BYTE);
545 nb = 0;
546 if (fd_type[0] != FLOPPY_DRIVE_TYPE_NONE) {
547 nb++;
548 }
549 if (fd_type[1] != FLOPPY_DRIVE_TYPE_NONE) {
550 nb++;
551 }
552 switch (nb) {
553 case 0:
554 break;
555 case 1:
556 val |= 0x01; /* 1 drive, ready for boot */
557 break;
558 case 2:
559 val |= 0x41; /* 2 drives, ready for boot */
560 break;
561 }
562 rtc_set_memory(rtc_state, REG_EQUIPMENT_BYTE, val);
563 }
564
565 typedef struct pc_cmos_init_late_arg {
566 ISADevice *rtc_state;
567 BusState *idebus[2];
568 } pc_cmos_init_late_arg;
569
570 typedef struct check_fdc_state {
571 ISADevice *floppy;
572 bool multiple;
573 } CheckFdcState;
574
575 static int check_fdc(Object *obj, void *opaque)
576 {
577 CheckFdcState *state = opaque;
578 Object *fdc;
579 uint32_t iobase;
580 Error *local_err = NULL;
581
582 fdc = object_dynamic_cast(obj, TYPE_ISA_FDC);
583 if (!fdc) {
584 return 0;
585 }
586
587 iobase = object_property_get_uint(obj, "iobase", &local_err);
588 if (local_err || iobase != 0x3f0) {
589 error_free(local_err);
590 return 0;
591 }
592
593 if (state->floppy) {
594 state->multiple = true;
595 } else {
596 state->floppy = ISA_DEVICE(obj);
597 }
598 return 0;
599 }
600
601 static const char * const fdc_container_path[] = {
602 "/unattached", "/peripheral", "/peripheral-anon"
603 };
604
605 /*
606 * Locate the FDC at IO address 0x3f0, in order to configure the CMOS registers
607 * and ACPI objects.
608 */
609 ISADevice *pc_find_fdc0(void)
610 {
611 int i;
612 Object *container;
613 CheckFdcState state = { 0 };
614
615 for (i = 0; i < ARRAY_SIZE(fdc_container_path); i++) {
616 container = container_get(qdev_get_machine(), fdc_container_path[i]);
617 object_child_foreach(container, check_fdc, &state);
618 }
619
620 if (state.multiple) {
621 warn_report("multiple floppy disk controllers with "
622 "iobase=0x3f0 have been found");
623 error_printf("the one being picked for CMOS setup might not reflect "
624 "your intent");
625 }
626
627 return state.floppy;
628 }
629
630 static void pc_cmos_init_late(void *opaque)
631 {
632 pc_cmos_init_late_arg *arg = opaque;
633 ISADevice *s = arg->rtc_state;
634 int16_t cylinders;
635 int8_t heads, sectors;
636 int val;
637 int i, trans;
638
639 val = 0;
640 if (arg->idebus[0] && ide_get_geometry(arg->idebus[0], 0,
641 &cylinders, &heads, &sectors) >= 0) {
642 cmos_init_hd(s, 0x19, 0x1b, cylinders, heads, sectors);
643 val |= 0xf0;
644 }
645 if (arg->idebus[0] && ide_get_geometry(arg->idebus[0], 1,
646 &cylinders, &heads, &sectors) >= 0) {
647 cmos_init_hd(s, 0x1a, 0x24, cylinders, heads, sectors);
648 val |= 0x0f;
649 }
650 rtc_set_memory(s, 0x12, val);
651
652 val = 0;
653 for (i = 0; i < 4; i++) {
654 /* NOTE: ide_get_geometry() returns the physical
655 geometry. It is always such that: 1 <= sects <= 63, 1
656 <= heads <= 16, 1 <= cylinders <= 16383. The BIOS
657 geometry can be different if a translation is done. */
658 if (arg->idebus[i / 2] &&
659 ide_get_geometry(arg->idebus[i / 2], i % 2,
660 &cylinders, &heads, &sectors) >= 0) {
661 trans = ide_get_bios_chs_trans(arg->idebus[i / 2], i % 2) - 1;
662 assert((trans & ~3) == 0);
663 val |= trans << (i * 2);
664 }
665 }
666 rtc_set_memory(s, 0x39, val);
667
668 pc_cmos_init_floppy(s, pc_find_fdc0());
669
670 qemu_unregister_reset(pc_cmos_init_late, opaque);
671 }
672
673 void pc_cmos_init(PCMachineState *pcms,
674 BusState *idebus0, BusState *idebus1,
675 ISADevice *s)
676 {
677 int val;
678 static pc_cmos_init_late_arg arg;
679 X86MachineState *x86ms = X86_MACHINE(pcms);
680
681 /* various important CMOS locations needed by PC/Bochs bios */
682
683 /* memory size */
684 /* base memory (first MiB) */
685 val = MIN(x86ms->below_4g_mem_size / KiB, 640);
686 rtc_set_memory(s, 0x15, val);
687 rtc_set_memory(s, 0x16, val >> 8);
688 /* extended memory (next 64MiB) */
689 if (x86ms->below_4g_mem_size > 1 * MiB) {
690 val = (x86ms->below_4g_mem_size - 1 * MiB) / KiB;
691 } else {
692 val = 0;
693 }
694 if (val > 65535)
695 val = 65535;
696 rtc_set_memory(s, 0x17, val);
697 rtc_set_memory(s, 0x18, val >> 8);
698 rtc_set_memory(s, 0x30, val);
699 rtc_set_memory(s, 0x31, val >> 8);
700 /* memory between 16MiB and 4GiB */
701 if (x86ms->below_4g_mem_size > 16 * MiB) {
702 val = (x86ms->below_4g_mem_size - 16 * MiB) / (64 * KiB);
703 } else {
704 val = 0;
705 }
706 if (val > 65535)
707 val = 65535;
708 rtc_set_memory(s, 0x34, val);
709 rtc_set_memory(s, 0x35, val >> 8);
710 /* memory above 4GiB */
711 val = x86ms->above_4g_mem_size / 65536;
712 rtc_set_memory(s, 0x5b, val);
713 rtc_set_memory(s, 0x5c, val >> 8);
714 rtc_set_memory(s, 0x5d, val >> 16);
715
716 object_property_add_link(OBJECT(pcms), "rtc_state",
717 TYPE_ISA_DEVICE,
718 (Object **)&x86ms->rtc,
719 object_property_allow_set_link,
720 OBJ_PROP_LINK_STRONG, &error_abort);
721 object_property_set_link(OBJECT(pcms), OBJECT(s),
722 "rtc_state", &error_abort);
723
724 set_boot_dev(s, MACHINE(pcms)->boot_order, &error_fatal);
725
726 val = 0;
727 val |= 0x02; /* FPU is there */
728 val |= 0x04; /* PS/2 mouse installed */
729 rtc_set_memory(s, REG_EQUIPMENT_BYTE, val);
730
731 /* hard drives and FDC */
732 arg.rtc_state = s;
733 arg.idebus[0] = idebus0;
734 arg.idebus[1] = idebus1;
735 qemu_register_reset(pc_cmos_init_late, &arg);
736 }
737
738 #define TYPE_PORT92 "port92"
739 #define PORT92(obj) OBJECT_CHECK(Port92State, (obj), TYPE_PORT92)
740
741 /* port 92 stuff: could be split off */
742 typedef struct Port92State {
743 ISADevice parent_obj;
744
745 MemoryRegion io;
746 uint8_t outport;
747 qemu_irq a20_out;
748 } Port92State;
749
750 static void port92_write(void *opaque, hwaddr addr, uint64_t val,
751 unsigned size)
752 {
753 Port92State *s = opaque;
754 int oldval = s->outport;
755
756 trace_port92_write(val);
757 s->outport = val;
758 qemu_set_irq(s->a20_out, (val >> 1) & 1);
759 if ((val & 1) && !(oldval & 1)) {
760 qemu_system_reset_request(SHUTDOWN_CAUSE_GUEST_RESET);
761 }
762 }
763
764 static uint64_t port92_read(void *opaque, hwaddr addr,
765 unsigned size)
766 {
767 Port92State *s = opaque;
768 uint32_t ret;
769
770 ret = s->outport;
771 trace_port92_read(ret);
772 return ret;
773 }
774
775 static void port92_init(ISADevice *dev, qemu_irq a20_out)
776 {
777 qdev_connect_gpio_out_named(DEVICE(dev), PORT92_A20_LINE, 0, a20_out);
778 }
779
780 static const VMStateDescription vmstate_port92_isa = {
781 .name = "port92",
782 .version_id = 1,
783 .minimum_version_id = 1,
784 .fields = (VMStateField[]) {
785 VMSTATE_UINT8(outport, Port92State),
786 VMSTATE_END_OF_LIST()
787 }
788 };
789
790 static void port92_reset(DeviceState *d)
791 {
792 Port92State *s = PORT92(d);
793
794 s->outport &= ~1;
795 }
796
797 static const MemoryRegionOps port92_ops = {
798 .read = port92_read,
799 .write = port92_write,
800 .impl = {
801 .min_access_size = 1,
802 .max_access_size = 1,
803 },
804 .endianness = DEVICE_LITTLE_ENDIAN,
805 };
806
807 static void port92_initfn(Object *obj)
808 {
809 Port92State *s = PORT92(obj);
810
811 memory_region_init_io(&s->io, OBJECT(s), &port92_ops, s, "port92", 1);
812
813 s->outport = 0;
814
815 qdev_init_gpio_out_named(DEVICE(obj), &s->a20_out, PORT92_A20_LINE, 1);
816 }
817
818 static void port92_realizefn(DeviceState *dev, Error **errp)
819 {
820 ISADevice *isadev = ISA_DEVICE(dev);
821 Port92State *s = PORT92(dev);
822
823 isa_register_ioport(isadev, &s->io, 0x92);
824 }
825
826 static void port92_class_initfn(ObjectClass *klass, void *data)
827 {
828 DeviceClass *dc = DEVICE_CLASS(klass);
829
830 dc->realize = port92_realizefn;
831 dc->reset = port92_reset;
832 dc->vmsd = &vmstate_port92_isa;
833 /*
834 * Reason: unlike ordinary ISA devices, this one needs additional
835 * wiring: its A20 output line needs to be wired up by
836 * port92_init().
837 */
838 dc->user_creatable = false;
839 }
840
841 static const TypeInfo port92_info = {
842 .name = TYPE_PORT92,
843 .parent = TYPE_ISA_DEVICE,
844 .instance_size = sizeof(Port92State),
845 .instance_init = port92_initfn,
846 .class_init = port92_class_initfn,
847 };
848
849 static void port92_register_types(void)
850 {
851 type_register_static(&port92_info);
852 }
853
854 type_init(port92_register_types)
855
856 static void handle_a20_line_change(void *opaque, int irq, int level)
857 {
858 X86CPU *cpu = opaque;
859
860 /* XXX: send to all CPUs ? */
861 /* XXX: add logic to handle multiple A20 line sources */
862 x86_cpu_set_a20(cpu, level);
863 }
864
865 #define NE2000_NB_MAX 6
866
867 static const int ne2000_io[NE2000_NB_MAX] = { 0x300, 0x320, 0x340, 0x360,
868 0x280, 0x380 };
869 static const int ne2000_irq[NE2000_NB_MAX] = { 9, 10, 11, 3, 4, 5 };
870
871 void pc_init_ne2k_isa(ISABus *bus, NICInfo *nd)
872 {
873 static int nb_ne2k = 0;
874
875 if (nb_ne2k == NE2000_NB_MAX)
876 return;
877 isa_ne2000_init(bus, ne2000_io[nb_ne2k],
878 ne2000_irq[nb_ne2k], nd);
879 nb_ne2k++;
880 }
881
882 DeviceState *cpu_get_current_apic(void)
883 {
884 if (current_cpu) {
885 X86CPU *cpu = X86_CPU(current_cpu);
886 return cpu->apic_state;
887 } else {
888 return NULL;
889 }
890 }
891
892 void pc_acpi_smi_interrupt(void *opaque, int irq, int level)
893 {
894 X86CPU *cpu = opaque;
895
896 if (level) {
897 cpu_interrupt(CPU(cpu), CPU_INTERRUPT_SMI);
898 }
899 }
900
901 /*
902 * This function is very similar to smp_parse()
903 * in hw/core/machine.c but includes CPU die support.
904 */
905 void pc_smp_parse(MachineState *ms, QemuOpts *opts)
906 {
907 X86MachineState *x86ms = X86_MACHINE(ms);
908
909 if (opts) {
910 unsigned cpus = qemu_opt_get_number(opts, "cpus", 0);
911 unsigned sockets = qemu_opt_get_number(opts, "sockets", 0);
912 unsigned dies = qemu_opt_get_number(opts, "dies", 1);
913 unsigned cores = qemu_opt_get_number(opts, "cores", 0);
914 unsigned threads = qemu_opt_get_number(opts, "threads", 0);
915
916 /* compute missing values, prefer sockets over cores over threads */
917 if (cpus == 0 || sockets == 0) {
918 cores = cores > 0 ? cores : 1;
919 threads = threads > 0 ? threads : 1;
920 if (cpus == 0) {
921 sockets = sockets > 0 ? sockets : 1;
922 cpus = cores * threads * dies * sockets;
923 } else {
924 ms->smp.max_cpus =
925 qemu_opt_get_number(opts, "maxcpus", cpus);
926 sockets = ms->smp.max_cpus / (cores * threads * dies);
927 }
928 } else if (cores == 0) {
929 threads = threads > 0 ? threads : 1;
930 cores = cpus / (sockets * dies * threads);
931 cores = cores > 0 ? cores : 1;
932 } else if (threads == 0) {
933 threads = cpus / (cores * dies * sockets);
934 threads = threads > 0 ? threads : 1;
935 } else if (sockets * dies * cores * threads < cpus) {
936 error_report("cpu topology: "
937 "sockets (%u) * dies (%u) * cores (%u) * threads (%u) < "
938 "smp_cpus (%u)",
939 sockets, dies, cores, threads, cpus);
940 exit(1);
941 }
942
943 ms->smp.max_cpus =
944 qemu_opt_get_number(opts, "maxcpus", cpus);
945
946 if (ms->smp.max_cpus < cpus) {
947 error_report("maxcpus must be equal to or greater than smp");
948 exit(1);
949 }
950
951 if (sockets * dies * cores * threads > ms->smp.max_cpus) {
952 error_report("cpu topology: "
953 "sockets (%u) * dies (%u) * cores (%u) * threads (%u) > "
954 "maxcpus (%u)",
955 sockets, dies, cores, threads,
956 ms->smp.max_cpus);
957 exit(1);
958 }
959
960 if (sockets * dies * cores * threads != ms->smp.max_cpus) {
961 warn_report("Invalid CPU topology deprecated: "
962 "sockets (%u) * dies (%u) * cores (%u) * threads (%u) "
963 "!= maxcpus (%u)",
964 sockets, dies, cores, threads,
965 ms->smp.max_cpus);
966 }
967
968 ms->smp.cpus = cpus;
969 ms->smp.cores = cores;
970 ms->smp.threads = threads;
971 x86ms->smp_dies = dies;
972 }
973
974 if (ms->smp.cpus > 1) {
975 Error *blocker = NULL;
976 error_setg(&blocker, QERR_REPLAY_NOT_SUPPORTED, "smp");
977 replay_add_blocker(blocker);
978 }
979 }
980
981 void pc_hot_add_cpu(MachineState *ms, const int64_t id, Error **errp)
982 {
983 X86MachineState *x86ms = X86_MACHINE(ms);
984 int64_t apic_id = x86_cpu_apic_id_from_index(x86ms, id);
985 Error *local_err = NULL;
986
987 if (id < 0) {
988 error_setg(errp, "Invalid CPU id: %" PRIi64, id);
989 return;
990 }
991
992 if (apic_id >= ACPI_CPU_HOTPLUG_ID_LIMIT) {
993 error_setg(errp, "Unable to add CPU: %" PRIi64
994 ", resulting APIC ID (%" PRIi64 ") is too large",
995 id, apic_id);
996 return;
997 }
998
999
1000 x86_cpu_new(X86_MACHINE(ms), apic_id, &local_err);
1001 if (local_err) {
1002 error_propagate(errp, local_err);
1003 return;
1004 }
1005 }
1006
1007 static void rtc_set_cpus_count(ISADevice *rtc, uint16_t cpus_count)
1008 {
1009 if (cpus_count > 0xff) {
1010 /* If the number of CPUs can't be represented in 8 bits, the
1011 * BIOS must use "FW_CFG_NB_CPUS". Set RTC field to 0 just
1012 * to make old BIOSes fail more predictably.
1013 */
1014 rtc_set_memory(rtc, 0x5f, 0);
1015 } else {
1016 rtc_set_memory(rtc, 0x5f, cpus_count - 1);
1017 }
1018 }
1019
1020 static
1021 void pc_machine_done(Notifier *notifier, void *data)
1022 {
1023 PCMachineState *pcms = container_of(notifier,
1024 PCMachineState, machine_done);
1025 X86MachineState *x86ms = X86_MACHINE(pcms);
1026 PCIBus *bus = pcms->bus;
1027
1028 /* set the number of CPUs */
1029 rtc_set_cpus_count(x86ms->rtc, x86ms->boot_cpus);
1030
1031 if (bus) {
1032 int extra_hosts = 0;
1033
1034 QLIST_FOREACH(bus, &bus->child, sibling) {
1035 /* look for expander root buses */
1036 if (pci_bus_is_root(bus)) {
1037 extra_hosts++;
1038 }
1039 }
1040 if (extra_hosts && x86ms->fw_cfg) {
1041 uint64_t *val = g_malloc(sizeof(*val));
1042 *val = cpu_to_le64(extra_hosts);
1043 fw_cfg_add_file(x86ms->fw_cfg,
1044 "etc/extra-pci-roots", val, sizeof(*val));
1045 }
1046 }
1047
1048 acpi_setup();
1049 if (x86ms->fw_cfg) {
1050 fw_cfg_build_smbios(MACHINE(pcms), x86ms->fw_cfg);
1051 fw_cfg_build_feature_control(MACHINE(pcms), x86ms->fw_cfg);
1052 /* update FW_CFG_NB_CPUS to account for -device added CPUs */
1053 fw_cfg_modify_i16(x86ms->fw_cfg, FW_CFG_NB_CPUS, x86ms->boot_cpus);
1054 }
1055
1056 if (x86ms->apic_id_limit > 255 && !xen_enabled()) {
1057 IntelIOMMUState *iommu = INTEL_IOMMU_DEVICE(x86_iommu_get_default());
1058
1059 if (!iommu || !x86_iommu_ir_supported(X86_IOMMU_DEVICE(iommu)) ||
1060 iommu->intr_eim != ON_OFF_AUTO_ON) {
1061 error_report("current -smp configuration requires "
1062 "Extended Interrupt Mode enabled. "
1063 "You can add an IOMMU using: "
1064 "-device intel-iommu,intremap=on,eim=on");
1065 exit(EXIT_FAILURE);
1066 }
1067 }
1068 }
1069
1070 void pc_guest_info_init(PCMachineState *pcms)
1071 {
1072 int i;
1073 MachineState *ms = MACHINE(pcms);
1074 X86MachineState *x86ms = X86_MACHINE(pcms);
1075
1076 x86ms->apic_xrupt_override = kvm_allows_irq0_override();
1077 pcms->numa_nodes = ms->numa_state->num_nodes;
1078 pcms->node_mem = g_malloc0(pcms->numa_nodes *
1079 sizeof *pcms->node_mem);
1080 for (i = 0; i < ms->numa_state->num_nodes; i++) {
1081 pcms->node_mem[i] = ms->numa_state->nodes[i].node_mem;
1082 }
1083
1084 pcms->machine_done.notify = pc_machine_done;
1085 qemu_add_machine_init_done_notifier(&pcms->machine_done);
1086 }
1087
1088 /* setup pci memory address space mapping into system address space */
1089 void pc_pci_as_mapping_init(Object *owner, MemoryRegion *system_memory,
1090 MemoryRegion *pci_address_space)
1091 {
1092 /* Set to lower priority than RAM */
1093 memory_region_add_subregion_overlap(system_memory, 0x0,
1094 pci_address_space, -1);
1095 }
1096
1097 void xen_load_linux(PCMachineState *pcms)
1098 {
1099 int i;
1100 FWCfgState *fw_cfg;
1101 PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1102 X86MachineState *x86ms = X86_MACHINE(pcms);
1103
1104 assert(MACHINE(pcms)->kernel_filename != NULL);
1105
1106 fw_cfg = fw_cfg_init_io(FW_CFG_IO_BASE);
1107 fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, x86ms->boot_cpus);
1108 rom_set_fw(fw_cfg);
1109
1110 x86_load_linux(x86ms, fw_cfg, pcmc->acpi_data_size,
1111 pcmc->pvh_enabled, pcmc->linuxboot_dma_enabled);
1112 for (i = 0; i < nb_option_roms; i++) {
1113 assert(!strcmp(option_rom[i].name, "linuxboot.bin") ||
1114 !strcmp(option_rom[i].name, "linuxboot_dma.bin") ||
1115 !strcmp(option_rom[i].name, "pvh.bin") ||
1116 !strcmp(option_rom[i].name, "multiboot.bin"));
1117 rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1118 }
1119 x86ms->fw_cfg = fw_cfg;
1120 }
1121
1122 void pc_memory_init(PCMachineState *pcms,
1123 MemoryRegion *system_memory,
1124 MemoryRegion *rom_memory,
1125 MemoryRegion **ram_memory)
1126 {
1127 int linux_boot, i;
1128 MemoryRegion *ram, *option_rom_mr;
1129 MemoryRegion *ram_below_4g, *ram_above_4g;
1130 FWCfgState *fw_cfg;
1131 MachineState *machine = MACHINE(pcms);
1132 MachineClass *mc = MACHINE_GET_CLASS(machine);
1133 PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1134 X86MachineState *x86ms = X86_MACHINE(pcms);
1135
1136 assert(machine->ram_size == x86ms->below_4g_mem_size +
1137 x86ms->above_4g_mem_size);
1138
1139 linux_boot = (machine->kernel_filename != NULL);
1140
1141 /* Allocate RAM. We allocate it as a single memory region and use
1142 * aliases to address portions of it, mostly for backwards compatibility
1143 * with older qemus that used qemu_ram_alloc().
1144 */
1145 ram = g_malloc(sizeof(*ram));
1146 memory_region_allocate_system_memory(ram, NULL, "pc.ram",
1147 machine->ram_size);
1148 *ram_memory = ram;
1149 ram_below_4g = g_malloc(sizeof(*ram_below_4g));
1150 memory_region_init_alias(ram_below_4g, NULL, "ram-below-4g", ram,
1151 0, x86ms->below_4g_mem_size);
1152 memory_region_add_subregion(system_memory, 0, ram_below_4g);
1153 e820_add_entry(0, x86ms->below_4g_mem_size, E820_RAM);
1154 if (x86ms->above_4g_mem_size > 0) {
1155 ram_above_4g = g_malloc(sizeof(*ram_above_4g));
1156 memory_region_init_alias(ram_above_4g, NULL, "ram-above-4g", ram,
1157 x86ms->below_4g_mem_size,
1158 x86ms->above_4g_mem_size);
1159 memory_region_add_subregion(system_memory, 0x100000000ULL,
1160 ram_above_4g);
1161 e820_add_entry(0x100000000ULL, x86ms->above_4g_mem_size, E820_RAM);
1162 }
1163
1164 if (!pcmc->has_reserved_memory &&
1165 (machine->ram_slots ||
1166 (machine->maxram_size > machine->ram_size))) {
1167
1168 error_report("\"-memory 'slots|maxmem'\" is not supported by: %s",
1169 mc->name);
1170 exit(EXIT_FAILURE);
1171 }
1172
1173 /* always allocate the device memory information */
1174 machine->device_memory = g_malloc0(sizeof(*machine->device_memory));
1175
1176 /* initialize device memory address space */
1177 if (pcmc->has_reserved_memory &&
1178 (machine->ram_size < machine->maxram_size)) {
1179 ram_addr_t device_mem_size = machine->maxram_size - machine->ram_size;
1180
1181 if (machine->ram_slots > ACPI_MAX_RAM_SLOTS) {
1182 error_report("unsupported amount of memory slots: %"PRIu64,
1183 machine->ram_slots);
1184 exit(EXIT_FAILURE);
1185 }
1186
1187 if (QEMU_ALIGN_UP(machine->maxram_size,
1188 TARGET_PAGE_SIZE) != machine->maxram_size) {
1189 error_report("maximum memory size must by aligned to multiple of "
1190 "%d bytes", TARGET_PAGE_SIZE);
1191 exit(EXIT_FAILURE);
1192 }
1193
1194 machine->device_memory->base =
1195 ROUND_UP(0x100000000ULL + x86ms->above_4g_mem_size, 1 * GiB);
1196
1197 if (pcmc->enforce_aligned_dimm) {
1198 /* size device region assuming 1G page max alignment per slot */
1199 device_mem_size += (1 * GiB) * machine->ram_slots;
1200 }
1201
1202 if ((machine->device_memory->base + device_mem_size) <
1203 device_mem_size) {
1204 error_report("unsupported amount of maximum memory: " RAM_ADDR_FMT,
1205 machine->maxram_size);
1206 exit(EXIT_FAILURE);
1207 }
1208
1209 memory_region_init(&machine->device_memory->mr, OBJECT(pcms),
1210 "device-memory", device_mem_size);
1211 memory_region_add_subregion(system_memory, machine->device_memory->base,
1212 &machine->device_memory->mr);
1213 }
1214
1215 /* Initialize PC system firmware */
1216 pc_system_firmware_init(pcms, rom_memory);
1217
1218 option_rom_mr = g_malloc(sizeof(*option_rom_mr));
1219 memory_region_init_ram(option_rom_mr, NULL, "pc.rom", PC_ROM_SIZE,
1220 &error_fatal);
1221 if (pcmc->pci_enabled) {
1222 memory_region_set_readonly(option_rom_mr, true);
1223 }
1224 memory_region_add_subregion_overlap(rom_memory,
1225 PC_ROM_MIN_VGA,
1226 option_rom_mr,
1227 1);
1228
1229 fw_cfg = fw_cfg_arch_create(machine,
1230 x86ms->boot_cpus, x86ms->apic_id_limit);
1231
1232 rom_set_fw(fw_cfg);
1233
1234 if (pcmc->has_reserved_memory && machine->device_memory->base) {
1235 uint64_t *val = g_malloc(sizeof(*val));
1236 PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1237 uint64_t res_mem_end = machine->device_memory->base;
1238
1239 if (!pcmc->broken_reserved_end) {
1240 res_mem_end += memory_region_size(&machine->device_memory->mr);
1241 }
1242 *val = cpu_to_le64(ROUND_UP(res_mem_end, 1 * GiB));
1243 fw_cfg_add_file(fw_cfg, "etc/reserved-memory-end", val, sizeof(*val));
1244 }
1245
1246 if (linux_boot) {
1247 x86_load_linux(x86ms, fw_cfg, pcmc->acpi_data_size,
1248 pcmc->pvh_enabled, pcmc->linuxboot_dma_enabled);
1249 }
1250
1251 for (i = 0; i < nb_option_roms; i++) {
1252 rom_add_option(option_rom[i].name, option_rom[i].bootindex);
1253 }
1254 x86ms->fw_cfg = fw_cfg;
1255
1256 /* Init default IOAPIC address space */
1257 x86ms->ioapic_as = &address_space_memory;
1258
1259 /* Init ACPI memory hotplug IO base address */
1260 pcms->memhp_io_base = ACPI_MEMORY_HOTPLUG_BASE;
1261 }
1262
1263 /*
1264 * The 64bit pci hole starts after "above 4G RAM" and
1265 * potentially the space reserved for memory hotplug.
1266 */
1267 uint64_t pc_pci_hole64_start(void)
1268 {
1269 PCMachineState *pcms = PC_MACHINE(qdev_get_machine());
1270 PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1271 MachineState *ms = MACHINE(pcms);
1272 X86MachineState *x86ms = X86_MACHINE(pcms);
1273 uint64_t hole64_start = 0;
1274
1275 if (pcmc->has_reserved_memory && ms->device_memory->base) {
1276 hole64_start = ms->device_memory->base;
1277 if (!pcmc->broken_reserved_end) {
1278 hole64_start += memory_region_size(&ms->device_memory->mr);
1279 }
1280 } else {
1281 hole64_start = 0x100000000ULL + x86ms->above_4g_mem_size;
1282 }
1283
1284 return ROUND_UP(hole64_start, 1 * GiB);
1285 }
1286
1287 qemu_irq pc_allocate_cpu_irq(void)
1288 {
1289 return qemu_allocate_irq(pic_irq_request, NULL, 0);
1290 }
1291
1292 DeviceState *pc_vga_init(ISABus *isa_bus, PCIBus *pci_bus)
1293 {
1294 DeviceState *dev = NULL;
1295
1296 rom_set_order_override(FW_CFG_ORDER_OVERRIDE_VGA);
1297 if (pci_bus) {
1298 PCIDevice *pcidev = pci_vga_init(pci_bus);
1299 dev = pcidev ? &pcidev->qdev : NULL;
1300 } else if (isa_bus) {
1301 ISADevice *isadev = isa_vga_init(isa_bus);
1302 dev = isadev ? DEVICE(isadev) : NULL;
1303 }
1304 rom_reset_order_override();
1305 return dev;
1306 }
1307
1308 static const MemoryRegionOps ioport80_io_ops = {
1309 .write = ioport80_write,
1310 .read = ioport80_read,
1311 .endianness = DEVICE_NATIVE_ENDIAN,
1312 .impl = {
1313 .min_access_size = 1,
1314 .max_access_size = 1,
1315 },
1316 };
1317
1318 static const MemoryRegionOps ioportF0_io_ops = {
1319 .write = ioportF0_write,
1320 .read = ioportF0_read,
1321 .endianness = DEVICE_NATIVE_ENDIAN,
1322 .impl = {
1323 .min_access_size = 1,
1324 .max_access_size = 1,
1325 },
1326 };
1327
1328 static void pc_superio_init(ISABus *isa_bus, bool create_fdctrl, bool no_vmport)
1329 {
1330 int i;
1331 DriveInfo *fd[MAX_FD];
1332 qemu_irq *a20_line;
1333 ISADevice *i8042, *port92, *vmmouse;
1334
1335 serial_hds_isa_init(isa_bus, 0, MAX_ISA_SERIAL_PORTS);
1336 parallel_hds_isa_init(isa_bus, MAX_PARALLEL_PORTS);
1337
1338 for (i = 0; i < MAX_FD; i++) {
1339 fd[i] = drive_get(IF_FLOPPY, 0, i);
1340 create_fdctrl |= !!fd[i];
1341 }
1342 if (create_fdctrl) {
1343 fdctrl_init_isa(isa_bus, fd);
1344 }
1345
1346 i8042 = isa_create_simple(isa_bus, "i8042");
1347 if (!no_vmport) {
1348 vmport_init(isa_bus);
1349 vmmouse = isa_try_create(isa_bus, "vmmouse");
1350 } else {
1351 vmmouse = NULL;
1352 }
1353 if (vmmouse) {
1354 DeviceState *dev = DEVICE(vmmouse);
1355 qdev_prop_set_ptr(dev, "ps2_mouse", i8042);
1356 qdev_init_nofail(dev);
1357 }
1358 port92 = isa_create_simple(isa_bus, "port92");
1359
1360 a20_line = qemu_allocate_irqs(handle_a20_line_change, first_cpu, 2);
1361 i8042_setup_a20_line(i8042, a20_line[0]);
1362 port92_init(port92, a20_line[1]);
1363 g_free(a20_line);
1364 }
1365
1366 void pc_basic_device_init(ISABus *isa_bus, qemu_irq *gsi,
1367 ISADevice **rtc_state,
1368 bool create_fdctrl,
1369 bool no_vmport,
1370 bool has_pit,
1371 uint32_t hpet_irqs)
1372 {
1373 int i;
1374 DeviceState *hpet = NULL;
1375 int pit_isa_irq = 0;
1376 qemu_irq pit_alt_irq = NULL;
1377 qemu_irq rtc_irq = NULL;
1378 ISADevice *pit = NULL;
1379 MemoryRegion *ioport80_io = g_new(MemoryRegion, 1);
1380 MemoryRegion *ioportF0_io = g_new(MemoryRegion, 1);
1381
1382 memory_region_init_io(ioport80_io, NULL, &ioport80_io_ops, NULL, "ioport80", 1);
1383 memory_region_add_subregion(isa_bus->address_space_io, 0x80, ioport80_io);
1384
1385 memory_region_init_io(ioportF0_io, NULL, &ioportF0_io_ops, NULL, "ioportF0", 1);
1386 memory_region_add_subregion(isa_bus->address_space_io, 0xf0, ioportF0_io);
1387
1388 /*
1389 * Check if an HPET shall be created.
1390 *
1391 * Without KVM_CAP_PIT_STATE2, we cannot switch off the in-kernel PIT
1392 * when the HPET wants to take over. Thus we have to disable the latter.
1393 */
1394 if (!no_hpet && (!kvm_irqchip_in_kernel() || kvm_has_pit_state2())) {
1395 /* In order to set property, here not using sysbus_try_create_simple */
1396 hpet = qdev_try_create(NULL, TYPE_HPET);
1397 if (hpet) {
1398 /* For pc-piix-*, hpet's intcap is always IRQ2. For pc-q35-1.7
1399 * and earlier, use IRQ2 for compat. Otherwise, use IRQ16~23,
1400 * IRQ8 and IRQ2.
1401 */
1402 uint8_t compat = object_property_get_uint(OBJECT(hpet),
1403 HPET_INTCAP, NULL);
1404 if (!compat) {
1405 qdev_prop_set_uint32(hpet, HPET_INTCAP, hpet_irqs);
1406 }
1407 qdev_init_nofail(hpet);
1408 sysbus_mmio_map(SYS_BUS_DEVICE(hpet), 0, HPET_BASE);
1409
1410 for (i = 0; i < GSI_NUM_PINS; i++) {
1411 sysbus_connect_irq(SYS_BUS_DEVICE(hpet), i, gsi[i]);
1412 }
1413 pit_isa_irq = -1;
1414 pit_alt_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_PIT_INT);
1415 rtc_irq = qdev_get_gpio_in(hpet, HPET_LEGACY_RTC_INT);
1416 }
1417 }
1418 *rtc_state = mc146818_rtc_init(isa_bus, 2000, rtc_irq);
1419
1420 qemu_register_boot_set(pc_boot_set, *rtc_state);
1421
1422 if (!xen_enabled() && has_pit) {
1423 if (kvm_pit_in_kernel()) {
1424 pit = kvm_pit_init(isa_bus, 0x40);
1425 } else {
1426 pit = i8254_pit_init(isa_bus, 0x40, pit_isa_irq, pit_alt_irq);
1427 }
1428 if (hpet) {
1429 /* connect PIT to output control line of the HPET */
1430 qdev_connect_gpio_out(hpet, 0, qdev_get_gpio_in(DEVICE(pit), 0));
1431 }
1432 pcspk_init(isa_bus, pit);
1433 }
1434
1435 i8257_dma_init(isa_bus, 0);
1436
1437 /* Super I/O */
1438 pc_superio_init(isa_bus, create_fdctrl, no_vmport);
1439 }
1440
1441 void pc_nic_init(PCMachineClass *pcmc, ISABus *isa_bus, PCIBus *pci_bus)
1442 {
1443 int i;
1444
1445 rom_set_order_override(FW_CFG_ORDER_OVERRIDE_NIC);
1446 for (i = 0; i < nb_nics; i++) {
1447 NICInfo *nd = &nd_table[i];
1448 const char *model = nd->model ? nd->model : pcmc->default_nic_model;
1449
1450 if (g_str_equal(model, "ne2k_isa")) {
1451 pc_init_ne2k_isa(isa_bus, nd);
1452 } else {
1453 pci_nic_init_nofail(nd, pci_bus, model, NULL);
1454 }
1455 }
1456 rom_reset_order_override();
1457 }
1458
1459 void pc_i8259_create(ISABus *isa_bus, qemu_irq *i8259_irqs)
1460 {
1461 qemu_irq *i8259;
1462
1463 if (kvm_pic_in_kernel()) {
1464 i8259 = kvm_i8259_init(isa_bus);
1465 } else if (xen_enabled()) {
1466 i8259 = xen_interrupt_controller_init();
1467 } else {
1468 i8259 = i8259_init(isa_bus, pc_allocate_cpu_irq());
1469 }
1470
1471 for (size_t i = 0; i < ISA_NUM_IRQS; i++) {
1472 i8259_irqs[i] = i8259[i];
1473 }
1474
1475 g_free(i8259);
1476 }
1477
1478 void ioapic_init_gsi(GSIState *gsi_state, const char *parent_name)
1479 {
1480 DeviceState *dev;
1481 SysBusDevice *d;
1482 unsigned int i;
1483
1484 if (kvm_ioapic_in_kernel()) {
1485 dev = qdev_create(NULL, TYPE_KVM_IOAPIC);
1486 } else {
1487 dev = qdev_create(NULL, TYPE_IOAPIC);
1488 }
1489 if (parent_name) {
1490 object_property_add_child(object_resolve_path(parent_name, NULL),
1491 "ioapic", OBJECT(dev), NULL);
1492 }
1493 qdev_init_nofail(dev);
1494 d = SYS_BUS_DEVICE(dev);
1495 sysbus_mmio_map(d, 0, IO_APIC_DEFAULT_ADDRESS);
1496
1497 for (i = 0; i < IOAPIC_NUM_PINS; i++) {
1498 gsi_state->ioapic_irq[i] = qdev_get_gpio_in(dev, i);
1499 }
1500 }
1501
1502 static void pc_memory_pre_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
1503 Error **errp)
1504 {
1505 const PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1506 const PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
1507 const MachineState *ms = MACHINE(hotplug_dev);
1508 const bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
1509 const uint64_t legacy_align = TARGET_PAGE_SIZE;
1510 Error *local_err = NULL;
1511
1512 /*
1513 * When -no-acpi is used with Q35 machine type, no ACPI is built,
1514 * but pcms->acpi_dev is still created. Check !acpi_enabled in
1515 * addition to cover this case.
1516 */
1517 if (!pcms->acpi_dev || !acpi_enabled) {
1518 error_setg(errp,
1519 "memory hotplug is not enabled: missing acpi device or acpi disabled");
1520 return;
1521 }
1522
1523 if (is_nvdimm && !ms->nvdimms_state->is_enabled) {
1524 error_setg(errp, "nvdimm is not enabled: missing 'nvdimm' in '-M'");
1525 return;
1526 }
1527
1528 hotplug_handler_pre_plug(pcms->acpi_dev, dev, &local_err);
1529 if (local_err) {
1530 error_propagate(errp, local_err);
1531 return;
1532 }
1533
1534 pc_dimm_pre_plug(PC_DIMM(dev), MACHINE(hotplug_dev),
1535 pcmc->enforce_aligned_dimm ? NULL : &legacy_align, errp);
1536 }
1537
1538 static void pc_memory_plug(HotplugHandler *hotplug_dev,
1539 DeviceState *dev, Error **errp)
1540 {
1541 Error *local_err = NULL;
1542 PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1543 MachineState *ms = MACHINE(hotplug_dev);
1544 bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
1545
1546 pc_dimm_plug(PC_DIMM(dev), MACHINE(pcms), &local_err);
1547 if (local_err) {
1548 goto out;
1549 }
1550
1551 if (is_nvdimm) {
1552 nvdimm_plug(ms->nvdimms_state);
1553 }
1554
1555 hotplug_handler_plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &error_abort);
1556 out:
1557 error_propagate(errp, local_err);
1558 }
1559
1560 static void pc_memory_unplug_request(HotplugHandler *hotplug_dev,
1561 DeviceState *dev, Error **errp)
1562 {
1563 Error *local_err = NULL;
1564 PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1565
1566 /*
1567 * When -no-acpi is used with Q35 machine type, no ACPI is built,
1568 * but pcms->acpi_dev is still created. Check !acpi_enabled in
1569 * addition to cover this case.
1570 */
1571 if (!pcms->acpi_dev || !acpi_enabled) {
1572 error_setg(&local_err,
1573 "memory hotplug is not enabled: missing acpi device or acpi disabled");
1574 goto out;
1575 }
1576
1577 if (object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM)) {
1578 error_setg(&local_err,
1579 "nvdimm device hot unplug is not supported yet.");
1580 goto out;
1581 }
1582
1583 hotplug_handler_unplug_request(HOTPLUG_HANDLER(pcms->acpi_dev), dev,
1584 &local_err);
1585 out:
1586 error_propagate(errp, local_err);
1587 }
1588
1589 static void pc_memory_unplug(HotplugHandler *hotplug_dev,
1590 DeviceState *dev, Error **errp)
1591 {
1592 PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1593 Error *local_err = NULL;
1594
1595 hotplug_handler_unplug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
1596 if (local_err) {
1597 goto out;
1598 }
1599
1600 pc_dimm_unplug(PC_DIMM(dev), MACHINE(pcms));
1601 object_property_set_bool(OBJECT(dev), false, "realized", NULL);
1602 out:
1603 error_propagate(errp, local_err);
1604 }
1605
1606 static int pc_apic_cmp(const void *a, const void *b)
1607 {
1608 CPUArchId *apic_a = (CPUArchId *)a;
1609 CPUArchId *apic_b = (CPUArchId *)b;
1610
1611 return apic_a->arch_id - apic_b->arch_id;
1612 }
1613
1614 /* returns pointer to CPUArchId descriptor that matches CPU's apic_id
1615 * in ms->possible_cpus->cpus, if ms->possible_cpus->cpus has no
1616 * entry corresponding to CPU's apic_id returns NULL.
1617 */
1618 static CPUArchId *pc_find_cpu_slot(MachineState *ms, uint32_t id, int *idx)
1619 {
1620 CPUArchId apic_id, *found_cpu;
1621
1622 apic_id.arch_id = id;
1623 found_cpu = bsearch(&apic_id, ms->possible_cpus->cpus,
1624 ms->possible_cpus->len, sizeof(*ms->possible_cpus->cpus),
1625 pc_apic_cmp);
1626 if (found_cpu && idx) {
1627 *idx = found_cpu - ms->possible_cpus->cpus;
1628 }
1629 return found_cpu;
1630 }
1631
1632 static void pc_cpu_plug(HotplugHandler *hotplug_dev,
1633 DeviceState *dev, Error **errp)
1634 {
1635 CPUArchId *found_cpu;
1636 Error *local_err = NULL;
1637 X86CPU *cpu = X86_CPU(dev);
1638 PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1639 X86MachineState *x86ms = X86_MACHINE(pcms);
1640
1641 if (pcms->acpi_dev) {
1642 hotplug_handler_plug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
1643 if (local_err) {
1644 goto out;
1645 }
1646 }
1647
1648 /* increment the number of CPUs */
1649 x86ms->boot_cpus++;
1650 if (x86ms->rtc) {
1651 rtc_set_cpus_count(x86ms->rtc, x86ms->boot_cpus);
1652 }
1653 if (x86ms->fw_cfg) {
1654 fw_cfg_modify_i16(x86ms->fw_cfg, FW_CFG_NB_CPUS, x86ms->boot_cpus);
1655 }
1656
1657 found_cpu = pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, NULL);
1658 found_cpu->cpu = OBJECT(dev);
1659 out:
1660 error_propagate(errp, local_err);
1661 }
1662 static void pc_cpu_unplug_request_cb(HotplugHandler *hotplug_dev,
1663 DeviceState *dev, Error **errp)
1664 {
1665 int idx = -1;
1666 Error *local_err = NULL;
1667 X86CPU *cpu = X86_CPU(dev);
1668 PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1669
1670 if (!pcms->acpi_dev) {
1671 error_setg(&local_err, "CPU hot unplug not supported without ACPI");
1672 goto out;
1673 }
1674
1675 pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, &idx);
1676 assert(idx != -1);
1677 if (idx == 0) {
1678 error_setg(&local_err, "Boot CPU is unpluggable");
1679 goto out;
1680 }
1681
1682 hotplug_handler_unplug_request(HOTPLUG_HANDLER(pcms->acpi_dev), dev,
1683 &local_err);
1684 if (local_err) {
1685 goto out;
1686 }
1687
1688 out:
1689 error_propagate(errp, local_err);
1690
1691 }
1692
1693 static void pc_cpu_unplug_cb(HotplugHandler *hotplug_dev,
1694 DeviceState *dev, Error **errp)
1695 {
1696 CPUArchId *found_cpu;
1697 Error *local_err = NULL;
1698 X86CPU *cpu = X86_CPU(dev);
1699 PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1700 X86MachineState *x86ms = X86_MACHINE(pcms);
1701
1702 hotplug_handler_unplug(HOTPLUG_HANDLER(pcms->acpi_dev), dev, &local_err);
1703 if (local_err) {
1704 goto out;
1705 }
1706
1707 found_cpu = pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, NULL);
1708 found_cpu->cpu = NULL;
1709 object_property_set_bool(OBJECT(dev), false, "realized", NULL);
1710
1711 /* decrement the number of CPUs */
1712 x86ms->boot_cpus--;
1713 /* Update the number of CPUs in CMOS */
1714 rtc_set_cpus_count(x86ms->rtc, x86ms->boot_cpus);
1715 fw_cfg_modify_i16(x86ms->fw_cfg, FW_CFG_NB_CPUS, x86ms->boot_cpus);
1716 out:
1717 error_propagate(errp, local_err);
1718 }
1719
1720 static void pc_cpu_pre_plug(HotplugHandler *hotplug_dev,
1721 DeviceState *dev, Error **errp)
1722 {
1723 int idx;
1724 CPUState *cs;
1725 CPUArchId *cpu_slot;
1726 X86CPUTopoInfo topo;
1727 X86CPU *cpu = X86_CPU(dev);
1728 CPUX86State *env = &cpu->env;
1729 MachineState *ms = MACHINE(hotplug_dev);
1730 PCMachineState *pcms = PC_MACHINE(hotplug_dev);
1731 X86MachineState *x86ms = X86_MACHINE(pcms);
1732 unsigned int smp_cores = ms->smp.cores;
1733 unsigned int smp_threads = ms->smp.threads;
1734
1735 if(!object_dynamic_cast(OBJECT(cpu), ms->cpu_type)) {
1736 error_setg(errp, "Invalid CPU type, expected cpu type: '%s'",
1737 ms->cpu_type);
1738 return;
1739 }
1740
1741 env->nr_dies = x86ms->smp_dies;
1742
1743 /*
1744 * If APIC ID is not set,
1745 * set it based on socket/die/core/thread properties.
1746 */
1747 if (cpu->apic_id == UNASSIGNED_APIC_ID) {
1748 int max_socket = (ms->smp.max_cpus - 1) /
1749 smp_threads / smp_cores / x86ms->smp_dies;
1750
1751 /*
1752 * die-id was optional in QEMU 4.0 and older, so keep it optional
1753 * if there's only one die per socket.
1754 */
1755 if (cpu->die_id < 0 && x86ms->smp_dies == 1) {
1756 cpu->die_id = 0;
1757 }
1758
1759 if (cpu->socket_id < 0) {
1760 error_setg(errp, "CPU socket-id is not set");
1761 return;
1762 } else if (cpu->socket_id > max_socket) {
1763 error_setg(errp, "Invalid CPU socket-id: %u must be in range 0:%u",
1764 cpu->socket_id, max_socket);
1765 return;
1766 }
1767 if (cpu->die_id < 0) {
1768 error_setg(errp, "CPU die-id is not set");
1769 return;
1770 } else if (cpu->die_id > x86ms->smp_dies - 1) {
1771 error_setg(errp, "Invalid CPU die-id: %u must be in range 0:%u",
1772 cpu->die_id, x86ms->smp_dies - 1);
1773 return;
1774 }
1775 if (cpu->core_id < 0) {
1776 error_setg(errp, "CPU core-id is not set");
1777 return;
1778 } else if (cpu->core_id > (smp_cores - 1)) {
1779 error_setg(errp, "Invalid CPU core-id: %u must be in range 0:%u",
1780 cpu->core_id, smp_cores - 1);
1781 return;
1782 }
1783 if (cpu->thread_id < 0) {
1784 error_setg(errp, "CPU thread-id is not set");
1785 return;
1786 } else if (cpu->thread_id > (smp_threads - 1)) {
1787 error_setg(errp, "Invalid CPU thread-id: %u must be in range 0:%u",
1788 cpu->thread_id, smp_threads - 1);
1789 return;
1790 }
1791
1792 topo.pkg_id = cpu->socket_id;
1793 topo.die_id = cpu->die_id;
1794 topo.core_id = cpu->core_id;
1795 topo.smt_id = cpu->thread_id;
1796 cpu->apic_id = apicid_from_topo_ids(x86ms->smp_dies, smp_cores,
1797 smp_threads, &topo);
1798 }
1799
1800 cpu_slot = pc_find_cpu_slot(MACHINE(pcms), cpu->apic_id, &idx);
1801 if (!cpu_slot) {
1802 MachineState *ms = MACHINE(pcms);
1803
1804 x86_topo_ids_from_apicid(cpu->apic_id, x86ms->smp_dies,
1805 smp_cores, smp_threads, &topo);
1806 error_setg(errp,
1807 "Invalid CPU [socket: %u, die: %u, core: %u, thread: %u] with"
1808 " APIC ID %" PRIu32 ", valid index range 0:%d",
1809 topo.pkg_id, topo.die_id, topo.core_id, topo.smt_id,
1810 cpu->apic_id, ms->possible_cpus->len - 1);
1811 return;
1812 }
1813
1814 if (cpu_slot->cpu) {
1815 error_setg(errp, "CPU[%d] with APIC ID %" PRIu32 " exists",
1816 idx, cpu->apic_id);
1817 return;
1818 }
1819
1820 /* if 'address' properties socket-id/core-id/thread-id are not set, set them
1821 * so that machine_query_hotpluggable_cpus would show correct values
1822 */
1823 /* TODO: move socket_id/core_id/thread_id checks into x86_cpu_realizefn()
1824 * once -smp refactoring is complete and there will be CPU private
1825 * CPUState::nr_cores and CPUState::nr_threads fields instead of globals */
1826 x86_topo_ids_from_apicid(cpu->apic_id, x86ms->smp_dies,
1827 smp_cores, smp_threads, &topo);
1828 if (cpu->socket_id != -1 && cpu->socket_id != topo.pkg_id) {
1829 error_setg(errp, "property socket-id: %u doesn't match set apic-id:"
1830 " 0x%x (socket-id: %u)", cpu->socket_id, cpu->apic_id, topo.pkg_id);
1831 return;
1832 }
1833 cpu->socket_id = topo.pkg_id;
1834
1835 if (cpu->die_id != -1 && cpu->die_id != topo.die_id) {
1836 error_setg(errp, "property die-id: %u doesn't match set apic-id:"
1837 " 0x%x (die-id: %u)", cpu->die_id, cpu->apic_id, topo.die_id);
1838 return;
1839 }
1840 cpu->die_id = topo.die_id;
1841
1842 if (cpu->core_id != -1 && cpu->core_id != topo.core_id) {
1843 error_setg(errp, "property core-id: %u doesn't match set apic-id:"
1844 " 0x%x (core-id: %u)", cpu->core_id, cpu->apic_id, topo.core_id);
1845 return;
1846 }
1847 cpu->core_id = topo.core_id;
1848
1849 if (cpu->thread_id != -1 && cpu->thread_id != topo.smt_id) {
1850 error_setg(errp, "property thread-id: %u doesn't match set apic-id:"
1851 " 0x%x (thread-id: %u)", cpu->thread_id, cpu->apic_id, topo.smt_id);
1852 return;
1853 }
1854 cpu->thread_id = topo.smt_id;
1855
1856 if (hyperv_feat_enabled(cpu, HYPERV_FEAT_VPINDEX) &&
1857 !kvm_hv_vpindex_settable()) {
1858 error_setg(errp, "kernel doesn't allow setting HyperV VP_INDEX");
1859 return;
1860 }
1861
1862 cs = CPU(cpu);
1863 cs->cpu_index = idx;
1864
1865 numa_cpu_pre_plug(cpu_slot, dev, errp);
1866 }
1867
1868 static void pc_virtio_pmem_pci_pre_plug(HotplugHandler *hotplug_dev,
1869 DeviceState *dev, Error **errp)
1870 {
1871 HotplugHandler *hotplug_dev2 = qdev_get_bus_hotplug_handler(dev);
1872 Error *local_err = NULL;
1873
1874 if (!hotplug_dev2) {
1875 /*
1876 * Without a bus hotplug handler, we cannot control the plug/unplug
1877 * order. This should never be the case on x86, however better add
1878 * a safety net.
1879 */
1880 error_setg(errp, "virtio-pmem-pci not supported on this bus.");
1881 return;
1882 }
1883 /*
1884 * First, see if we can plug this memory device at all. If that
1885 * succeeds, branch of to the actual hotplug handler.
1886 */
1887 memory_device_pre_plug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev), NULL,
1888 &local_err);
1889 if (!local_err) {
1890 hotplug_handler_pre_plug(hotplug_dev2, dev, &local_err);
1891 }
1892 error_propagate(errp, local_err);
1893 }
1894
1895 static void pc_virtio_pmem_pci_plug(HotplugHandler *hotplug_dev,
1896 DeviceState *dev, Error **errp)
1897 {
1898 HotplugHandler *hotplug_dev2 = qdev_get_bus_hotplug_handler(dev);
1899 Error *local_err = NULL;
1900
1901 /*
1902 * Plug the memory device first and then branch off to the actual
1903 * hotplug handler. If that one fails, we can easily undo the memory
1904 * device bits.
1905 */
1906 memory_device_plug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev));
1907 hotplug_handler_plug(hotplug_dev2, dev, &local_err);
1908 if (local_err) {
1909 memory_device_unplug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev));
1910 }
1911 error_propagate(errp, local_err);
1912 }
1913
1914 static void pc_virtio_pmem_pci_unplug_request(HotplugHandler *hotplug_dev,
1915 DeviceState *dev, Error **errp)
1916 {
1917 /* We don't support virtio pmem hot unplug */
1918 error_setg(errp, "virtio pmem device unplug not supported.");
1919 }
1920
1921 static void pc_virtio_pmem_pci_unplug(HotplugHandler *hotplug_dev,
1922 DeviceState *dev, Error **errp)
1923 {
1924 /* We don't support virtio pmem hot unplug */
1925 }
1926
1927 static void pc_machine_device_pre_plug_cb(HotplugHandler *hotplug_dev,
1928 DeviceState *dev, Error **errp)
1929 {
1930 if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1931 pc_memory_pre_plug(hotplug_dev, dev, errp);
1932 } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1933 pc_cpu_pre_plug(hotplug_dev, dev, errp);
1934 } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI)) {
1935 pc_virtio_pmem_pci_pre_plug(hotplug_dev, dev, errp);
1936 }
1937 }
1938
1939 static void pc_machine_device_plug_cb(HotplugHandler *hotplug_dev,
1940 DeviceState *dev, Error **errp)
1941 {
1942 if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1943 pc_memory_plug(hotplug_dev, dev, errp);
1944 } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1945 pc_cpu_plug(hotplug_dev, dev, errp);
1946 } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI)) {
1947 pc_virtio_pmem_pci_plug(hotplug_dev, dev, errp);
1948 }
1949 }
1950
1951 static void pc_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
1952 DeviceState *dev, Error **errp)
1953 {
1954 if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1955 pc_memory_unplug_request(hotplug_dev, dev, errp);
1956 } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1957 pc_cpu_unplug_request_cb(hotplug_dev, dev, errp);
1958 } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI)) {
1959 pc_virtio_pmem_pci_unplug_request(hotplug_dev, dev, errp);
1960 } else {
1961 error_setg(errp, "acpi: device unplug request for not supported device"
1962 " type: %s", object_get_typename(OBJECT(dev)));
1963 }
1964 }
1965
1966 static void pc_machine_device_unplug_cb(HotplugHandler *hotplug_dev,
1967 DeviceState *dev, Error **errp)
1968 {
1969 if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
1970 pc_memory_unplug(hotplug_dev, dev, errp);
1971 } else if (object_dynamic_cast(OBJECT(dev), TYPE_CPU)) {
1972 pc_cpu_unplug_cb(hotplug_dev, dev, errp);
1973 } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI)) {
1974 pc_virtio_pmem_pci_unplug(hotplug_dev, dev, errp);
1975 } else {
1976 error_setg(errp, "acpi: device unplug for not supported device"
1977 " type: %s", object_get_typename(OBJECT(dev)));
1978 }
1979 }
1980
1981 static HotplugHandler *pc_get_hotplug_handler(MachineState *machine,
1982 DeviceState *dev)
1983 {
1984 if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) ||
1985 object_dynamic_cast(OBJECT(dev), TYPE_CPU) ||
1986 object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_PMEM_PCI)) {
1987 return HOTPLUG_HANDLER(machine);
1988 }
1989
1990 return NULL;
1991 }
1992
1993 static void
1994 pc_machine_get_device_memory_region_size(Object *obj, Visitor *v,
1995 const char *name, void *opaque,
1996 Error **errp)
1997 {
1998 MachineState *ms = MACHINE(obj);
1999 int64_t value = 0;
2000
2001 if (ms->device_memory) {
2002 value = memory_region_size(&ms->device_memory->mr);
2003 }
2004
2005 visit_type_int(v, name, &value, errp);
2006 }
2007
2008 static void pc_machine_get_vmport(Object *obj, Visitor *v, const char *name,
2009 void *opaque, Error **errp)
2010 {
2011 PCMachineState *pcms = PC_MACHINE(obj);
2012 OnOffAuto vmport = pcms->vmport;
2013
2014 visit_type_OnOffAuto(v, name, &vmport, errp);
2015 }
2016
2017 static void pc_machine_set_vmport(Object *obj, Visitor *v, const char *name,
2018 void *opaque, Error **errp)
2019 {
2020 PCMachineState *pcms = PC_MACHINE(obj);
2021
2022 visit_type_OnOffAuto(v, name, &pcms->vmport, errp);
2023 }
2024
2025 static bool pc_machine_get_smbus(Object *obj, Error **errp)
2026 {
2027 PCMachineState *pcms = PC_MACHINE(obj);
2028
2029 return pcms->smbus_enabled;
2030 }
2031
2032 static void pc_machine_set_smbus(Object *obj, bool value, Error **errp)
2033 {
2034 PCMachineState *pcms = PC_MACHINE(obj);
2035
2036 pcms->smbus_enabled = value;
2037 }
2038
2039 static bool pc_machine_get_sata(Object *obj, Error **errp)
2040 {
2041 PCMachineState *pcms = PC_MACHINE(obj);
2042
2043 return pcms->sata_enabled;
2044 }
2045
2046 static void pc_machine_set_sata(Object *obj, bool value, Error **errp)
2047 {
2048 PCMachineState *pcms = PC_MACHINE(obj);
2049
2050 pcms->sata_enabled = value;
2051 }
2052
2053 static bool pc_machine_get_pit(Object *obj, Error **errp)
2054 {
2055 PCMachineState *pcms = PC_MACHINE(obj);
2056
2057 return pcms->pit_enabled;
2058 }
2059
2060 static void pc_machine_set_pit(Object *obj, bool value, Error **errp)
2061 {
2062 PCMachineState *pcms = PC_MACHINE(obj);
2063
2064 pcms->pit_enabled = value;
2065 }
2066
2067 static void pc_machine_initfn(Object *obj)
2068 {
2069 PCMachineState *pcms = PC_MACHINE(obj);
2070
2071 #ifdef CONFIG_VMPORT
2072 pcms->vmport = ON_OFF_AUTO_AUTO;
2073 #else
2074 pcms->vmport = ON_OFF_AUTO_OFF;
2075 #endif /* CONFIG_VMPORT */
2076 /* acpi build is enabled by default if machine supports it */
2077 pcms->acpi_build_enabled = PC_MACHINE_GET_CLASS(pcms)->has_acpi_build;
2078 pcms->smbus_enabled = true;
2079 pcms->sata_enabled = true;
2080 pcms->pit_enabled = true;
2081
2082 pc_system_flash_create(pcms);
2083 }
2084
2085 static void pc_machine_reset(MachineState *machine)
2086 {
2087 CPUState *cs;
2088 X86CPU *cpu;
2089
2090 qemu_devices_reset();
2091
2092 /* Reset APIC after devices have been reset to cancel
2093 * any changes that qemu_devices_reset() might have done.
2094 */
2095 CPU_FOREACH(cs) {
2096 cpu = X86_CPU(cs);
2097
2098 if (cpu->apic_state) {
2099 device_reset(cpu->apic_state);
2100 }
2101 }
2102 }
2103
2104 static void pc_machine_wakeup(MachineState *machine)
2105 {
2106 cpu_synchronize_all_states();
2107 pc_machine_reset(machine);
2108 cpu_synchronize_all_post_reset();
2109 }
2110
2111 static bool pc_hotplug_allowed(MachineState *ms, DeviceState *dev, Error **errp)
2112 {
2113 X86IOMMUState *iommu = x86_iommu_get_default();
2114 IntelIOMMUState *intel_iommu;
2115
2116 if (iommu &&
2117 object_dynamic_cast((Object *)iommu, TYPE_INTEL_IOMMU_DEVICE) &&
2118 object_dynamic_cast((Object *)dev, "vfio-pci")) {
2119 intel_iommu = INTEL_IOMMU_DEVICE(iommu);
2120 if (!intel_iommu->caching_mode) {
2121 error_setg(errp, "Device assignment is not allowed without "
2122 "enabling caching-mode=on for Intel IOMMU.");
2123 return false;
2124 }
2125 }
2126
2127 return true;
2128 }
2129
2130 static void pc_machine_class_init(ObjectClass *oc, void *data)
2131 {
2132 MachineClass *mc = MACHINE_CLASS(oc);
2133 PCMachineClass *pcmc = PC_MACHINE_CLASS(oc);
2134 HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
2135
2136 pcmc->pci_enabled = true;
2137 pcmc->has_acpi_build = true;
2138 pcmc->rsdp_in_ram = true;
2139 pcmc->smbios_defaults = true;
2140 pcmc->smbios_uuid_encoded = true;
2141 pcmc->gigabyte_align = true;
2142 pcmc->has_reserved_memory = true;
2143 pcmc->kvmclock_enabled = true;
2144 pcmc->enforce_aligned_dimm = true;
2145 /* BIOS ACPI tables: 128K. Other BIOS datastructures: less than 4K reported
2146 * to be used at the moment, 32K should be enough for a while. */
2147 pcmc->acpi_data_size = 0x20000 + 0x8000;
2148 pcmc->linuxboot_dma_enabled = true;
2149 pcmc->pvh_enabled = true;
2150 assert(!mc->get_hotplug_handler);
2151 mc->get_hotplug_handler = pc_get_hotplug_handler;
2152 mc->hotplug_allowed = pc_hotplug_allowed;
2153 mc->cpu_index_to_instance_props = x86_cpu_index_to_props;
2154 mc->get_default_cpu_node_id = x86_get_default_cpu_node_id;
2155 mc->possible_cpu_arch_ids = x86_possible_cpu_arch_ids;
2156 mc->auto_enable_numa_with_memhp = true;
2157 mc->has_hotpluggable_cpus = true;
2158 mc->default_boot_order = "cad";
2159 mc->hot_add_cpu = pc_hot_add_cpu;
2160 mc->smp_parse = pc_smp_parse;
2161 mc->block_default_type = IF_IDE;
2162 mc->max_cpus = 255;
2163 mc->reset = pc_machine_reset;
2164 mc->wakeup = pc_machine_wakeup;
2165 hc->pre_plug = pc_machine_device_pre_plug_cb;
2166 hc->plug = pc_machine_device_plug_cb;
2167 hc->unplug_request = pc_machine_device_unplug_request_cb;
2168 hc->unplug = pc_machine_device_unplug_cb;
2169 mc->default_cpu_type = TARGET_DEFAULT_CPU_TYPE;
2170 mc->nvdimm_supported = true;
2171 mc->numa_mem_supported = true;
2172
2173 object_class_property_add(oc, PC_MACHINE_DEVMEM_REGION_SIZE, "int",
2174 pc_machine_get_device_memory_region_size, NULL,
2175 NULL, NULL, &error_abort);
2176
2177 object_class_property_add(oc, PC_MACHINE_VMPORT, "OnOffAuto",
2178 pc_machine_get_vmport, pc_machine_set_vmport,
2179 NULL, NULL, &error_abort);
2180 object_class_property_set_description(oc, PC_MACHINE_VMPORT,
2181 "Enable vmport (pc & q35)", &error_abort);
2182
2183 object_class_property_add_bool(oc, PC_MACHINE_SMBUS,
2184 pc_machine_get_smbus, pc_machine_set_smbus, &error_abort);
2185
2186 object_class_property_add_bool(oc, PC_MACHINE_SATA,
2187 pc_machine_get_sata, pc_machine_set_sata, &error_abort);
2188
2189 object_class_property_add_bool(oc, PC_MACHINE_PIT,
2190 pc_machine_get_pit, pc_machine_set_pit, &error_abort);
2191 }
2192
2193 static const TypeInfo pc_machine_info = {
2194 .name = TYPE_PC_MACHINE,
2195 .parent = TYPE_X86_MACHINE,
2196 .abstract = true,
2197 .instance_size = sizeof(PCMachineState),
2198 .instance_init = pc_machine_initfn,
2199 .class_size = sizeof(PCMachineClass),
2200 .class_init = pc_machine_class_init,
2201 .interfaces = (InterfaceInfo[]) {
2202 { TYPE_HOTPLUG_HANDLER },
2203 { }
2204 },
2205 };
2206
2207 static void pc_machine_register_types(void)
2208 {
2209 type_register_static(&pc_machine_info);
2210 }
2211
2212 type_init(pc_machine_register_types)