xref: /openbmc/qemu/hw/i386/acpi-build.c (revision f1f7e4bf)
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 "sysemu/tpm.h"
43 #include "hw/acpi/tpm.h"
44 #include "sysemu/tpm_backend.h"
45 
46 /* Supported chipsets: */
47 #include "hw/acpi/piix4.h"
48 #include "hw/acpi/pcihp.h"
49 #include "hw/i386/ich9.h"
50 #include "hw/pci/pci_bus.h"
51 #include "hw/pci-host/q35.h"
52 #include "hw/i386/intel_iommu.h"
53 
54 #include "hw/i386/q35-acpi-dsdt.hex"
55 #include "hw/i386/acpi-dsdt.hex"
56 
57 #include "hw/acpi/aml-build.h"
58 
59 #include "qapi/qmp/qint.h"
60 #include "qom/qom-qobject.h"
61 
62 /* These are used to size the ACPI tables for -M pc-i440fx-1.7 and
63  * -M pc-i440fx-2.0.  Even if the actual amount of AML generated grows
64  * a little bit, there should be plenty of free space since the DSDT
65  * shrunk by ~1.5k between QEMU 2.0 and QEMU 2.1.
66  */
67 #define ACPI_BUILD_LEGACY_CPU_AML_SIZE    97
68 #define ACPI_BUILD_ALIGN_SIZE             0x1000
69 
70 #define ACPI_BUILD_TABLE_SIZE             0x20000
71 
72 /* #define DEBUG_ACPI_BUILD */
73 #ifdef DEBUG_ACPI_BUILD
74 #define ACPI_BUILD_DPRINTF(fmt, ...)        \
75     do {printf("ACPI_BUILD: " fmt, ## __VA_ARGS__); } while (0)
76 #else
77 #define ACPI_BUILD_DPRINTF(fmt, ...)
78 #endif
79 
80 typedef struct AcpiCpuInfo {
81     DECLARE_BITMAP(found_cpus, ACPI_CPU_HOTPLUG_ID_LIMIT);
82 } AcpiCpuInfo;
83 
84 typedef struct AcpiMcfgInfo {
85     uint64_t mcfg_base;
86     uint32_t mcfg_size;
87 } AcpiMcfgInfo;
88 
89 typedef struct AcpiPmInfo {
90     bool s3_disabled;
91     bool s4_disabled;
92     bool pcihp_bridge_en;
93     uint8_t s4_val;
94     uint16_t sci_int;
95     uint8_t acpi_enable_cmd;
96     uint8_t acpi_disable_cmd;
97     uint32_t gpe0_blk;
98     uint32_t gpe0_blk_len;
99     uint32_t io_base;
100     uint16_t cpu_hp_io_base;
101     uint16_t cpu_hp_io_len;
102     uint16_t mem_hp_io_base;
103     uint16_t mem_hp_io_len;
104     uint16_t pcihp_io_base;
105     uint16_t pcihp_io_len;
106 } AcpiPmInfo;
107 
108 typedef struct AcpiMiscInfo {
109     bool has_hpet;
110     TPMVersion tpm_version;
111     const unsigned char *dsdt_code;
112     unsigned dsdt_size;
113     uint16_t pvpanic_port;
114     uint16_t applesmc_io_base;
115 } AcpiMiscInfo;
116 
117 typedef struct AcpiBuildPciBusHotplugState {
118     GArray *device_table;
119     GArray *notify_table;
120     struct AcpiBuildPciBusHotplugState *parent;
121     bool pcihp_bridge_en;
122 } AcpiBuildPciBusHotplugState;
123 
124 static void acpi_get_dsdt(AcpiMiscInfo *info)
125 {
126     Object *piix = piix4_pm_find();
127     Object *lpc = ich9_lpc_find();
128     assert(!!piix != !!lpc);
129 
130     if (piix) {
131         info->dsdt_code = AcpiDsdtAmlCode;
132         info->dsdt_size = sizeof AcpiDsdtAmlCode;
133     }
134     if (lpc) {
135         info->dsdt_code = Q35AcpiDsdtAmlCode;
136         info->dsdt_size = sizeof Q35AcpiDsdtAmlCode;
137     }
138 }
139 
140 static
141 int acpi_add_cpu_info(Object *o, void *opaque)
142 {
143     AcpiCpuInfo *cpu = opaque;
144     uint64_t apic_id;
145 
146     if (object_dynamic_cast(o, TYPE_CPU)) {
147         apic_id = object_property_get_int(o, "apic-id", NULL);
148         assert(apic_id < ACPI_CPU_HOTPLUG_ID_LIMIT);
149 
150         set_bit(apic_id, cpu->found_cpus);
151     }
152 
153     object_child_foreach(o, acpi_add_cpu_info, opaque);
154     return 0;
155 }
156 
157 static void acpi_get_cpu_info(AcpiCpuInfo *cpu)
158 {
159     Object *root = object_get_root();
160 
161     memset(cpu->found_cpus, 0, sizeof cpu->found_cpus);
162     object_child_foreach(root, acpi_add_cpu_info, cpu);
163 }
164 
165 static void acpi_get_pm_info(AcpiPmInfo *pm)
166 {
167     Object *piix = piix4_pm_find();
168     Object *lpc = ich9_lpc_find();
169     Object *obj = NULL;
170     QObject *o;
171 
172     pm->cpu_hp_io_base = 0;
173     pm->pcihp_io_base = 0;
174     pm->pcihp_io_len = 0;
175     if (piix) {
176         obj = piix;
177         pm->cpu_hp_io_base = PIIX4_CPU_HOTPLUG_IO_BASE;
178         pm->pcihp_io_base =
179             object_property_get_int(obj, ACPI_PCIHP_IO_BASE_PROP, NULL);
180         pm->pcihp_io_len =
181             object_property_get_int(obj, ACPI_PCIHP_IO_LEN_PROP, NULL);
182     }
183     if (lpc) {
184         obj = lpc;
185         pm->cpu_hp_io_base = ICH9_CPU_HOTPLUG_IO_BASE;
186     }
187     assert(obj);
188 
189     pm->cpu_hp_io_len = ACPI_GPE_PROC_LEN;
190     pm->mem_hp_io_base = ACPI_MEMORY_HOTPLUG_BASE;
191     pm->mem_hp_io_len = ACPI_MEMORY_HOTPLUG_IO_LEN;
192 
193     /* Fill in optional s3/s4 related properties */
194     o = object_property_get_qobject(obj, ACPI_PM_PROP_S3_DISABLED, NULL);
195     if (o) {
196         pm->s3_disabled = qint_get_int(qobject_to_qint(o));
197     } else {
198         pm->s3_disabled = false;
199     }
200     qobject_decref(o);
201     o = object_property_get_qobject(obj, ACPI_PM_PROP_S4_DISABLED, NULL);
202     if (o) {
203         pm->s4_disabled = qint_get_int(qobject_to_qint(o));
204     } else {
205         pm->s4_disabled = false;
206     }
207     qobject_decref(o);
208     o = object_property_get_qobject(obj, ACPI_PM_PROP_S4_VAL, NULL);
209     if (o) {
210         pm->s4_val = qint_get_int(qobject_to_qint(o));
211     } else {
212         pm->s4_val = false;
213     }
214     qobject_decref(o);
215 
216     /* Fill in mandatory properties */
217     pm->sci_int = object_property_get_int(obj, ACPI_PM_PROP_SCI_INT, NULL);
218 
219     pm->acpi_enable_cmd = object_property_get_int(obj,
220                                                   ACPI_PM_PROP_ACPI_ENABLE_CMD,
221                                                   NULL);
222     pm->acpi_disable_cmd = object_property_get_int(obj,
223                                                   ACPI_PM_PROP_ACPI_DISABLE_CMD,
224                                                   NULL);
225     pm->io_base = object_property_get_int(obj, ACPI_PM_PROP_PM_IO_BASE,
226                                           NULL);
227     pm->gpe0_blk = object_property_get_int(obj, ACPI_PM_PROP_GPE0_BLK,
228                                            NULL);
229     pm->gpe0_blk_len = object_property_get_int(obj, ACPI_PM_PROP_GPE0_BLK_LEN,
230                                                NULL);
231     pm->pcihp_bridge_en =
232         object_property_get_bool(obj, "acpi-pci-hotplug-with-bridge-support",
233                                  NULL);
234 }
235 
236 static void acpi_get_misc_info(AcpiMiscInfo *info)
237 {
238     info->has_hpet = hpet_find();
239     info->tpm_version = tpm_get_version();
240     info->pvpanic_port = pvpanic_port();
241     info->applesmc_io_base = applesmc_port();
242 }
243 
244 /*
245  * Because of the PXB hosts we cannot simply query TYPE_PCI_HOST_BRIDGE.
246  * On i386 arch we only have two pci hosts, so we can look only for them.
247  */
248 static Object *acpi_get_i386_pci_host(void)
249 {
250     PCIHostState *host;
251 
252     host = OBJECT_CHECK(PCIHostState,
253                         object_resolve_path("/machine/i440fx", NULL),
254                         TYPE_PCI_HOST_BRIDGE);
255     if (!host) {
256         host = OBJECT_CHECK(PCIHostState,
257                             object_resolve_path("/machine/q35", NULL),
258                             TYPE_PCI_HOST_BRIDGE);
259     }
260 
261     return OBJECT(host);
262 }
263 
264 static void acpi_get_pci_info(PcPciInfo *info)
265 {
266     Object *pci_host;
267 
268 
269     pci_host = acpi_get_i386_pci_host();
270     g_assert(pci_host);
271 
272     info->w32.begin = object_property_get_int(pci_host,
273                                               PCI_HOST_PROP_PCI_HOLE_START,
274                                               NULL);
275     info->w32.end = object_property_get_int(pci_host,
276                                             PCI_HOST_PROP_PCI_HOLE_END,
277                                             NULL);
278     info->w64.begin = object_property_get_int(pci_host,
279                                               PCI_HOST_PROP_PCI_HOLE64_START,
280                                               NULL);
281     info->w64.end = object_property_get_int(pci_host,
282                                             PCI_HOST_PROP_PCI_HOLE64_END,
283                                             NULL);
284 }
285 
286 #define ACPI_PORT_SMI_CMD           0x00b2 /* TODO: this is APM_CNT_IOPORT */
287 
288 static void acpi_align_size(GArray *blob, unsigned align)
289 {
290     /* Align size to multiple of given size. This reduces the chance
291      * we need to change size in the future (breaking cross version migration).
292      */
293     g_array_set_size(blob, ROUND_UP(acpi_data_len(blob), align));
294 }
295 
296 /* FACS */
297 static void
298 build_facs(GArray *table_data, GArray *linker, PcGuestInfo *guest_info)
299 {
300     AcpiFacsDescriptorRev1 *facs = acpi_data_push(table_data, sizeof *facs);
301     memcpy(&facs->signature, "FACS", 4);
302     facs->length = cpu_to_le32(sizeof(*facs));
303 }
304 
305 /* Load chipset information in FADT */
306 static void fadt_setup(AcpiFadtDescriptorRev1 *fadt, AcpiPmInfo *pm)
307 {
308     fadt->model = 1;
309     fadt->reserved1 = 0;
310     fadt->sci_int = cpu_to_le16(pm->sci_int);
311     fadt->smi_cmd = cpu_to_le32(ACPI_PORT_SMI_CMD);
312     fadt->acpi_enable = pm->acpi_enable_cmd;
313     fadt->acpi_disable = pm->acpi_disable_cmd;
314     /* EVT, CNT, TMR offset matches hw/acpi/core.c */
315     fadt->pm1a_evt_blk = cpu_to_le32(pm->io_base);
316     fadt->pm1a_cnt_blk = cpu_to_le32(pm->io_base + 0x04);
317     fadt->pm_tmr_blk = cpu_to_le32(pm->io_base + 0x08);
318     fadt->gpe0_blk = cpu_to_le32(pm->gpe0_blk);
319     /* EVT, CNT, TMR length matches hw/acpi/core.c */
320     fadt->pm1_evt_len = 4;
321     fadt->pm1_cnt_len = 2;
322     fadt->pm_tmr_len = 4;
323     fadt->gpe0_blk_len = pm->gpe0_blk_len;
324     fadt->plvl2_lat = cpu_to_le16(0xfff); /* C2 state not supported */
325     fadt->plvl3_lat = cpu_to_le16(0xfff); /* C3 state not supported */
326     fadt->flags = cpu_to_le32((1 << ACPI_FADT_F_WBINVD) |
327                               (1 << ACPI_FADT_F_PROC_C1) |
328                               (1 << ACPI_FADT_F_SLP_BUTTON) |
329                               (1 << ACPI_FADT_F_RTC_S4));
330     fadt->flags |= cpu_to_le32(1 << ACPI_FADT_F_USE_PLATFORM_CLOCK);
331     /* APIC destination mode ("Flat Logical") has an upper limit of 8 CPUs
332      * For more than 8 CPUs, "Clustered Logical" mode has to be used
333      */
334     if (max_cpus > 8) {
335         fadt->flags |= cpu_to_le32(1 << ACPI_FADT_F_FORCE_APIC_CLUSTER_MODEL);
336     }
337 }
338 
339 
340 /* FADT */
341 static void
342 build_fadt(GArray *table_data, GArray *linker, AcpiPmInfo *pm,
343            unsigned facs, unsigned dsdt)
344 {
345     AcpiFadtDescriptorRev1 *fadt = acpi_data_push(table_data, sizeof(*fadt));
346 
347     fadt->firmware_ctrl = cpu_to_le32(facs);
348     /* FACS address to be filled by Guest linker */
349     bios_linker_loader_add_pointer(linker, ACPI_BUILD_TABLE_FILE,
350                                    ACPI_BUILD_TABLE_FILE,
351                                    table_data, &fadt->firmware_ctrl,
352                                    sizeof fadt->firmware_ctrl);
353 
354     fadt->dsdt = cpu_to_le32(dsdt);
355     /* DSDT address to be filled by Guest linker */
356     bios_linker_loader_add_pointer(linker, ACPI_BUILD_TABLE_FILE,
357                                    ACPI_BUILD_TABLE_FILE,
358                                    table_data, &fadt->dsdt,
359                                    sizeof fadt->dsdt);
360 
361     fadt_setup(fadt, pm);
362 
363     build_header(linker, table_data,
364                  (void *)fadt, "FACP", sizeof(*fadt), 1);
365 }
366 
367 static void
368 build_madt(GArray *table_data, GArray *linker, AcpiCpuInfo *cpu,
369            PcGuestInfo *guest_info)
370 {
371     int madt_start = table_data->len;
372 
373     AcpiMultipleApicTable *madt;
374     AcpiMadtIoApic *io_apic;
375     AcpiMadtIntsrcovr *intsrcovr;
376     AcpiMadtLocalNmi *local_nmi;
377     int i;
378 
379     madt = acpi_data_push(table_data, sizeof *madt);
380     madt->local_apic_address = cpu_to_le32(APIC_DEFAULT_ADDRESS);
381     madt->flags = cpu_to_le32(1);
382 
383     for (i = 0; i < guest_info->apic_id_limit; i++) {
384         AcpiMadtProcessorApic *apic = acpi_data_push(table_data, sizeof *apic);
385         apic->type = ACPI_APIC_PROCESSOR;
386         apic->length = sizeof(*apic);
387         apic->processor_id = i;
388         apic->local_apic_id = i;
389         if (test_bit(i, cpu->found_cpus)) {
390             apic->flags = cpu_to_le32(1);
391         } else {
392             apic->flags = cpu_to_le32(0);
393         }
394     }
395     io_apic = acpi_data_push(table_data, sizeof *io_apic);
396     io_apic->type = ACPI_APIC_IO;
397     io_apic->length = sizeof(*io_apic);
398 #define ACPI_BUILD_IOAPIC_ID 0x0
399     io_apic->io_apic_id = ACPI_BUILD_IOAPIC_ID;
400     io_apic->address = cpu_to_le32(IO_APIC_DEFAULT_ADDRESS);
401     io_apic->interrupt = cpu_to_le32(0);
402 
403     if (guest_info->apic_xrupt_override) {
404         intsrcovr = acpi_data_push(table_data, sizeof *intsrcovr);
405         intsrcovr->type   = ACPI_APIC_XRUPT_OVERRIDE;
406         intsrcovr->length = sizeof(*intsrcovr);
407         intsrcovr->source = 0;
408         intsrcovr->gsi    = cpu_to_le32(2);
409         intsrcovr->flags  = cpu_to_le16(0); /* conforms to bus specifications */
410     }
411     for (i = 1; i < 16; i++) {
412 #define ACPI_BUILD_PCI_IRQS ((1<<5) | (1<<9) | (1<<10) | (1<<11))
413         if (!(ACPI_BUILD_PCI_IRQS & (1 << i))) {
414             /* No need for a INT source override structure. */
415             continue;
416         }
417         intsrcovr = acpi_data_push(table_data, sizeof *intsrcovr);
418         intsrcovr->type   = ACPI_APIC_XRUPT_OVERRIDE;
419         intsrcovr->length = sizeof(*intsrcovr);
420         intsrcovr->source = i;
421         intsrcovr->gsi    = cpu_to_le32(i);
422         intsrcovr->flags  = cpu_to_le16(0xd); /* active high, level triggered */
423     }
424 
425     local_nmi = acpi_data_push(table_data, sizeof *local_nmi);
426     local_nmi->type         = ACPI_APIC_LOCAL_NMI;
427     local_nmi->length       = sizeof(*local_nmi);
428     local_nmi->processor_id = 0xff; /* all processors */
429     local_nmi->flags        = cpu_to_le16(0);
430     local_nmi->lint         = 1; /* ACPI_LINT1 */
431 
432     build_header(linker, table_data,
433                  (void *)(table_data->data + madt_start), "APIC",
434                  table_data->len - madt_start, 1);
435 }
436 
437 /* Assign BSEL property to all buses.  In the future, this can be changed
438  * to only assign to buses that support hotplug.
439  */
440 static void *acpi_set_bsel(PCIBus *bus, void *opaque)
441 {
442     unsigned *bsel_alloc = opaque;
443     unsigned *bus_bsel;
444 
445     if (qbus_is_hotpluggable(BUS(bus))) {
446         bus_bsel = g_malloc(sizeof *bus_bsel);
447 
448         *bus_bsel = (*bsel_alloc)++;
449         object_property_add_uint32_ptr(OBJECT(bus), ACPI_PCIHP_PROP_BSEL,
450                                        bus_bsel, NULL);
451     }
452 
453     return bsel_alloc;
454 }
455 
456 static void acpi_set_pci_info(void)
457 {
458     PCIBus *bus = find_i440fx(); /* TODO: Q35 support */
459     unsigned bsel_alloc = 0;
460 
461     if (bus) {
462         /* Scan all PCI buses. Set property to enable acpi based hotplug. */
463         pci_for_each_bus_depth_first(bus, acpi_set_bsel, NULL, &bsel_alloc);
464     }
465 }
466 
467 static void build_append_pcihp_notify_entry(Aml *method, int slot)
468 {
469     Aml *if_ctx;
470     int32_t devfn = PCI_DEVFN(slot, 0);
471 
472     if_ctx = aml_if(aml_and(aml_arg(0), aml_int(0x1U << slot)));
473     aml_append(if_ctx, aml_notify(aml_name("S%.02X", devfn), aml_arg(1)));
474     aml_append(method, if_ctx);
475 }
476 
477 static void build_append_pci_bus_devices(Aml *parent_scope, PCIBus *bus,
478                                          bool pcihp_bridge_en)
479 {
480     Aml *dev, *notify_method, *method;
481     QObject *bsel;
482     PCIBus *sec;
483     int i;
484 
485     bsel = object_property_get_qobject(OBJECT(bus), ACPI_PCIHP_PROP_BSEL, NULL);
486     if (bsel) {
487         int64_t bsel_val = qint_get_int(qobject_to_qint(bsel));
488 
489         aml_append(parent_scope, aml_name_decl("BSEL", aml_int(bsel_val)));
490         notify_method = aml_method("DVNT", 2);
491     }
492 
493     for (i = 0; i < ARRAY_SIZE(bus->devices); i += PCI_FUNC_MAX) {
494         DeviceClass *dc;
495         PCIDeviceClass *pc;
496         PCIDevice *pdev = bus->devices[i];
497         int slot = PCI_SLOT(i);
498         bool hotplug_enabled_dev;
499         bool bridge_in_acpi;
500 
501         if (!pdev) {
502             if (bsel) { /* add hotplug slots for non present devices */
503                 dev = aml_device("S%.02X", PCI_DEVFN(slot, 0));
504                 aml_append(dev, aml_name_decl("_SUN", aml_int(slot)));
505                 aml_append(dev, aml_name_decl("_ADR", aml_int(slot << 16)));
506                 method = aml_method("_EJ0", 1);
507                 aml_append(method,
508                     aml_call2("PCEJ", aml_name("BSEL"), aml_name("_SUN"))
509                 );
510                 aml_append(dev, method);
511                 aml_append(parent_scope, dev);
512 
513                 build_append_pcihp_notify_entry(notify_method, slot);
514             }
515             continue;
516         }
517 
518         pc = PCI_DEVICE_GET_CLASS(pdev);
519         dc = DEVICE_GET_CLASS(pdev);
520 
521         /* When hotplug for bridges is enabled, bridges are
522          * described in ACPI separately (see build_pci_bus_end).
523          * In this case they aren't themselves hot-pluggable.
524          * Hotplugged bridges *are* hot-pluggable.
525          */
526         bridge_in_acpi = pc->is_bridge && pcihp_bridge_en &&
527             !DEVICE(pdev)->hotplugged;
528 
529         hotplug_enabled_dev = bsel && dc->hotpluggable && !bridge_in_acpi;
530 
531         if (pc->class_id == PCI_CLASS_BRIDGE_ISA) {
532             continue;
533         }
534 
535         /* start to compose PCI slot descriptor */
536         dev = aml_device("S%.02X", PCI_DEVFN(slot, 0));
537         aml_append(dev, aml_name_decl("_ADR", aml_int(slot << 16)));
538 
539         if (pc->class_id == PCI_CLASS_DISPLAY_VGA) {
540             /* add VGA specific AML methods */
541             int s3d;
542 
543             if (object_dynamic_cast(OBJECT(pdev), "qxl-vga")) {
544                 s3d = 3;
545             } else {
546                 s3d = 0;
547             }
548 
549             method = aml_method("_S1D", 0);
550             aml_append(method, aml_return(aml_int(0)));
551             aml_append(dev, method);
552 
553             method = aml_method("_S2D", 0);
554             aml_append(method, aml_return(aml_int(0)));
555             aml_append(dev, method);
556 
557             method = aml_method("_S3D", 0);
558             aml_append(method, aml_return(aml_int(s3d)));
559             aml_append(dev, method);
560         } else if (hotplug_enabled_dev) {
561             /* add _SUN/_EJ0 to make slot hotpluggable  */
562             aml_append(dev, aml_name_decl("_SUN", aml_int(slot)));
563 
564             method = aml_method("_EJ0", 1);
565             aml_append(method,
566                 aml_call2("PCEJ", aml_name("BSEL"), aml_name("_SUN"))
567             );
568             aml_append(dev, method);
569 
570             if (bsel) {
571                 build_append_pcihp_notify_entry(notify_method, slot);
572             }
573         } else if (bridge_in_acpi) {
574             /*
575              * device is coldplugged bridge,
576              * add child device descriptions into its scope
577              */
578             PCIBus *sec_bus = pci_bridge_get_sec_bus(PCI_BRIDGE(pdev));
579 
580             build_append_pci_bus_devices(dev, sec_bus, pcihp_bridge_en);
581         }
582         /* slot descriptor has been composed, add it into parent context */
583         aml_append(parent_scope, dev);
584     }
585 
586     if (bsel) {
587         aml_append(parent_scope, notify_method);
588     }
589 
590     /* Append PCNT method to notify about events on local and child buses.
591      * Add unconditionally for root since DSDT expects it.
592      */
593     method = aml_method("PCNT", 0);
594 
595     /* If bus supports hotplug select it and notify about local events */
596     if (bsel) {
597         int64_t bsel_val = qint_get_int(qobject_to_qint(bsel));
598         aml_append(method, aml_store(aml_int(bsel_val), aml_name("BNUM")));
599         aml_append(method,
600             aml_call2("DVNT", aml_name("PCIU"), aml_int(1) /* Device Check */)
601         );
602         aml_append(method,
603             aml_call2("DVNT", aml_name("PCID"), aml_int(3)/* Eject Request */)
604         );
605     }
606 
607     /* Notify about child bus events in any case */
608     if (pcihp_bridge_en) {
609         QLIST_FOREACH(sec, &bus->child, sibling) {
610             int32_t devfn = sec->parent_dev->devfn;
611 
612             aml_append(method, aml_name("^S%.02X.PCNT", devfn));
613         }
614     }
615     aml_append(parent_scope, method);
616     qobject_decref(bsel);
617 }
618 
619 /*
620  * initialize_route - Initialize the interrupt routing rule
621  * through a specific LINK:
622  *  if (lnk_idx == idx)
623  *      route using link 'link_name'
624  */
625 static Aml *initialize_route(Aml *route, const char *link_name,
626                              Aml *lnk_idx, int idx)
627 {
628     Aml *if_ctx = aml_if(aml_equal(lnk_idx, aml_int(idx)));
629     Aml *pkg = aml_package(4);
630 
631     aml_append(pkg, aml_int(0));
632     aml_append(pkg, aml_int(0));
633     aml_append(pkg, aml_name("%s", link_name));
634     aml_append(pkg, aml_int(0));
635     aml_append(if_ctx, aml_store(pkg, route));
636 
637     return if_ctx;
638 }
639 
640 /*
641  * build_prt - Define interrupt rounting rules
642  *
643  * Returns an array of 128 routes, one for each device,
644  * based on device location.
645  * The main goal is to equaly distribute the interrupts
646  * over the 4 existing ACPI links (works only for i440fx).
647  * The hash function is  (slot + pin) & 3 -> "LNK[D|A|B|C]".
648  *
649  */
650 static Aml *build_prt(void)
651 {
652     Aml *method, *while_ctx, *pin, *res;
653 
654     method = aml_method("_PRT", 0);
655     res = aml_local(0);
656     pin = aml_local(1);
657     aml_append(method, aml_store(aml_package(128), res));
658     aml_append(method, aml_store(aml_int(0), pin));
659 
660     /* while (pin < 128) */
661     while_ctx = aml_while(aml_lless(pin, aml_int(128)));
662     {
663         Aml *slot = aml_local(2);
664         Aml *lnk_idx = aml_local(3);
665         Aml *route = aml_local(4);
666 
667         /* slot = pin >> 2 */
668         aml_append(while_ctx,
669                    aml_store(aml_shiftright(pin, aml_int(2)), slot));
670         /* lnk_idx = (slot + pin) & 3 */
671         aml_append(while_ctx,
672                    aml_store(aml_and(aml_add(pin, slot), aml_int(3)), lnk_idx));
673 
674         /* route[2] = "LNK[D|A|B|C]", selection based on pin % 3  */
675         aml_append(while_ctx, initialize_route(route, "LNKD", lnk_idx, 0));
676         aml_append(while_ctx, initialize_route(route, "LNKA", lnk_idx, 1));
677         aml_append(while_ctx, initialize_route(route, "LNKB", lnk_idx, 2));
678         aml_append(while_ctx, initialize_route(route, "LNKC", lnk_idx, 3));
679 
680         /* route[0] = 0x[slot]FFFF */
681         aml_append(while_ctx,
682             aml_store(aml_or(aml_shiftleft(slot, aml_int(16)), aml_int(0xFFFF)),
683                       aml_index(route, aml_int(0))));
684         /* route[1] = pin & 3 */
685         aml_append(while_ctx,
686             aml_store(aml_and(pin, aml_int(3)), aml_index(route, aml_int(1))));
687         /* res[pin] = route */
688         aml_append(while_ctx, aml_store(route, aml_index(res, pin)));
689         /* pin++ */
690         aml_append(while_ctx, aml_increment(pin));
691     }
692     aml_append(method, while_ctx);
693     /* return res*/
694     aml_append(method, aml_return(res));
695 
696     return method;
697 }
698 
699 typedef struct CrsRangeEntry {
700     uint64_t base;
701     uint64_t limit;
702 } CrsRangeEntry;
703 
704 static void crs_range_insert(GPtrArray *ranges, uint64_t base, uint64_t limit)
705 {
706     CrsRangeEntry *entry;
707 
708     entry = g_malloc(sizeof(*entry));
709     entry->base = base;
710     entry->limit = limit;
711 
712     g_ptr_array_add(ranges, entry);
713 }
714 
715 static void crs_range_free(gpointer data)
716 {
717     CrsRangeEntry *entry = (CrsRangeEntry *)data;
718     g_free(entry);
719 }
720 
721 static gint crs_range_compare(gconstpointer a, gconstpointer b)
722 {
723      CrsRangeEntry *entry_a = *(CrsRangeEntry **)a;
724      CrsRangeEntry *entry_b = *(CrsRangeEntry **)b;
725 
726      return (int64_t)entry_a->base - (int64_t)entry_b->base;
727 }
728 
729 /*
730  * crs_replace_with_free_ranges - given the 'used' ranges within [start - end]
731  * interval, computes the 'free' ranges from the same interval.
732  * Example: If the input array is { [a1 - a2],[b1 - b2] }, the function
733  * will return { [base - a1], [a2 - b1], [b2 - limit] }.
734  */
735 static void crs_replace_with_free_ranges(GPtrArray *ranges,
736                                          uint64_t start, uint64_t end)
737 {
738     GPtrArray *free_ranges = g_ptr_array_new_with_free_func(crs_range_free);
739     uint64_t free_base = start;
740     int i;
741 
742     g_ptr_array_sort(ranges, crs_range_compare);
743     for (i = 0; i < ranges->len; i++) {
744         CrsRangeEntry *used = g_ptr_array_index(ranges, i);
745 
746         if (free_base < used->base) {
747             crs_range_insert(free_ranges, free_base, used->base - 1);
748         }
749 
750         free_base = used->limit + 1;
751     }
752 
753     if (free_base < end) {
754         crs_range_insert(free_ranges, free_base, end);
755     }
756 
757     g_ptr_array_set_size(ranges, 0);
758     for (i = 0; i < free_ranges->len; i++) {
759         g_ptr_array_add(ranges, g_ptr_array_index(free_ranges, i));
760     }
761 
762     g_ptr_array_free(free_ranges, false);
763 }
764 
765 static Aml *build_crs(PCIHostState *host,
766                       GPtrArray *io_ranges, GPtrArray *mem_ranges)
767 {
768     Aml *crs = aml_resource_template();
769     uint8_t max_bus = pci_bus_num(host->bus);
770     uint8_t type;
771     int devfn;
772 
773     for (devfn = 0; devfn < ARRAY_SIZE(host->bus->devices); devfn++) {
774         int i;
775         uint64_t range_base, range_limit;
776         PCIDevice *dev = host->bus->devices[devfn];
777 
778         if (!dev) {
779             continue;
780         }
781 
782         for (i = 0; i < PCI_NUM_REGIONS; i++) {
783             PCIIORegion *r = &dev->io_regions[i];
784 
785             range_base = r->addr;
786             range_limit = r->addr + r->size - 1;
787 
788             /*
789              * Work-around for old bioses
790              * that do not support multiple root buses
791              */
792             if (!range_base || range_base > range_limit) {
793                 continue;
794             }
795 
796             if (r->type & PCI_BASE_ADDRESS_SPACE_IO) {
797                 aml_append(crs,
798                     aml_word_io(AML_MIN_FIXED, AML_MAX_FIXED,
799                                 AML_POS_DECODE, AML_ENTIRE_RANGE,
800                                 0,
801                                 range_base,
802                                 range_limit,
803                                 0,
804                                 range_limit - range_base + 1));
805                 crs_range_insert(io_ranges, range_base, range_limit);
806             } else { /* "memory" */
807                 aml_append(crs,
808                     aml_dword_memory(AML_POS_DECODE, AML_MIN_FIXED,
809                                      AML_MAX_FIXED, AML_NON_CACHEABLE,
810                                      AML_READ_WRITE,
811                                      0,
812                                      range_base,
813                                      range_limit,
814                                      0,
815                                      range_limit - range_base + 1));
816                 crs_range_insert(mem_ranges, range_base, range_limit);
817             }
818         }
819 
820         type = dev->config[PCI_HEADER_TYPE] & ~PCI_HEADER_TYPE_MULTI_FUNCTION;
821         if (type == PCI_HEADER_TYPE_BRIDGE) {
822             uint8_t subordinate = dev->config[PCI_SUBORDINATE_BUS];
823             if (subordinate > max_bus) {
824                 max_bus = subordinate;
825             }
826 
827             range_base = pci_bridge_get_base(dev, PCI_BASE_ADDRESS_SPACE_IO);
828             range_limit = pci_bridge_get_limit(dev, PCI_BASE_ADDRESS_SPACE_IO);
829 
830             /*
831              * Work-around for old bioses
832              * that do not support multiple root buses
833              */
834             if (range_base && range_base <= range_limit) {
835                 aml_append(crs,
836                            aml_word_io(AML_MIN_FIXED, AML_MAX_FIXED,
837                                        AML_POS_DECODE, AML_ENTIRE_RANGE,
838                                        0,
839                                        range_base,
840                                        range_limit,
841                                        0,
842                                        range_limit - range_base + 1));
843                 crs_range_insert(io_ranges, range_base, range_limit);
844             }
845 
846             range_base =
847                 pci_bridge_get_base(dev, PCI_BASE_ADDRESS_SPACE_MEMORY);
848             range_limit =
849                 pci_bridge_get_limit(dev, PCI_BASE_ADDRESS_SPACE_MEMORY);
850 
851             /*
852              * Work-around for old bioses
853              * that do not support multiple root buses
854              */
855             if (range_base && range_base <= range_limit) {
856                 aml_append(crs,
857                            aml_dword_memory(AML_POS_DECODE, AML_MIN_FIXED,
858                                             AML_MAX_FIXED, AML_NON_CACHEABLE,
859                                             AML_READ_WRITE,
860                                             0,
861                                             range_base,
862                                             range_limit,
863                                             0,
864                                             range_limit - range_base + 1));
865                 crs_range_insert(mem_ranges, range_base, range_limit);
866             }
867 
868             range_base =
869                 pci_bridge_get_base(dev, PCI_BASE_ADDRESS_MEM_PREFETCH);
870             range_limit =
871                 pci_bridge_get_limit(dev, PCI_BASE_ADDRESS_MEM_PREFETCH);
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                 aml_append(crs,
879                            aml_dword_memory(AML_POS_DECODE, AML_MIN_FIXED,
880                                             AML_MAX_FIXED, AML_NON_CACHEABLE,
881                                             AML_READ_WRITE,
882                                             0,
883                                             range_base,
884                                             range_limit,
885                                             0,
886                                             range_limit - range_base + 1));
887                 crs_range_insert(mem_ranges, range_base, range_limit);
888             }
889         }
890     }
891 
892     aml_append(crs,
893         aml_word_bus_number(AML_MIN_FIXED, AML_MAX_FIXED, AML_POS_DECODE,
894                             0,
895                             pci_bus_num(host->bus),
896                             max_bus,
897                             0,
898                             max_bus - pci_bus_num(host->bus) + 1));
899 
900     return crs;
901 }
902 
903 static void
904 build_ssdt(GArray *table_data, GArray *linker,
905            AcpiCpuInfo *cpu, AcpiPmInfo *pm, AcpiMiscInfo *misc,
906            PcPciInfo *pci, PcGuestInfo *guest_info)
907 {
908     MachineState *machine = MACHINE(qdev_get_machine());
909     uint32_t nr_mem = machine->ram_slots;
910     unsigned acpi_cpus = guest_info->apic_id_limit;
911     Aml *ssdt, *sb_scope, *scope, *pkg, *dev, *method, *crs, *field, *ifctx;
912     PCIBus *bus = NULL;
913     GPtrArray *io_ranges = g_ptr_array_new_with_free_func(crs_range_free);
914     GPtrArray *mem_ranges = g_ptr_array_new_with_free_func(crs_range_free);
915     CrsRangeEntry *entry;
916     int root_bus_limit = 0xFF;
917     int i;
918 
919     ssdt = init_aml_allocator();
920     /* The current AML generator can cover the APIC ID range [0..255],
921      * inclusive, for VCPU hotplug. */
922     QEMU_BUILD_BUG_ON(ACPI_CPU_HOTPLUG_ID_LIMIT > 256);
923     g_assert(acpi_cpus <= ACPI_CPU_HOTPLUG_ID_LIMIT);
924 
925     /* Reserve space for header */
926     acpi_data_push(ssdt->buf, sizeof(AcpiTableHeader));
927 
928     /* Extra PCI root buses are implemented  only for i440fx */
929     bus = find_i440fx();
930     if (bus) {
931         QLIST_FOREACH(bus, &bus->child, sibling) {
932             uint8_t bus_num = pci_bus_num(bus);
933             uint8_t numa_node = pci_bus_numa_node(bus);
934 
935             /* look only for expander root buses */
936             if (!pci_bus_is_root(bus)) {
937                 continue;
938             }
939 
940             if (bus_num < root_bus_limit) {
941                 root_bus_limit = bus_num - 1;
942             }
943 
944             scope = aml_scope("\\_SB");
945             dev = aml_device("PC%.02X", bus_num);
946             aml_append(dev, aml_name_decl("_UID", aml_int(bus_num)));
947             aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0A03")));
948             aml_append(dev, aml_name_decl("_BBN", aml_int(bus_num)));
949 
950             if (numa_node != NUMA_NODE_UNASSIGNED) {
951                 aml_append(dev, aml_name_decl("_PXM", aml_int(numa_node)));
952             }
953 
954             aml_append(dev, build_prt());
955             crs = build_crs(PCI_HOST_BRIDGE(BUS(bus)->parent),
956                             io_ranges, mem_ranges);
957             aml_append(dev, aml_name_decl("_CRS", crs));
958             aml_append(scope, dev);
959             aml_append(ssdt, scope);
960         }
961     }
962 
963     scope = aml_scope("\\_SB.PCI0");
964     /* build PCI0._CRS */
965     crs = aml_resource_template();
966     aml_append(crs,
967         aml_word_bus_number(AML_MIN_FIXED, AML_MAX_FIXED, AML_POS_DECODE,
968                             0x0000, 0x0, root_bus_limit,
969                             0x0000, root_bus_limit + 1));
970     aml_append(crs, aml_io(AML_DECODE16, 0x0CF8, 0x0CF8, 0x01, 0x08));
971 
972     aml_append(crs,
973         aml_word_io(AML_MIN_FIXED, AML_MAX_FIXED,
974                     AML_POS_DECODE, AML_ENTIRE_RANGE,
975                     0x0000, 0x0000, 0x0CF7, 0x0000, 0x0CF8));
976 
977     crs_replace_with_free_ranges(io_ranges, 0x0D00, 0xFFFF);
978     for (i = 0; i < io_ranges->len; i++) {
979         entry = g_ptr_array_index(io_ranges, i);
980         aml_append(crs,
981             aml_word_io(AML_MIN_FIXED, AML_MAX_FIXED,
982                         AML_POS_DECODE, AML_ENTIRE_RANGE,
983                         0x0000, entry->base, entry->limit,
984                         0x0000, entry->limit - entry->base + 1));
985     }
986 
987     aml_append(crs,
988         aml_dword_memory(AML_POS_DECODE, AML_MIN_FIXED, AML_MAX_FIXED,
989                          AML_CACHEABLE, AML_READ_WRITE,
990                          0, 0x000A0000, 0x000BFFFF, 0, 0x00020000));
991 
992     crs_replace_with_free_ranges(mem_ranges, pci->w32.begin, pci->w32.end - 1);
993     for (i = 0; i < mem_ranges->len; i++) {
994         entry = g_ptr_array_index(mem_ranges, i);
995         aml_append(crs,
996             aml_dword_memory(AML_POS_DECODE, AML_MIN_FIXED, AML_MAX_FIXED,
997                              AML_NON_CACHEABLE, AML_READ_WRITE,
998                              0, entry->base, entry->limit,
999                              0, entry->limit - entry->base + 1));
1000     }
1001 
1002     if (pci->w64.begin) {
1003         aml_append(crs,
1004             aml_qword_memory(AML_POS_DECODE, AML_MIN_FIXED, AML_MAX_FIXED,
1005                              AML_CACHEABLE, AML_READ_WRITE,
1006                              0, pci->w64.begin, pci->w64.end - 1, 0,
1007                              pci->w64.end - pci->w64.begin));
1008     }
1009     aml_append(scope, aml_name_decl("_CRS", crs));
1010 
1011     /* reserve GPE0 block resources */
1012     dev = aml_device("GPE0");
1013     aml_append(dev, aml_name_decl("_HID", aml_string("PNP0A06")));
1014     aml_append(dev, aml_name_decl("_UID", aml_string("GPE0 resources")));
1015     /* device present, functioning, decoding, not shown in UI */
1016     aml_append(dev, aml_name_decl("_STA", aml_int(0xB)));
1017     crs = aml_resource_template();
1018     aml_append(crs,
1019         aml_io(AML_DECODE16, pm->gpe0_blk, pm->gpe0_blk, 1, pm->gpe0_blk_len)
1020     );
1021     aml_append(dev, aml_name_decl("_CRS", crs));
1022     aml_append(scope, dev);
1023 
1024     g_ptr_array_free(io_ranges, true);
1025     g_ptr_array_free(mem_ranges, true);
1026 
1027     /* reserve PCIHP resources */
1028     if (pm->pcihp_io_len) {
1029         dev = aml_device("PHPR");
1030         aml_append(dev, aml_name_decl("_HID", aml_string("PNP0A06")));
1031         aml_append(dev,
1032             aml_name_decl("_UID", aml_string("PCI Hotplug resources")));
1033         /* device present, functioning, decoding, not shown in UI */
1034         aml_append(dev, aml_name_decl("_STA", aml_int(0xB)));
1035         crs = aml_resource_template();
1036         aml_append(crs,
1037             aml_io(AML_DECODE16, pm->pcihp_io_base, pm->pcihp_io_base, 1,
1038                    pm->pcihp_io_len)
1039         );
1040         aml_append(dev, aml_name_decl("_CRS", crs));
1041         aml_append(scope, dev);
1042     }
1043     aml_append(ssdt, scope);
1044 
1045     /*  create S3_ / S4_ / S5_ packages if necessary */
1046     scope = aml_scope("\\");
1047     if (!pm->s3_disabled) {
1048         pkg = aml_package(4);
1049         aml_append(pkg, aml_int(1)); /* PM1a_CNT.SLP_TYP */
1050         aml_append(pkg, aml_int(1)); /* PM1b_CNT.SLP_TYP, FIXME: not impl. */
1051         aml_append(pkg, aml_int(0)); /* reserved */
1052         aml_append(pkg, aml_int(0)); /* reserved */
1053         aml_append(scope, aml_name_decl("_S3", pkg));
1054     }
1055 
1056     if (!pm->s4_disabled) {
1057         pkg = aml_package(4);
1058         aml_append(pkg, aml_int(pm->s4_val)); /* PM1a_CNT.SLP_TYP */
1059         /* PM1b_CNT.SLP_TYP, FIXME: not impl. */
1060         aml_append(pkg, aml_int(pm->s4_val));
1061         aml_append(pkg, aml_int(0)); /* reserved */
1062         aml_append(pkg, aml_int(0)); /* reserved */
1063         aml_append(scope, aml_name_decl("_S4", pkg));
1064     }
1065 
1066     pkg = aml_package(4);
1067     aml_append(pkg, aml_int(0)); /* PM1a_CNT.SLP_TYP */
1068     aml_append(pkg, aml_int(0)); /* PM1b_CNT.SLP_TYP not impl. */
1069     aml_append(pkg, aml_int(0)); /* reserved */
1070     aml_append(pkg, aml_int(0)); /* reserved */
1071     aml_append(scope, aml_name_decl("_S5", pkg));
1072     aml_append(ssdt, scope);
1073 
1074     if (misc->applesmc_io_base) {
1075         scope = aml_scope("\\_SB.PCI0.ISA");
1076         dev = aml_device("SMC");
1077 
1078         aml_append(dev, aml_name_decl("_HID", aml_eisaid("APP0001")));
1079         /* device present, functioning, decoding, not shown in UI */
1080         aml_append(dev, aml_name_decl("_STA", aml_int(0xB)));
1081 
1082         crs = aml_resource_template();
1083         aml_append(crs,
1084             aml_io(AML_DECODE16, misc->applesmc_io_base, misc->applesmc_io_base,
1085                    0x01, APPLESMC_MAX_DATA_LENGTH)
1086         );
1087         aml_append(crs, aml_irq_no_flags(6));
1088         aml_append(dev, aml_name_decl("_CRS", crs));
1089 
1090         aml_append(scope, dev);
1091         aml_append(ssdt, scope);
1092     }
1093 
1094     if (misc->pvpanic_port) {
1095         scope = aml_scope("\\_SB.PCI0.ISA");
1096 
1097         dev = aml_device("PEVT");
1098         aml_append(dev, aml_name_decl("_HID", aml_string("QEMU0001")));
1099 
1100         crs = aml_resource_template();
1101         aml_append(crs,
1102             aml_io(AML_DECODE16, misc->pvpanic_port, misc->pvpanic_port, 1, 1)
1103         );
1104         aml_append(dev, aml_name_decl("_CRS", crs));
1105 
1106         aml_append(dev, aml_operation_region("PEOR", AML_SYSTEM_IO,
1107                                               misc->pvpanic_port, 1));
1108         field = aml_field("PEOR", AML_BYTE_ACC, AML_PRESERVE);
1109         aml_append(field, aml_named_field("PEPT", 8));
1110         aml_append(dev, field);
1111 
1112         /* device present, functioning, decoding, shown in UI */
1113         aml_append(dev, aml_name_decl("_STA", aml_int(0xF)));
1114 
1115         method = aml_method("RDPT", 0);
1116         aml_append(method, aml_store(aml_name("PEPT"), aml_local(0)));
1117         aml_append(method, aml_return(aml_local(0)));
1118         aml_append(dev, method);
1119 
1120         method = aml_method("WRPT", 1);
1121         aml_append(method, aml_store(aml_arg(0), aml_name("PEPT")));
1122         aml_append(dev, method);
1123 
1124         aml_append(scope, dev);
1125         aml_append(ssdt, scope);
1126     }
1127 
1128     sb_scope = aml_scope("\\_SB");
1129     {
1130         /* create PCI0.PRES device and its _CRS to reserve CPU hotplug MMIO */
1131         dev = aml_device("PCI0." stringify(CPU_HOTPLUG_RESOURCE_DEVICE));
1132         aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0A06")));
1133         aml_append(dev,
1134             aml_name_decl("_UID", aml_string("CPU Hotplug resources"))
1135         );
1136         /* device present, functioning, decoding, not shown in UI */
1137         aml_append(dev, aml_name_decl("_STA", aml_int(0xB)));
1138         crs = aml_resource_template();
1139         aml_append(crs,
1140             aml_io(AML_DECODE16, pm->cpu_hp_io_base, pm->cpu_hp_io_base, 1,
1141                    pm->cpu_hp_io_len)
1142         );
1143         aml_append(dev, aml_name_decl("_CRS", crs));
1144         aml_append(sb_scope, dev);
1145         /* declare CPU hotplug MMIO region and PRS field to access it */
1146         aml_append(sb_scope, aml_operation_region(
1147             "PRST", AML_SYSTEM_IO, pm->cpu_hp_io_base, pm->cpu_hp_io_len));
1148         field = aml_field("PRST", AML_BYTE_ACC, AML_PRESERVE);
1149         aml_append(field, aml_named_field("PRS", 256));
1150         aml_append(sb_scope, field);
1151 
1152         /* build Processor object for each processor */
1153         for (i = 0; i < acpi_cpus; i++) {
1154             dev = aml_processor(i, 0, 0, "CP%.02X", i);
1155 
1156             method = aml_method("_MAT", 0);
1157             aml_append(method, aml_return(aml_call1("CPMA", aml_int(i))));
1158             aml_append(dev, method);
1159 
1160             method = aml_method("_STA", 0);
1161             aml_append(method, aml_return(aml_call1("CPST", aml_int(i))));
1162             aml_append(dev, method);
1163 
1164             method = aml_method("_EJ0", 1);
1165             aml_append(method,
1166                 aml_return(aml_call2("CPEJ", aml_int(i), aml_arg(0)))
1167             );
1168             aml_append(dev, method);
1169 
1170             aml_append(sb_scope, dev);
1171         }
1172 
1173         /* build this code:
1174          *   Method(NTFY, 2) {If (LEqual(Arg0, 0x00)) {Notify(CP00, Arg1)} ...}
1175          */
1176         /* Arg0 = Processor ID = APIC ID */
1177         method = aml_method("NTFY", 2);
1178         for (i = 0; i < acpi_cpus; i++) {
1179             ifctx = aml_if(aml_equal(aml_arg(0), aml_int(i)));
1180             aml_append(ifctx,
1181                 aml_notify(aml_name("CP%.02X", i), aml_arg(1))
1182             );
1183             aml_append(method, ifctx);
1184         }
1185         aml_append(sb_scope, method);
1186 
1187         /* build "Name(CPON, Package() { One, One, ..., Zero, Zero, ... })"
1188          *
1189          * Note: The ability to create variable-sized packages was first
1190          * introduced in ACPI 2.0. ACPI 1.0 only allowed fixed-size packages
1191          * ith up to 255 elements. Windows guests up to win2k8 fail when
1192          * VarPackageOp is used.
1193          */
1194         pkg = acpi_cpus <= 255 ? aml_package(acpi_cpus) :
1195                                  aml_varpackage(acpi_cpus);
1196 
1197         for (i = 0; i < acpi_cpus; i++) {
1198             uint8_t b = test_bit(i, cpu->found_cpus) ? 0x01 : 0x00;
1199             aml_append(pkg, aml_int(b));
1200         }
1201         aml_append(sb_scope, aml_name_decl("CPON", pkg));
1202 
1203         /* build memory devices */
1204         assert(nr_mem <= ACPI_MAX_RAM_SLOTS);
1205         scope = aml_scope("\\_SB.PCI0." stringify(MEMORY_HOTPLUG_DEVICE));
1206         aml_append(scope,
1207             aml_name_decl(stringify(MEMORY_SLOTS_NUMBER), aml_int(nr_mem))
1208         );
1209 
1210         crs = aml_resource_template();
1211         aml_append(crs,
1212             aml_io(AML_DECODE16, pm->mem_hp_io_base, pm->mem_hp_io_base, 0,
1213                    pm->mem_hp_io_len)
1214         );
1215         aml_append(scope, aml_name_decl("_CRS", crs));
1216 
1217         aml_append(scope, aml_operation_region(
1218             stringify(MEMORY_HOTPLUG_IO_REGION), AML_SYSTEM_IO,
1219             pm->mem_hp_io_base, pm->mem_hp_io_len)
1220         );
1221 
1222         field = aml_field(stringify(MEMORY_HOTPLUG_IO_REGION), AML_DWORD_ACC,
1223                           AML_PRESERVE);
1224         aml_append(field, /* read only */
1225             aml_named_field(stringify(MEMORY_SLOT_ADDR_LOW), 32));
1226         aml_append(field, /* read only */
1227             aml_named_field(stringify(MEMORY_SLOT_ADDR_HIGH), 32));
1228         aml_append(field, /* read only */
1229             aml_named_field(stringify(MEMORY_SLOT_SIZE_LOW), 32));
1230         aml_append(field, /* read only */
1231             aml_named_field(stringify(MEMORY_SLOT_SIZE_HIGH), 32));
1232         aml_append(field, /* read only */
1233             aml_named_field(stringify(MEMORY_SLOT_PROXIMITY), 32));
1234         aml_append(scope, field);
1235 
1236         field = aml_field(stringify(MEMORY_HOTPLUG_IO_REGION), AML_BYTE_ACC,
1237                           AML_WRITE_AS_ZEROS);
1238         aml_append(field, aml_reserved_field(160 /* bits, Offset(20) */));
1239         aml_append(field, /* 1 if enabled, read only */
1240             aml_named_field(stringify(MEMORY_SLOT_ENABLED), 1));
1241         aml_append(field,
1242             /*(read) 1 if has a insert event. (write) 1 to clear event */
1243             aml_named_field(stringify(MEMORY_SLOT_INSERT_EVENT), 1));
1244         aml_append(field,
1245             /* (read) 1 if has a remove event. (write) 1 to clear event */
1246             aml_named_field(stringify(MEMORY_SLOT_REMOVE_EVENT), 1));
1247         aml_append(field,
1248             /* initiates device eject, write only */
1249             aml_named_field(stringify(MEMORY_SLOT_EJECT), 1));
1250         aml_append(scope, field);
1251 
1252         field = aml_field(stringify(MEMORY_HOTPLUG_IO_REGION), AML_DWORD_ACC,
1253                           AML_PRESERVE);
1254         aml_append(field, /* DIMM selector, write only */
1255             aml_named_field(stringify(MEMORY_SLOT_SLECTOR), 32));
1256         aml_append(field, /* _OST event code, write only */
1257             aml_named_field(stringify(MEMORY_SLOT_OST_EVENT), 32));
1258         aml_append(field, /* _OST status code, write only */
1259             aml_named_field(stringify(MEMORY_SLOT_OST_STATUS), 32));
1260         aml_append(scope, field);
1261 
1262         aml_append(sb_scope, scope);
1263 
1264         for (i = 0; i < nr_mem; i++) {
1265             #define BASEPATH "\\_SB.PCI0." stringify(MEMORY_HOTPLUG_DEVICE) "."
1266             const char *s;
1267 
1268             dev = aml_device("MP%02X", i);
1269             aml_append(dev, aml_name_decl("_UID", aml_string("0x%02X", i)));
1270             aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0C80")));
1271 
1272             method = aml_method("_CRS", 0);
1273             s = BASEPATH stringify(MEMORY_SLOT_CRS_METHOD);
1274             aml_append(method, aml_return(aml_call1(s, aml_name("_UID"))));
1275             aml_append(dev, method);
1276 
1277             method = aml_method("_STA", 0);
1278             s = BASEPATH stringify(MEMORY_SLOT_STATUS_METHOD);
1279             aml_append(method, aml_return(aml_call1(s, aml_name("_UID"))));
1280             aml_append(dev, method);
1281 
1282             method = aml_method("_PXM", 0);
1283             s = BASEPATH stringify(MEMORY_SLOT_PROXIMITY_METHOD);
1284             aml_append(method, aml_return(aml_call1(s, aml_name("_UID"))));
1285             aml_append(dev, method);
1286 
1287             method = aml_method("_OST", 3);
1288             s = BASEPATH stringify(MEMORY_SLOT_OST_METHOD);
1289             aml_append(method, aml_return(aml_call4(
1290                 s, aml_name("_UID"), aml_arg(0), aml_arg(1), aml_arg(2)
1291             )));
1292             aml_append(dev, method);
1293 
1294             method = aml_method("_EJ0", 1);
1295             s = BASEPATH stringify(MEMORY_SLOT_EJECT_METHOD);
1296             aml_append(method, aml_return(aml_call2(
1297                        s, aml_name("_UID"), aml_arg(0))));
1298             aml_append(dev, method);
1299 
1300             aml_append(sb_scope, dev);
1301         }
1302 
1303         /* build Method(MEMORY_SLOT_NOTIFY_METHOD, 2) {
1304          *     If (LEqual(Arg0, 0x00)) {Notify(MP00, Arg1)} ... }
1305          */
1306         method = aml_method(stringify(MEMORY_SLOT_NOTIFY_METHOD), 2);
1307         for (i = 0; i < nr_mem; i++) {
1308             ifctx = aml_if(aml_equal(aml_arg(0), aml_int(i)));
1309             aml_append(ifctx,
1310                 aml_notify(aml_name("MP%.02X", i), aml_arg(1))
1311             );
1312             aml_append(method, ifctx);
1313         }
1314         aml_append(sb_scope, method);
1315 
1316         {
1317             Object *pci_host;
1318             PCIBus *bus = NULL;
1319 
1320             pci_host = acpi_get_i386_pci_host();
1321             if (pci_host) {
1322                 bus = PCI_HOST_BRIDGE(pci_host)->bus;
1323             }
1324 
1325             if (bus) {
1326                 Aml *scope = aml_scope("PCI0");
1327                 /* Scan all PCI buses. Generate tables to support hotplug. */
1328                 build_append_pci_bus_devices(scope, bus, pm->pcihp_bridge_en);
1329 
1330                 if (misc->tpm_version != TPM_VERSION_UNSPEC) {
1331                     dev = aml_device("ISA.TPM");
1332                     aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0C31")));
1333                     aml_append(dev, aml_name_decl("_STA", aml_int(0xF)));
1334                     crs = aml_resource_template();
1335                     aml_append(crs, aml_memory32_fixed(TPM_TIS_ADDR_BASE,
1336                                TPM_TIS_ADDR_SIZE, AML_READ_WRITE));
1337                     aml_append(crs, aml_irq_no_flags(TPM_TIS_IRQ));
1338                     aml_append(dev, aml_name_decl("_CRS", crs));
1339                     aml_append(scope, dev);
1340                 }
1341 
1342                 aml_append(sb_scope, scope);
1343             }
1344         }
1345         aml_append(ssdt, sb_scope);
1346     }
1347 
1348     /* copy AML table into ACPI tables blob and patch header there */
1349     g_array_append_vals(table_data, ssdt->buf->data, ssdt->buf->len);
1350     build_header(linker, table_data,
1351         (void *)(table_data->data + table_data->len - ssdt->buf->len),
1352         "SSDT", ssdt->buf->len, 1);
1353     free_aml_allocator();
1354 }
1355 
1356 static void
1357 build_hpet(GArray *table_data, GArray *linker)
1358 {
1359     Acpi20Hpet *hpet;
1360 
1361     hpet = acpi_data_push(table_data, sizeof(*hpet));
1362     /* Note timer_block_id value must be kept in sync with value advertised by
1363      * emulated hpet
1364      */
1365     hpet->timer_block_id = cpu_to_le32(0x8086a201);
1366     hpet->addr.address = cpu_to_le64(HPET_BASE);
1367     build_header(linker, table_data,
1368                  (void *)hpet, "HPET", sizeof(*hpet), 1);
1369 }
1370 
1371 static void
1372 build_tpm_tcpa(GArray *table_data, GArray *linker, GArray *tcpalog)
1373 {
1374     Acpi20Tcpa *tcpa = acpi_data_push(table_data, sizeof *tcpa);
1375     uint64_t log_area_start_address = acpi_data_len(tcpalog);
1376 
1377     tcpa->platform_class = cpu_to_le16(TPM_TCPA_ACPI_CLASS_CLIENT);
1378     tcpa->log_area_minimum_length = cpu_to_le32(TPM_LOG_AREA_MINIMUM_SIZE);
1379     tcpa->log_area_start_address = cpu_to_le64(log_area_start_address);
1380 
1381     bios_linker_loader_alloc(linker, ACPI_BUILD_TPMLOG_FILE, 1,
1382                              false /* high memory */);
1383 
1384     /* log area start address to be filled by Guest linker */
1385     bios_linker_loader_add_pointer(linker, ACPI_BUILD_TABLE_FILE,
1386                                    ACPI_BUILD_TPMLOG_FILE,
1387                                    table_data, &tcpa->log_area_start_address,
1388                                    sizeof(tcpa->log_area_start_address));
1389 
1390     build_header(linker, table_data,
1391                  (void *)tcpa, "TCPA", sizeof(*tcpa), 2);
1392 
1393     acpi_data_push(tcpalog, TPM_LOG_AREA_MINIMUM_SIZE);
1394 }
1395 
1396 static void
1397 build_tpm2(GArray *table_data, GArray *linker)
1398 {
1399     Acpi20TPM2 *tpm2_ptr;
1400 
1401     tpm2_ptr = acpi_data_push(table_data, sizeof *tpm2_ptr);
1402 
1403     tpm2_ptr->platform_class = cpu_to_le16(TPM2_ACPI_CLASS_CLIENT);
1404     tpm2_ptr->control_area_address = cpu_to_le64(0);
1405     tpm2_ptr->start_method = cpu_to_le32(TPM2_START_METHOD_MMIO);
1406 
1407     build_header(linker, table_data,
1408                  (void *)tpm2_ptr, "TPM2", sizeof(*tpm2_ptr), 4);
1409 }
1410 
1411 typedef enum {
1412     MEM_AFFINITY_NOFLAGS      = 0,
1413     MEM_AFFINITY_ENABLED      = (1 << 0),
1414     MEM_AFFINITY_HOTPLUGGABLE = (1 << 1),
1415     MEM_AFFINITY_NON_VOLATILE = (1 << 2),
1416 } MemoryAffinityFlags;
1417 
1418 static void
1419 acpi_build_srat_memory(AcpiSratMemoryAffinity *numamem, uint64_t base,
1420                        uint64_t len, int node, MemoryAffinityFlags flags)
1421 {
1422     numamem->type = ACPI_SRAT_MEMORY;
1423     numamem->length = sizeof(*numamem);
1424     memset(numamem->proximity, 0, 4);
1425     numamem->proximity[0] = node;
1426     numamem->flags = cpu_to_le32(flags);
1427     numamem->base_addr = cpu_to_le64(base);
1428     numamem->range_length = cpu_to_le64(len);
1429 }
1430 
1431 static void
1432 build_srat(GArray *table_data, GArray *linker, PcGuestInfo *guest_info)
1433 {
1434     AcpiSystemResourceAffinityTable *srat;
1435     AcpiSratProcessorAffinity *core;
1436     AcpiSratMemoryAffinity *numamem;
1437 
1438     int i;
1439     uint64_t curnode;
1440     int srat_start, numa_start, slots;
1441     uint64_t mem_len, mem_base, next_base;
1442     PCMachineState *pcms = PC_MACHINE(qdev_get_machine());
1443     ram_addr_t hotplugabble_address_space_size =
1444         object_property_get_int(OBJECT(pcms), PC_MACHINE_MEMHP_REGION_SIZE,
1445                                 NULL);
1446 
1447     srat_start = table_data->len;
1448 
1449     srat = acpi_data_push(table_data, sizeof *srat);
1450     srat->reserved1 = cpu_to_le32(1);
1451     core = (void *)(srat + 1);
1452 
1453     for (i = 0; i < guest_info->apic_id_limit; ++i) {
1454         core = acpi_data_push(table_data, sizeof *core);
1455         core->type = ACPI_SRAT_PROCESSOR;
1456         core->length = sizeof(*core);
1457         core->local_apic_id = i;
1458         curnode = guest_info->node_cpu[i];
1459         core->proximity_lo = curnode;
1460         memset(core->proximity_hi, 0, 3);
1461         core->local_sapic_eid = 0;
1462         core->flags = cpu_to_le32(1);
1463     }
1464 
1465 
1466     /* the memory map is a bit tricky, it contains at least one hole
1467      * from 640k-1M and possibly another one from 3.5G-4G.
1468      */
1469     next_base = 0;
1470     numa_start = table_data->len;
1471 
1472     numamem = acpi_data_push(table_data, sizeof *numamem);
1473     acpi_build_srat_memory(numamem, 0, 640*1024, 0, MEM_AFFINITY_ENABLED);
1474     next_base = 1024 * 1024;
1475     for (i = 1; i < guest_info->numa_nodes + 1; ++i) {
1476         mem_base = next_base;
1477         mem_len = guest_info->node_mem[i - 1];
1478         if (i == 1) {
1479             mem_len -= 1024 * 1024;
1480         }
1481         next_base = mem_base + mem_len;
1482 
1483         /* Cut out the ACPI_PCI hole */
1484         if (mem_base <= guest_info->ram_size_below_4g &&
1485             next_base > guest_info->ram_size_below_4g) {
1486             mem_len -= next_base - guest_info->ram_size_below_4g;
1487             if (mem_len > 0) {
1488                 numamem = acpi_data_push(table_data, sizeof *numamem);
1489                 acpi_build_srat_memory(numamem, mem_base, mem_len, i - 1,
1490                                        MEM_AFFINITY_ENABLED);
1491             }
1492             mem_base = 1ULL << 32;
1493             mem_len = next_base - guest_info->ram_size_below_4g;
1494             next_base += (1ULL << 32) - guest_info->ram_size_below_4g;
1495         }
1496         numamem = acpi_data_push(table_data, sizeof *numamem);
1497         acpi_build_srat_memory(numamem, mem_base, mem_len, i - 1,
1498                                MEM_AFFINITY_ENABLED);
1499     }
1500     slots = (table_data->len - numa_start) / sizeof *numamem;
1501     for (; slots < guest_info->numa_nodes + 2; slots++) {
1502         numamem = acpi_data_push(table_data, sizeof *numamem);
1503         acpi_build_srat_memory(numamem, 0, 0, 0, MEM_AFFINITY_NOFLAGS);
1504     }
1505 
1506     /*
1507      * Entry is required for Windows to enable memory hotplug in OS.
1508      * Memory devices may override proximity set by this entry,
1509      * providing _PXM method if necessary.
1510      */
1511     if (hotplugabble_address_space_size) {
1512         numamem = acpi_data_push(table_data, sizeof *numamem);
1513         acpi_build_srat_memory(numamem, pcms->hotplug_memory.base,
1514                                hotplugabble_address_space_size, 0,
1515                                MEM_AFFINITY_HOTPLUGGABLE |
1516                                MEM_AFFINITY_ENABLED);
1517     }
1518 
1519     build_header(linker, table_data,
1520                  (void *)(table_data->data + srat_start),
1521                  "SRAT",
1522                  table_data->len - srat_start, 1);
1523 }
1524 
1525 static void
1526 build_mcfg_q35(GArray *table_data, GArray *linker, AcpiMcfgInfo *info)
1527 {
1528     AcpiTableMcfg *mcfg;
1529     const char *sig;
1530     int len = sizeof(*mcfg) + 1 * sizeof(mcfg->allocation[0]);
1531 
1532     mcfg = acpi_data_push(table_data, len);
1533     mcfg->allocation[0].address = cpu_to_le64(info->mcfg_base);
1534     /* Only a single allocation so no need to play with segments */
1535     mcfg->allocation[0].pci_segment = cpu_to_le16(0);
1536     mcfg->allocation[0].start_bus_number = 0;
1537     mcfg->allocation[0].end_bus_number = PCIE_MMCFG_BUS(info->mcfg_size - 1);
1538 
1539     /* MCFG is used for ECAM which can be enabled or disabled by guest.
1540      * To avoid table size changes (which create migration issues),
1541      * always create the table even if there are no allocations,
1542      * but set the signature to a reserved value in this case.
1543      * ACPI spec requires OSPMs to ignore such tables.
1544      */
1545     if (info->mcfg_base == PCIE_BASE_ADDR_UNMAPPED) {
1546         /* Reserved signature: ignored by OSPM */
1547         sig = "QEMU";
1548     } else {
1549         sig = "MCFG";
1550     }
1551     build_header(linker, table_data, (void *)mcfg, sig, len, 1);
1552 }
1553 
1554 static void
1555 build_dmar_q35(GArray *table_data, GArray *linker)
1556 {
1557     int dmar_start = table_data->len;
1558 
1559     AcpiTableDmar *dmar;
1560     AcpiDmarHardwareUnit *drhd;
1561 
1562     dmar = acpi_data_push(table_data, sizeof(*dmar));
1563     dmar->host_address_width = VTD_HOST_ADDRESS_WIDTH - 1;
1564     dmar->flags = 0;    /* No intr_remap for now */
1565 
1566     /* DMAR Remapping Hardware Unit Definition structure */
1567     drhd = acpi_data_push(table_data, sizeof(*drhd));
1568     drhd->type = cpu_to_le16(ACPI_DMAR_TYPE_HARDWARE_UNIT);
1569     drhd->length = cpu_to_le16(sizeof(*drhd));   /* No device scope now */
1570     drhd->flags = ACPI_DMAR_INCLUDE_PCI_ALL;
1571     drhd->pci_segment = cpu_to_le16(0);
1572     drhd->address = cpu_to_le64(Q35_HOST_BRIDGE_IOMMU_ADDR);
1573 
1574     build_header(linker, table_data, (void *)(table_data->data + dmar_start),
1575                  "DMAR", table_data->len - dmar_start, 1);
1576 }
1577 
1578 static void
1579 build_dsdt(GArray *table_data, GArray *linker, AcpiMiscInfo *misc)
1580 {
1581     AcpiTableHeader *dsdt;
1582 
1583     assert(misc->dsdt_code && misc->dsdt_size);
1584 
1585     dsdt = acpi_data_push(table_data, misc->dsdt_size);
1586     memcpy(dsdt, misc->dsdt_code, misc->dsdt_size);
1587 
1588     memset(dsdt, 0, sizeof *dsdt);
1589     build_header(linker, table_data, dsdt, "DSDT",
1590                  misc->dsdt_size, 1);
1591 }
1592 
1593 static GArray *
1594 build_rsdp(GArray *rsdp_table, GArray *linker, unsigned rsdt)
1595 {
1596     AcpiRsdpDescriptor *rsdp = acpi_data_push(rsdp_table, sizeof *rsdp);
1597 
1598     bios_linker_loader_alloc(linker, ACPI_BUILD_RSDP_FILE, 16,
1599                              true /* fseg memory */);
1600 
1601     memcpy(&rsdp->signature, "RSD PTR ", 8);
1602     memcpy(rsdp->oem_id, ACPI_BUILD_APPNAME6, 6);
1603     rsdp->rsdt_physical_address = cpu_to_le32(rsdt);
1604     /* Address to be filled by Guest linker */
1605     bios_linker_loader_add_pointer(linker, ACPI_BUILD_RSDP_FILE,
1606                                    ACPI_BUILD_TABLE_FILE,
1607                                    rsdp_table, &rsdp->rsdt_physical_address,
1608                                    sizeof rsdp->rsdt_physical_address);
1609     rsdp->checksum = 0;
1610     /* Checksum to be filled by Guest linker */
1611     bios_linker_loader_add_checksum(linker, ACPI_BUILD_RSDP_FILE,
1612                                     rsdp, rsdp, sizeof *rsdp, &rsdp->checksum);
1613 
1614     return rsdp_table;
1615 }
1616 
1617 typedef
1618 struct AcpiBuildState {
1619     /* Copy of table in RAM (for patching). */
1620     MemoryRegion *table_mr;
1621     /* Is table patched? */
1622     uint8_t patched;
1623     PcGuestInfo *guest_info;
1624     void *rsdp;
1625     MemoryRegion *rsdp_mr;
1626     MemoryRegion *linker_mr;
1627 } AcpiBuildState;
1628 
1629 static bool acpi_get_mcfg(AcpiMcfgInfo *mcfg)
1630 {
1631     Object *pci_host;
1632     QObject *o;
1633 
1634     pci_host = acpi_get_i386_pci_host();
1635     g_assert(pci_host);
1636 
1637     o = object_property_get_qobject(pci_host, PCIE_HOST_MCFG_BASE, NULL);
1638     if (!o) {
1639         return false;
1640     }
1641     mcfg->mcfg_base = qint_get_int(qobject_to_qint(o));
1642     qobject_decref(o);
1643 
1644     o = object_property_get_qobject(pci_host, PCIE_HOST_MCFG_SIZE, NULL);
1645     assert(o);
1646     mcfg->mcfg_size = qint_get_int(qobject_to_qint(o));
1647     qobject_decref(o);
1648     return true;
1649 }
1650 
1651 static bool acpi_has_iommu(void)
1652 {
1653     bool ambiguous;
1654     Object *intel_iommu;
1655 
1656     intel_iommu = object_resolve_path_type("", TYPE_INTEL_IOMMU_DEVICE,
1657                                            &ambiguous);
1658     return intel_iommu && !ambiguous;
1659 }
1660 
1661 static
1662 void acpi_build(PcGuestInfo *guest_info, AcpiBuildTables *tables)
1663 {
1664     GArray *table_offsets;
1665     unsigned facs, ssdt, dsdt, rsdt;
1666     AcpiCpuInfo cpu;
1667     AcpiPmInfo pm;
1668     AcpiMiscInfo misc;
1669     AcpiMcfgInfo mcfg;
1670     PcPciInfo pci;
1671     uint8_t *u;
1672     size_t aml_len = 0;
1673     GArray *tables_blob = tables->table_data;
1674 
1675     acpi_get_cpu_info(&cpu);
1676     acpi_get_pm_info(&pm);
1677     acpi_get_dsdt(&misc);
1678     acpi_get_misc_info(&misc);
1679     acpi_get_pci_info(&pci);
1680 
1681     table_offsets = g_array_new(false, true /* clear */,
1682                                         sizeof(uint32_t));
1683     ACPI_BUILD_DPRINTF("init ACPI tables\n");
1684 
1685     bios_linker_loader_alloc(tables->linker, ACPI_BUILD_TABLE_FILE,
1686                              64 /* Ensure FACS is aligned */,
1687                              false /* high memory */);
1688 
1689     /*
1690      * FACS is pointed to by FADT.
1691      * We place it first since it's the only table that has alignment
1692      * requirements.
1693      */
1694     facs = tables_blob->len;
1695     build_facs(tables_blob, tables->linker, guest_info);
1696 
1697     /* DSDT is pointed to by FADT */
1698     dsdt = tables_blob->len;
1699     build_dsdt(tables_blob, tables->linker, &misc);
1700 
1701     /* Count the size of the DSDT and SSDT, we will need it for legacy
1702      * sizing of ACPI tables.
1703      */
1704     aml_len += tables_blob->len - dsdt;
1705 
1706     /* ACPI tables pointed to by RSDT */
1707     acpi_add_table(table_offsets, tables_blob);
1708     build_fadt(tables_blob, tables->linker, &pm, facs, dsdt);
1709 
1710     ssdt = tables_blob->len;
1711     acpi_add_table(table_offsets, tables_blob);
1712     build_ssdt(tables_blob, tables->linker, &cpu, &pm, &misc, &pci,
1713                guest_info);
1714     aml_len += tables_blob->len - ssdt;
1715 
1716     acpi_add_table(table_offsets, tables_blob);
1717     build_madt(tables_blob, tables->linker, &cpu, guest_info);
1718 
1719     if (misc.has_hpet) {
1720         acpi_add_table(table_offsets, tables_blob);
1721         build_hpet(tables_blob, tables->linker);
1722     }
1723     if (misc.tpm_version != TPM_VERSION_UNSPEC) {
1724         acpi_add_table(table_offsets, tables_blob);
1725         build_tpm_tcpa(tables_blob, tables->linker, tables->tcpalog);
1726 
1727         if (misc.tpm_version == TPM_VERSION_2_0) {
1728             acpi_add_table(table_offsets, tables_blob);
1729             build_tpm2(tables_blob, tables->linker);
1730         }
1731     }
1732     if (guest_info->numa_nodes) {
1733         acpi_add_table(table_offsets, tables_blob);
1734         build_srat(tables_blob, tables->linker, guest_info);
1735     }
1736     if (acpi_get_mcfg(&mcfg)) {
1737         acpi_add_table(table_offsets, tables_blob);
1738         build_mcfg_q35(tables_blob, tables->linker, &mcfg);
1739     }
1740     if (acpi_has_iommu()) {
1741         acpi_add_table(table_offsets, tables_blob);
1742         build_dmar_q35(tables_blob, tables->linker);
1743     }
1744 
1745     /* Add tables supplied by user (if any) */
1746     for (u = acpi_table_first(); u; u = acpi_table_next(u)) {
1747         unsigned len = acpi_table_len(u);
1748 
1749         acpi_add_table(table_offsets, tables_blob);
1750         g_array_append_vals(tables_blob, u, len);
1751     }
1752 
1753     /* RSDT is pointed to by RSDP */
1754     rsdt = tables_blob->len;
1755     build_rsdt(tables_blob, tables->linker, table_offsets);
1756 
1757     /* RSDP is in FSEG memory, so allocate it separately */
1758     build_rsdp(tables->rsdp, tables->linker, rsdt);
1759 
1760     /* We'll expose it all to Guest so we want to reduce
1761      * chance of size changes.
1762      *
1763      * We used to align the tables to 4k, but of course this would
1764      * too simple to be enough.  4k turned out to be too small an
1765      * alignment very soon, and in fact it is almost impossible to
1766      * keep the table size stable for all (max_cpus, max_memory_slots)
1767      * combinations.  So the table size is always 64k for pc-i440fx-2.1
1768      * and we give an error if the table grows beyond that limit.
1769      *
1770      * We still have the problem of migrating from "-M pc-i440fx-2.0".  For
1771      * that, we exploit the fact that QEMU 2.1 generates _smaller_ tables
1772      * than 2.0 and we can always pad the smaller tables with zeros.  We can
1773      * then use the exact size of the 2.0 tables.
1774      *
1775      * All this is for PIIX4, since QEMU 2.0 didn't support Q35 migration.
1776      */
1777     if (guest_info->legacy_acpi_table_size) {
1778         /* Subtracting aml_len gives the size of fixed tables.  Then add the
1779          * size of the PIIX4 DSDT/SSDT in QEMU 2.0.
1780          */
1781         int legacy_aml_len =
1782             guest_info->legacy_acpi_table_size +
1783             ACPI_BUILD_LEGACY_CPU_AML_SIZE * max_cpus;
1784         int legacy_table_size =
1785             ROUND_UP(tables_blob->len - aml_len + legacy_aml_len,
1786                      ACPI_BUILD_ALIGN_SIZE);
1787         if (tables_blob->len > legacy_table_size) {
1788             /* Should happen only with PCI bridges and -M pc-i440fx-2.0.  */
1789             error_report("Warning: migration may not work.");
1790         }
1791         g_array_set_size(tables_blob, legacy_table_size);
1792     } else {
1793         /* Make sure we have a buffer in case we need to resize the tables. */
1794         if (tables_blob->len > ACPI_BUILD_TABLE_SIZE / 2) {
1795             /* As of QEMU 2.1, this fires with 160 VCPUs and 255 memory slots.  */
1796             error_report("Warning: ACPI tables are larger than 64k.");
1797             error_report("Warning: migration may not work.");
1798             error_report("Warning: please remove CPUs, NUMA nodes, "
1799                          "memory slots or PCI bridges.");
1800         }
1801         acpi_align_size(tables_blob, ACPI_BUILD_TABLE_SIZE);
1802     }
1803 
1804     acpi_align_size(tables->linker, ACPI_BUILD_ALIGN_SIZE);
1805 
1806     /* Cleanup memory that's no longer used. */
1807     g_array_free(table_offsets, true);
1808 }
1809 
1810 static void acpi_ram_update(MemoryRegion *mr, GArray *data)
1811 {
1812     uint32_t size = acpi_data_len(data);
1813 
1814     /* Make sure RAM size is correct - in case it got changed e.g. by migration */
1815     memory_region_ram_resize(mr, size, &error_abort);
1816 
1817     memcpy(memory_region_get_ram_ptr(mr), data->data, size);
1818     memory_region_set_dirty(mr, 0, size);
1819 }
1820 
1821 static void acpi_build_update(void *build_opaque)
1822 {
1823     AcpiBuildState *build_state = build_opaque;
1824     AcpiBuildTables tables;
1825 
1826     /* No state to update or already patched? Nothing to do. */
1827     if (!build_state || build_state->patched) {
1828         return;
1829     }
1830     build_state->patched = 1;
1831 
1832     acpi_build_tables_init(&tables);
1833 
1834     acpi_build(build_state->guest_info, &tables);
1835 
1836     acpi_ram_update(build_state->table_mr, tables.table_data);
1837 
1838     if (build_state->rsdp) {
1839         memcpy(build_state->rsdp, tables.rsdp->data, acpi_data_len(tables.rsdp));
1840     } else {
1841         acpi_ram_update(build_state->rsdp_mr, tables.rsdp);
1842     }
1843 
1844     acpi_ram_update(build_state->linker_mr, tables.linker);
1845     acpi_build_tables_cleanup(&tables, true);
1846 }
1847 
1848 static void acpi_build_reset(void *build_opaque)
1849 {
1850     AcpiBuildState *build_state = build_opaque;
1851     build_state->patched = 0;
1852 }
1853 
1854 static MemoryRegion *acpi_add_rom_blob(AcpiBuildState *build_state,
1855                                        GArray *blob, const char *name,
1856                                        uint64_t max_size)
1857 {
1858     return rom_add_blob(name, blob->data, acpi_data_len(blob), max_size, -1,
1859                         name, acpi_build_update, build_state);
1860 }
1861 
1862 static const VMStateDescription vmstate_acpi_build = {
1863     .name = "acpi_build",
1864     .version_id = 1,
1865     .minimum_version_id = 1,
1866     .fields = (VMStateField[]) {
1867         VMSTATE_UINT8(patched, AcpiBuildState),
1868         VMSTATE_END_OF_LIST()
1869     },
1870 };
1871 
1872 void acpi_setup(PcGuestInfo *guest_info)
1873 {
1874     AcpiBuildTables tables;
1875     AcpiBuildState *build_state;
1876 
1877     if (!guest_info->fw_cfg) {
1878         ACPI_BUILD_DPRINTF("No fw cfg. Bailing out.\n");
1879         return;
1880     }
1881 
1882     if (!guest_info->has_acpi_build) {
1883         ACPI_BUILD_DPRINTF("ACPI build disabled. Bailing out.\n");
1884         return;
1885     }
1886 
1887     if (!acpi_enabled) {
1888         ACPI_BUILD_DPRINTF("ACPI disabled. Bailing out.\n");
1889         return;
1890     }
1891 
1892     build_state = g_malloc0(sizeof *build_state);
1893 
1894     build_state->guest_info = guest_info;
1895 
1896     acpi_set_pci_info();
1897 
1898     acpi_build_tables_init(&tables);
1899     acpi_build(build_state->guest_info, &tables);
1900 
1901     /* Now expose it all to Guest */
1902     build_state->table_mr = acpi_add_rom_blob(build_state, tables.table_data,
1903                                                ACPI_BUILD_TABLE_FILE,
1904                                                ACPI_BUILD_TABLE_MAX_SIZE);
1905     assert(build_state->table_mr != NULL);
1906 
1907     build_state->linker_mr =
1908         acpi_add_rom_blob(build_state, tables.linker, "etc/table-loader", 0);
1909 
1910     fw_cfg_add_file(guest_info->fw_cfg, ACPI_BUILD_TPMLOG_FILE,
1911                     tables.tcpalog->data, acpi_data_len(tables.tcpalog));
1912 
1913     if (!guest_info->rsdp_in_ram) {
1914         /*
1915          * Keep for compatibility with old machine types.
1916          * Though RSDP is small, its contents isn't immutable, so
1917          * we'll update it along with the rest of tables on guest access.
1918          */
1919         uint32_t rsdp_size = acpi_data_len(tables.rsdp);
1920 
1921         build_state->rsdp = g_memdup(tables.rsdp->data, rsdp_size);
1922         fw_cfg_add_file_callback(guest_info->fw_cfg, ACPI_BUILD_RSDP_FILE,
1923                                  acpi_build_update, build_state,
1924                                  build_state->rsdp, rsdp_size);
1925         build_state->rsdp_mr = NULL;
1926     } else {
1927         build_state->rsdp = NULL;
1928         build_state->rsdp_mr = acpi_add_rom_blob(build_state, tables.rsdp,
1929                                                   ACPI_BUILD_RSDP_FILE, 0);
1930     }
1931 
1932     qemu_register_reset(acpi_build_reset, build_state);
1933     acpi_build_reset(build_state);
1934     vmstate_register(NULL, 0, &vmstate_acpi_build, build_state);
1935 
1936     /* Cleanup tables but don't free the memory: we track it
1937      * in build_state.
1938      */
1939     acpi_build_tables_cleanup(&tables, false);
1940 }
1941