xref: /openbmc/qemu/hw/i386/acpi-build.c (revision ff80dc7fa8045e2b2531888d965424d2b0e1d1b6)
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 
45 /* Supported chipsets: */
46 #include "hw/acpi/piix4.h"
47 #include "hw/acpi/pcihp.h"
48 #include "hw/i386/ich9.h"
49 #include "hw/pci/pci_bus.h"
50 #include "hw/pci-host/q35.h"
51 #include "hw/i386/intel_iommu.h"
52 
53 #include "hw/i386/q35-acpi-dsdt.hex"
54 #include "hw/i386/acpi-dsdt.hex"
55 
56 #include "hw/acpi/aml-build.h"
57 
58 #include "qapi/qmp/qint.h"
59 #include "qom/qom-qobject.h"
60 
61 /* These are used to size the ACPI tables for -M pc-i440fx-1.7 and
62  * -M pc-i440fx-2.0.  Even if the actual amount of AML generated grows
63  * a little bit, there should be plenty of free space since the DSDT
64  * shrunk by ~1.5k between QEMU 2.0 and QEMU 2.1.
65  */
66 #define ACPI_BUILD_LEGACY_CPU_AML_SIZE    97
67 #define ACPI_BUILD_ALIGN_SIZE             0x1000
68 
69 #define ACPI_BUILD_TABLE_SIZE             0x20000
70 
71 /* #define DEBUG_ACPI_BUILD */
72 #ifdef DEBUG_ACPI_BUILD
73 #define ACPI_BUILD_DPRINTF(fmt, ...)        \
74     do {printf("ACPI_BUILD: " fmt, ## __VA_ARGS__); } while (0)
75 #else
76 #define ACPI_BUILD_DPRINTF(fmt, ...)
77 #endif
78 
79 typedef struct AcpiCpuInfo {
80     DECLARE_BITMAP(found_cpus, ACPI_CPU_HOTPLUG_ID_LIMIT);
81 } AcpiCpuInfo;
82 
83 typedef struct AcpiMcfgInfo {
84     uint64_t mcfg_base;
85     uint32_t mcfg_size;
86 } AcpiMcfgInfo;
87 
88 typedef struct AcpiPmInfo {
89     bool s3_disabled;
90     bool s4_disabled;
91     bool pcihp_bridge_en;
92     uint8_t s4_val;
93     uint16_t sci_int;
94     uint8_t acpi_enable_cmd;
95     uint8_t acpi_disable_cmd;
96     uint32_t gpe0_blk;
97     uint32_t gpe0_blk_len;
98     uint32_t io_base;
99     uint16_t cpu_hp_io_base;
100     uint16_t cpu_hp_io_len;
101     uint16_t mem_hp_io_base;
102     uint16_t mem_hp_io_len;
103     uint16_t pcihp_io_base;
104     uint16_t pcihp_io_len;
105 } AcpiPmInfo;
106 
107 typedef struct AcpiMiscInfo {
108     bool has_hpet;
109     bool has_tpm;
110     const unsigned char *dsdt_code;
111     unsigned dsdt_size;
112     uint16_t pvpanic_port;
113     uint16_t applesmc_io_base;
114 } AcpiMiscInfo;
115 
116 typedef struct AcpiBuildPciBusHotplugState {
117     GArray *device_table;
118     GArray *notify_table;
119     struct AcpiBuildPciBusHotplugState *parent;
120     bool pcihp_bridge_en;
121 } AcpiBuildPciBusHotplugState;
122 
123 static void acpi_get_dsdt(AcpiMiscInfo *info)
124 {
125     Object *piix = piix4_pm_find();
126     Object *lpc = ich9_lpc_find();
127     assert(!!piix != !!lpc);
128 
129     if (piix) {
130         info->dsdt_code = AcpiDsdtAmlCode;
131         info->dsdt_size = sizeof AcpiDsdtAmlCode;
132     }
133     if (lpc) {
134         info->dsdt_code = Q35AcpiDsdtAmlCode;
135         info->dsdt_size = sizeof Q35AcpiDsdtAmlCode;
136     }
137 }
138 
139 static
140 int acpi_add_cpu_info(Object *o, void *opaque)
141 {
142     AcpiCpuInfo *cpu = opaque;
143     uint64_t apic_id;
144 
145     if (object_dynamic_cast(o, TYPE_CPU)) {
146         apic_id = object_property_get_int(o, "apic-id", NULL);
147         assert(apic_id < ACPI_CPU_HOTPLUG_ID_LIMIT);
148 
149         set_bit(apic_id, cpu->found_cpus);
150     }
151 
152     object_child_foreach(o, acpi_add_cpu_info, opaque);
153     return 0;
154 }
155 
156 static void acpi_get_cpu_info(AcpiCpuInfo *cpu)
157 {
158     Object *root = object_get_root();
159 
160     memset(cpu->found_cpus, 0, sizeof cpu->found_cpus);
161     object_child_foreach(root, acpi_add_cpu_info, cpu);
162 }
163 
164 static void acpi_get_pm_info(AcpiPmInfo *pm)
165 {
166     Object *piix = piix4_pm_find();
167     Object *lpc = ich9_lpc_find();
168     Object *obj = NULL;
169     QObject *o;
170 
171     pm->pcihp_io_base = 0;
172     pm->pcihp_io_len = 0;
173     if (piix) {
174         obj = piix;
175         pm->cpu_hp_io_base = PIIX4_CPU_HOTPLUG_IO_BASE;
176         pm->pcihp_io_base =
177             object_property_get_int(obj, ACPI_PCIHP_IO_BASE_PROP, NULL);
178         pm->pcihp_io_len =
179             object_property_get_int(obj, ACPI_PCIHP_IO_LEN_PROP, NULL);
180     }
181     if (lpc) {
182         obj = lpc;
183         pm->cpu_hp_io_base = ICH9_CPU_HOTPLUG_IO_BASE;
184     }
185     assert(obj);
186 
187     pm->cpu_hp_io_len = ACPI_GPE_PROC_LEN;
188     pm->mem_hp_io_base = ACPI_MEMORY_HOTPLUG_BASE;
189     pm->mem_hp_io_len = ACPI_MEMORY_HOTPLUG_IO_LEN;
190 
191     /* Fill in optional s3/s4 related properties */
192     o = object_property_get_qobject(obj, ACPI_PM_PROP_S3_DISABLED, NULL);
193     if (o) {
194         pm->s3_disabled = qint_get_int(qobject_to_qint(o));
195     } else {
196         pm->s3_disabled = false;
197     }
198     qobject_decref(o);
199     o = object_property_get_qobject(obj, ACPI_PM_PROP_S4_DISABLED, NULL);
200     if (o) {
201         pm->s4_disabled = qint_get_int(qobject_to_qint(o));
202     } else {
203         pm->s4_disabled = false;
204     }
205     qobject_decref(o);
206     o = object_property_get_qobject(obj, ACPI_PM_PROP_S4_VAL, NULL);
207     if (o) {
208         pm->s4_val = qint_get_int(qobject_to_qint(o));
209     } else {
210         pm->s4_val = false;
211     }
212     qobject_decref(o);
213 
214     /* Fill in mandatory properties */
215     pm->sci_int = object_property_get_int(obj, ACPI_PM_PROP_SCI_INT, NULL);
216 
217     pm->acpi_enable_cmd = object_property_get_int(obj,
218                                                   ACPI_PM_PROP_ACPI_ENABLE_CMD,
219                                                   NULL);
220     pm->acpi_disable_cmd = object_property_get_int(obj,
221                                                   ACPI_PM_PROP_ACPI_DISABLE_CMD,
222                                                   NULL);
223     pm->io_base = object_property_get_int(obj, ACPI_PM_PROP_PM_IO_BASE,
224                                           NULL);
225     pm->gpe0_blk = object_property_get_int(obj, ACPI_PM_PROP_GPE0_BLK,
226                                            NULL);
227     pm->gpe0_blk_len = object_property_get_int(obj, ACPI_PM_PROP_GPE0_BLK_LEN,
228                                                NULL);
229     pm->pcihp_bridge_en =
230         object_property_get_bool(obj, "acpi-pci-hotplug-with-bridge-support",
231                                  NULL);
232 }
233 
234 static void acpi_get_misc_info(AcpiMiscInfo *info)
235 {
236     info->has_hpet = hpet_find();
237     info->has_tpm = tpm_find();
238     info->pvpanic_port = pvpanic_port();
239     info->applesmc_io_base = applesmc_port();
240 }
241 
242 static void acpi_get_pci_info(PcPciInfo *info)
243 {
244     Object *pci_host;
245     bool ambiguous;
246 
247     pci_host = object_resolve_path_type("", TYPE_PCI_HOST_BRIDGE, &ambiguous);
248     g_assert(!ambiguous);
249     g_assert(pci_host);
250 
251     info->w32.begin = object_property_get_int(pci_host,
252                                               PCI_HOST_PROP_PCI_HOLE_START,
253                                               NULL);
254     info->w32.end = object_property_get_int(pci_host,
255                                             PCI_HOST_PROP_PCI_HOLE_END,
256                                             NULL);
257     info->w64.begin = object_property_get_int(pci_host,
258                                               PCI_HOST_PROP_PCI_HOLE64_START,
259                                               NULL);
260     info->w64.end = object_property_get_int(pci_host,
261                                             PCI_HOST_PROP_PCI_HOLE64_END,
262                                             NULL);
263 }
264 
265 #define ACPI_PORT_SMI_CMD           0x00b2 /* TODO: this is APM_CNT_IOPORT */
266 
267 static void acpi_align_size(GArray *blob, unsigned align)
268 {
269     /* Align size to multiple of given size. This reduces the chance
270      * we need to change size in the future (breaking cross version migration).
271      */
272     g_array_set_size(blob, ROUND_UP(acpi_data_len(blob), align));
273 }
274 
275 /* FACS */
276 static void
277 build_facs(GArray *table_data, GArray *linker, PcGuestInfo *guest_info)
278 {
279     AcpiFacsDescriptorRev1 *facs = acpi_data_push(table_data, sizeof *facs);
280     memcpy(&facs->signature, "FACS", 4);
281     facs->length = cpu_to_le32(sizeof(*facs));
282 }
283 
284 /* Load chipset information in FADT */
285 static void fadt_setup(AcpiFadtDescriptorRev1 *fadt, AcpiPmInfo *pm)
286 {
287     fadt->model = 1;
288     fadt->reserved1 = 0;
289     fadt->sci_int = cpu_to_le16(pm->sci_int);
290     fadt->smi_cmd = cpu_to_le32(ACPI_PORT_SMI_CMD);
291     fadt->acpi_enable = pm->acpi_enable_cmd;
292     fadt->acpi_disable = pm->acpi_disable_cmd;
293     /* EVT, CNT, TMR offset matches hw/acpi/core.c */
294     fadt->pm1a_evt_blk = cpu_to_le32(pm->io_base);
295     fadt->pm1a_cnt_blk = cpu_to_le32(pm->io_base + 0x04);
296     fadt->pm_tmr_blk = cpu_to_le32(pm->io_base + 0x08);
297     fadt->gpe0_blk = cpu_to_le32(pm->gpe0_blk);
298     /* EVT, CNT, TMR length matches hw/acpi/core.c */
299     fadt->pm1_evt_len = 4;
300     fadt->pm1_cnt_len = 2;
301     fadt->pm_tmr_len = 4;
302     fadt->gpe0_blk_len = pm->gpe0_blk_len;
303     fadt->plvl2_lat = cpu_to_le16(0xfff); /* C2 state not supported */
304     fadt->plvl3_lat = cpu_to_le16(0xfff); /* C3 state not supported */
305     fadt->flags = cpu_to_le32((1 << ACPI_FADT_F_WBINVD) |
306                               (1 << ACPI_FADT_F_PROC_C1) |
307                               (1 << ACPI_FADT_F_SLP_BUTTON) |
308                               (1 << ACPI_FADT_F_RTC_S4));
309     fadt->flags |= cpu_to_le32(1 << ACPI_FADT_F_USE_PLATFORM_CLOCK);
310     /* APIC destination mode ("Flat Logical") has an upper limit of 8 CPUs
311      * For more than 8 CPUs, "Clustered Logical" mode has to be used
312      */
313     if (max_cpus > 8) {
314         fadt->flags |= cpu_to_le32(1 << ACPI_FADT_F_FORCE_APIC_CLUSTER_MODEL);
315     }
316 }
317 
318 
319 /* FADT */
320 static void
321 build_fadt(GArray *table_data, GArray *linker, AcpiPmInfo *pm,
322            unsigned facs, unsigned dsdt)
323 {
324     AcpiFadtDescriptorRev1 *fadt = acpi_data_push(table_data, sizeof(*fadt));
325 
326     fadt->firmware_ctrl = cpu_to_le32(facs);
327     /* FACS address to be filled by Guest linker */
328     bios_linker_loader_add_pointer(linker, ACPI_BUILD_TABLE_FILE,
329                                    ACPI_BUILD_TABLE_FILE,
330                                    table_data, &fadt->firmware_ctrl,
331                                    sizeof fadt->firmware_ctrl);
332 
333     fadt->dsdt = cpu_to_le32(dsdt);
334     /* DSDT address to be filled by Guest linker */
335     bios_linker_loader_add_pointer(linker, ACPI_BUILD_TABLE_FILE,
336                                    ACPI_BUILD_TABLE_FILE,
337                                    table_data, &fadt->dsdt,
338                                    sizeof fadt->dsdt);
339 
340     fadt_setup(fadt, pm);
341 
342     build_header(linker, table_data,
343                  (void *)fadt, "FACP", sizeof(*fadt), 1);
344 }
345 
346 static void
347 build_madt(GArray *table_data, GArray *linker, AcpiCpuInfo *cpu,
348            PcGuestInfo *guest_info)
349 {
350     int madt_start = table_data->len;
351 
352     AcpiMultipleApicTable *madt;
353     AcpiMadtIoApic *io_apic;
354     AcpiMadtIntsrcovr *intsrcovr;
355     AcpiMadtLocalNmi *local_nmi;
356     int i;
357 
358     madt = acpi_data_push(table_data, sizeof *madt);
359     madt->local_apic_address = cpu_to_le32(APIC_DEFAULT_ADDRESS);
360     madt->flags = cpu_to_le32(1);
361 
362     for (i = 0; i < guest_info->apic_id_limit; i++) {
363         AcpiMadtProcessorApic *apic = acpi_data_push(table_data, sizeof *apic);
364         apic->type = ACPI_APIC_PROCESSOR;
365         apic->length = sizeof(*apic);
366         apic->processor_id = i;
367         apic->local_apic_id = i;
368         if (test_bit(i, cpu->found_cpus)) {
369             apic->flags = cpu_to_le32(1);
370         } else {
371             apic->flags = cpu_to_le32(0);
372         }
373     }
374     io_apic = acpi_data_push(table_data, sizeof *io_apic);
375     io_apic->type = ACPI_APIC_IO;
376     io_apic->length = sizeof(*io_apic);
377 #define ACPI_BUILD_IOAPIC_ID 0x0
378     io_apic->io_apic_id = ACPI_BUILD_IOAPIC_ID;
379     io_apic->address = cpu_to_le32(IO_APIC_DEFAULT_ADDRESS);
380     io_apic->interrupt = cpu_to_le32(0);
381 
382     if (guest_info->apic_xrupt_override) {
383         intsrcovr = acpi_data_push(table_data, sizeof *intsrcovr);
384         intsrcovr->type   = ACPI_APIC_XRUPT_OVERRIDE;
385         intsrcovr->length = sizeof(*intsrcovr);
386         intsrcovr->source = 0;
387         intsrcovr->gsi    = cpu_to_le32(2);
388         intsrcovr->flags  = cpu_to_le16(0); /* conforms to bus specifications */
389     }
390     for (i = 1; i < 16; i++) {
391 #define ACPI_BUILD_PCI_IRQS ((1<<5) | (1<<9) | (1<<10) | (1<<11))
392         if (!(ACPI_BUILD_PCI_IRQS & (1 << i))) {
393             /* No need for a INT source override structure. */
394             continue;
395         }
396         intsrcovr = acpi_data_push(table_data, sizeof *intsrcovr);
397         intsrcovr->type   = ACPI_APIC_XRUPT_OVERRIDE;
398         intsrcovr->length = sizeof(*intsrcovr);
399         intsrcovr->source = i;
400         intsrcovr->gsi    = cpu_to_le32(i);
401         intsrcovr->flags  = cpu_to_le16(0xd); /* active high, level triggered */
402     }
403 
404     local_nmi = acpi_data_push(table_data, sizeof *local_nmi);
405     local_nmi->type         = ACPI_APIC_LOCAL_NMI;
406     local_nmi->length       = sizeof(*local_nmi);
407     local_nmi->processor_id = 0xff; /* all processors */
408     local_nmi->flags        = cpu_to_le16(0);
409     local_nmi->lint         = 1; /* ACPI_LINT1 */
410 
411     build_header(linker, table_data,
412                  (void *)(table_data->data + madt_start), "APIC",
413                  table_data->len - madt_start, 1);
414 }
415 
416 #include "hw/i386/ssdt-tpm.hex"
417 
418 /* Assign BSEL property to all buses.  In the future, this can be changed
419  * to only assign to buses that support hotplug.
420  */
421 static void *acpi_set_bsel(PCIBus *bus, void *opaque)
422 {
423     unsigned *bsel_alloc = opaque;
424     unsigned *bus_bsel;
425 
426     if (qbus_is_hotpluggable(BUS(bus))) {
427         bus_bsel = g_malloc(sizeof *bus_bsel);
428 
429         *bus_bsel = (*bsel_alloc)++;
430         object_property_add_uint32_ptr(OBJECT(bus), ACPI_PCIHP_PROP_BSEL,
431                                        bus_bsel, NULL);
432     }
433 
434     return bsel_alloc;
435 }
436 
437 static void acpi_set_pci_info(void)
438 {
439     PCIBus *bus = find_i440fx(); /* TODO: Q35 support */
440     unsigned bsel_alloc = 0;
441 
442     if (bus) {
443         /* Scan all PCI buses. Set property to enable acpi based hotplug. */
444         pci_for_each_bus_depth_first(bus, acpi_set_bsel, NULL, &bsel_alloc);
445     }
446 }
447 
448 static void build_append_pcihp_notify_entry(Aml *method, int slot)
449 {
450     Aml *if_ctx;
451     int32_t devfn = PCI_DEVFN(slot, 0);
452 
453     if_ctx = aml_if(aml_and(aml_arg(0), aml_int(0x1U << slot)));
454     aml_append(if_ctx, aml_notify(aml_name("S%.02X", devfn), aml_arg(1)));
455     aml_append(method, if_ctx);
456 }
457 
458 static void build_append_pci_bus_devices(Aml *parent_scope, PCIBus *bus,
459                                          bool pcihp_bridge_en)
460 {
461     Aml *dev, *notify_method, *method;
462     QObject *bsel;
463     PCIBus *sec;
464     int i;
465 
466     bsel = object_property_get_qobject(OBJECT(bus), ACPI_PCIHP_PROP_BSEL, NULL);
467     if (bsel) {
468         int64_t bsel_val = qint_get_int(qobject_to_qint(bsel));
469 
470         aml_append(parent_scope, aml_name_decl("BSEL", aml_int(bsel_val)));
471         notify_method = aml_method("DVNT", 2);
472     }
473 
474     for (i = 0; i < ARRAY_SIZE(bus->devices); i += PCI_FUNC_MAX) {
475         DeviceClass *dc;
476         PCIDeviceClass *pc;
477         PCIDevice *pdev = bus->devices[i];
478         int slot = PCI_SLOT(i);
479         bool hotplug_enabled_dev;
480         bool bridge_in_acpi;
481 
482         if (!pdev) {
483             if (bsel) { /* add hotplug slots for non present devices */
484                 dev = aml_device("S%.02X", PCI_DEVFN(slot, 0));
485                 aml_append(dev, aml_name_decl("_SUN", aml_int(slot)));
486                 aml_append(dev, aml_name_decl("_ADR", aml_int(slot << 16)));
487                 method = aml_method("_EJ0", 1);
488                 aml_append(method,
489                     aml_call2("PCEJ", aml_name("BSEL"), aml_name("_SUN"))
490                 );
491                 aml_append(dev, method);
492                 aml_append(parent_scope, dev);
493 
494                 build_append_pcihp_notify_entry(notify_method, slot);
495             }
496             continue;
497         }
498 
499         pc = PCI_DEVICE_GET_CLASS(pdev);
500         dc = DEVICE_GET_CLASS(pdev);
501 
502         /* When hotplug for bridges is enabled, bridges are
503          * described in ACPI separately (see build_pci_bus_end).
504          * In this case they aren't themselves hot-pluggable.
505          * Hotplugged bridges *are* hot-pluggable.
506          */
507         bridge_in_acpi = pc->is_bridge && pcihp_bridge_en &&
508             !DEVICE(pdev)->hotplugged;
509 
510         hotplug_enabled_dev = bsel && dc->hotpluggable && !bridge_in_acpi;
511 
512         if (pc->class_id == PCI_CLASS_BRIDGE_ISA) {
513             continue;
514         }
515 
516         /* start to compose PCI slot descriptor */
517         dev = aml_device("S%.02X", PCI_DEVFN(slot, 0));
518         aml_append(dev, aml_name_decl("_ADR", aml_int(slot << 16)));
519 
520         if (pc->class_id == PCI_CLASS_DISPLAY_VGA) {
521             /* add VGA specific AML methods */
522             int s3d;
523 
524             if (object_dynamic_cast(OBJECT(pdev), "qxl-vga")) {
525                 s3d = 3;
526             } else {
527                 s3d = 0;
528             }
529 
530             method = aml_method("_S1D", 0);
531             aml_append(method, aml_return(aml_int(0)));
532             aml_append(dev, method);
533 
534             method = aml_method("_S2D", 0);
535             aml_append(method, aml_return(aml_int(0)));
536             aml_append(dev, method);
537 
538             method = aml_method("_S3D", 0);
539             aml_append(method, aml_return(aml_int(s3d)));
540             aml_append(dev, method);
541         } else if (hotplug_enabled_dev) {
542             /* add _SUN/_EJ0 to make slot hotpluggable  */
543             aml_append(dev, aml_name_decl("_SUN", aml_int(slot)));
544 
545             method = aml_method("_EJ0", 1);
546             aml_append(method,
547                 aml_call2("PCEJ", aml_name("BSEL"), aml_name("_SUN"))
548             );
549             aml_append(dev, method);
550 
551             if (bsel) {
552                 build_append_pcihp_notify_entry(notify_method, slot);
553             }
554         } else if (bridge_in_acpi) {
555             /*
556              * device is coldplugged bridge,
557              * add child device descriptions into its scope
558              */
559             PCIBus *sec_bus = pci_bridge_get_sec_bus(PCI_BRIDGE(pdev));
560 
561             build_append_pci_bus_devices(dev, sec_bus, pcihp_bridge_en);
562         }
563         /* slot descriptor has been composed, add it into parent context */
564         aml_append(parent_scope, dev);
565     }
566 
567     if (bsel) {
568         aml_append(parent_scope, notify_method);
569     }
570 
571     /* Append PCNT method to notify about events on local and child buses.
572      * Add unconditionally for root since DSDT expects it.
573      */
574     method = aml_method("PCNT", 0);
575 
576     /* If bus supports hotplug select it and notify about local events */
577     if (bsel) {
578         int64_t bsel_val = qint_get_int(qobject_to_qint(bsel));
579         aml_append(method, aml_store(aml_int(bsel_val), aml_name("BNUM")));
580         aml_append(method,
581             aml_call2("DVNT", aml_name("PCIU"), aml_int(1) /* Device Check */)
582         );
583         aml_append(method,
584             aml_call2("DVNT", aml_name("PCID"), aml_int(3)/* Eject Request */)
585         );
586     }
587 
588     /* Notify about child bus events in any case */
589     if (pcihp_bridge_en) {
590         QLIST_FOREACH(sec, &bus->child, sibling) {
591             int32_t devfn = sec->parent_dev->devfn;
592 
593             aml_append(method, aml_name("^S%.02X.PCNT", devfn));
594         }
595     }
596     aml_append(parent_scope, method);
597 }
598 
599 static void
600 build_ssdt(GArray *table_data, GArray *linker,
601            AcpiCpuInfo *cpu, AcpiPmInfo *pm, AcpiMiscInfo *misc,
602            PcPciInfo *pci, PcGuestInfo *guest_info)
603 {
604     MachineState *machine = MACHINE(qdev_get_machine());
605     uint32_t nr_mem = machine->ram_slots;
606     unsigned acpi_cpus = guest_info->apic_id_limit;
607     Aml *ssdt, *sb_scope, *scope, *pkg, *dev, *method, *crs, *field, *ifctx;
608     int i;
609 
610     ssdt = init_aml_allocator();
611     /* The current AML generator can cover the APIC ID range [0..255],
612      * inclusive, for VCPU hotplug. */
613     QEMU_BUILD_BUG_ON(ACPI_CPU_HOTPLUG_ID_LIMIT > 256);
614     g_assert(acpi_cpus <= ACPI_CPU_HOTPLUG_ID_LIMIT);
615 
616     /* Reserve space for header */
617     acpi_data_push(ssdt->buf, sizeof(AcpiTableHeader));
618 
619     scope = aml_scope("\\_SB.PCI0");
620     /* build PCI0._CRS */
621     crs = aml_resource_template();
622     aml_append(crs,
623         aml_word_bus_number(AML_MIN_FIXED, AML_MAX_FIXED, AML_POS_DECODE,
624                             0x0000, 0x0000, 0x00FF, 0x0000, 0x0100));
625     aml_append(crs, aml_io(AML_DECODE16, 0x0CF8, 0x0CF8, 0x01, 0x08));
626 
627     aml_append(crs,
628         aml_word_io(AML_MIN_FIXED, AML_MAX_FIXED,
629                     AML_POS_DECODE, AML_ENTIRE_RANGE,
630                     0x0000, 0x0000, 0x0CF7, 0x0000, 0x0CF8));
631     aml_append(crs,
632         aml_word_io(AML_MIN_FIXED, AML_MAX_FIXED,
633                     AML_POS_DECODE, AML_ENTIRE_RANGE,
634                     0x0000, 0x0D00, 0xFFFF, 0x0000, 0xF300));
635     aml_append(crs,
636         aml_dword_memory(AML_POS_DECODE, AML_MIN_FIXED, AML_MAX_FIXED,
637                          AML_CACHEABLE, AML_READ_WRITE,
638                          0, 0x000A0000, 0x000BFFFF, 0, 0x00020000));
639     aml_append(crs,
640         aml_dword_memory(AML_POS_DECODE, AML_MIN_FIXED, AML_MAX_FIXED,
641                          AML_NON_CACHEABLE, AML_READ_WRITE,
642                          0, pci->w32.begin, pci->w32.end - 1, 0,
643                          pci->w32.end - pci->w32.begin));
644     if (pci->w64.begin) {
645         aml_append(crs,
646             aml_qword_memory(AML_POS_DECODE, AML_MIN_FIXED, AML_MAX_FIXED,
647                              AML_CACHEABLE, AML_READ_WRITE,
648                              0, pci->w64.begin, pci->w64.end - 1, 0,
649                              pci->w64.end - pci->w64.begin));
650     }
651     aml_append(scope, aml_name_decl("_CRS", crs));
652 
653     /* reserve GPE0 block resources */
654     dev = aml_device("GPE0");
655     aml_append(dev, aml_name_decl("_HID", aml_string("PNP0A06")));
656     aml_append(dev, aml_name_decl("_UID", aml_string("GPE0 resources")));
657     /* device present, functioning, decoding, not shown in UI */
658     aml_append(dev, aml_name_decl("_STA", aml_int(0xB)));
659     crs = aml_resource_template();
660     aml_append(crs,
661         aml_io(AML_DECODE16, pm->gpe0_blk, pm->gpe0_blk, 1, pm->gpe0_blk_len)
662     );
663     aml_append(dev, aml_name_decl("_CRS", crs));
664     aml_append(scope, dev);
665 
666     /* reserve PCIHP resources */
667     if (pm->pcihp_io_len) {
668         dev = aml_device("PHPR");
669         aml_append(dev, aml_name_decl("_HID", aml_string("PNP0A06")));
670         aml_append(dev,
671             aml_name_decl("_UID", aml_string("PCI Hotplug resources")));
672         /* device present, functioning, decoding, not shown in UI */
673         aml_append(dev, aml_name_decl("_STA", aml_int(0xB)));
674         crs = aml_resource_template();
675         aml_append(crs,
676             aml_io(AML_DECODE16, pm->pcihp_io_base, pm->pcihp_io_base, 1,
677                    pm->pcihp_io_len)
678         );
679         aml_append(dev, aml_name_decl("_CRS", crs));
680         aml_append(scope, dev);
681     }
682     aml_append(ssdt, scope);
683 
684     /*  create S3_ / S4_ / S5_ packages if necessary */
685     scope = aml_scope("\\");
686     if (!pm->s3_disabled) {
687         pkg = aml_package(4);
688         aml_append(pkg, aml_int(1)); /* PM1a_CNT.SLP_TYP */
689         aml_append(pkg, aml_int(1)); /* PM1b_CNT.SLP_TYP, FIXME: not impl. */
690         aml_append(pkg, aml_int(0)); /* reserved */
691         aml_append(pkg, aml_int(0)); /* reserved */
692         aml_append(scope, aml_name_decl("_S3", pkg));
693     }
694 
695     if (!pm->s4_disabled) {
696         pkg = aml_package(4);
697         aml_append(pkg, aml_int(pm->s4_val)); /* PM1a_CNT.SLP_TYP */
698         /* PM1b_CNT.SLP_TYP, FIXME: not impl. */
699         aml_append(pkg, aml_int(pm->s4_val));
700         aml_append(pkg, aml_int(0)); /* reserved */
701         aml_append(pkg, aml_int(0)); /* reserved */
702         aml_append(scope, aml_name_decl("_S4", pkg));
703     }
704 
705     pkg = aml_package(4);
706     aml_append(pkg, aml_int(0)); /* PM1a_CNT.SLP_TYP */
707     aml_append(pkg, aml_int(0)); /* PM1b_CNT.SLP_TYP not impl. */
708     aml_append(pkg, aml_int(0)); /* reserved */
709     aml_append(pkg, aml_int(0)); /* reserved */
710     aml_append(scope, aml_name_decl("_S5", pkg));
711     aml_append(ssdt, scope);
712 
713     if (misc->applesmc_io_base) {
714         scope = aml_scope("\\_SB.PCI0.ISA");
715         dev = aml_device("SMC");
716 
717         aml_append(dev, aml_name_decl("_HID", aml_eisaid("APP0001")));
718         /* device present, functioning, decoding, not shown in UI */
719         aml_append(dev, aml_name_decl("_STA", aml_int(0xB)));
720 
721         crs = aml_resource_template();
722         aml_append(crs,
723             aml_io(AML_DECODE16, misc->applesmc_io_base, misc->applesmc_io_base,
724                    0x01, APPLESMC_MAX_DATA_LENGTH)
725         );
726         aml_append(crs, aml_irq_no_flags(6));
727         aml_append(dev, aml_name_decl("_CRS", crs));
728 
729         aml_append(scope, dev);
730         aml_append(ssdt, scope);
731     }
732 
733     if (misc->pvpanic_port) {
734         scope = aml_scope("\\_SB.PCI0.ISA");
735 
736         dev = aml_device("PEVR");
737         aml_append(dev, aml_name_decl("_HID", aml_string("QEMU0001")));
738 
739         crs = aml_resource_template();
740         aml_append(crs,
741             aml_io(AML_DECODE16, misc->pvpanic_port, misc->pvpanic_port, 1, 1)
742         );
743         aml_append(dev, aml_name_decl("_CRS", crs));
744 
745         aml_append(dev, aml_operation_region("PEOR", AML_SYSTEM_IO,
746                                               misc->pvpanic_port, 1));
747         field = aml_field("PEOR", AML_BYTE_ACC, AML_PRESERVE);
748         aml_append(field, aml_named_field("PEPT", 8));
749         aml_append(dev, field);
750 
751         method = aml_method("RDPT", 0);
752         aml_append(method, aml_store(aml_name("PEPT"), aml_local(0)));
753         aml_append(method, aml_return(aml_local(0)));
754         aml_append(dev, method);
755 
756         method = aml_method("WRPT", 1);
757         aml_append(method, aml_store(aml_arg(0), aml_name("PEPT")));
758         aml_append(dev, method);
759 
760         aml_append(scope, dev);
761         aml_append(ssdt, scope);
762     }
763 
764     sb_scope = aml_scope("\\_SB");
765     {
766         /* create PCI0.PRES device and its _CRS to reserve CPU hotplug MMIO */
767         dev = aml_device("PCI0." stringify(CPU_HOTPLUG_RESOURCE_DEVICE));
768         aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0A06")));
769         aml_append(dev,
770             aml_name_decl("_UID", aml_string("CPU Hotplug resources"))
771         );
772         /* device present, functioning, decoding, not shown in UI */
773         aml_append(dev, aml_name_decl("_STA", aml_int(0xB)));
774         crs = aml_resource_template();
775         aml_append(crs,
776             aml_io(AML_DECODE16, pm->cpu_hp_io_base, pm->cpu_hp_io_base, 1,
777                    pm->cpu_hp_io_len)
778         );
779         aml_append(dev, aml_name_decl("_CRS", crs));
780         aml_append(sb_scope, dev);
781         /* declare CPU hotplug MMIO region and PRS field to access it */
782         aml_append(sb_scope, aml_operation_region(
783             "PRST", AML_SYSTEM_IO, pm->cpu_hp_io_base, pm->cpu_hp_io_len));
784         field = aml_field("PRST", AML_BYTE_ACC, AML_PRESERVE);
785         aml_append(field, aml_named_field("PRS", 256));
786         aml_append(sb_scope, field);
787 
788         /* build Processor object for each processor */
789         for (i = 0; i < acpi_cpus; i++) {
790             dev = aml_processor(i, 0, 0, "CP%.02X", i);
791 
792             method = aml_method("_MAT", 0);
793             aml_append(method, aml_return(aml_call1("CPMA", aml_int(i))));
794             aml_append(dev, method);
795 
796             method = aml_method("_STA", 0);
797             aml_append(method, aml_return(aml_call1("CPST", aml_int(i))));
798             aml_append(dev, method);
799 
800             method = aml_method("_EJ0", 1);
801             aml_append(method,
802                 aml_return(aml_call2("CPEJ", aml_int(i), aml_arg(0)))
803             );
804             aml_append(dev, method);
805 
806             aml_append(sb_scope, dev);
807         }
808 
809         /* build this code:
810          *   Method(NTFY, 2) {If (LEqual(Arg0, 0x00)) {Notify(CP00, Arg1)} ...}
811          */
812         /* Arg0 = Processor ID = APIC ID */
813         method = aml_method("NTFY", 2);
814         for (i = 0; i < acpi_cpus; i++) {
815             ifctx = aml_if(aml_equal(aml_arg(0), aml_int(i)));
816             aml_append(ifctx,
817                 aml_notify(aml_name("CP%.02X", i), aml_arg(1))
818             );
819             aml_append(method, ifctx);
820         }
821         aml_append(sb_scope, method);
822 
823         /* build "Name(CPON, Package() { One, One, ..., Zero, Zero, ... })"
824          *
825          * Note: The ability to create variable-sized packages was first
826          * introduced in ACPI 2.0. ACPI 1.0 only allowed fixed-size packages
827          * ith up to 255 elements. Windows guests up to win2k8 fail when
828          * VarPackageOp is used.
829          */
830         pkg = acpi_cpus <= 255 ? aml_package(acpi_cpus) :
831                                  aml_varpackage(acpi_cpus);
832 
833         for (i = 0; i < acpi_cpus; i++) {
834             uint8_t b = test_bit(i, cpu->found_cpus) ? 0x01 : 0x00;
835             aml_append(pkg, aml_int(b));
836         }
837         aml_append(sb_scope, aml_name_decl("CPON", pkg));
838 
839         /* build memory devices */
840         assert(nr_mem <= ACPI_MAX_RAM_SLOTS);
841         scope = aml_scope("\\_SB.PCI0." stringify(MEMORY_HOTPLUG_DEVICE));
842         aml_append(scope,
843             aml_name_decl(stringify(MEMORY_SLOTS_NUMBER), aml_int(nr_mem))
844         );
845 
846         crs = aml_resource_template();
847         aml_append(crs,
848             aml_io(AML_DECODE16, pm->mem_hp_io_base, pm->mem_hp_io_base, 0,
849                    pm->mem_hp_io_len)
850         );
851         aml_append(scope, aml_name_decl("_CRS", crs));
852 
853         aml_append(scope, aml_operation_region(
854             stringify(MEMORY_HOTPLUG_IO_REGION), AML_SYSTEM_IO,
855             pm->mem_hp_io_base, pm->mem_hp_io_len)
856         );
857 
858         field = aml_field(stringify(MEMORY_HOTPLUG_IO_REGION), AML_DWORD_ACC,
859                           AML_PRESERVE);
860         aml_append(field, /* read only */
861             aml_named_field(stringify(MEMORY_SLOT_ADDR_LOW), 32));
862         aml_append(field, /* read only */
863             aml_named_field(stringify(MEMORY_SLOT_ADDR_HIGH), 32));
864         aml_append(field, /* read only */
865             aml_named_field(stringify(MEMORY_SLOT_SIZE_LOW), 32));
866         aml_append(field, /* read only */
867             aml_named_field(stringify(MEMORY_SLOT_SIZE_HIGH), 32));
868         aml_append(field, /* read only */
869             aml_named_field(stringify(MEMORY_SLOT_PROXIMITY), 32));
870         aml_append(scope, field);
871 
872         field = aml_field(stringify(MEMORY_HOTPLUG_IO_REGION), AML_BYTE_ACC,
873                           AML_WRITE_AS_ZEROS);
874         aml_append(field, aml_reserved_field(160 /* bits, Offset(20) */));
875         aml_append(field, /* 1 if enabled, read only */
876             aml_named_field(stringify(MEMORY_SLOT_ENABLED), 1));
877         aml_append(field,
878             /*(read) 1 if has a insert event. (write) 1 to clear event */
879             aml_named_field(stringify(MEMORY_SLOT_INSERT_EVENT), 1));
880         aml_append(field,
881             /* (read) 1 if has a remove event. (write) 1 to clear event */
882             aml_named_field(stringify(MEMORY_SLOT_REMOVE_EVENT), 1));
883         aml_append(field,
884             /* initiates device eject, write only */
885             aml_named_field(stringify(MEMORY_SLOT_EJECT), 1));
886         aml_append(scope, field);
887 
888         field = aml_field(stringify(MEMORY_HOTPLUG_IO_REGION), AML_DWORD_ACC,
889                           AML_PRESERVE);
890         aml_append(field, /* DIMM selector, write only */
891             aml_named_field(stringify(MEMORY_SLOT_SLECTOR), 32));
892         aml_append(field, /* _OST event code, write only */
893             aml_named_field(stringify(MEMORY_SLOT_OST_EVENT), 32));
894         aml_append(field, /* _OST status code, write only */
895             aml_named_field(stringify(MEMORY_SLOT_OST_STATUS), 32));
896         aml_append(scope, field);
897 
898         aml_append(sb_scope, scope);
899 
900         for (i = 0; i < nr_mem; i++) {
901             #define BASEPATH "\\_SB.PCI0." stringify(MEMORY_HOTPLUG_DEVICE) "."
902             const char *s;
903 
904             dev = aml_device("MP%02X", i);
905             aml_append(dev, aml_name_decl("_UID", aml_string("0x%02X", i)));
906             aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0C80")));
907 
908             method = aml_method("_CRS", 0);
909             s = BASEPATH stringify(MEMORY_SLOT_CRS_METHOD);
910             aml_append(method, aml_return(aml_call1(s, aml_name("_UID"))));
911             aml_append(dev, method);
912 
913             method = aml_method("_STA", 0);
914             s = BASEPATH stringify(MEMORY_SLOT_STATUS_METHOD);
915             aml_append(method, aml_return(aml_call1(s, aml_name("_UID"))));
916             aml_append(dev, method);
917 
918             method = aml_method("_PXM", 0);
919             s = BASEPATH stringify(MEMORY_SLOT_PROXIMITY_METHOD);
920             aml_append(method, aml_return(aml_call1(s, aml_name("_UID"))));
921             aml_append(dev, method);
922 
923             method = aml_method("_OST", 3);
924             s = BASEPATH stringify(MEMORY_SLOT_OST_METHOD);
925             aml_append(method, aml_return(aml_call4(
926                 s, aml_name("_UID"), aml_arg(0), aml_arg(1), aml_arg(2)
927             )));
928             aml_append(dev, method);
929 
930             method = aml_method("_EJ0", 1);
931             s = BASEPATH stringify(MEMORY_SLOT_EJECT_METHOD);
932             aml_append(method, aml_return(aml_call2(
933                        s, aml_name("_UID"), aml_arg(0))));
934             aml_append(dev, method);
935 
936             aml_append(sb_scope, dev);
937         }
938 
939         /* build Method(MEMORY_SLOT_NOTIFY_METHOD, 2) {
940          *     If (LEqual(Arg0, 0x00)) {Notify(MP00, Arg1)} ... }
941          */
942         method = aml_method(stringify(MEMORY_SLOT_NOTIFY_METHOD), 2);
943         for (i = 0; i < nr_mem; i++) {
944             ifctx = aml_if(aml_equal(aml_arg(0), aml_int(i)));
945             aml_append(ifctx,
946                 aml_notify(aml_name("MP%.02X", i), aml_arg(1))
947             );
948             aml_append(method, ifctx);
949         }
950         aml_append(sb_scope, method);
951 
952         {
953             Object *pci_host;
954             PCIBus *bus = NULL;
955             bool ambiguous;
956 
957             pci_host = object_resolve_path_type("", TYPE_PCI_HOST_BRIDGE, &ambiguous);
958             if (!ambiguous && pci_host) {
959                 bus = PCI_HOST_BRIDGE(pci_host)->bus;
960             }
961 
962             if (bus) {
963                 Aml *scope = aml_scope("PCI0");
964                 /* Scan all PCI buses. Generate tables to support hotplug. */
965                 build_append_pci_bus_devices(scope, bus, pm->pcihp_bridge_en);
966                 aml_append(sb_scope, scope);
967             }
968         }
969         aml_append(ssdt, sb_scope);
970     }
971 
972     /* copy AML table into ACPI tables blob and patch header there */
973     g_array_append_vals(table_data, ssdt->buf->data, ssdt->buf->len);
974     build_header(linker, table_data,
975         (void *)(table_data->data + table_data->len - ssdt->buf->len),
976         "SSDT", ssdt->buf->len, 1);
977     free_aml_allocator();
978 }
979 
980 static void
981 build_hpet(GArray *table_data, GArray *linker)
982 {
983     Acpi20Hpet *hpet;
984 
985     hpet = acpi_data_push(table_data, sizeof(*hpet));
986     /* Note timer_block_id value must be kept in sync with value advertised by
987      * emulated hpet
988      */
989     hpet->timer_block_id = cpu_to_le32(0x8086a201);
990     hpet->addr.address = cpu_to_le64(HPET_BASE);
991     build_header(linker, table_data,
992                  (void *)hpet, "HPET", sizeof(*hpet), 1);
993 }
994 
995 static void
996 build_tpm_tcpa(GArray *table_data, GArray *linker, GArray *tcpalog)
997 {
998     Acpi20Tcpa *tcpa = acpi_data_push(table_data, sizeof *tcpa);
999     uint64_t log_area_start_address = acpi_data_len(tcpalog);
1000 
1001     tcpa->platform_class = cpu_to_le16(TPM_TCPA_ACPI_CLASS_CLIENT);
1002     tcpa->log_area_minimum_length = cpu_to_le32(TPM_LOG_AREA_MINIMUM_SIZE);
1003     tcpa->log_area_start_address = cpu_to_le64(log_area_start_address);
1004 
1005     bios_linker_loader_alloc(linker, ACPI_BUILD_TPMLOG_FILE, 1,
1006                              false /* high memory */);
1007 
1008     /* log area start address to be filled by Guest linker */
1009     bios_linker_loader_add_pointer(linker, ACPI_BUILD_TABLE_FILE,
1010                                    ACPI_BUILD_TPMLOG_FILE,
1011                                    table_data, &tcpa->log_area_start_address,
1012                                    sizeof(tcpa->log_area_start_address));
1013 
1014     build_header(linker, table_data,
1015                  (void *)tcpa, "TCPA", sizeof(*tcpa), 2);
1016 
1017     acpi_data_push(tcpalog, TPM_LOG_AREA_MINIMUM_SIZE);
1018 }
1019 
1020 static void
1021 build_tpm_ssdt(GArray *table_data, GArray *linker)
1022 {
1023     void *tpm_ptr;
1024 
1025     tpm_ptr = acpi_data_push(table_data, sizeof(ssdt_tpm_aml));
1026     memcpy(tpm_ptr, ssdt_tpm_aml, sizeof(ssdt_tpm_aml));
1027 }
1028 
1029 typedef enum {
1030     MEM_AFFINITY_NOFLAGS      = 0,
1031     MEM_AFFINITY_ENABLED      = (1 << 0),
1032     MEM_AFFINITY_HOTPLUGGABLE = (1 << 1),
1033     MEM_AFFINITY_NON_VOLATILE = (1 << 2),
1034 } MemoryAffinityFlags;
1035 
1036 static void
1037 acpi_build_srat_memory(AcpiSratMemoryAffinity *numamem, uint64_t base,
1038                        uint64_t len, int node, MemoryAffinityFlags flags)
1039 {
1040     numamem->type = ACPI_SRAT_MEMORY;
1041     numamem->length = sizeof(*numamem);
1042     memset(numamem->proximity, 0, 4);
1043     numamem->proximity[0] = node;
1044     numamem->flags = cpu_to_le32(flags);
1045     numamem->base_addr = cpu_to_le64(base);
1046     numamem->range_length = cpu_to_le64(len);
1047 }
1048 
1049 static void
1050 build_srat(GArray *table_data, GArray *linker, PcGuestInfo *guest_info)
1051 {
1052     AcpiSystemResourceAffinityTable *srat;
1053     AcpiSratProcessorAffinity *core;
1054     AcpiSratMemoryAffinity *numamem;
1055 
1056     int i;
1057     uint64_t curnode;
1058     int srat_start, numa_start, slots;
1059     uint64_t mem_len, mem_base, next_base;
1060     PCMachineState *pcms = PC_MACHINE(qdev_get_machine());
1061     ram_addr_t hotplugabble_address_space_size =
1062         object_property_get_int(OBJECT(pcms), PC_MACHINE_MEMHP_REGION_SIZE,
1063                                 NULL);
1064 
1065     srat_start = table_data->len;
1066 
1067     srat = acpi_data_push(table_data, sizeof *srat);
1068     srat->reserved1 = cpu_to_le32(1);
1069     core = (void *)(srat + 1);
1070 
1071     for (i = 0; i < guest_info->apic_id_limit; ++i) {
1072         core = acpi_data_push(table_data, sizeof *core);
1073         core->type = ACPI_SRAT_PROCESSOR;
1074         core->length = sizeof(*core);
1075         core->local_apic_id = i;
1076         curnode = guest_info->node_cpu[i];
1077         core->proximity_lo = curnode;
1078         memset(core->proximity_hi, 0, 3);
1079         core->local_sapic_eid = 0;
1080         core->flags = cpu_to_le32(1);
1081     }
1082 
1083 
1084     /* the memory map is a bit tricky, it contains at least one hole
1085      * from 640k-1M and possibly another one from 3.5G-4G.
1086      */
1087     next_base = 0;
1088     numa_start = table_data->len;
1089 
1090     numamem = acpi_data_push(table_data, sizeof *numamem);
1091     acpi_build_srat_memory(numamem, 0, 640*1024, 0, MEM_AFFINITY_ENABLED);
1092     next_base = 1024 * 1024;
1093     for (i = 1; i < guest_info->numa_nodes + 1; ++i) {
1094         mem_base = next_base;
1095         mem_len = guest_info->node_mem[i - 1];
1096         if (i == 1) {
1097             mem_len -= 1024 * 1024;
1098         }
1099         next_base = mem_base + mem_len;
1100 
1101         /* Cut out the ACPI_PCI hole */
1102         if (mem_base <= guest_info->ram_size_below_4g &&
1103             next_base > guest_info->ram_size_below_4g) {
1104             mem_len -= next_base - guest_info->ram_size_below_4g;
1105             if (mem_len > 0) {
1106                 numamem = acpi_data_push(table_data, sizeof *numamem);
1107                 acpi_build_srat_memory(numamem, mem_base, mem_len, i - 1,
1108                                        MEM_AFFINITY_ENABLED);
1109             }
1110             mem_base = 1ULL << 32;
1111             mem_len = next_base - guest_info->ram_size_below_4g;
1112             next_base += (1ULL << 32) - guest_info->ram_size_below_4g;
1113         }
1114         numamem = acpi_data_push(table_data, sizeof *numamem);
1115         acpi_build_srat_memory(numamem, mem_base, mem_len, i - 1,
1116                                MEM_AFFINITY_ENABLED);
1117     }
1118     slots = (table_data->len - numa_start) / sizeof *numamem;
1119     for (; slots < guest_info->numa_nodes + 2; slots++) {
1120         numamem = acpi_data_push(table_data, sizeof *numamem);
1121         acpi_build_srat_memory(numamem, 0, 0, 0, MEM_AFFINITY_NOFLAGS);
1122     }
1123 
1124     /*
1125      * Entry is required for Windows to enable memory hotplug in OS.
1126      * Memory devices may override proximity set by this entry,
1127      * providing _PXM method if necessary.
1128      */
1129     if (hotplugabble_address_space_size) {
1130         numamem = acpi_data_push(table_data, sizeof *numamem);
1131         acpi_build_srat_memory(numamem, pcms->hotplug_memory_base,
1132                                hotplugabble_address_space_size, 0,
1133                                MEM_AFFINITY_HOTPLUGGABLE |
1134                                MEM_AFFINITY_ENABLED);
1135     }
1136 
1137     build_header(linker, table_data,
1138                  (void *)(table_data->data + srat_start),
1139                  "SRAT",
1140                  table_data->len - srat_start, 1);
1141 }
1142 
1143 static void
1144 build_mcfg_q35(GArray *table_data, GArray *linker, AcpiMcfgInfo *info)
1145 {
1146     AcpiTableMcfg *mcfg;
1147     const char *sig;
1148     int len = sizeof(*mcfg) + 1 * sizeof(mcfg->allocation[0]);
1149 
1150     mcfg = acpi_data_push(table_data, len);
1151     mcfg->allocation[0].address = cpu_to_le64(info->mcfg_base);
1152     /* Only a single allocation so no need to play with segments */
1153     mcfg->allocation[0].pci_segment = cpu_to_le16(0);
1154     mcfg->allocation[0].start_bus_number = 0;
1155     mcfg->allocation[0].end_bus_number = PCIE_MMCFG_BUS(info->mcfg_size - 1);
1156 
1157     /* MCFG is used for ECAM which can be enabled or disabled by guest.
1158      * To avoid table size changes (which create migration issues),
1159      * always create the table even if there are no allocations,
1160      * but set the signature to a reserved value in this case.
1161      * ACPI spec requires OSPMs to ignore such tables.
1162      */
1163     if (info->mcfg_base == PCIE_BASE_ADDR_UNMAPPED) {
1164         /* Reserved signature: ignored by OSPM */
1165         sig = "QEMU";
1166     } else {
1167         sig = "MCFG";
1168     }
1169     build_header(linker, table_data, (void *)mcfg, sig, len, 1);
1170 }
1171 
1172 static void
1173 build_dmar_q35(GArray *table_data, GArray *linker)
1174 {
1175     int dmar_start = table_data->len;
1176 
1177     AcpiTableDmar *dmar;
1178     AcpiDmarHardwareUnit *drhd;
1179 
1180     dmar = acpi_data_push(table_data, sizeof(*dmar));
1181     dmar->host_address_width = VTD_HOST_ADDRESS_WIDTH - 1;
1182     dmar->flags = 0;    /* No intr_remap for now */
1183 
1184     /* DMAR Remapping Hardware Unit Definition structure */
1185     drhd = acpi_data_push(table_data, sizeof(*drhd));
1186     drhd->type = cpu_to_le16(ACPI_DMAR_TYPE_HARDWARE_UNIT);
1187     drhd->length = cpu_to_le16(sizeof(*drhd));   /* No device scope now */
1188     drhd->flags = ACPI_DMAR_INCLUDE_PCI_ALL;
1189     drhd->pci_segment = cpu_to_le16(0);
1190     drhd->address = cpu_to_le64(Q35_HOST_BRIDGE_IOMMU_ADDR);
1191 
1192     build_header(linker, table_data, (void *)(table_data->data + dmar_start),
1193                  "DMAR", table_data->len - dmar_start, 1);
1194 }
1195 
1196 static void
1197 build_dsdt(GArray *table_data, GArray *linker, AcpiMiscInfo *misc)
1198 {
1199     AcpiTableHeader *dsdt;
1200 
1201     assert(misc->dsdt_code && misc->dsdt_size);
1202 
1203     dsdt = acpi_data_push(table_data, misc->dsdt_size);
1204     memcpy(dsdt, misc->dsdt_code, misc->dsdt_size);
1205 
1206     memset(dsdt, 0, sizeof *dsdt);
1207     build_header(linker, table_data, dsdt, "DSDT",
1208                  misc->dsdt_size, 1);
1209 }
1210 
1211 /* Build final rsdt table */
1212 static void
1213 build_rsdt(GArray *table_data, GArray *linker, GArray *table_offsets)
1214 {
1215     AcpiRsdtDescriptorRev1 *rsdt;
1216     size_t rsdt_len;
1217     int i;
1218 
1219     rsdt_len = sizeof(*rsdt) + sizeof(uint32_t) * table_offsets->len;
1220     rsdt = acpi_data_push(table_data, rsdt_len);
1221     memcpy(rsdt->table_offset_entry, table_offsets->data,
1222            sizeof(uint32_t) * table_offsets->len);
1223     for (i = 0; i < table_offsets->len; ++i) {
1224         /* rsdt->table_offset_entry to be filled by Guest linker */
1225         bios_linker_loader_add_pointer(linker,
1226                                        ACPI_BUILD_TABLE_FILE,
1227                                        ACPI_BUILD_TABLE_FILE,
1228                                        table_data, &rsdt->table_offset_entry[i],
1229                                        sizeof(uint32_t));
1230     }
1231     build_header(linker, table_data,
1232                  (void *)rsdt, "RSDT", rsdt_len, 1);
1233 }
1234 
1235 static GArray *
1236 build_rsdp(GArray *rsdp_table, GArray *linker, unsigned rsdt)
1237 {
1238     AcpiRsdpDescriptor *rsdp = acpi_data_push(rsdp_table, sizeof *rsdp);
1239 
1240     bios_linker_loader_alloc(linker, ACPI_BUILD_RSDP_FILE, 16,
1241                              true /* fseg memory */);
1242 
1243     memcpy(&rsdp->signature, "RSD PTR ", 8);
1244     memcpy(rsdp->oem_id, ACPI_BUILD_APPNAME6, 6);
1245     rsdp->rsdt_physical_address = cpu_to_le32(rsdt);
1246     /* Address to be filled by Guest linker */
1247     bios_linker_loader_add_pointer(linker, ACPI_BUILD_RSDP_FILE,
1248                                    ACPI_BUILD_TABLE_FILE,
1249                                    rsdp_table, &rsdp->rsdt_physical_address,
1250                                    sizeof rsdp->rsdt_physical_address);
1251     rsdp->checksum = 0;
1252     /* Checksum to be filled by Guest linker */
1253     bios_linker_loader_add_checksum(linker, ACPI_BUILD_RSDP_FILE,
1254                                     rsdp, rsdp, sizeof *rsdp, &rsdp->checksum);
1255 
1256     return rsdp_table;
1257 }
1258 
1259 typedef
1260 struct AcpiBuildState {
1261     /* Copy of table in RAM (for patching). */
1262     MemoryRegion *table_mr;
1263     /* Is table patched? */
1264     uint8_t patched;
1265     PcGuestInfo *guest_info;
1266     void *rsdp;
1267     MemoryRegion *rsdp_mr;
1268     MemoryRegion *linker_mr;
1269 } AcpiBuildState;
1270 
1271 static bool acpi_get_mcfg(AcpiMcfgInfo *mcfg)
1272 {
1273     Object *pci_host;
1274     QObject *o;
1275     bool ambiguous;
1276 
1277     pci_host = object_resolve_path_type("", TYPE_PCI_HOST_BRIDGE, &ambiguous);
1278     g_assert(!ambiguous);
1279     g_assert(pci_host);
1280 
1281     o = object_property_get_qobject(pci_host, PCIE_HOST_MCFG_BASE, NULL);
1282     if (!o) {
1283         return false;
1284     }
1285     mcfg->mcfg_base = qint_get_int(qobject_to_qint(o));
1286     qobject_decref(o);
1287 
1288     o = object_property_get_qobject(pci_host, PCIE_HOST_MCFG_SIZE, NULL);
1289     assert(o);
1290     mcfg->mcfg_size = qint_get_int(qobject_to_qint(o));
1291     qobject_decref(o);
1292     return true;
1293 }
1294 
1295 static bool acpi_has_iommu(void)
1296 {
1297     bool ambiguous;
1298     Object *intel_iommu;
1299 
1300     intel_iommu = object_resolve_path_type("", TYPE_INTEL_IOMMU_DEVICE,
1301                                            &ambiguous);
1302     return intel_iommu && !ambiguous;
1303 }
1304 
1305 static
1306 void acpi_build(PcGuestInfo *guest_info, AcpiBuildTables *tables)
1307 {
1308     GArray *table_offsets;
1309     unsigned facs, ssdt, dsdt, rsdt;
1310     AcpiCpuInfo cpu;
1311     AcpiPmInfo pm;
1312     AcpiMiscInfo misc;
1313     AcpiMcfgInfo mcfg;
1314     PcPciInfo pci;
1315     uint8_t *u;
1316     size_t aml_len = 0;
1317     GArray *tables_blob = tables->table_data;
1318 
1319     acpi_get_cpu_info(&cpu);
1320     acpi_get_pm_info(&pm);
1321     acpi_get_dsdt(&misc);
1322     acpi_get_misc_info(&misc);
1323     acpi_get_pci_info(&pci);
1324 
1325     table_offsets = g_array_new(false, true /* clear */,
1326                                         sizeof(uint32_t));
1327     ACPI_BUILD_DPRINTF("init ACPI tables\n");
1328 
1329     bios_linker_loader_alloc(tables->linker, ACPI_BUILD_TABLE_FILE,
1330                              64 /* Ensure FACS is aligned */,
1331                              false /* high memory */);
1332 
1333     /*
1334      * FACS is pointed to by FADT.
1335      * We place it first since it's the only table that has alignment
1336      * requirements.
1337      */
1338     facs = tables_blob->len;
1339     build_facs(tables_blob, tables->linker, guest_info);
1340 
1341     /* DSDT is pointed to by FADT */
1342     dsdt = tables_blob->len;
1343     build_dsdt(tables_blob, tables->linker, &misc);
1344 
1345     /* Count the size of the DSDT and SSDT, we will need it for legacy
1346      * sizing of ACPI tables.
1347      */
1348     aml_len += tables_blob->len - dsdt;
1349 
1350     /* ACPI tables pointed to by RSDT */
1351     acpi_add_table(table_offsets, tables_blob);
1352     build_fadt(tables_blob, tables->linker, &pm, facs, dsdt);
1353 
1354     ssdt = tables_blob->len;
1355     acpi_add_table(table_offsets, tables_blob);
1356     build_ssdt(tables_blob, tables->linker, &cpu, &pm, &misc, &pci,
1357                guest_info);
1358     aml_len += tables_blob->len - ssdt;
1359 
1360     acpi_add_table(table_offsets, tables_blob);
1361     build_madt(tables_blob, tables->linker, &cpu, guest_info);
1362 
1363     if (misc.has_hpet) {
1364         acpi_add_table(table_offsets, tables_blob);
1365         build_hpet(tables_blob, tables->linker);
1366     }
1367     if (misc.has_tpm) {
1368         acpi_add_table(table_offsets, tables_blob);
1369         build_tpm_tcpa(tables_blob, tables->linker, tables->tcpalog);
1370 
1371         acpi_add_table(table_offsets, tables_blob);
1372         build_tpm_ssdt(tables_blob, tables->linker);
1373     }
1374     if (guest_info->numa_nodes) {
1375         acpi_add_table(table_offsets, tables_blob);
1376         build_srat(tables_blob, tables->linker, guest_info);
1377     }
1378     if (acpi_get_mcfg(&mcfg)) {
1379         acpi_add_table(table_offsets, tables_blob);
1380         build_mcfg_q35(tables_blob, tables->linker, &mcfg);
1381     }
1382     if (acpi_has_iommu()) {
1383         acpi_add_table(table_offsets, tables_blob);
1384         build_dmar_q35(tables_blob, tables->linker);
1385     }
1386 
1387     /* Add tables supplied by user (if any) */
1388     for (u = acpi_table_first(); u; u = acpi_table_next(u)) {
1389         unsigned len = acpi_table_len(u);
1390 
1391         acpi_add_table(table_offsets, tables_blob);
1392         g_array_append_vals(tables_blob, u, len);
1393     }
1394 
1395     /* RSDT is pointed to by RSDP */
1396     rsdt = tables_blob->len;
1397     build_rsdt(tables_blob, tables->linker, table_offsets);
1398 
1399     /* RSDP is in FSEG memory, so allocate it separately */
1400     build_rsdp(tables->rsdp, tables->linker, rsdt);
1401 
1402     /* We'll expose it all to Guest so we want to reduce
1403      * chance of size changes.
1404      *
1405      * We used to align the tables to 4k, but of course this would
1406      * too simple to be enough.  4k turned out to be too small an
1407      * alignment very soon, and in fact it is almost impossible to
1408      * keep the table size stable for all (max_cpus, max_memory_slots)
1409      * combinations.  So the table size is always 64k for pc-i440fx-2.1
1410      * and we give an error if the table grows beyond that limit.
1411      *
1412      * We still have the problem of migrating from "-M pc-i440fx-2.0".  For
1413      * that, we exploit the fact that QEMU 2.1 generates _smaller_ tables
1414      * than 2.0 and we can always pad the smaller tables with zeros.  We can
1415      * then use the exact size of the 2.0 tables.
1416      *
1417      * All this is for PIIX4, since QEMU 2.0 didn't support Q35 migration.
1418      */
1419     if (guest_info->legacy_acpi_table_size) {
1420         /* Subtracting aml_len gives the size of fixed tables.  Then add the
1421          * size of the PIIX4 DSDT/SSDT in QEMU 2.0.
1422          */
1423         int legacy_aml_len =
1424             guest_info->legacy_acpi_table_size +
1425             ACPI_BUILD_LEGACY_CPU_AML_SIZE * max_cpus;
1426         int legacy_table_size =
1427             ROUND_UP(tables_blob->len - aml_len + legacy_aml_len,
1428                      ACPI_BUILD_ALIGN_SIZE);
1429         if (tables_blob->len > legacy_table_size) {
1430             /* Should happen only with PCI bridges and -M pc-i440fx-2.0.  */
1431             error_report("Warning: migration may not work.");
1432         }
1433         g_array_set_size(tables_blob, legacy_table_size);
1434     } else {
1435         /* Make sure we have a buffer in case we need to resize the tables. */
1436         if (tables_blob->len > ACPI_BUILD_TABLE_SIZE / 2) {
1437             /* As of QEMU 2.1, this fires with 160 VCPUs and 255 memory slots.  */
1438             error_report("Warning: ACPI tables are larger than 64k.");
1439             error_report("Warning: migration may not work.");
1440             error_report("Warning: please remove CPUs, NUMA nodes, "
1441                          "memory slots or PCI bridges.");
1442         }
1443         acpi_align_size(tables_blob, ACPI_BUILD_TABLE_SIZE);
1444     }
1445 
1446     acpi_align_size(tables->linker, ACPI_BUILD_ALIGN_SIZE);
1447 
1448     /* Cleanup memory that's no longer used. */
1449     g_array_free(table_offsets, true);
1450 }
1451 
1452 static void acpi_ram_update(MemoryRegion *mr, GArray *data)
1453 {
1454     uint32_t size = acpi_data_len(data);
1455 
1456     /* Make sure RAM size is correct - in case it got changed e.g. by migration */
1457     memory_region_ram_resize(mr, size, &error_abort);
1458 
1459     memcpy(memory_region_get_ram_ptr(mr), data->data, size);
1460     memory_region_set_dirty(mr, 0, size);
1461 }
1462 
1463 static void acpi_build_update(void *build_opaque, uint32_t offset)
1464 {
1465     AcpiBuildState *build_state = build_opaque;
1466     AcpiBuildTables tables;
1467 
1468     /* No state to update or already patched? Nothing to do. */
1469     if (!build_state || build_state->patched) {
1470         return;
1471     }
1472     build_state->patched = 1;
1473 
1474     acpi_build_tables_init(&tables);
1475 
1476     acpi_build(build_state->guest_info, &tables);
1477 
1478     acpi_ram_update(build_state->table_mr, tables.table_data);
1479 
1480     if (build_state->rsdp) {
1481         memcpy(build_state->rsdp, tables.rsdp->data, acpi_data_len(tables.rsdp));
1482     } else {
1483         acpi_ram_update(build_state->rsdp_mr, tables.rsdp);
1484     }
1485 
1486     acpi_ram_update(build_state->linker_mr, tables.linker);
1487     acpi_build_tables_cleanup(&tables, true);
1488 }
1489 
1490 static void acpi_build_reset(void *build_opaque)
1491 {
1492     AcpiBuildState *build_state = build_opaque;
1493     build_state->patched = 0;
1494 }
1495 
1496 static MemoryRegion *acpi_add_rom_blob(AcpiBuildState *build_state,
1497                                        GArray *blob, const char *name,
1498                                        uint64_t max_size)
1499 {
1500     return rom_add_blob(name, blob->data, acpi_data_len(blob), max_size, -1,
1501                         name, acpi_build_update, build_state);
1502 }
1503 
1504 static const VMStateDescription vmstate_acpi_build = {
1505     .name = "acpi_build",
1506     .version_id = 1,
1507     .minimum_version_id = 1,
1508     .fields = (VMStateField[]) {
1509         VMSTATE_UINT8(patched, AcpiBuildState),
1510         VMSTATE_END_OF_LIST()
1511     },
1512 };
1513 
1514 void acpi_setup(PcGuestInfo *guest_info)
1515 {
1516     AcpiBuildTables tables;
1517     AcpiBuildState *build_state;
1518 
1519     if (!guest_info->fw_cfg) {
1520         ACPI_BUILD_DPRINTF("No fw cfg. Bailing out.\n");
1521         return;
1522     }
1523 
1524     if (!guest_info->has_acpi_build) {
1525         ACPI_BUILD_DPRINTF("ACPI build disabled. Bailing out.\n");
1526         return;
1527     }
1528 
1529     if (!acpi_enabled) {
1530         ACPI_BUILD_DPRINTF("ACPI disabled. Bailing out.\n");
1531         return;
1532     }
1533 
1534     build_state = g_malloc0(sizeof *build_state);
1535 
1536     build_state->guest_info = guest_info;
1537 
1538     acpi_set_pci_info();
1539 
1540     acpi_build_tables_init(&tables);
1541     acpi_build(build_state->guest_info, &tables);
1542 
1543     /* Now expose it all to Guest */
1544     build_state->table_mr = acpi_add_rom_blob(build_state, tables.table_data,
1545                                                ACPI_BUILD_TABLE_FILE,
1546                                                ACPI_BUILD_TABLE_MAX_SIZE);
1547     assert(build_state->table_mr != NULL);
1548 
1549     build_state->linker_mr =
1550         acpi_add_rom_blob(build_state, tables.linker, "etc/table-loader", 0);
1551 
1552     fw_cfg_add_file(guest_info->fw_cfg, ACPI_BUILD_TPMLOG_FILE,
1553                     tables.tcpalog->data, acpi_data_len(tables.tcpalog));
1554 
1555     if (!guest_info->rsdp_in_ram) {
1556         /*
1557          * Keep for compatibility with old machine types.
1558          * Though RSDP is small, its contents isn't immutable, so
1559          * we'll update it along with the rest of tables on guest access.
1560          */
1561         uint32_t rsdp_size = acpi_data_len(tables.rsdp);
1562 
1563         build_state->rsdp = g_memdup(tables.rsdp->data, rsdp_size);
1564         fw_cfg_add_file_callback(guest_info->fw_cfg, ACPI_BUILD_RSDP_FILE,
1565                                  acpi_build_update, build_state,
1566                                  build_state->rsdp, rsdp_size);
1567         build_state->rsdp_mr = NULL;
1568     } else {
1569         build_state->rsdp = NULL;
1570         build_state->rsdp_mr = acpi_add_rom_blob(build_state, tables.rsdp,
1571                                                   ACPI_BUILD_RSDP_FILE, 0);
1572     }
1573 
1574     qemu_register_reset(acpi_build_reset, build_state);
1575     acpi_build_reset(build_state);
1576     vmstate_register(NULL, 0, &vmstate_acpi_build, build_state);
1577 
1578     /* Cleanup tables but don't free the memory: we track it
1579      * in build_state.
1580      */
1581     acpi_build_tables_cleanup(&tables, false);
1582 }
1583