xref: /openbmc/qemu/hw/i386/acpi-build.c (revision 9967e4fec0dc6c0157f37574dd33ddd03ca3bee8)
1 /* Support for generating ACPI tables and passing them to Guests
2  *
3  * Copyright (C) 2008-2010  Kevin O'Connor <kevin@koconnor.net>
4  * Copyright (C) 2006 Fabrice Bellard
5  * Copyright (C) 2013 Red Hat Inc
6  *
7  * Author: Michael S. Tsirkin <mst@redhat.com>
8  *
9  * This program is free software; you can redistribute it and/or modify
10  * it under the terms of the GNU General Public License as published by
11  * the Free Software Foundation; either version 2 of the License, or
12  * (at your option) any later version.
13 
14  * This program is distributed in the hope that it will be useful,
15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
17  * GNU General Public License for more details.
18 
19  * You should have received a copy of the GNU General Public License along
20  * with this program; if not, see <http://www.gnu.org/licenses/>.
21  */
22 
23 #include "acpi-build.h"
24 #include <stddef.h>
25 #include <glib.h>
26 #include "qemu-common.h"
27 #include "qemu/bitmap.h"
28 #include "qemu/osdep.h"
29 #include "qemu/error-report.h"
30 #include "hw/pci/pci.h"
31 #include "qom/cpu.h"
32 #include "hw/i386/pc.h"
33 #include "target-i386/cpu.h"
34 #include "hw/timer/hpet.h"
35 #include "hw/acpi/acpi-defs.h"
36 #include "hw/acpi/acpi.h"
37 #include "hw/nvram/fw_cfg.h"
38 #include "hw/acpi/bios-linker-loader.h"
39 #include "hw/loader.h"
40 #include "hw/isa/isa.h"
41 #include "hw/acpi/memory_hotplug.h"
42 #include "hw/mem/nvdimm.h"
43 #include "sysemu/tpm.h"
44 #include "hw/acpi/tpm.h"
45 #include "sysemu/tpm_backend.h"
46 
47 /* Supported chipsets: */
48 #include "hw/acpi/piix4.h"
49 #include "hw/acpi/pcihp.h"
50 #include "hw/i386/ich9.h"
51 #include "hw/pci/pci_bus.h"
52 #include "hw/pci-host/q35.h"
53 #include "hw/i386/intel_iommu.h"
54 
55 #include "hw/i386/q35-acpi-dsdt.hex"
56 #include "hw/i386/acpi-dsdt.hex"
57 
58 #include "hw/acpi/aml-build.h"
59 
60 #include "qapi/qmp/qint.h"
61 #include "qom/qom-qobject.h"
62 
63 /* These are used to size the ACPI tables for -M pc-i440fx-1.7 and
64  * -M pc-i440fx-2.0.  Even if the actual amount of AML generated grows
65  * a little bit, there should be plenty of free space since the DSDT
66  * shrunk by ~1.5k between QEMU 2.0 and QEMU 2.1.
67  */
68 #define ACPI_BUILD_LEGACY_CPU_AML_SIZE    97
69 #define ACPI_BUILD_ALIGN_SIZE             0x1000
70 
71 #define ACPI_BUILD_TABLE_SIZE             0x20000
72 
73 /* #define DEBUG_ACPI_BUILD */
74 #ifdef DEBUG_ACPI_BUILD
75 #define ACPI_BUILD_DPRINTF(fmt, ...)        \
76     do {printf("ACPI_BUILD: " fmt, ## __VA_ARGS__); } while (0)
77 #else
78 #define ACPI_BUILD_DPRINTF(fmt, ...)
79 #endif
80 
81 typedef struct AcpiCpuInfo {
82     DECLARE_BITMAP(found_cpus, ACPI_CPU_HOTPLUG_ID_LIMIT);
83 } AcpiCpuInfo;
84 
85 typedef struct AcpiMcfgInfo {
86     uint64_t mcfg_base;
87     uint32_t mcfg_size;
88 } AcpiMcfgInfo;
89 
90 typedef struct AcpiPmInfo {
91     bool s3_disabled;
92     bool s4_disabled;
93     bool pcihp_bridge_en;
94     uint8_t s4_val;
95     uint16_t sci_int;
96     uint8_t acpi_enable_cmd;
97     uint8_t acpi_disable_cmd;
98     uint32_t gpe0_blk;
99     uint32_t gpe0_blk_len;
100     uint32_t io_base;
101     uint16_t cpu_hp_io_base;
102     uint16_t cpu_hp_io_len;
103     uint16_t mem_hp_io_base;
104     uint16_t mem_hp_io_len;
105     uint16_t pcihp_io_base;
106     uint16_t pcihp_io_len;
107 } AcpiPmInfo;
108 
109 typedef struct AcpiMiscInfo {
110     bool has_hpet;
111     TPMVersion tpm_version;
112     const unsigned char *dsdt_code;
113     unsigned dsdt_size;
114     uint16_t pvpanic_port;
115     uint16_t applesmc_io_base;
116 } AcpiMiscInfo;
117 
118 typedef struct AcpiBuildPciBusHotplugState {
119     GArray *device_table;
120     GArray *notify_table;
121     struct AcpiBuildPciBusHotplugState *parent;
122     bool pcihp_bridge_en;
123 } AcpiBuildPciBusHotplugState;
124 
125 static void acpi_get_dsdt(AcpiMiscInfo *info)
126 {
127     Object *piix = piix4_pm_find();
128     Object *lpc = ich9_lpc_find();
129     assert(!!piix != !!lpc);
130 
131     if (piix) {
132         info->dsdt_code = AcpiDsdtAmlCode;
133         info->dsdt_size = sizeof AcpiDsdtAmlCode;
134     }
135     if (lpc) {
136         info->dsdt_code = Q35AcpiDsdtAmlCode;
137         info->dsdt_size = sizeof Q35AcpiDsdtAmlCode;
138     }
139 }
140 
141 static
142 int acpi_add_cpu_info(Object *o, void *opaque)
143 {
144     AcpiCpuInfo *cpu = opaque;
145     uint64_t apic_id;
146 
147     if (object_dynamic_cast(o, TYPE_CPU)) {
148         apic_id = object_property_get_int(o, "apic-id", NULL);
149         assert(apic_id < ACPI_CPU_HOTPLUG_ID_LIMIT);
150 
151         set_bit(apic_id, cpu->found_cpus);
152     }
153 
154     object_child_foreach(o, acpi_add_cpu_info, opaque);
155     return 0;
156 }
157 
158 static void acpi_get_cpu_info(AcpiCpuInfo *cpu)
159 {
160     Object *root = object_get_root();
161 
162     memset(cpu->found_cpus, 0, sizeof cpu->found_cpus);
163     object_child_foreach(root, acpi_add_cpu_info, cpu);
164 }
165 
166 static void acpi_get_pm_info(AcpiPmInfo *pm)
167 {
168     Object *piix = piix4_pm_find();
169     Object *lpc = ich9_lpc_find();
170     Object *obj = NULL;
171     QObject *o;
172 
173     pm->cpu_hp_io_base = 0;
174     pm->pcihp_io_base = 0;
175     pm->pcihp_io_len = 0;
176     if (piix) {
177         obj = piix;
178         pm->cpu_hp_io_base = PIIX4_CPU_HOTPLUG_IO_BASE;
179         pm->pcihp_io_base =
180             object_property_get_int(obj, ACPI_PCIHP_IO_BASE_PROP, NULL);
181         pm->pcihp_io_len =
182             object_property_get_int(obj, ACPI_PCIHP_IO_LEN_PROP, NULL);
183     }
184     if (lpc) {
185         obj = lpc;
186         pm->cpu_hp_io_base = ICH9_CPU_HOTPLUG_IO_BASE;
187     }
188     assert(obj);
189 
190     pm->cpu_hp_io_len = ACPI_GPE_PROC_LEN;
191     pm->mem_hp_io_base = ACPI_MEMORY_HOTPLUG_BASE;
192     pm->mem_hp_io_len = ACPI_MEMORY_HOTPLUG_IO_LEN;
193 
194     /* Fill in optional s3/s4 related properties */
195     o = object_property_get_qobject(obj, ACPI_PM_PROP_S3_DISABLED, NULL);
196     if (o) {
197         pm->s3_disabled = qint_get_int(qobject_to_qint(o));
198     } else {
199         pm->s3_disabled = false;
200     }
201     qobject_decref(o);
202     o = object_property_get_qobject(obj, ACPI_PM_PROP_S4_DISABLED, NULL);
203     if (o) {
204         pm->s4_disabled = qint_get_int(qobject_to_qint(o));
205     } else {
206         pm->s4_disabled = false;
207     }
208     qobject_decref(o);
209     o = object_property_get_qobject(obj, ACPI_PM_PROP_S4_VAL, NULL);
210     if (o) {
211         pm->s4_val = qint_get_int(qobject_to_qint(o));
212     } else {
213         pm->s4_val = false;
214     }
215     qobject_decref(o);
216 
217     /* Fill in mandatory properties */
218     pm->sci_int = object_property_get_int(obj, ACPI_PM_PROP_SCI_INT, NULL);
219 
220     pm->acpi_enable_cmd = object_property_get_int(obj,
221                                                   ACPI_PM_PROP_ACPI_ENABLE_CMD,
222                                                   NULL);
223     pm->acpi_disable_cmd = object_property_get_int(obj,
224                                                   ACPI_PM_PROP_ACPI_DISABLE_CMD,
225                                                   NULL);
226     pm->io_base = object_property_get_int(obj, ACPI_PM_PROP_PM_IO_BASE,
227                                           NULL);
228     pm->gpe0_blk = object_property_get_int(obj, ACPI_PM_PROP_GPE0_BLK,
229                                            NULL);
230     pm->gpe0_blk_len = object_property_get_int(obj, ACPI_PM_PROP_GPE0_BLK_LEN,
231                                                NULL);
232     pm->pcihp_bridge_en =
233         object_property_get_bool(obj, "acpi-pci-hotplug-with-bridge-support",
234                                  NULL);
235 }
236 
237 static void acpi_get_misc_info(AcpiMiscInfo *info)
238 {
239     info->has_hpet = hpet_find();
240     info->tpm_version = tpm_get_version();
241     info->pvpanic_port = pvpanic_port();
242     info->applesmc_io_base = applesmc_port();
243 }
244 
245 /*
246  * Because of the PXB hosts we cannot simply query TYPE_PCI_HOST_BRIDGE.
247  * On i386 arch we only have two pci hosts, so we can look only for them.
248  */
249 static Object *acpi_get_i386_pci_host(void)
250 {
251     PCIHostState *host;
252 
253     host = OBJECT_CHECK(PCIHostState,
254                         object_resolve_path("/machine/i440fx", NULL),
255                         TYPE_PCI_HOST_BRIDGE);
256     if (!host) {
257         host = OBJECT_CHECK(PCIHostState,
258                             object_resolve_path("/machine/q35", NULL),
259                             TYPE_PCI_HOST_BRIDGE);
260     }
261 
262     return OBJECT(host);
263 }
264 
265 static void acpi_get_pci_info(PcPciInfo *info)
266 {
267     Object *pci_host;
268 
269 
270     pci_host = acpi_get_i386_pci_host();
271     g_assert(pci_host);
272 
273     info->w32.begin = object_property_get_int(pci_host,
274                                               PCI_HOST_PROP_PCI_HOLE_START,
275                                               NULL);
276     info->w32.end = object_property_get_int(pci_host,
277                                             PCI_HOST_PROP_PCI_HOLE_END,
278                                             NULL);
279     info->w64.begin = object_property_get_int(pci_host,
280                                               PCI_HOST_PROP_PCI_HOLE64_START,
281                                               NULL);
282     info->w64.end = object_property_get_int(pci_host,
283                                             PCI_HOST_PROP_PCI_HOLE64_END,
284                                             NULL);
285 }
286 
287 #define ACPI_PORT_SMI_CMD           0x00b2 /* TODO: this is APM_CNT_IOPORT */
288 
289 static void acpi_align_size(GArray *blob, unsigned align)
290 {
291     /* Align size to multiple of given size. This reduces the chance
292      * we need to change size in the future (breaking cross version migration).
293      */
294     g_array_set_size(blob, ROUND_UP(acpi_data_len(blob), align));
295 }
296 
297 /* FACS */
298 static void
299 build_facs(GArray *table_data, GArray *linker, PcGuestInfo *guest_info)
300 {
301     AcpiFacsDescriptorRev1 *facs = acpi_data_push(table_data, sizeof *facs);
302     memcpy(&facs->signature, "FACS", 4);
303     facs->length = cpu_to_le32(sizeof(*facs));
304 }
305 
306 /* Load chipset information in FADT */
307 static void fadt_setup(AcpiFadtDescriptorRev1 *fadt, AcpiPmInfo *pm)
308 {
309     fadt->model = 1;
310     fadt->reserved1 = 0;
311     fadt->sci_int = cpu_to_le16(pm->sci_int);
312     fadt->smi_cmd = cpu_to_le32(ACPI_PORT_SMI_CMD);
313     fadt->acpi_enable = pm->acpi_enable_cmd;
314     fadt->acpi_disable = pm->acpi_disable_cmd;
315     /* EVT, CNT, TMR offset matches hw/acpi/core.c */
316     fadt->pm1a_evt_blk = cpu_to_le32(pm->io_base);
317     fadt->pm1a_cnt_blk = cpu_to_le32(pm->io_base + 0x04);
318     fadt->pm_tmr_blk = cpu_to_le32(pm->io_base + 0x08);
319     fadt->gpe0_blk = cpu_to_le32(pm->gpe0_blk);
320     /* EVT, CNT, TMR length matches hw/acpi/core.c */
321     fadt->pm1_evt_len = 4;
322     fadt->pm1_cnt_len = 2;
323     fadt->pm_tmr_len = 4;
324     fadt->gpe0_blk_len = pm->gpe0_blk_len;
325     fadt->plvl2_lat = cpu_to_le16(0xfff); /* C2 state not supported */
326     fadt->plvl3_lat = cpu_to_le16(0xfff); /* C3 state not supported */
327     fadt->flags = cpu_to_le32((1 << ACPI_FADT_F_WBINVD) |
328                               (1 << ACPI_FADT_F_PROC_C1) |
329                               (1 << ACPI_FADT_F_SLP_BUTTON) |
330                               (1 << ACPI_FADT_F_RTC_S4));
331     fadt->flags |= cpu_to_le32(1 << ACPI_FADT_F_USE_PLATFORM_CLOCK);
332     /* APIC destination mode ("Flat Logical") has an upper limit of 8 CPUs
333      * For more than 8 CPUs, "Clustered Logical" mode has to be used
334      */
335     if (max_cpus > 8) {
336         fadt->flags |= cpu_to_le32(1 << ACPI_FADT_F_FORCE_APIC_CLUSTER_MODEL);
337     }
338 }
339 
340 
341 /* FADT */
342 static void
343 build_fadt(GArray *table_data, GArray *linker, AcpiPmInfo *pm,
344            unsigned facs, unsigned dsdt)
345 {
346     AcpiFadtDescriptorRev1 *fadt = acpi_data_push(table_data, sizeof(*fadt));
347 
348     fadt->firmware_ctrl = cpu_to_le32(facs);
349     /* FACS address to be filled by Guest linker */
350     bios_linker_loader_add_pointer(linker, ACPI_BUILD_TABLE_FILE,
351                                    ACPI_BUILD_TABLE_FILE,
352                                    table_data, &fadt->firmware_ctrl,
353                                    sizeof fadt->firmware_ctrl);
354 
355     fadt->dsdt = cpu_to_le32(dsdt);
356     /* DSDT address to be filled by Guest linker */
357     bios_linker_loader_add_pointer(linker, ACPI_BUILD_TABLE_FILE,
358                                    ACPI_BUILD_TABLE_FILE,
359                                    table_data, &fadt->dsdt,
360                                    sizeof fadt->dsdt);
361 
362     fadt_setup(fadt, pm);
363 
364     build_header(linker, table_data,
365                  (void *)fadt, "FACP", sizeof(*fadt), 1, NULL);
366 }
367 
368 static void
369 build_madt(GArray *table_data, GArray *linker, AcpiCpuInfo *cpu,
370            PcGuestInfo *guest_info)
371 {
372     int madt_start = table_data->len;
373 
374     AcpiMultipleApicTable *madt;
375     AcpiMadtIoApic *io_apic;
376     AcpiMadtIntsrcovr *intsrcovr;
377     AcpiMadtLocalNmi *local_nmi;
378     int i;
379 
380     madt = acpi_data_push(table_data, sizeof *madt);
381     madt->local_apic_address = cpu_to_le32(APIC_DEFAULT_ADDRESS);
382     madt->flags = cpu_to_le32(1);
383 
384     for (i = 0; i < guest_info->apic_id_limit; i++) {
385         AcpiMadtProcessorApic *apic = acpi_data_push(table_data, sizeof *apic);
386         apic->type = ACPI_APIC_PROCESSOR;
387         apic->length = sizeof(*apic);
388         apic->processor_id = i;
389         apic->local_apic_id = i;
390         if (test_bit(i, cpu->found_cpus)) {
391             apic->flags = cpu_to_le32(1);
392         } else {
393             apic->flags = cpu_to_le32(0);
394         }
395     }
396     io_apic = acpi_data_push(table_data, sizeof *io_apic);
397     io_apic->type = ACPI_APIC_IO;
398     io_apic->length = sizeof(*io_apic);
399 #define ACPI_BUILD_IOAPIC_ID 0x0
400     io_apic->io_apic_id = ACPI_BUILD_IOAPIC_ID;
401     io_apic->address = cpu_to_le32(IO_APIC_DEFAULT_ADDRESS);
402     io_apic->interrupt = cpu_to_le32(0);
403 
404     if (guest_info->apic_xrupt_override) {
405         intsrcovr = acpi_data_push(table_data, sizeof *intsrcovr);
406         intsrcovr->type   = ACPI_APIC_XRUPT_OVERRIDE;
407         intsrcovr->length = sizeof(*intsrcovr);
408         intsrcovr->source = 0;
409         intsrcovr->gsi    = cpu_to_le32(2);
410         intsrcovr->flags  = cpu_to_le16(0); /* conforms to bus specifications */
411     }
412     for (i = 1; i < 16; i++) {
413 #define ACPI_BUILD_PCI_IRQS ((1<<5) | (1<<9) | (1<<10) | (1<<11))
414         if (!(ACPI_BUILD_PCI_IRQS & (1 << i))) {
415             /* No need for a INT source override structure. */
416             continue;
417         }
418         intsrcovr = acpi_data_push(table_data, sizeof *intsrcovr);
419         intsrcovr->type   = ACPI_APIC_XRUPT_OVERRIDE;
420         intsrcovr->length = sizeof(*intsrcovr);
421         intsrcovr->source = i;
422         intsrcovr->gsi    = cpu_to_le32(i);
423         intsrcovr->flags  = cpu_to_le16(0xd); /* active high, level triggered */
424     }
425 
426     local_nmi = acpi_data_push(table_data, sizeof *local_nmi);
427     local_nmi->type         = ACPI_APIC_LOCAL_NMI;
428     local_nmi->length       = sizeof(*local_nmi);
429     local_nmi->processor_id = 0xff; /* all processors */
430     local_nmi->flags        = cpu_to_le16(0);
431     local_nmi->lint         = 1; /* ACPI_LINT1 */
432 
433     build_header(linker, table_data,
434                  (void *)(table_data->data + madt_start), "APIC",
435                  table_data->len - madt_start, 1, NULL);
436 }
437 
438 /* Assign BSEL property to all buses.  In the future, this can be changed
439  * to only assign to buses that support hotplug.
440  */
441 static void *acpi_set_bsel(PCIBus *bus, void *opaque)
442 {
443     unsigned *bsel_alloc = opaque;
444     unsigned *bus_bsel;
445 
446     if (qbus_is_hotpluggable(BUS(bus))) {
447         bus_bsel = g_malloc(sizeof *bus_bsel);
448 
449         *bus_bsel = (*bsel_alloc)++;
450         object_property_add_uint32_ptr(OBJECT(bus), ACPI_PCIHP_PROP_BSEL,
451                                        bus_bsel, NULL);
452     }
453 
454     return bsel_alloc;
455 }
456 
457 static void acpi_set_pci_info(void)
458 {
459     PCIBus *bus = find_i440fx(); /* TODO: Q35 support */
460     unsigned bsel_alloc = 0;
461 
462     if (bus) {
463         /* Scan all PCI buses. Set property to enable acpi based hotplug. */
464         pci_for_each_bus_depth_first(bus, acpi_set_bsel, NULL, &bsel_alloc);
465     }
466 }
467 
468 static void build_append_pcihp_notify_entry(Aml *method, int slot)
469 {
470     Aml *if_ctx;
471     int32_t devfn = PCI_DEVFN(slot, 0);
472 
473     if_ctx = aml_if(aml_and(aml_arg(0), aml_int(0x1U << slot), NULL));
474     aml_append(if_ctx, aml_notify(aml_name("S%.02X", devfn), aml_arg(1)));
475     aml_append(method, if_ctx);
476 }
477 
478 static void build_append_pci_bus_devices(Aml *parent_scope, PCIBus *bus,
479                                          bool pcihp_bridge_en)
480 {
481     Aml *dev, *notify_method, *method;
482     QObject *bsel;
483     PCIBus *sec;
484     int i;
485 
486     bsel = object_property_get_qobject(OBJECT(bus), ACPI_PCIHP_PROP_BSEL, NULL);
487     if (bsel) {
488         int64_t bsel_val = qint_get_int(qobject_to_qint(bsel));
489 
490         aml_append(parent_scope, aml_name_decl("BSEL", aml_int(bsel_val)));
491         notify_method = aml_method("DVNT", 2, AML_NOTSERIALIZED);
492     }
493 
494     for (i = 0; i < ARRAY_SIZE(bus->devices); i += PCI_FUNC_MAX) {
495         DeviceClass *dc;
496         PCIDeviceClass *pc;
497         PCIDevice *pdev = bus->devices[i];
498         int slot = PCI_SLOT(i);
499         bool hotplug_enabled_dev;
500         bool bridge_in_acpi;
501 
502         if (!pdev) {
503             if (bsel) { /* add hotplug slots for non present devices */
504                 dev = aml_device("S%.02X", PCI_DEVFN(slot, 0));
505                 aml_append(dev, aml_name_decl("_SUN", aml_int(slot)));
506                 aml_append(dev, aml_name_decl("_ADR", aml_int(slot << 16)));
507                 method = aml_method("_EJ0", 1, AML_NOTSERIALIZED);
508                 aml_append(method,
509                     aml_call2("PCEJ", aml_name("BSEL"), aml_name("_SUN"))
510                 );
511                 aml_append(dev, method);
512                 aml_append(parent_scope, dev);
513 
514                 build_append_pcihp_notify_entry(notify_method, slot);
515             }
516             continue;
517         }
518 
519         pc = PCI_DEVICE_GET_CLASS(pdev);
520         dc = DEVICE_GET_CLASS(pdev);
521 
522         /* When hotplug for bridges is enabled, bridges are
523          * described in ACPI separately (see build_pci_bus_end).
524          * In this case they aren't themselves hot-pluggable.
525          * Hotplugged bridges *are* hot-pluggable.
526          */
527         bridge_in_acpi = pc->is_bridge && pcihp_bridge_en &&
528             !DEVICE(pdev)->hotplugged;
529 
530         hotplug_enabled_dev = bsel && dc->hotpluggable && !bridge_in_acpi;
531 
532         if (pc->class_id == PCI_CLASS_BRIDGE_ISA) {
533             continue;
534         }
535 
536         /* start to compose PCI slot descriptor */
537         dev = aml_device("S%.02X", PCI_DEVFN(slot, 0));
538         aml_append(dev, aml_name_decl("_ADR", aml_int(slot << 16)));
539 
540         if (pc->class_id == PCI_CLASS_DISPLAY_VGA) {
541             /* add VGA specific AML methods */
542             int s3d;
543 
544             if (object_dynamic_cast(OBJECT(pdev), "qxl-vga")) {
545                 s3d = 3;
546             } else {
547                 s3d = 0;
548             }
549 
550             method = aml_method("_S1D", 0, AML_NOTSERIALIZED);
551             aml_append(method, aml_return(aml_int(0)));
552             aml_append(dev, method);
553 
554             method = aml_method("_S2D", 0, AML_NOTSERIALIZED);
555             aml_append(method, aml_return(aml_int(0)));
556             aml_append(dev, method);
557 
558             method = aml_method("_S3D", 0, AML_NOTSERIALIZED);
559             aml_append(method, aml_return(aml_int(s3d)));
560             aml_append(dev, method);
561         } else if (hotplug_enabled_dev) {
562             /* add _SUN/_EJ0 to make slot hotpluggable  */
563             aml_append(dev, aml_name_decl("_SUN", aml_int(slot)));
564 
565             method = aml_method("_EJ0", 1, AML_NOTSERIALIZED);
566             aml_append(method,
567                 aml_call2("PCEJ", aml_name("BSEL"), aml_name("_SUN"))
568             );
569             aml_append(dev, method);
570 
571             if (bsel) {
572                 build_append_pcihp_notify_entry(notify_method, slot);
573             }
574         } else if (bridge_in_acpi) {
575             /*
576              * device is coldplugged bridge,
577              * add child device descriptions into its scope
578              */
579             PCIBus *sec_bus = pci_bridge_get_sec_bus(PCI_BRIDGE(pdev));
580 
581             build_append_pci_bus_devices(dev, sec_bus, pcihp_bridge_en);
582         }
583         /* slot descriptor has been composed, add it into parent context */
584         aml_append(parent_scope, dev);
585     }
586 
587     if (bsel) {
588         aml_append(parent_scope, notify_method);
589     }
590 
591     /* Append PCNT method to notify about events on local and child buses.
592      * Add unconditionally for root since DSDT expects it.
593      */
594     method = aml_method("PCNT", 0, AML_NOTSERIALIZED);
595 
596     /* If bus supports hotplug select it and notify about local events */
597     if (bsel) {
598         int64_t bsel_val = qint_get_int(qobject_to_qint(bsel));
599         aml_append(method, aml_store(aml_int(bsel_val), aml_name("BNUM")));
600         aml_append(method,
601             aml_call2("DVNT", aml_name("PCIU"), aml_int(1) /* Device Check */)
602         );
603         aml_append(method,
604             aml_call2("DVNT", aml_name("PCID"), aml_int(3)/* Eject Request */)
605         );
606     }
607 
608     /* Notify about child bus events in any case */
609     if (pcihp_bridge_en) {
610         QLIST_FOREACH(sec, &bus->child, sibling) {
611             int32_t devfn = sec->parent_dev->devfn;
612 
613             aml_append(method, aml_name("^S%.02X.PCNT", devfn));
614         }
615     }
616     aml_append(parent_scope, method);
617     qobject_decref(bsel);
618 }
619 
620 /*
621  * initialize_route - Initialize the interrupt routing rule
622  * through a specific LINK:
623  *  if (lnk_idx == idx)
624  *      route using link 'link_name'
625  */
626 static Aml *initialize_route(Aml *route, const char *link_name,
627                              Aml *lnk_idx, int idx)
628 {
629     Aml *if_ctx = aml_if(aml_equal(lnk_idx, aml_int(idx)));
630     Aml *pkg = aml_package(4);
631 
632     aml_append(pkg, aml_int(0));
633     aml_append(pkg, aml_int(0));
634     aml_append(pkg, aml_name("%s", link_name));
635     aml_append(pkg, aml_int(0));
636     aml_append(if_ctx, aml_store(pkg, route));
637 
638     return if_ctx;
639 }
640 
641 /*
642  * build_prt - Define interrupt rounting rules
643  *
644  * Returns an array of 128 routes, one for each device,
645  * based on device location.
646  * The main goal is to equaly distribute the interrupts
647  * over the 4 existing ACPI links (works only for i440fx).
648  * The hash function is  (slot + pin) & 3 -> "LNK[D|A|B|C]".
649  *
650  */
651 static Aml *build_prt(void)
652 {
653     Aml *method, *while_ctx, *pin, *res;
654 
655     method = aml_method("_PRT", 0, AML_NOTSERIALIZED);
656     res = aml_local(0);
657     pin = aml_local(1);
658     aml_append(method, aml_store(aml_package(128), res));
659     aml_append(method, aml_store(aml_int(0), pin));
660 
661     /* while (pin < 128) */
662     while_ctx = aml_while(aml_lless(pin, aml_int(128)));
663     {
664         Aml *slot = aml_local(2);
665         Aml *lnk_idx = aml_local(3);
666         Aml *route = aml_local(4);
667 
668         /* slot = pin >> 2 */
669         aml_append(while_ctx,
670                    aml_store(aml_shiftright(pin, aml_int(2), NULL), slot));
671         /* lnk_idx = (slot + pin) & 3 */
672         aml_append(while_ctx,
673             aml_store(aml_and(aml_add(pin, slot, NULL), aml_int(3), NULL),
674                       lnk_idx));
675 
676         /* route[2] = "LNK[D|A|B|C]", selection based on pin % 3  */
677         aml_append(while_ctx, initialize_route(route, "LNKD", lnk_idx, 0));
678         aml_append(while_ctx, initialize_route(route, "LNKA", lnk_idx, 1));
679         aml_append(while_ctx, initialize_route(route, "LNKB", lnk_idx, 2));
680         aml_append(while_ctx, initialize_route(route, "LNKC", lnk_idx, 3));
681 
682         /* route[0] = 0x[slot]FFFF */
683         aml_append(while_ctx,
684             aml_store(aml_or(aml_shiftleft(slot, aml_int(16)), aml_int(0xFFFF),
685                              NULL),
686                       aml_index(route, aml_int(0))));
687         /* route[1] = pin & 3 */
688         aml_append(while_ctx,
689             aml_store(aml_and(pin, aml_int(3), NULL),
690                       aml_index(route, aml_int(1))));
691         /* res[pin] = route */
692         aml_append(while_ctx, aml_store(route, aml_index(res, pin)));
693         /* pin++ */
694         aml_append(while_ctx, aml_increment(pin));
695     }
696     aml_append(method, while_ctx);
697     /* return res*/
698     aml_append(method, aml_return(res));
699 
700     return method;
701 }
702 
703 typedef struct CrsRangeEntry {
704     uint64_t base;
705     uint64_t limit;
706 } CrsRangeEntry;
707 
708 static void crs_range_insert(GPtrArray *ranges, uint64_t base, uint64_t limit)
709 {
710     CrsRangeEntry *entry;
711 
712     entry = g_malloc(sizeof(*entry));
713     entry->base = base;
714     entry->limit = limit;
715 
716     g_ptr_array_add(ranges, entry);
717 }
718 
719 static void crs_range_free(gpointer data)
720 {
721     CrsRangeEntry *entry = (CrsRangeEntry *)data;
722     g_free(entry);
723 }
724 
725 static gint crs_range_compare(gconstpointer a, gconstpointer b)
726 {
727      CrsRangeEntry *entry_a = *(CrsRangeEntry **)a;
728      CrsRangeEntry *entry_b = *(CrsRangeEntry **)b;
729 
730      return (int64_t)entry_a->base - (int64_t)entry_b->base;
731 }
732 
733 /*
734  * crs_replace_with_free_ranges - given the 'used' ranges within [start - end]
735  * interval, computes the 'free' ranges from the same interval.
736  * Example: If the input array is { [a1 - a2],[b1 - b2] }, the function
737  * will return { [base - a1], [a2 - b1], [b2 - limit] }.
738  */
739 static void crs_replace_with_free_ranges(GPtrArray *ranges,
740                                          uint64_t start, uint64_t end)
741 {
742     GPtrArray *free_ranges = g_ptr_array_new_with_free_func(crs_range_free);
743     uint64_t free_base = start;
744     int i;
745 
746     g_ptr_array_sort(ranges, crs_range_compare);
747     for (i = 0; i < ranges->len; i++) {
748         CrsRangeEntry *used = g_ptr_array_index(ranges, i);
749 
750         if (free_base < used->base) {
751             crs_range_insert(free_ranges, free_base, used->base - 1);
752         }
753 
754         free_base = used->limit + 1;
755     }
756 
757     if (free_base < end) {
758         crs_range_insert(free_ranges, free_base, end);
759     }
760 
761     g_ptr_array_set_size(ranges, 0);
762     for (i = 0; i < free_ranges->len; i++) {
763         g_ptr_array_add(ranges, g_ptr_array_index(free_ranges, i));
764     }
765 
766     g_ptr_array_free(free_ranges, false);
767 }
768 
769 /*
770  * crs_range_merge - merges adjacent ranges in the given array.
771  * Array elements are deleted and replaced with the merged ranges.
772  */
773 static void crs_range_merge(GPtrArray *range)
774 {
775     GPtrArray *tmp =  g_ptr_array_new_with_free_func(crs_range_free);
776     CrsRangeEntry *entry;
777     uint64_t range_base, range_limit;
778     int i;
779 
780     if (!range->len) {
781         return;
782     }
783 
784     g_ptr_array_sort(range, crs_range_compare);
785 
786     entry = g_ptr_array_index(range, 0);
787     range_base = entry->base;
788     range_limit = entry->limit;
789     for (i = 1; i < range->len; i++) {
790         entry = g_ptr_array_index(range, i);
791         if (entry->base - 1 == range_limit) {
792             range_limit = entry->limit;
793         } else {
794             crs_range_insert(tmp, range_base, range_limit);
795             range_base = entry->base;
796             range_limit = entry->limit;
797         }
798     }
799     crs_range_insert(tmp, range_base, range_limit);
800 
801     g_ptr_array_set_size(range, 0);
802     for (i = 0; i < tmp->len; i++) {
803         entry = g_ptr_array_index(tmp, i);
804         crs_range_insert(range, entry->base, entry->limit);
805     }
806     g_ptr_array_free(tmp, true);
807 }
808 
809 static Aml *build_crs(PCIHostState *host,
810                       GPtrArray *io_ranges, GPtrArray *mem_ranges)
811 {
812     Aml *crs = aml_resource_template();
813     GPtrArray *host_io_ranges = g_ptr_array_new_with_free_func(crs_range_free);
814     GPtrArray *host_mem_ranges = g_ptr_array_new_with_free_func(crs_range_free);
815     CrsRangeEntry *entry;
816     uint8_t max_bus = pci_bus_num(host->bus);
817     uint8_t type;
818     int devfn;
819     int i;
820 
821     for (devfn = 0; devfn < ARRAY_SIZE(host->bus->devices); devfn++) {
822         uint64_t range_base, range_limit;
823         PCIDevice *dev = host->bus->devices[devfn];
824 
825         if (!dev) {
826             continue;
827         }
828 
829         for (i = 0; i < PCI_NUM_REGIONS; i++) {
830             PCIIORegion *r = &dev->io_regions[i];
831 
832             range_base = r->addr;
833             range_limit = r->addr + r->size - 1;
834 
835             /*
836              * Work-around for old bioses
837              * that do not support multiple root buses
838              */
839             if (!range_base || range_base > range_limit) {
840                 continue;
841             }
842 
843             if (r->type & PCI_BASE_ADDRESS_SPACE_IO) {
844                 crs_range_insert(host_io_ranges, range_base, range_limit);
845             } else { /* "memory" */
846                 crs_range_insert(host_mem_ranges, range_base, range_limit);
847             }
848         }
849 
850         type = dev->config[PCI_HEADER_TYPE] & ~PCI_HEADER_TYPE_MULTI_FUNCTION;
851         if (type == PCI_HEADER_TYPE_BRIDGE) {
852             uint8_t subordinate = dev->config[PCI_SUBORDINATE_BUS];
853             if (subordinate > max_bus) {
854                 max_bus = subordinate;
855             }
856 
857             range_base = pci_bridge_get_base(dev, PCI_BASE_ADDRESS_SPACE_IO);
858             range_limit = pci_bridge_get_limit(dev, PCI_BASE_ADDRESS_SPACE_IO);
859 
860             /*
861              * Work-around for old bioses
862              * that do not support multiple root buses
863              */
864             if (range_base && range_base <= range_limit) {
865                 crs_range_insert(host_io_ranges, range_base, range_limit);
866             }
867 
868             range_base =
869                 pci_bridge_get_base(dev, PCI_BASE_ADDRESS_SPACE_MEMORY);
870             range_limit =
871                 pci_bridge_get_limit(dev, PCI_BASE_ADDRESS_SPACE_MEMORY);
872 
873             /*
874              * Work-around for old bioses
875              * that do not support multiple root buses
876              */
877             if (range_base && range_base <= range_limit) {
878                 crs_range_insert(host_mem_ranges, range_base, range_limit);
879             }
880 
881             range_base =
882                 pci_bridge_get_base(dev, PCI_BASE_ADDRESS_MEM_PREFETCH);
883             range_limit =
884                 pci_bridge_get_limit(dev, PCI_BASE_ADDRESS_MEM_PREFETCH);
885 
886             /*
887              * Work-around for old bioses
888              * that do not support multiple root buses
889              */
890             if (range_base && range_base <= range_limit) {
891                 crs_range_insert(host_mem_ranges, range_base, range_limit);
892             }
893         }
894     }
895 
896     crs_range_merge(host_io_ranges);
897     for (i = 0; i < host_io_ranges->len; i++) {
898         entry = g_ptr_array_index(host_io_ranges, i);
899         aml_append(crs,
900                    aml_word_io(AML_MIN_FIXED, AML_MAX_FIXED,
901                                AML_POS_DECODE, AML_ENTIRE_RANGE,
902                                0, entry->base, entry->limit, 0,
903                                entry->limit - entry->base + 1));
904         crs_range_insert(io_ranges, entry->base, entry->limit);
905     }
906     g_ptr_array_free(host_io_ranges, true);
907 
908     crs_range_merge(host_mem_ranges);
909     for (i = 0; i < host_mem_ranges->len; i++) {
910         entry = g_ptr_array_index(host_mem_ranges, i);
911         aml_append(crs,
912                    aml_dword_memory(AML_POS_DECODE, AML_MIN_FIXED,
913                                     AML_MAX_FIXED, AML_NON_CACHEABLE,
914                                     AML_READ_WRITE,
915                                     0, entry->base, entry->limit, 0,
916                                     entry->limit - entry->base + 1));
917         crs_range_insert(mem_ranges, entry->base, entry->limit);
918     }
919     g_ptr_array_free(host_mem_ranges, true);
920 
921     aml_append(crs,
922         aml_word_bus_number(AML_MIN_FIXED, AML_MAX_FIXED, AML_POS_DECODE,
923                             0,
924                             pci_bus_num(host->bus),
925                             max_bus,
926                             0,
927                             max_bus - pci_bus_num(host->bus) + 1));
928 
929     return crs;
930 }
931 
932 static void
933 build_ssdt(GArray *table_data, GArray *linker,
934            AcpiCpuInfo *cpu, AcpiPmInfo *pm, AcpiMiscInfo *misc,
935            PcPciInfo *pci, PcGuestInfo *guest_info)
936 {
937     MachineState *machine = MACHINE(qdev_get_machine());
938     uint32_t nr_mem = machine->ram_slots;
939     unsigned acpi_cpus = guest_info->apic_id_limit;
940     Aml *ssdt, *sb_scope, *scope, *pkg, *dev, *method, *crs, *field, *ifctx;
941     PCIBus *bus = NULL;
942     GPtrArray *io_ranges = g_ptr_array_new_with_free_func(crs_range_free);
943     GPtrArray *mem_ranges = g_ptr_array_new_with_free_func(crs_range_free);
944     CrsRangeEntry *entry;
945     int root_bus_limit = 0xFF;
946     int i;
947 
948     ssdt = init_aml_allocator();
949     /* The current AML generator can cover the APIC ID range [0..255],
950      * inclusive, for VCPU hotplug. */
951     QEMU_BUILD_BUG_ON(ACPI_CPU_HOTPLUG_ID_LIMIT > 256);
952     g_assert(acpi_cpus <= ACPI_CPU_HOTPLUG_ID_LIMIT);
953 
954     /* Reserve space for header */
955     acpi_data_push(ssdt->buf, sizeof(AcpiTableHeader));
956 
957     bus = PC_MACHINE(machine)->bus;
958     if (bus) {
959         QLIST_FOREACH(bus, &bus->child, sibling) {
960             uint8_t bus_num = pci_bus_num(bus);
961             uint8_t numa_node = pci_bus_numa_node(bus);
962 
963             /* look only for expander root buses */
964             if (!pci_bus_is_root(bus)) {
965                 continue;
966             }
967 
968             if (bus_num < root_bus_limit) {
969                 root_bus_limit = bus_num - 1;
970             }
971 
972             scope = aml_scope("\\_SB");
973             dev = aml_device("PC%.02X", bus_num);
974             aml_append(dev, aml_name_decl("_UID", aml_int(bus_num)));
975             aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0A03")));
976             aml_append(dev, aml_name_decl("_BBN", aml_int(bus_num)));
977 
978             if (numa_node != NUMA_NODE_UNASSIGNED) {
979                 aml_append(dev, aml_name_decl("_PXM", aml_int(numa_node)));
980             }
981 
982             aml_append(dev, build_prt());
983             crs = build_crs(PCI_HOST_BRIDGE(BUS(bus)->parent),
984                             io_ranges, mem_ranges);
985             aml_append(dev, aml_name_decl("_CRS", crs));
986             aml_append(scope, dev);
987             aml_append(ssdt, scope);
988         }
989     }
990 
991     scope = aml_scope("\\_SB.PCI0");
992     /* build PCI0._CRS */
993     crs = aml_resource_template();
994     aml_append(crs,
995         aml_word_bus_number(AML_MIN_FIXED, AML_MAX_FIXED, AML_POS_DECODE,
996                             0x0000, 0x0, root_bus_limit,
997                             0x0000, root_bus_limit + 1));
998     aml_append(crs, aml_io(AML_DECODE16, 0x0CF8, 0x0CF8, 0x01, 0x08));
999 
1000     aml_append(crs,
1001         aml_word_io(AML_MIN_FIXED, AML_MAX_FIXED,
1002                     AML_POS_DECODE, AML_ENTIRE_RANGE,
1003                     0x0000, 0x0000, 0x0CF7, 0x0000, 0x0CF8));
1004 
1005     crs_replace_with_free_ranges(io_ranges, 0x0D00, 0xFFFF);
1006     for (i = 0; i < io_ranges->len; i++) {
1007         entry = g_ptr_array_index(io_ranges, i);
1008         aml_append(crs,
1009             aml_word_io(AML_MIN_FIXED, AML_MAX_FIXED,
1010                         AML_POS_DECODE, AML_ENTIRE_RANGE,
1011                         0x0000, entry->base, entry->limit,
1012                         0x0000, entry->limit - entry->base + 1));
1013     }
1014 
1015     aml_append(crs,
1016         aml_dword_memory(AML_POS_DECODE, AML_MIN_FIXED, AML_MAX_FIXED,
1017                          AML_CACHEABLE, AML_READ_WRITE,
1018                          0, 0x000A0000, 0x000BFFFF, 0, 0x00020000));
1019 
1020     crs_replace_with_free_ranges(mem_ranges, pci->w32.begin, pci->w32.end - 1);
1021     for (i = 0; i < mem_ranges->len; i++) {
1022         entry = g_ptr_array_index(mem_ranges, i);
1023         aml_append(crs,
1024             aml_dword_memory(AML_POS_DECODE, AML_MIN_FIXED, AML_MAX_FIXED,
1025                              AML_NON_CACHEABLE, AML_READ_WRITE,
1026                              0, entry->base, entry->limit,
1027                              0, entry->limit - entry->base + 1));
1028     }
1029 
1030     if (pci->w64.begin) {
1031         aml_append(crs,
1032             aml_qword_memory(AML_POS_DECODE, AML_MIN_FIXED, AML_MAX_FIXED,
1033                              AML_CACHEABLE, AML_READ_WRITE,
1034                              0, pci->w64.begin, pci->w64.end - 1, 0,
1035                              pci->w64.end - pci->w64.begin));
1036     }
1037     aml_append(scope, aml_name_decl("_CRS", crs));
1038 
1039     /* reserve GPE0 block resources */
1040     dev = aml_device("GPE0");
1041     aml_append(dev, aml_name_decl("_HID", aml_string("PNP0A06")));
1042     aml_append(dev, aml_name_decl("_UID", aml_string("GPE0 resources")));
1043     /* device present, functioning, decoding, not shown in UI */
1044     aml_append(dev, aml_name_decl("_STA", aml_int(0xB)));
1045     crs = aml_resource_template();
1046     aml_append(crs,
1047         aml_io(AML_DECODE16, pm->gpe0_blk, pm->gpe0_blk, 1, pm->gpe0_blk_len)
1048     );
1049     aml_append(dev, aml_name_decl("_CRS", crs));
1050     aml_append(scope, dev);
1051 
1052     g_ptr_array_free(io_ranges, true);
1053     g_ptr_array_free(mem_ranges, true);
1054 
1055     /* reserve PCIHP resources */
1056     if (pm->pcihp_io_len) {
1057         dev = aml_device("PHPR");
1058         aml_append(dev, aml_name_decl("_HID", aml_string("PNP0A06")));
1059         aml_append(dev,
1060             aml_name_decl("_UID", aml_string("PCI Hotplug resources")));
1061         /* device present, functioning, decoding, not shown in UI */
1062         aml_append(dev, aml_name_decl("_STA", aml_int(0xB)));
1063         crs = aml_resource_template();
1064         aml_append(crs,
1065             aml_io(AML_DECODE16, pm->pcihp_io_base, pm->pcihp_io_base, 1,
1066                    pm->pcihp_io_len)
1067         );
1068         aml_append(dev, aml_name_decl("_CRS", crs));
1069         aml_append(scope, dev);
1070     }
1071     aml_append(ssdt, scope);
1072 
1073     /*  create S3_ / S4_ / S5_ packages if necessary */
1074     scope = aml_scope("\\");
1075     if (!pm->s3_disabled) {
1076         pkg = aml_package(4);
1077         aml_append(pkg, aml_int(1)); /* PM1a_CNT.SLP_TYP */
1078         aml_append(pkg, aml_int(1)); /* PM1b_CNT.SLP_TYP, FIXME: not impl. */
1079         aml_append(pkg, aml_int(0)); /* reserved */
1080         aml_append(pkg, aml_int(0)); /* reserved */
1081         aml_append(scope, aml_name_decl("_S3", pkg));
1082     }
1083 
1084     if (!pm->s4_disabled) {
1085         pkg = aml_package(4);
1086         aml_append(pkg, aml_int(pm->s4_val)); /* PM1a_CNT.SLP_TYP */
1087         /* PM1b_CNT.SLP_TYP, FIXME: not impl. */
1088         aml_append(pkg, aml_int(pm->s4_val));
1089         aml_append(pkg, aml_int(0)); /* reserved */
1090         aml_append(pkg, aml_int(0)); /* reserved */
1091         aml_append(scope, aml_name_decl("_S4", pkg));
1092     }
1093 
1094     pkg = aml_package(4);
1095     aml_append(pkg, aml_int(0)); /* PM1a_CNT.SLP_TYP */
1096     aml_append(pkg, aml_int(0)); /* PM1b_CNT.SLP_TYP not impl. */
1097     aml_append(pkg, aml_int(0)); /* reserved */
1098     aml_append(pkg, aml_int(0)); /* reserved */
1099     aml_append(scope, aml_name_decl("_S5", pkg));
1100     aml_append(ssdt, scope);
1101 
1102     if (misc->applesmc_io_base) {
1103         scope = aml_scope("\\_SB.PCI0.ISA");
1104         dev = aml_device("SMC");
1105 
1106         aml_append(dev, aml_name_decl("_HID", aml_eisaid("APP0001")));
1107         /* device present, functioning, decoding, not shown in UI */
1108         aml_append(dev, aml_name_decl("_STA", aml_int(0xB)));
1109 
1110         crs = aml_resource_template();
1111         aml_append(crs,
1112             aml_io(AML_DECODE16, misc->applesmc_io_base, misc->applesmc_io_base,
1113                    0x01, APPLESMC_MAX_DATA_LENGTH)
1114         );
1115         aml_append(crs, aml_irq_no_flags(6));
1116         aml_append(dev, aml_name_decl("_CRS", crs));
1117 
1118         aml_append(scope, dev);
1119         aml_append(ssdt, scope);
1120     }
1121 
1122     if (misc->pvpanic_port) {
1123         scope = aml_scope("\\_SB.PCI0.ISA");
1124 
1125         dev = aml_device("PEVT");
1126         aml_append(dev, aml_name_decl("_HID", aml_string("QEMU0001")));
1127 
1128         crs = aml_resource_template();
1129         aml_append(crs,
1130             aml_io(AML_DECODE16, misc->pvpanic_port, misc->pvpanic_port, 1, 1)
1131         );
1132         aml_append(dev, aml_name_decl("_CRS", crs));
1133 
1134         aml_append(dev, aml_operation_region("PEOR", AML_SYSTEM_IO,
1135                                               misc->pvpanic_port, 1));
1136         field = aml_field("PEOR", AML_BYTE_ACC, AML_NOLOCK, AML_PRESERVE);
1137         aml_append(field, aml_named_field("PEPT", 8));
1138         aml_append(dev, field);
1139 
1140         /* device present, functioning, decoding, shown in UI */
1141         aml_append(dev, aml_name_decl("_STA", aml_int(0xF)));
1142 
1143         method = aml_method("RDPT", 0, AML_NOTSERIALIZED);
1144         aml_append(method, aml_store(aml_name("PEPT"), aml_local(0)));
1145         aml_append(method, aml_return(aml_local(0)));
1146         aml_append(dev, method);
1147 
1148         method = aml_method("WRPT", 1, AML_NOTSERIALIZED);
1149         aml_append(method, aml_store(aml_arg(0), aml_name("PEPT")));
1150         aml_append(dev, method);
1151 
1152         aml_append(scope, dev);
1153         aml_append(ssdt, scope);
1154     }
1155 
1156     sb_scope = aml_scope("\\_SB");
1157     {
1158         /* create PCI0.PRES device and its _CRS to reserve CPU hotplug MMIO */
1159         dev = aml_device("PCI0." stringify(CPU_HOTPLUG_RESOURCE_DEVICE));
1160         aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0A06")));
1161         aml_append(dev,
1162             aml_name_decl("_UID", aml_string("CPU Hotplug resources"))
1163         );
1164         /* device present, functioning, decoding, not shown in UI */
1165         aml_append(dev, aml_name_decl("_STA", aml_int(0xB)));
1166         crs = aml_resource_template();
1167         aml_append(crs,
1168             aml_io(AML_DECODE16, pm->cpu_hp_io_base, pm->cpu_hp_io_base, 1,
1169                    pm->cpu_hp_io_len)
1170         );
1171         aml_append(dev, aml_name_decl("_CRS", crs));
1172         aml_append(sb_scope, dev);
1173         /* declare CPU hotplug MMIO region and PRS field to access it */
1174         aml_append(sb_scope, aml_operation_region(
1175             "PRST", AML_SYSTEM_IO, pm->cpu_hp_io_base, pm->cpu_hp_io_len));
1176         field = aml_field("PRST", AML_BYTE_ACC, AML_NOLOCK, AML_PRESERVE);
1177         aml_append(field, aml_named_field("PRS", 256));
1178         aml_append(sb_scope, field);
1179 
1180         /* build Processor object for each processor */
1181         for (i = 0; i < acpi_cpus; i++) {
1182             dev = aml_processor(i, 0, 0, "CP%.02X", i);
1183 
1184             method = aml_method("_MAT", 0, AML_NOTSERIALIZED);
1185             aml_append(method, aml_return(aml_call1("CPMA", aml_int(i))));
1186             aml_append(dev, method);
1187 
1188             method = aml_method("_STA", 0, AML_NOTSERIALIZED);
1189             aml_append(method, aml_return(aml_call1("CPST", aml_int(i))));
1190             aml_append(dev, method);
1191 
1192             method = aml_method("_EJ0", 1, AML_NOTSERIALIZED);
1193             aml_append(method,
1194                 aml_return(aml_call2("CPEJ", aml_int(i), aml_arg(0)))
1195             );
1196             aml_append(dev, method);
1197 
1198             aml_append(sb_scope, dev);
1199         }
1200 
1201         /* build this code:
1202          *   Method(NTFY, 2) {If (LEqual(Arg0, 0x00)) {Notify(CP00, Arg1)} ...}
1203          */
1204         /* Arg0 = Processor ID = APIC ID */
1205         method = aml_method("NTFY", 2, AML_NOTSERIALIZED);
1206         for (i = 0; i < acpi_cpus; i++) {
1207             ifctx = aml_if(aml_equal(aml_arg(0), aml_int(i)));
1208             aml_append(ifctx,
1209                 aml_notify(aml_name("CP%.02X", i), aml_arg(1))
1210             );
1211             aml_append(method, ifctx);
1212         }
1213         aml_append(sb_scope, method);
1214 
1215         /* build "Name(CPON, Package() { One, One, ..., Zero, Zero, ... })"
1216          *
1217          * Note: The ability to create variable-sized packages was first
1218          * introduced in ACPI 2.0. ACPI 1.0 only allowed fixed-size packages
1219          * ith up to 255 elements. Windows guests up to win2k8 fail when
1220          * VarPackageOp is used.
1221          */
1222         pkg = acpi_cpus <= 255 ? aml_package(acpi_cpus) :
1223                                  aml_varpackage(acpi_cpus);
1224 
1225         for (i = 0; i < acpi_cpus; i++) {
1226             uint8_t b = test_bit(i, cpu->found_cpus) ? 0x01 : 0x00;
1227             aml_append(pkg, aml_int(b));
1228         }
1229         aml_append(sb_scope, aml_name_decl("CPON", pkg));
1230 
1231         /* build memory devices */
1232         assert(nr_mem <= ACPI_MAX_RAM_SLOTS);
1233         scope = aml_scope("\\_SB.PCI0." stringify(MEMORY_HOTPLUG_DEVICE));
1234         aml_append(scope,
1235             aml_name_decl(stringify(MEMORY_SLOTS_NUMBER), aml_int(nr_mem))
1236         );
1237 
1238         crs = aml_resource_template();
1239         aml_append(crs,
1240             aml_io(AML_DECODE16, pm->mem_hp_io_base, pm->mem_hp_io_base, 0,
1241                    pm->mem_hp_io_len)
1242         );
1243         aml_append(scope, aml_name_decl("_CRS", crs));
1244 
1245         aml_append(scope, aml_operation_region(
1246             stringify(MEMORY_HOTPLUG_IO_REGION), AML_SYSTEM_IO,
1247             pm->mem_hp_io_base, pm->mem_hp_io_len)
1248         );
1249 
1250         field = aml_field(stringify(MEMORY_HOTPLUG_IO_REGION), AML_DWORD_ACC,
1251                           AML_NOLOCK, AML_PRESERVE);
1252         aml_append(field, /* read only */
1253             aml_named_field(stringify(MEMORY_SLOT_ADDR_LOW), 32));
1254         aml_append(field, /* read only */
1255             aml_named_field(stringify(MEMORY_SLOT_ADDR_HIGH), 32));
1256         aml_append(field, /* read only */
1257             aml_named_field(stringify(MEMORY_SLOT_SIZE_LOW), 32));
1258         aml_append(field, /* read only */
1259             aml_named_field(stringify(MEMORY_SLOT_SIZE_HIGH), 32));
1260         aml_append(field, /* read only */
1261             aml_named_field(stringify(MEMORY_SLOT_PROXIMITY), 32));
1262         aml_append(scope, field);
1263 
1264         field = aml_field(stringify(MEMORY_HOTPLUG_IO_REGION), AML_BYTE_ACC,
1265                           AML_NOLOCK, AML_WRITE_AS_ZEROS);
1266         aml_append(field, aml_reserved_field(160 /* bits, Offset(20) */));
1267         aml_append(field, /* 1 if enabled, read only */
1268             aml_named_field(stringify(MEMORY_SLOT_ENABLED), 1));
1269         aml_append(field,
1270             /*(read) 1 if has a insert event. (write) 1 to clear event */
1271             aml_named_field(stringify(MEMORY_SLOT_INSERT_EVENT), 1));
1272         aml_append(field,
1273             /* (read) 1 if has a remove event. (write) 1 to clear event */
1274             aml_named_field(stringify(MEMORY_SLOT_REMOVE_EVENT), 1));
1275         aml_append(field,
1276             /* initiates device eject, write only */
1277             aml_named_field(stringify(MEMORY_SLOT_EJECT), 1));
1278         aml_append(scope, field);
1279 
1280         field = aml_field(stringify(MEMORY_HOTPLUG_IO_REGION), AML_DWORD_ACC,
1281                           AML_NOLOCK, AML_PRESERVE);
1282         aml_append(field, /* DIMM selector, write only */
1283             aml_named_field(stringify(MEMORY_SLOT_SLECTOR), 32));
1284         aml_append(field, /* _OST event code, write only */
1285             aml_named_field(stringify(MEMORY_SLOT_OST_EVENT), 32));
1286         aml_append(field, /* _OST status code, write only */
1287             aml_named_field(stringify(MEMORY_SLOT_OST_STATUS), 32));
1288         aml_append(scope, field);
1289 
1290         aml_append(sb_scope, scope);
1291 
1292         for (i = 0; i < nr_mem; i++) {
1293             #define BASEPATH "\\_SB.PCI0." stringify(MEMORY_HOTPLUG_DEVICE) "."
1294             const char *s;
1295 
1296             dev = aml_device("MP%02X", i);
1297             aml_append(dev, aml_name_decl("_UID", aml_string("0x%02X", i)));
1298             aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0C80")));
1299 
1300             method = aml_method("_CRS", 0, AML_NOTSERIALIZED);
1301             s = BASEPATH stringify(MEMORY_SLOT_CRS_METHOD);
1302             aml_append(method, aml_return(aml_call1(s, aml_name("_UID"))));
1303             aml_append(dev, method);
1304 
1305             method = aml_method("_STA", 0, AML_NOTSERIALIZED);
1306             s = BASEPATH stringify(MEMORY_SLOT_STATUS_METHOD);
1307             aml_append(method, aml_return(aml_call1(s, aml_name("_UID"))));
1308             aml_append(dev, method);
1309 
1310             method = aml_method("_PXM", 0, AML_NOTSERIALIZED);
1311             s = BASEPATH stringify(MEMORY_SLOT_PROXIMITY_METHOD);
1312             aml_append(method, aml_return(aml_call1(s, aml_name("_UID"))));
1313             aml_append(dev, method);
1314 
1315             method = aml_method("_OST", 3, AML_NOTSERIALIZED);
1316             s = BASEPATH stringify(MEMORY_SLOT_OST_METHOD);
1317             aml_append(method, aml_return(aml_call4(
1318                 s, aml_name("_UID"), aml_arg(0), aml_arg(1), aml_arg(2)
1319             )));
1320             aml_append(dev, method);
1321 
1322             method = aml_method("_EJ0", 1, AML_NOTSERIALIZED);
1323             s = BASEPATH stringify(MEMORY_SLOT_EJECT_METHOD);
1324             aml_append(method, aml_return(aml_call2(
1325                        s, aml_name("_UID"), aml_arg(0))));
1326             aml_append(dev, method);
1327 
1328             aml_append(sb_scope, dev);
1329         }
1330 
1331         /* build Method(MEMORY_SLOT_NOTIFY_METHOD, 2) {
1332          *     If (LEqual(Arg0, 0x00)) {Notify(MP00, Arg1)} ... }
1333          */
1334         method = aml_method(stringify(MEMORY_SLOT_NOTIFY_METHOD), 2,
1335                             AML_NOTSERIALIZED);
1336         for (i = 0; i < nr_mem; i++) {
1337             ifctx = aml_if(aml_equal(aml_arg(0), aml_int(i)));
1338             aml_append(ifctx,
1339                 aml_notify(aml_name("MP%.02X", i), aml_arg(1))
1340             );
1341             aml_append(method, ifctx);
1342         }
1343         aml_append(sb_scope, method);
1344 
1345         {
1346             Object *pci_host;
1347             PCIBus *bus = NULL;
1348 
1349             pci_host = acpi_get_i386_pci_host();
1350             if (pci_host) {
1351                 bus = PCI_HOST_BRIDGE(pci_host)->bus;
1352             }
1353 
1354             if (bus) {
1355                 Aml *scope = aml_scope("PCI0");
1356                 /* Scan all PCI buses. Generate tables to support hotplug. */
1357                 build_append_pci_bus_devices(scope, bus, pm->pcihp_bridge_en);
1358 
1359                 if (misc->tpm_version != TPM_VERSION_UNSPEC) {
1360                     dev = aml_device("ISA.TPM");
1361                     aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0C31")));
1362                     aml_append(dev, aml_name_decl("_STA", aml_int(0xF)));
1363                     crs = aml_resource_template();
1364                     aml_append(crs, aml_memory32_fixed(TPM_TIS_ADDR_BASE,
1365                                TPM_TIS_ADDR_SIZE, AML_READ_WRITE));
1366                     aml_append(crs, aml_irq_no_flags(TPM_TIS_IRQ));
1367                     aml_append(dev, aml_name_decl("_CRS", crs));
1368                     aml_append(scope, dev);
1369                 }
1370 
1371                 aml_append(sb_scope, scope);
1372             }
1373         }
1374         aml_append(ssdt, sb_scope);
1375     }
1376 
1377     /* copy AML table into ACPI tables blob and patch header there */
1378     g_array_append_vals(table_data, ssdt->buf->data, ssdt->buf->len);
1379     build_header(linker, table_data,
1380         (void *)(table_data->data + table_data->len - ssdt->buf->len),
1381         "SSDT", ssdt->buf->len, 1, NULL);
1382     free_aml_allocator();
1383 }
1384 
1385 static void
1386 build_hpet(GArray *table_data, GArray *linker)
1387 {
1388     Acpi20Hpet *hpet;
1389 
1390     hpet = acpi_data_push(table_data, sizeof(*hpet));
1391     /* Note timer_block_id value must be kept in sync with value advertised by
1392      * emulated hpet
1393      */
1394     hpet->timer_block_id = cpu_to_le32(0x8086a201);
1395     hpet->addr.address = cpu_to_le64(HPET_BASE);
1396     build_header(linker, table_data,
1397                  (void *)hpet, "HPET", sizeof(*hpet), 1, NULL);
1398 }
1399 
1400 static void
1401 build_tpm_tcpa(GArray *table_data, GArray *linker, GArray *tcpalog)
1402 {
1403     Acpi20Tcpa *tcpa = acpi_data_push(table_data, sizeof *tcpa);
1404     uint64_t log_area_start_address = acpi_data_len(tcpalog);
1405 
1406     tcpa->platform_class = cpu_to_le16(TPM_TCPA_ACPI_CLASS_CLIENT);
1407     tcpa->log_area_minimum_length = cpu_to_le32(TPM_LOG_AREA_MINIMUM_SIZE);
1408     tcpa->log_area_start_address = cpu_to_le64(log_area_start_address);
1409 
1410     bios_linker_loader_alloc(linker, ACPI_BUILD_TPMLOG_FILE, 1,
1411                              false /* high memory */);
1412 
1413     /* log area start address to be filled by Guest linker */
1414     bios_linker_loader_add_pointer(linker, ACPI_BUILD_TABLE_FILE,
1415                                    ACPI_BUILD_TPMLOG_FILE,
1416                                    table_data, &tcpa->log_area_start_address,
1417                                    sizeof(tcpa->log_area_start_address));
1418 
1419     build_header(linker, table_data,
1420                  (void *)tcpa, "TCPA", sizeof(*tcpa), 2, NULL);
1421 
1422     acpi_data_push(tcpalog, TPM_LOG_AREA_MINIMUM_SIZE);
1423 }
1424 
1425 static void
1426 build_tpm2(GArray *table_data, GArray *linker)
1427 {
1428     Acpi20TPM2 *tpm2_ptr;
1429 
1430     tpm2_ptr = acpi_data_push(table_data, sizeof *tpm2_ptr);
1431 
1432     tpm2_ptr->platform_class = cpu_to_le16(TPM2_ACPI_CLASS_CLIENT);
1433     tpm2_ptr->control_area_address = cpu_to_le64(0);
1434     tpm2_ptr->start_method = cpu_to_le32(TPM2_START_METHOD_MMIO);
1435 
1436     build_header(linker, table_data,
1437                  (void *)tpm2_ptr, "TPM2", sizeof(*tpm2_ptr), 4, NULL);
1438 }
1439 
1440 typedef enum {
1441     MEM_AFFINITY_NOFLAGS      = 0,
1442     MEM_AFFINITY_ENABLED      = (1 << 0),
1443     MEM_AFFINITY_HOTPLUGGABLE = (1 << 1),
1444     MEM_AFFINITY_NON_VOLATILE = (1 << 2),
1445 } MemoryAffinityFlags;
1446 
1447 static void
1448 acpi_build_srat_memory(AcpiSratMemoryAffinity *numamem, uint64_t base,
1449                        uint64_t len, int node, MemoryAffinityFlags flags)
1450 {
1451     numamem->type = ACPI_SRAT_MEMORY;
1452     numamem->length = sizeof(*numamem);
1453     memset(numamem->proximity, 0, 4);
1454     numamem->proximity[0] = node;
1455     numamem->flags = cpu_to_le32(flags);
1456     numamem->base_addr = cpu_to_le64(base);
1457     numamem->range_length = cpu_to_le64(len);
1458 }
1459 
1460 static void
1461 build_srat(GArray *table_data, GArray *linker, PcGuestInfo *guest_info)
1462 {
1463     AcpiSystemResourceAffinityTable *srat;
1464     AcpiSratProcessorAffinity *core;
1465     AcpiSratMemoryAffinity *numamem;
1466 
1467     int i;
1468     uint64_t curnode;
1469     int srat_start, numa_start, slots;
1470     uint64_t mem_len, mem_base, next_base;
1471     PCMachineState *pcms = PC_MACHINE(qdev_get_machine());
1472     ram_addr_t hotplugabble_address_space_size =
1473         object_property_get_int(OBJECT(pcms), PC_MACHINE_MEMHP_REGION_SIZE,
1474                                 NULL);
1475 
1476     srat_start = table_data->len;
1477 
1478     srat = acpi_data_push(table_data, sizeof *srat);
1479     srat->reserved1 = cpu_to_le32(1);
1480     core = (void *)(srat + 1);
1481 
1482     for (i = 0; i < guest_info->apic_id_limit; ++i) {
1483         core = acpi_data_push(table_data, sizeof *core);
1484         core->type = ACPI_SRAT_PROCESSOR;
1485         core->length = sizeof(*core);
1486         core->local_apic_id = i;
1487         curnode = guest_info->node_cpu[i];
1488         core->proximity_lo = curnode;
1489         memset(core->proximity_hi, 0, 3);
1490         core->local_sapic_eid = 0;
1491         core->flags = cpu_to_le32(1);
1492     }
1493 
1494 
1495     /* the memory map is a bit tricky, it contains at least one hole
1496      * from 640k-1M and possibly another one from 3.5G-4G.
1497      */
1498     next_base = 0;
1499     numa_start = table_data->len;
1500 
1501     numamem = acpi_data_push(table_data, sizeof *numamem);
1502     acpi_build_srat_memory(numamem, 0, 640*1024, 0, MEM_AFFINITY_ENABLED);
1503     next_base = 1024 * 1024;
1504     for (i = 1; i < guest_info->numa_nodes + 1; ++i) {
1505         mem_base = next_base;
1506         mem_len = guest_info->node_mem[i - 1];
1507         if (i == 1) {
1508             mem_len -= 1024 * 1024;
1509         }
1510         next_base = mem_base + mem_len;
1511 
1512         /* Cut out the ACPI_PCI hole */
1513         if (mem_base <= guest_info->ram_size_below_4g &&
1514             next_base > guest_info->ram_size_below_4g) {
1515             mem_len -= next_base - guest_info->ram_size_below_4g;
1516             if (mem_len > 0) {
1517                 numamem = acpi_data_push(table_data, sizeof *numamem);
1518                 acpi_build_srat_memory(numamem, mem_base, mem_len, i - 1,
1519                                        MEM_AFFINITY_ENABLED);
1520             }
1521             mem_base = 1ULL << 32;
1522             mem_len = next_base - guest_info->ram_size_below_4g;
1523             next_base += (1ULL << 32) - guest_info->ram_size_below_4g;
1524         }
1525         numamem = acpi_data_push(table_data, sizeof *numamem);
1526         acpi_build_srat_memory(numamem, mem_base, mem_len, i - 1,
1527                                MEM_AFFINITY_ENABLED);
1528     }
1529     slots = (table_data->len - numa_start) / sizeof *numamem;
1530     for (; slots < guest_info->numa_nodes + 2; slots++) {
1531         numamem = acpi_data_push(table_data, sizeof *numamem);
1532         acpi_build_srat_memory(numamem, 0, 0, 0, MEM_AFFINITY_NOFLAGS);
1533     }
1534 
1535     /*
1536      * Entry is required for Windows to enable memory hotplug in OS.
1537      * Memory devices may override proximity set by this entry,
1538      * providing _PXM method if necessary.
1539      */
1540     if (hotplugabble_address_space_size) {
1541         numamem = acpi_data_push(table_data, sizeof *numamem);
1542         acpi_build_srat_memory(numamem, pcms->hotplug_memory.base,
1543                                hotplugabble_address_space_size, 0,
1544                                MEM_AFFINITY_HOTPLUGGABLE |
1545                                MEM_AFFINITY_ENABLED);
1546     }
1547 
1548     build_header(linker, table_data,
1549                  (void *)(table_data->data + srat_start),
1550                  "SRAT",
1551                  table_data->len - srat_start, 1, NULL);
1552 }
1553 
1554 static void
1555 build_mcfg_q35(GArray *table_data, GArray *linker, AcpiMcfgInfo *info)
1556 {
1557     AcpiTableMcfg *mcfg;
1558     const char *sig;
1559     int len = sizeof(*mcfg) + 1 * sizeof(mcfg->allocation[0]);
1560 
1561     mcfg = acpi_data_push(table_data, len);
1562     mcfg->allocation[0].address = cpu_to_le64(info->mcfg_base);
1563     /* Only a single allocation so no need to play with segments */
1564     mcfg->allocation[0].pci_segment = cpu_to_le16(0);
1565     mcfg->allocation[0].start_bus_number = 0;
1566     mcfg->allocation[0].end_bus_number = PCIE_MMCFG_BUS(info->mcfg_size - 1);
1567 
1568     /* MCFG is used for ECAM which can be enabled or disabled by guest.
1569      * To avoid table size changes (which create migration issues),
1570      * always create the table even if there are no allocations,
1571      * but set the signature to a reserved value in this case.
1572      * ACPI spec requires OSPMs to ignore such tables.
1573      */
1574     if (info->mcfg_base == PCIE_BASE_ADDR_UNMAPPED) {
1575         /* Reserved signature: ignored by OSPM */
1576         sig = "QEMU";
1577     } else {
1578         sig = "MCFG";
1579     }
1580     build_header(linker, table_data, (void *)mcfg, sig, len, 1, NULL);
1581 }
1582 
1583 static void
1584 build_dmar_q35(GArray *table_data, GArray *linker)
1585 {
1586     int dmar_start = table_data->len;
1587 
1588     AcpiTableDmar *dmar;
1589     AcpiDmarHardwareUnit *drhd;
1590 
1591     dmar = acpi_data_push(table_data, sizeof(*dmar));
1592     dmar->host_address_width = VTD_HOST_ADDRESS_WIDTH - 1;
1593     dmar->flags = 0;    /* No intr_remap for now */
1594 
1595     /* DMAR Remapping Hardware Unit Definition structure */
1596     drhd = acpi_data_push(table_data, sizeof(*drhd));
1597     drhd->type = cpu_to_le16(ACPI_DMAR_TYPE_HARDWARE_UNIT);
1598     drhd->length = cpu_to_le16(sizeof(*drhd));   /* No device scope now */
1599     drhd->flags = ACPI_DMAR_INCLUDE_PCI_ALL;
1600     drhd->pci_segment = cpu_to_le16(0);
1601     drhd->address = cpu_to_le64(Q35_HOST_BRIDGE_IOMMU_ADDR);
1602 
1603     build_header(linker, table_data, (void *)(table_data->data + dmar_start),
1604                  "DMAR", table_data->len - dmar_start, 1, NULL);
1605 }
1606 
1607 static void
1608 build_dsdt(GArray *table_data, GArray *linker, AcpiMiscInfo *misc)
1609 {
1610     AcpiTableHeader *dsdt;
1611 
1612     assert(misc->dsdt_code && misc->dsdt_size);
1613 
1614     dsdt = acpi_data_push(table_data, misc->dsdt_size);
1615     memcpy(dsdt, misc->dsdt_code, misc->dsdt_size);
1616 
1617     memset(dsdt, 0, sizeof *dsdt);
1618     build_header(linker, table_data, dsdt, "DSDT",
1619                  misc->dsdt_size, 1, NULL);
1620 }
1621 
1622 static GArray *
1623 build_rsdp(GArray *rsdp_table, GArray *linker, unsigned rsdt)
1624 {
1625     AcpiRsdpDescriptor *rsdp = acpi_data_push(rsdp_table, sizeof *rsdp);
1626 
1627     bios_linker_loader_alloc(linker, ACPI_BUILD_RSDP_FILE, 16,
1628                              true /* fseg memory */);
1629 
1630     memcpy(&rsdp->signature, "RSD PTR ", 8);
1631     memcpy(rsdp->oem_id, ACPI_BUILD_APPNAME6, 6);
1632     rsdp->rsdt_physical_address = cpu_to_le32(rsdt);
1633     /* Address to be filled by Guest linker */
1634     bios_linker_loader_add_pointer(linker, ACPI_BUILD_RSDP_FILE,
1635                                    ACPI_BUILD_TABLE_FILE,
1636                                    rsdp_table, &rsdp->rsdt_physical_address,
1637                                    sizeof rsdp->rsdt_physical_address);
1638     rsdp->checksum = 0;
1639     /* Checksum to be filled by Guest linker */
1640     bios_linker_loader_add_checksum(linker, ACPI_BUILD_RSDP_FILE,
1641                                     rsdp, rsdp, sizeof *rsdp, &rsdp->checksum);
1642 
1643     return rsdp_table;
1644 }
1645 
1646 typedef
1647 struct AcpiBuildState {
1648     /* Copy of table in RAM (for patching). */
1649     MemoryRegion *table_mr;
1650     /* Is table patched? */
1651     uint8_t patched;
1652     PcGuestInfo *guest_info;
1653     void *rsdp;
1654     MemoryRegion *rsdp_mr;
1655     MemoryRegion *linker_mr;
1656 } AcpiBuildState;
1657 
1658 static bool acpi_get_mcfg(AcpiMcfgInfo *mcfg)
1659 {
1660     Object *pci_host;
1661     QObject *o;
1662 
1663     pci_host = acpi_get_i386_pci_host();
1664     g_assert(pci_host);
1665 
1666     o = object_property_get_qobject(pci_host, PCIE_HOST_MCFG_BASE, NULL);
1667     if (!o) {
1668         return false;
1669     }
1670     mcfg->mcfg_base = qint_get_int(qobject_to_qint(o));
1671     qobject_decref(o);
1672 
1673     o = object_property_get_qobject(pci_host, PCIE_HOST_MCFG_SIZE, NULL);
1674     assert(o);
1675     mcfg->mcfg_size = qint_get_int(qobject_to_qint(o));
1676     qobject_decref(o);
1677     return true;
1678 }
1679 
1680 static bool acpi_has_iommu(void)
1681 {
1682     bool ambiguous;
1683     Object *intel_iommu;
1684 
1685     intel_iommu = object_resolve_path_type("", TYPE_INTEL_IOMMU_DEVICE,
1686                                            &ambiguous);
1687     return intel_iommu && !ambiguous;
1688 }
1689 
1690 static bool acpi_has_nvdimm(void)
1691 {
1692     PCMachineState *pcms = PC_MACHINE(qdev_get_machine());
1693 
1694     return pcms->nvdimm;
1695 }
1696 
1697 static
1698 void acpi_build(PcGuestInfo *guest_info, AcpiBuildTables *tables)
1699 {
1700     GArray *table_offsets;
1701     unsigned facs, ssdt, dsdt, rsdt;
1702     AcpiCpuInfo cpu;
1703     AcpiPmInfo pm;
1704     AcpiMiscInfo misc;
1705     AcpiMcfgInfo mcfg;
1706     PcPciInfo pci;
1707     uint8_t *u;
1708     size_t aml_len = 0;
1709     GArray *tables_blob = tables->table_data;
1710 
1711     acpi_get_cpu_info(&cpu);
1712     acpi_get_pm_info(&pm);
1713     acpi_get_dsdt(&misc);
1714     acpi_get_misc_info(&misc);
1715     acpi_get_pci_info(&pci);
1716 
1717     table_offsets = g_array_new(false, true /* clear */,
1718                                         sizeof(uint32_t));
1719     ACPI_BUILD_DPRINTF("init ACPI tables\n");
1720 
1721     bios_linker_loader_alloc(tables->linker, ACPI_BUILD_TABLE_FILE,
1722                              64 /* Ensure FACS is aligned */,
1723                              false /* high memory */);
1724 
1725     /*
1726      * FACS is pointed to by FADT.
1727      * We place it first since it's the only table that has alignment
1728      * requirements.
1729      */
1730     facs = tables_blob->len;
1731     build_facs(tables_blob, tables->linker, guest_info);
1732 
1733     /* DSDT is pointed to by FADT */
1734     dsdt = tables_blob->len;
1735     build_dsdt(tables_blob, tables->linker, &misc);
1736 
1737     /* Count the size of the DSDT and SSDT, we will need it for legacy
1738      * sizing of ACPI tables.
1739      */
1740     aml_len += tables_blob->len - dsdt;
1741 
1742     /* ACPI tables pointed to by RSDT */
1743     acpi_add_table(table_offsets, tables_blob);
1744     build_fadt(tables_blob, tables->linker, &pm, facs, dsdt);
1745 
1746     ssdt = tables_blob->len;
1747     acpi_add_table(table_offsets, tables_blob);
1748     build_ssdt(tables_blob, tables->linker, &cpu, &pm, &misc, &pci,
1749                guest_info);
1750     aml_len += tables_blob->len - ssdt;
1751 
1752     acpi_add_table(table_offsets, tables_blob);
1753     build_madt(tables_blob, tables->linker, &cpu, guest_info);
1754 
1755     if (misc.has_hpet) {
1756         acpi_add_table(table_offsets, tables_blob);
1757         build_hpet(tables_blob, tables->linker);
1758     }
1759     if (misc.tpm_version != TPM_VERSION_UNSPEC) {
1760         acpi_add_table(table_offsets, tables_blob);
1761         build_tpm_tcpa(tables_blob, tables->linker, tables->tcpalog);
1762 
1763         if (misc.tpm_version == TPM_VERSION_2_0) {
1764             acpi_add_table(table_offsets, tables_blob);
1765             build_tpm2(tables_blob, tables->linker);
1766         }
1767     }
1768     if (guest_info->numa_nodes) {
1769         acpi_add_table(table_offsets, tables_blob);
1770         build_srat(tables_blob, tables->linker, guest_info);
1771     }
1772     if (acpi_get_mcfg(&mcfg)) {
1773         acpi_add_table(table_offsets, tables_blob);
1774         build_mcfg_q35(tables_blob, tables->linker, &mcfg);
1775     }
1776     if (acpi_has_iommu()) {
1777         acpi_add_table(table_offsets, tables_blob);
1778         build_dmar_q35(tables_blob, tables->linker);
1779     }
1780 
1781     if (acpi_has_nvdimm()) {
1782         nvdimm_build_acpi(table_offsets, tables_blob, tables->linker);
1783     }
1784 
1785     /* Add tables supplied by user (if any) */
1786     for (u = acpi_table_first(); u; u = acpi_table_next(u)) {
1787         unsigned len = acpi_table_len(u);
1788 
1789         acpi_add_table(table_offsets, tables_blob);
1790         g_array_append_vals(tables_blob, u, len);
1791     }
1792 
1793     /* RSDT is pointed to by RSDP */
1794     rsdt = tables_blob->len;
1795     build_rsdt(tables_blob, tables->linker, table_offsets);
1796 
1797     /* RSDP is in FSEG memory, so allocate it separately */
1798     build_rsdp(tables->rsdp, tables->linker, rsdt);
1799 
1800     /* We'll expose it all to Guest so we want to reduce
1801      * chance of size changes.
1802      *
1803      * We used to align the tables to 4k, but of course this would
1804      * too simple to be enough.  4k turned out to be too small an
1805      * alignment very soon, and in fact it is almost impossible to
1806      * keep the table size stable for all (max_cpus, max_memory_slots)
1807      * combinations.  So the table size is always 64k for pc-i440fx-2.1
1808      * and we give an error if the table grows beyond that limit.
1809      *
1810      * We still have the problem of migrating from "-M pc-i440fx-2.0".  For
1811      * that, we exploit the fact that QEMU 2.1 generates _smaller_ tables
1812      * than 2.0 and we can always pad the smaller tables with zeros.  We can
1813      * then use the exact size of the 2.0 tables.
1814      *
1815      * All this is for PIIX4, since QEMU 2.0 didn't support Q35 migration.
1816      */
1817     if (guest_info->legacy_acpi_table_size) {
1818         /* Subtracting aml_len gives the size of fixed tables.  Then add the
1819          * size of the PIIX4 DSDT/SSDT in QEMU 2.0.
1820          */
1821         int legacy_aml_len =
1822             guest_info->legacy_acpi_table_size +
1823             ACPI_BUILD_LEGACY_CPU_AML_SIZE * max_cpus;
1824         int legacy_table_size =
1825             ROUND_UP(tables_blob->len - aml_len + legacy_aml_len,
1826                      ACPI_BUILD_ALIGN_SIZE);
1827         if (tables_blob->len > legacy_table_size) {
1828             /* Should happen only with PCI bridges and -M pc-i440fx-2.0.  */
1829             error_report("Warning: migration may not work.");
1830         }
1831         g_array_set_size(tables_blob, legacy_table_size);
1832     } else {
1833         /* Make sure we have a buffer in case we need to resize the tables. */
1834         if (tables_blob->len > ACPI_BUILD_TABLE_SIZE / 2) {
1835             /* As of QEMU 2.1, this fires with 160 VCPUs and 255 memory slots.  */
1836             error_report("Warning: ACPI tables are larger than 64k.");
1837             error_report("Warning: migration may not work.");
1838             error_report("Warning: please remove CPUs, NUMA nodes, "
1839                          "memory slots or PCI bridges.");
1840         }
1841         acpi_align_size(tables_blob, ACPI_BUILD_TABLE_SIZE);
1842     }
1843 
1844     acpi_align_size(tables->linker, ACPI_BUILD_ALIGN_SIZE);
1845 
1846     /* Cleanup memory that's no longer used. */
1847     g_array_free(table_offsets, true);
1848 }
1849 
1850 static void acpi_ram_update(MemoryRegion *mr, GArray *data)
1851 {
1852     uint32_t size = acpi_data_len(data);
1853 
1854     /* Make sure RAM size is correct - in case it got changed e.g. by migration */
1855     memory_region_ram_resize(mr, size, &error_abort);
1856 
1857     memcpy(memory_region_get_ram_ptr(mr), data->data, size);
1858     memory_region_set_dirty(mr, 0, size);
1859 }
1860 
1861 static void acpi_build_update(void *build_opaque)
1862 {
1863     AcpiBuildState *build_state = build_opaque;
1864     AcpiBuildTables tables;
1865 
1866     /* No state to update or already patched? Nothing to do. */
1867     if (!build_state || build_state->patched) {
1868         return;
1869     }
1870     build_state->patched = 1;
1871 
1872     acpi_build_tables_init(&tables);
1873 
1874     acpi_build(build_state->guest_info, &tables);
1875 
1876     acpi_ram_update(build_state->table_mr, tables.table_data);
1877 
1878     if (build_state->rsdp) {
1879         memcpy(build_state->rsdp, tables.rsdp->data, acpi_data_len(tables.rsdp));
1880     } else {
1881         acpi_ram_update(build_state->rsdp_mr, tables.rsdp);
1882     }
1883 
1884     acpi_ram_update(build_state->linker_mr, tables.linker);
1885     acpi_build_tables_cleanup(&tables, true);
1886 }
1887 
1888 static void acpi_build_reset(void *build_opaque)
1889 {
1890     AcpiBuildState *build_state = build_opaque;
1891     build_state->patched = 0;
1892 }
1893 
1894 static MemoryRegion *acpi_add_rom_blob(AcpiBuildState *build_state,
1895                                        GArray *blob, const char *name,
1896                                        uint64_t max_size)
1897 {
1898     return rom_add_blob(name, blob->data, acpi_data_len(blob), max_size, -1,
1899                         name, acpi_build_update, build_state);
1900 }
1901 
1902 static const VMStateDescription vmstate_acpi_build = {
1903     .name = "acpi_build",
1904     .version_id = 1,
1905     .minimum_version_id = 1,
1906     .fields = (VMStateField[]) {
1907         VMSTATE_UINT8(patched, AcpiBuildState),
1908         VMSTATE_END_OF_LIST()
1909     },
1910 };
1911 
1912 void acpi_setup(PcGuestInfo *guest_info)
1913 {
1914     AcpiBuildTables tables;
1915     AcpiBuildState *build_state;
1916 
1917     if (!guest_info->fw_cfg) {
1918         ACPI_BUILD_DPRINTF("No fw cfg. Bailing out.\n");
1919         return;
1920     }
1921 
1922     if (!guest_info->has_acpi_build) {
1923         ACPI_BUILD_DPRINTF("ACPI build disabled. Bailing out.\n");
1924         return;
1925     }
1926 
1927     if (!acpi_enabled) {
1928         ACPI_BUILD_DPRINTF("ACPI disabled. Bailing out.\n");
1929         return;
1930     }
1931 
1932     build_state = g_malloc0(sizeof *build_state);
1933 
1934     build_state->guest_info = guest_info;
1935 
1936     acpi_set_pci_info();
1937 
1938     acpi_build_tables_init(&tables);
1939     acpi_build(build_state->guest_info, &tables);
1940 
1941     /* Now expose it all to Guest */
1942     build_state->table_mr = acpi_add_rom_blob(build_state, tables.table_data,
1943                                                ACPI_BUILD_TABLE_FILE,
1944                                                ACPI_BUILD_TABLE_MAX_SIZE);
1945     assert(build_state->table_mr != NULL);
1946 
1947     build_state->linker_mr =
1948         acpi_add_rom_blob(build_state, tables.linker, "etc/table-loader", 0);
1949 
1950     fw_cfg_add_file(guest_info->fw_cfg, ACPI_BUILD_TPMLOG_FILE,
1951                     tables.tcpalog->data, acpi_data_len(tables.tcpalog));
1952 
1953     if (!guest_info->rsdp_in_ram) {
1954         /*
1955          * Keep for compatibility with old machine types.
1956          * Though RSDP is small, its contents isn't immutable, so
1957          * we'll update it along with the rest of tables on guest access.
1958          */
1959         uint32_t rsdp_size = acpi_data_len(tables.rsdp);
1960 
1961         build_state->rsdp = g_memdup(tables.rsdp->data, rsdp_size);
1962         fw_cfg_add_file_callback(guest_info->fw_cfg, ACPI_BUILD_RSDP_FILE,
1963                                  acpi_build_update, build_state,
1964                                  build_state->rsdp, rsdp_size);
1965         build_state->rsdp_mr = NULL;
1966     } else {
1967         build_state->rsdp = NULL;
1968         build_state->rsdp_mr = acpi_add_rom_blob(build_state, tables.rsdp,
1969                                                   ACPI_BUILD_RSDP_FILE, 0);
1970     }
1971 
1972     qemu_register_reset(acpi_build_reset, build_state);
1973     acpi_build_reset(build_state);
1974     vmstate_register(NULL, 0, &vmstate_acpi_build, build_state);
1975 
1976     /* Cleanup tables but don't free the memory: we track it
1977      * in build_state.
1978      */
1979     acpi_build_tables_cleanup(&tables, false);
1980 }
1981