xref: /openbmc/qemu/hw/i386/acpi-build.c (revision c5dee204b0ae45c92b1a9640055ec068c6748b3c)
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 "qemu/osdep.h"
24 #include "qapi/error.h"
25 #include "qobject/qnum.h"
26 #include "acpi-build.h"
27 #include "acpi-common.h"
28 #include "qemu/bitmap.h"
29 #include "qemu/error-report.h"
30 #include "hw/pci/pci_bridge.h"
31 #include "hw/cxl/cxl.h"
32 #include "hw/core/cpu.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/acpi/cpu.h"
38 #include "hw/nvram/fw_cfg.h"
39 #include "hw/acpi/bios-linker-loader.h"
40 #include "hw/acpi/acpi_aml_interface.h"
41 #include "hw/input/i8042.h"
42 #include "hw/acpi/memory_hotplug.h"
43 #include "system/tpm.h"
44 #include "hw/acpi/tpm.h"
45 #include "hw/acpi/vmgenid.h"
46 #include "hw/acpi/vmclock.h"
47 #include "hw/acpi/erst.h"
48 #include "hw/acpi/piix4.h"
49 #include "system/tpm_backend.h"
50 #include "hw/rtc/mc146818rtc_regs.h"
51 #include "migration/vmstate.h"
52 #include "hw/mem/memory-device.h"
53 #include "hw/mem/nvdimm.h"
54 #include "system/numa.h"
55 #include "system/reset.h"
56 #include "hw/hyperv/vmbus-bridge.h"
57 
58 /* Supported chipsets: */
59 #include "hw/southbridge/ich9.h"
60 #include "hw/acpi/pcihp.h"
61 #include "hw/i386/fw_cfg.h"
62 #include "hw/i386/pc.h"
63 #include "hw/pci/pci_bus.h"
64 #include "hw/pci-host/i440fx.h"
65 #include "hw/pci-host/q35.h"
66 #include "hw/i386/x86-iommu.h"
67 
68 #include "hw/acpi/aml-build.h"
69 #include "hw/acpi/utils.h"
70 #include "hw/acpi/pci.h"
71 #include "hw/acpi/cxl.h"
72 
73 #include "qom/qom-qobject.h"
74 #include "hw/i386/amd_iommu.h"
75 #include "hw/i386/intel_iommu.h"
76 #include "hw/virtio/virtio-iommu.h"
77 
78 #include "hw/acpi/hmat.h"
79 #include "hw/acpi/viot.h"
80 
81 #include CONFIG_DEVICES
82 
83 /* These are used to size the ACPI tables for -M pc-i440fx-1.7 and
84  * -M pc-i440fx-2.0.  Even if the actual amount of AML generated grows
85  * a little bit, there should be plenty of free space since the DSDT
86  * shrunk by ~1.5k between QEMU 2.0 and QEMU 2.1.
87  */
88 #define ACPI_BUILD_ALIGN_SIZE             0x1000
89 
90 #define ACPI_BUILD_TABLE_SIZE             0x20000
91 
92 /* #define DEBUG_ACPI_BUILD */
93 #ifdef DEBUG_ACPI_BUILD
94 #define ACPI_BUILD_DPRINTF(fmt, ...)        \
95     do {printf("ACPI_BUILD: " fmt, ## __VA_ARGS__); } while (0)
96 #else
97 #define ACPI_BUILD_DPRINTF(fmt, ...)
98 #endif
99 
100 typedef struct AcpiPmInfo {
101     bool s3_disabled;
102     bool s4_disabled;
103     bool pcihp_bridge_en;
104     bool smi_on_cpuhp;
105     bool smi_on_cpu_unplug;
106     bool pcihp_root_en;
107     uint8_t s4_val;
108     AcpiFadtData fadt;
109     uint16_t cpu_hp_io_base;
110     uint16_t pcihp_io_base;
111     uint16_t pcihp_io_len;
112 } AcpiPmInfo;
113 
114 typedef struct AcpiMiscInfo {
115     bool has_hpet;
116 #ifdef CONFIG_TPM
117     TPMVersion tpm_version;
118 #endif
119 } AcpiMiscInfo;
120 
121 typedef struct FwCfgTPMConfig {
122     uint32_t tpmppi_address;
123     uint8_t tpm_version;
124     uint8_t tpmppi_version;
125 } QEMU_PACKED FwCfgTPMConfig;
126 
127 static bool acpi_get_mcfg(AcpiMcfgInfo *mcfg);
128 
129 const struct AcpiGenericAddress x86_nvdimm_acpi_dsmio = {
130     .space_id = AML_AS_SYSTEM_IO,
131     .address = NVDIMM_ACPI_IO_BASE,
132     .bit_width = NVDIMM_ACPI_IO_LEN << 3
133 };
134 
135 static void init_common_fadt_data(MachineState *ms, Object *o,
136                                   AcpiFadtData *data)
137 {
138     X86MachineState *x86ms = X86_MACHINE(ms);
139     /*
140      * "ICH9-LPC" or "PIIX4_PM" has "smm-compat" property to keep the old
141      * behavior for compatibility irrelevant to smm_enabled, which doesn't
142      * conform to the ACPI spec.
143      */
144     bool smm_enabled = object_property_get_bool(o, "smm-compat", NULL) ?
145         true : x86_machine_is_smm_enabled(x86ms);
146     uint32_t io = object_property_get_uint(o, ACPI_PM_PROP_PM_IO_BASE, NULL);
147     AmlAddressSpace as = AML_AS_SYSTEM_IO;
148     AcpiFadtData fadt = {
149         .rev = 3,
150         .flags =
151             (1 << ACPI_FADT_F_WBINVD) |
152             (1 << ACPI_FADT_F_PROC_C1) |
153             (1 << ACPI_FADT_F_SLP_BUTTON) |
154             (1 << ACPI_FADT_F_RTC_S4) |
155             (1 << ACPI_FADT_F_USE_PLATFORM_CLOCK) |
156             /* APIC destination mode ("Flat Logical") has an upper limit of 8
157              * CPUs for more than 8 CPUs, "Clustered Logical" mode has to be
158              * used
159              */
160             ((ms->smp.max_cpus > 8) ?
161                         (1 << ACPI_FADT_F_FORCE_APIC_CLUSTER_MODEL) : 0),
162         .int_model = 1 /* Multiple APIC */,
163         .rtc_century = RTC_CENTURY,
164         .plvl2_lat = 0xfff /* C2 state not supported */,
165         .plvl3_lat = 0xfff /* C3 state not supported */,
166         .smi_cmd = smm_enabled ? ACPI_PORT_SMI_CMD : 0,
167         .sci_int = object_property_get_uint(o, ACPI_PM_PROP_SCI_INT, NULL),
168         .acpi_enable_cmd =
169             smm_enabled ?
170             object_property_get_uint(o, ACPI_PM_PROP_ACPI_ENABLE_CMD, NULL) :
171             0,
172         .acpi_disable_cmd =
173             smm_enabled ?
174             object_property_get_uint(o, ACPI_PM_PROP_ACPI_DISABLE_CMD, NULL) :
175             0,
176         .pm1a_evt = { .space_id = as, .bit_width = 4 * 8, .address = io },
177         .pm1a_cnt = { .space_id = as, .bit_width = 2 * 8,
178                       .address = io + 0x04 },
179         .pm_tmr = { .space_id = as, .bit_width = 4 * 8, .address = io + 0x08 },
180         .gpe0_blk = { .space_id = as, .bit_width =
181             object_property_get_uint(o, ACPI_PM_PROP_GPE0_BLK_LEN, NULL) * 8,
182             .address = object_property_get_uint(o, ACPI_PM_PROP_GPE0_BLK, NULL)
183         },
184     };
185 
186     /*
187      * ACPI v2, Table 5-10 - Fixed ACPI Description Table Boot Architecture
188      * Flags, bit offset 1 - 8042.
189      */
190     fadt.iapc_boot_arch = iapc_boot_arch_8042();
191 
192     *data = fadt;
193 }
194 
195 static void acpi_get_pm_info(MachineState *machine, AcpiPmInfo *pm)
196 {
197     Object *piix = object_resolve_type_unambiguous(TYPE_PIIX4_PM, NULL);
198     Object *lpc = object_resolve_type_unambiguous(TYPE_ICH9_LPC_DEVICE, NULL);
199     Object *obj = piix ? piix : lpc;
200     QObject *o;
201     pm->cpu_hp_io_base = 0;
202     pm->pcihp_io_base = 0;
203     pm->pcihp_io_len = 0;
204     pm->smi_on_cpuhp = false;
205     pm->smi_on_cpu_unplug = false;
206 
207     assert(obj);
208     init_common_fadt_data(machine, obj, &pm->fadt);
209     if (piix) {
210         /* w2k requires FADT(rev1) or it won't boot, keep PC compatible */
211         pm->fadt.rev = 1;
212         pm->cpu_hp_io_base = PIIX4_CPU_HOTPLUG_IO_BASE;
213     }
214     if (lpc) {
215         uint64_t smi_features = object_property_get_uint(lpc,
216             ICH9_LPC_SMI_NEGOTIATED_FEAT_PROP, NULL);
217         struct AcpiGenericAddress r = { .space_id = AML_AS_SYSTEM_IO,
218             .bit_width = 8, .address = ICH9_RST_CNT_IOPORT };
219         pm->fadt.reset_reg = r;
220         pm->fadt.reset_val = 0xf;
221         pm->fadt.flags |= 1 << ACPI_FADT_F_RESET_REG_SUP;
222         pm->cpu_hp_io_base = ICH9_CPU_HOTPLUG_IO_BASE;
223         pm->smi_on_cpuhp =
224             !!(smi_features & BIT_ULL(ICH9_LPC_SMI_F_CPU_HOTPLUG_BIT));
225         pm->smi_on_cpu_unplug =
226             !!(smi_features & BIT_ULL(ICH9_LPC_SMI_F_CPU_HOT_UNPLUG_BIT));
227     }
228     pm->pcihp_io_base =
229         object_property_get_uint(obj, ACPI_PCIHP_IO_BASE_PROP, NULL);
230     pm->pcihp_io_len =
231         object_property_get_uint(obj, ACPI_PCIHP_IO_LEN_PROP, NULL);
232 
233     /* Fill in optional s3/s4 related properties */
234     o = object_property_get_qobject(obj, ACPI_PM_PROP_S3_DISABLED, NULL);
235     if (o) {
236         pm->s3_disabled = qnum_get_uint(qobject_to(QNum, o));
237     } else {
238         pm->s3_disabled = false;
239     }
240     qobject_unref(o);
241     o = object_property_get_qobject(obj, ACPI_PM_PROP_S4_DISABLED, NULL);
242     if (o) {
243         pm->s4_disabled = qnum_get_uint(qobject_to(QNum, o));
244     } else {
245         pm->s4_disabled = false;
246     }
247     qobject_unref(o);
248     o = object_property_get_qobject(obj, ACPI_PM_PROP_S4_VAL, NULL);
249     if (o) {
250         pm->s4_val = qnum_get_uint(qobject_to(QNum, o));
251     } else {
252         pm->s4_val = false;
253     }
254     qobject_unref(o);
255 
256     pm->pcihp_bridge_en =
257         object_property_get_bool(obj, ACPI_PM_PROP_ACPI_PCIHP_BRIDGE,
258                                  NULL);
259     pm->pcihp_root_en =
260         object_property_get_bool(obj, ACPI_PM_PROP_ACPI_PCI_ROOTHP,
261                                  NULL);
262 }
263 
264 static void acpi_get_misc_info(AcpiMiscInfo *info)
265 {
266     info->has_hpet = hpet_find();
267 #ifdef CONFIG_TPM
268     info->tpm_version = tpm_get_version(tpm_find());
269 #endif
270 }
271 
272 /*
273  * Because of the PXB hosts we cannot simply query TYPE_PCI_HOST_BRIDGE.
274  * On i386 arch we only have two pci hosts, so we can look only for them.
275  */
276 Object *acpi_get_i386_pci_host(void)
277 {
278     PCIHostState *host;
279 
280     host = PCI_HOST_BRIDGE(object_resolve_path("/machine/i440fx", NULL));
281     if (!host) {
282         host = PCI_HOST_BRIDGE(object_resolve_path("/machine/q35", NULL));
283     }
284 
285     return OBJECT(host);
286 }
287 
288 static void acpi_get_pci_holes(Range *hole, Range *hole64)
289 {
290     Object *pci_host;
291 
292     pci_host = acpi_get_i386_pci_host();
293 
294     if (!pci_host) {
295         return;
296     }
297 
298     range_set_bounds1(hole,
299                       object_property_get_uint(pci_host,
300                                                PCI_HOST_PROP_PCI_HOLE_START,
301                                                NULL),
302                       object_property_get_uint(pci_host,
303                                                PCI_HOST_PROP_PCI_HOLE_END,
304                                                NULL));
305     range_set_bounds1(hole64,
306                       object_property_get_uint(pci_host,
307                                                PCI_HOST_PROP_PCI_HOLE64_START,
308                                                NULL),
309                       object_property_get_uint(pci_host,
310                                                PCI_HOST_PROP_PCI_HOLE64_END,
311                                                NULL));
312 }
313 
314 static void acpi_align_size(GArray *blob, unsigned align)
315 {
316     /* Align size to multiple of given size. This reduces the chance
317      * we need to change size in the future (breaking cross version migration).
318      */
319     g_array_set_size(blob, ROUND_UP(acpi_data_len(blob), align));
320 }
321 
322 /*
323  * ACPI spec 1.0b,
324  * 5.2.6 Firmware ACPI Control Structure
325  */
326 static void
327 build_facs(GArray *table_data)
328 {
329     const char *sig = "FACS";
330     const uint8_t reserved[40] = {};
331 
332     g_array_append_vals(table_data, sig, 4); /* Signature */
333     build_append_int_noprefix(table_data, 64, 4); /* Length */
334     build_append_int_noprefix(table_data, 0, 4); /* Hardware Signature */
335     build_append_int_noprefix(table_data, 0, 4); /* Firmware Waking Vector */
336     build_append_int_noprefix(table_data, 0, 4); /* Global Lock */
337     build_append_int_noprefix(table_data, 0, 4); /* Flags */
338     g_array_append_vals(table_data, reserved, 40); /* Reserved */
339 }
340 
341 static Aml *aml_pci_device_dsm(void)
342 {
343     Aml *method;
344 
345     method = aml_method("_DSM", 4, AML_SERIALIZED);
346     {
347         Aml *params = aml_local(0);
348         Aml *pkg = aml_package(2);
349         aml_append(pkg, aml_int(0));
350         aml_append(pkg, aml_int(0));
351         aml_append(method, aml_store(pkg, params));
352         aml_append(method,
353             aml_store(aml_name("BSEL"), aml_index(params, aml_int(0))));
354         aml_append(method,
355             aml_store(aml_name("ASUN"), aml_index(params, aml_int(1))));
356         aml_append(method,
357             aml_return(aml_call5("PDSM", aml_arg(0), aml_arg(1),
358                                  aml_arg(2), aml_arg(3), params))
359         );
360     }
361     return method;
362 }
363 
364 static Aml *aml_pci_edsm(void)
365 {
366     Aml *method, *ifctx;
367     Aml *zero = aml_int(0);
368     Aml *func = aml_arg(2);
369     Aml *ret = aml_local(0);
370     Aml *aidx = aml_local(1);
371     Aml *params = aml_arg(4);
372 
373     method = aml_method("EDSM", 5, AML_SERIALIZED);
374 
375     /* get supported functions */
376     ifctx = aml_if(aml_equal(func, zero));
377     {
378         /* 1: have supported functions */
379         /* 7: support for function 7 */
380         const uint8_t caps = 1 | BIT(7);
381         build_append_pci_dsm_func0_common(ifctx, ret);
382         aml_append(ifctx, aml_store(aml_int(caps), aml_index(ret, zero)));
383         aml_append(ifctx, aml_return(ret));
384     }
385     aml_append(method, ifctx);
386 
387     /* handle specific functions requests */
388     /*
389      * PCI Firmware Specification 3.1
390      * 4.6.7. _DSM for Naming a PCI or PCI Express Device Under
391      *        Operating Systems
392      */
393     ifctx = aml_if(aml_equal(func, aml_int(7)));
394     {
395        Aml *pkg = aml_package(2);
396        aml_append(pkg, zero);
397        /* optional, if not impl. should return null string */
398        aml_append(pkg, aml_string("%s", ""));
399        aml_append(ifctx, aml_store(pkg, ret));
400 
401        /*
402         * IASL is fine when initializing Package with computational data,
403         * however it makes guest unhappy /it fails to process such AML/.
404         * So use runtime assignment to set acpi-index after initializer
405         * to make OSPM happy.
406         */
407        aml_append(ifctx,
408            aml_store(aml_derefof(aml_index(params, aml_int(0))), aidx));
409        aml_append(ifctx, aml_store(aidx, aml_index(ret, zero)));
410        aml_append(ifctx, aml_return(ret));
411     }
412     aml_append(method, ifctx);
413 
414     return method;
415 }
416 
417 static Aml *aml_pci_static_endpoint_dsm(PCIDevice *pdev)
418 {
419     Aml *method;
420 
421     g_assert(pdev->acpi_index != 0);
422     method = aml_method("_DSM", 4, AML_SERIALIZED);
423     {
424         Aml *params = aml_local(0);
425         Aml *pkg = aml_package(1);
426         aml_append(pkg, aml_int(pdev->acpi_index));
427         aml_append(method, aml_store(pkg, params));
428         aml_append(method,
429             aml_return(aml_call5("EDSM", aml_arg(0), aml_arg(1),
430                                  aml_arg(2), aml_arg(3), params))
431         );
432     }
433     return method;
434 }
435 
436 static void build_append_pcihp_notify_entry(Aml *method, int slot)
437 {
438     Aml *if_ctx;
439     int32_t devfn = PCI_DEVFN(slot, 0);
440 
441     if_ctx = aml_if(aml_and(aml_arg(0), aml_int(0x1U << slot), NULL));
442     aml_append(if_ctx, aml_notify(aml_name("S%.02X", devfn), aml_arg(1)));
443     aml_append(method, if_ctx);
444 }
445 
446 static bool is_devfn_ignored_generic(const int devfn, const PCIBus *bus)
447 {
448     const PCIDevice *pdev = bus->devices[devfn];
449 
450     if (PCI_FUNC(devfn)) {
451         if (IS_PCI_BRIDGE(pdev)) {
452             /*
453              * Ignore only hotplugged PCI bridges on !0 functions, but
454              * allow describing cold plugged bridges on all functions
455              */
456             if (DEVICE(pdev)->hotplugged) {
457                 return true;
458             }
459         }
460     }
461     return false;
462 }
463 
464 static bool is_devfn_ignored_hotplug(const int devfn, const PCIBus *bus)
465 {
466     PCIDevice *pdev = bus->devices[devfn];
467     if (pdev) {
468         return is_devfn_ignored_generic(devfn, bus) ||
469                !DEVICE_GET_CLASS(pdev)->hotpluggable ||
470                /* Cold plugged bridges aren't themselves hot-pluggable */
471                (IS_PCI_BRIDGE(pdev) && !DEVICE(pdev)->hotplugged);
472     } else { /* non populated slots */
473          /*
474          * hotplug is supported only for non-multifunction device
475          * so generate device description only for function 0
476          */
477         if (PCI_FUNC(devfn) ||
478             (pci_bus_is_express(bus) && PCI_SLOT(devfn) > 0)) {
479             return true;
480         }
481     }
482     return false;
483 }
484 
485 void build_append_pcihp_slots(Aml *parent_scope, PCIBus *bus)
486 {
487     int devfn;
488     Aml *dev, *notify_method = NULL, *method;
489     QObject *bsel = object_property_get_qobject(OBJECT(bus),
490                         ACPI_PCIHP_PROP_BSEL, NULL);
491     uint64_t bsel_val = qnum_get_uint(qobject_to(QNum, bsel));
492     qobject_unref(bsel);
493 
494     aml_append(parent_scope, aml_name_decl("BSEL", aml_int(bsel_val)));
495     notify_method = aml_method("DVNT", 2, AML_NOTSERIALIZED);
496 
497     for (devfn = 0; devfn < ARRAY_SIZE(bus->devices); devfn++) {
498         int slot = PCI_SLOT(devfn);
499         int adr = slot << 16 | PCI_FUNC(devfn);
500 
501         if (is_devfn_ignored_hotplug(devfn, bus)) {
502             continue;
503         }
504 
505         if (bus->devices[devfn]) {
506             dev = aml_scope("S%.02X", devfn);
507         } else {
508             dev = aml_device("S%.02X", devfn);
509             aml_append(dev, aml_name_decl("_ADR", aml_int(adr)));
510         }
511 
512         /*
513          * Can't declare _SUN here for every device as it changes 'slot'
514          * enumeration order in linux kernel, so use another variable for it
515          */
516         aml_append(dev, aml_name_decl("ASUN", aml_int(slot)));
517         aml_append(dev, aml_pci_device_dsm());
518 
519         aml_append(dev, aml_name_decl("_SUN", aml_int(slot)));
520         /* add _EJ0 to make slot hotpluggable  */
521         method = aml_method("_EJ0", 1, AML_NOTSERIALIZED);
522         aml_append(method,
523             aml_call2("PCEJ", aml_name("BSEL"), aml_name("_SUN"))
524         );
525         aml_append(dev, method);
526 
527         build_append_pcihp_notify_entry(notify_method, slot);
528 
529         /* device descriptor has been composed, add it into parent context */
530         aml_append(parent_scope, dev);
531     }
532     aml_append(parent_scope, notify_method);
533 }
534 
535 void build_append_pci_bus_devices(Aml *parent_scope, PCIBus *bus)
536 {
537     int devfn;
538     Aml *dev;
539 
540     for (devfn = 0; devfn < ARRAY_SIZE(bus->devices); devfn++) {
541         /* ACPI spec: 1.0b: Table 6-2 _ADR Object Bus Types, PCI type */
542         int adr = PCI_SLOT(devfn) << 16 | PCI_FUNC(devfn);
543         PCIDevice *pdev = bus->devices[devfn];
544 
545         if (!pdev || is_devfn_ignored_generic(devfn, bus)) {
546             continue;
547         }
548 
549         /* start to compose PCI device descriptor */
550         dev = aml_device("S%.02X", devfn);
551         aml_append(dev, aml_name_decl("_ADR", aml_int(adr)));
552 
553         call_dev_aml_func(DEVICE(bus->devices[devfn]), dev);
554         /* add _DSM if device has acpi-index set */
555         if (pdev->acpi_index &&
556             !object_property_get_bool(OBJECT(pdev), "hotpluggable",
557                                       &error_abort)) {
558             aml_append(dev, aml_pci_static_endpoint_dsm(pdev));
559         }
560 
561         /* device descriptor has been composed, add it into parent context */
562         aml_append(parent_scope, dev);
563     }
564 }
565 
566 static bool build_append_notification_callback(Aml *parent_scope,
567                                                const PCIBus *bus)
568 {
569     Aml *method;
570     PCIBus *sec;
571     QObject *bsel;
572     int nr_notifiers = 0;
573     GQueue *pcnt_bus_list = g_queue_new();
574 
575     QLIST_FOREACH(sec, &bus->child, sibling) {
576         Aml *br_scope = aml_scope("S%.02X", sec->parent_dev->devfn);
577         if (pci_bus_is_root(sec)) {
578             continue;
579         }
580         nr_notifiers = nr_notifiers +
581                        build_append_notification_callback(br_scope, sec);
582         /*
583          * add new child scope to parent
584          * and keep track of bus that have PCNT,
585          * bus list is used later to call children PCNTs from this level PCNT
586          */
587         if (nr_notifiers) {
588             g_queue_push_tail(pcnt_bus_list, sec);
589             aml_append(parent_scope, br_scope);
590         }
591     }
592 
593     /*
594      * Append PCNT method to notify about events on local and child buses.
595      * ps: hostbridge might not have hotplug (bsel) enabled but might have
596      * child bridges that do have bsel.
597      */
598     method = aml_method("PCNT", 0, AML_NOTSERIALIZED);
599 
600     /* If bus supports hotplug select it and notify about local events */
601     bsel = object_property_get_qobject(OBJECT(bus), ACPI_PCIHP_PROP_BSEL, NULL);
602     if (bsel) {
603         uint64_t bsel_val = qnum_get_uint(qobject_to(QNum, bsel));
604 
605         aml_append(method, aml_store(aml_int(bsel_val), aml_name("BNUM")));
606         aml_append(method, aml_call2("DVNT", aml_name("PCIU"),
607                                      aml_int(1))); /* Device Check */
608         aml_append(method, aml_call2("DVNT", aml_name("PCID"),
609                                      aml_int(3))); /* Eject Request */
610         nr_notifiers++;
611     }
612 
613     /* Notify about child bus events in any case */
614     while ((sec = g_queue_pop_head(pcnt_bus_list))) {
615         aml_append(method, aml_name("^S%.02X.PCNT", sec->parent_dev->devfn));
616     }
617 
618     aml_append(parent_scope, method);
619     qobject_unref(bsel);
620     g_queue_free(pcnt_bus_list);
621     return !!nr_notifiers;
622 }
623 
624 /*
625  * build_prt - Define interrupt routing rules
626  *
627  * Returns an array of 128 routes, one for each device,
628  * based on device location.
629  * The main goal is to equally distribute the interrupts
630  * over the 4 existing ACPI links (works only for i440fx).
631  * The hash function is: (slot + pin) & 3 -> "LNK[D|A|B|C]".
632  *
633  */
634 static Aml *build_prt(bool is_pci0_prt)
635 {
636     const int nroutes = 128;
637     Aml *rt_pkg, *method;
638     int pin;
639 
640     method = aml_method("_PRT", 0, AML_NOTSERIALIZED);
641     assert(nroutes < 256);
642     rt_pkg = aml_package(nroutes);
643 
644     for (pin = 0; pin < nroutes; pin++) {
645         Aml *pkg = aml_package(4);
646         int slot = pin >> 2;
647 
648         aml_append(pkg, aml_int((slot << 16) | 0xFFFF));
649         aml_append(pkg, aml_int(pin & 3));
650         /* device 1 is the power-management device, needs SCI */
651         if (is_pci0_prt && pin == 4) {
652             aml_append(pkg, aml_name("%s", "LNKS"));
653         } else {
654             static const char link_name[][5] = {"LNKD", "LNKA", "LNKB", "LNKC"};
655             int hash = (slot + pin) & 3;
656             aml_append(pkg, aml_name("%s", link_name[hash]));
657         }
658         aml_append(pkg, aml_int(0));
659         aml_append(rt_pkg, pkg);
660     }
661 
662     aml_append(method, aml_return(rt_pkg));
663 
664     return method;
665 }
666 
667 static void build_hpet_aml(Aml *table)
668 {
669     Aml *crs;
670     Aml *field;
671     Aml *method;
672     Aml *if_ctx;
673     Aml *scope = aml_scope("_SB");
674     Aml *dev = aml_device("HPET");
675     Aml *zero = aml_int(0);
676     Aml *id = aml_local(0);
677     Aml *period = aml_local(1);
678 
679     aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0103")));
680     aml_append(dev, aml_name_decl("_UID", zero));
681 
682     aml_append(dev,
683         aml_operation_region("HPTM", AML_SYSTEM_MEMORY, aml_int(HPET_BASE),
684                              HPET_LEN));
685     field = aml_field("HPTM", AML_DWORD_ACC, AML_LOCK, AML_PRESERVE);
686     aml_append(field, aml_named_field("VEND", 32));
687     aml_append(field, aml_named_field("PRD", 32));
688     aml_append(dev, field);
689 
690     method = aml_method("_STA", 0, AML_NOTSERIALIZED);
691     aml_append(method, aml_store(aml_name("VEND"), id));
692     aml_append(method, aml_store(aml_name("PRD"), period));
693     aml_append(method, aml_shiftright(id, aml_int(16), id));
694     if_ctx = aml_if(aml_lor(aml_equal(id, zero),
695                             aml_equal(id, aml_int(0xffff))));
696     {
697         aml_append(if_ctx, aml_return(zero));
698     }
699     aml_append(method, if_ctx);
700 
701     if_ctx = aml_if(aml_lor(aml_equal(period, zero),
702                             aml_lgreater(period, aml_int(100000000))));
703     {
704         aml_append(if_ctx, aml_return(zero));
705     }
706     aml_append(method, if_ctx);
707 
708     aml_append(method, aml_return(aml_int(0x0F)));
709     aml_append(dev, method);
710 
711     crs = aml_resource_template();
712     aml_append(crs, aml_memory32_fixed(HPET_BASE, HPET_LEN, AML_READ_ONLY));
713     aml_append(dev, aml_name_decl("_CRS", crs));
714 
715     aml_append(scope, dev);
716     aml_append(table, scope);
717 }
718 
719 static Aml *build_vmbus_device_aml(VMBusBridge *vmbus_bridge)
720 {
721     Aml *dev;
722     Aml *method;
723     Aml *crs;
724 
725     dev = aml_device("VMBS");
726     aml_append(dev, aml_name_decl("STA", aml_int(0xF)));
727     aml_append(dev, aml_name_decl("_HID", aml_string("VMBus")));
728     aml_append(dev, aml_name_decl("_UID", aml_int(0x0)));
729     aml_append(dev, aml_name_decl("_DDN", aml_string("VMBUS")));
730 
731     method = aml_method("_DIS", 0, AML_NOTSERIALIZED);
732     aml_append(method, aml_store(aml_and(aml_name("STA"), aml_int(0xD), NULL),
733                                      aml_name("STA")));
734     aml_append(dev, method);
735 
736     method = aml_method("_PS0", 0, AML_NOTSERIALIZED);
737     aml_append(method, aml_store(aml_or(aml_name("STA"), aml_int(0xF), NULL),
738                                      aml_name("STA")));
739     aml_append(dev, method);
740 
741     method = aml_method("_STA", 0, AML_NOTSERIALIZED);
742     aml_append(method, aml_return(aml_name("STA")));
743     aml_append(dev, method);
744 
745     aml_append(dev, aml_name_decl("_PS3", aml_int(0x0)));
746 
747     crs = aml_resource_template();
748     aml_append(crs, aml_irq_no_flags(vmbus_bridge->irq));
749     aml_append(dev, aml_name_decl("_CRS", crs));
750 
751     return dev;
752 }
753 
754 static void build_dbg_aml(Aml *table)
755 {
756     Aml *field;
757     Aml *method;
758     Aml *while_ctx;
759     Aml *scope = aml_scope("\\");
760     Aml *buf = aml_local(0);
761     Aml *len = aml_local(1);
762     Aml *idx = aml_local(2);
763 
764     aml_append(scope,
765        aml_operation_region("DBG", AML_SYSTEM_IO, aml_int(0x0402), 0x01));
766     field = aml_field("DBG", AML_BYTE_ACC, AML_NOLOCK, AML_PRESERVE);
767     aml_append(field, aml_named_field("DBGB", 8));
768     aml_append(scope, field);
769 
770     method = aml_method("DBUG", 1, AML_NOTSERIALIZED);
771 
772     aml_append(method, aml_to_hexstring(aml_arg(0), buf));
773     aml_append(method, aml_to_buffer(buf, buf));
774     aml_append(method, aml_subtract(aml_sizeof(buf), aml_int(1), len));
775     aml_append(method, aml_store(aml_int(0), idx));
776 
777     while_ctx = aml_while(aml_lless(idx, len));
778     aml_append(while_ctx,
779         aml_store(aml_derefof(aml_index(buf, idx)), aml_name("DBGB")));
780     aml_append(while_ctx, aml_increment(idx));
781     aml_append(method, while_ctx);
782 
783     aml_append(method, aml_store(aml_int(0x0A), aml_name("DBGB")));
784     aml_append(scope, method);
785 
786     aml_append(table, scope);
787 }
788 
789 static Aml *build_link_dev(const char *name, uint8_t uid, Aml *reg)
790 {
791     Aml *dev;
792     Aml *crs;
793     Aml *method;
794     uint32_t irqs[] = {5, 10, 11};
795 
796     dev = aml_device("%s", name);
797     aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0C0F")));
798     aml_append(dev, aml_name_decl("_UID", aml_int(uid)));
799 
800     crs = aml_resource_template();
801     aml_append(crs, aml_interrupt(AML_CONSUMER, AML_LEVEL, AML_ACTIVE_HIGH,
802                                   AML_SHARED, irqs, ARRAY_SIZE(irqs)));
803     aml_append(dev, aml_name_decl("_PRS", crs));
804 
805     method = aml_method("_STA", 0, AML_NOTSERIALIZED);
806     aml_append(method, aml_return(aml_call1("IQST", reg)));
807     aml_append(dev, method);
808 
809     method = aml_method("_DIS", 0, AML_NOTSERIALIZED);
810     aml_append(method, aml_or(reg, aml_int(0x80), reg));
811     aml_append(dev, method);
812 
813     method = aml_method("_CRS", 0, AML_NOTSERIALIZED);
814     aml_append(method, aml_return(aml_call1("IQCR", reg)));
815     aml_append(dev, method);
816 
817     method = aml_method("_SRS", 1, AML_NOTSERIALIZED);
818     aml_append(method, aml_create_dword_field(aml_arg(0), aml_int(5), "PRRI"));
819     aml_append(method, aml_store(aml_name("PRRI"), reg));
820     aml_append(dev, method);
821 
822     return dev;
823  }
824 
825 static Aml *build_gsi_link_dev(const char *name, uint8_t uid, uint8_t gsi)
826 {
827     Aml *dev;
828     Aml *crs;
829     Aml *method;
830     uint32_t irqs;
831 
832     dev = aml_device("%s", name);
833     aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0C0F")));
834     aml_append(dev, aml_name_decl("_UID", aml_int(uid)));
835 
836     crs = aml_resource_template();
837     irqs = gsi;
838     aml_append(crs, aml_interrupt(AML_CONSUMER, AML_LEVEL, AML_ACTIVE_HIGH,
839                                   AML_SHARED, &irqs, 1));
840     aml_append(dev, aml_name_decl("_PRS", crs));
841 
842     aml_append(dev, aml_name_decl("_CRS", crs));
843 
844     /*
845      * _DIS can be no-op because the interrupt cannot be disabled.
846      */
847     method = aml_method("_DIS", 0, AML_NOTSERIALIZED);
848     aml_append(dev, method);
849 
850     method = aml_method("_SRS", 1, AML_NOTSERIALIZED);
851     aml_append(dev, method);
852 
853     return dev;
854 }
855 
856 /* _CRS method - get current settings */
857 static Aml *build_iqcr_method(bool is_piix4)
858 {
859     Aml *if_ctx;
860     uint32_t irqs;
861     Aml *method = aml_method("IQCR", 1, AML_SERIALIZED);
862     Aml *crs = aml_resource_template();
863 
864     irqs = 0;
865     aml_append(crs, aml_interrupt(AML_CONSUMER, AML_LEVEL,
866                                   AML_ACTIVE_HIGH, AML_SHARED, &irqs, 1));
867     aml_append(method, aml_name_decl("PRR0", crs));
868 
869     aml_append(method,
870         aml_create_dword_field(aml_name("PRR0"), aml_int(5), "PRRI"));
871 
872     if (is_piix4) {
873         if_ctx = aml_if(aml_lless(aml_arg(0), aml_int(0x80)));
874         aml_append(if_ctx, aml_store(aml_arg(0), aml_name("PRRI")));
875         aml_append(method, if_ctx);
876     } else {
877         aml_append(method,
878             aml_store(aml_and(aml_arg(0), aml_int(0xF), NULL),
879                       aml_name("PRRI")));
880     }
881 
882     aml_append(method, aml_return(aml_name("PRR0")));
883     return method;
884 }
885 
886 /* _STA method - get status */
887 static Aml *build_irq_status_method(void)
888 {
889     Aml *if_ctx;
890     Aml *method = aml_method("IQST", 1, AML_NOTSERIALIZED);
891 
892     if_ctx = aml_if(aml_and(aml_int(0x80), aml_arg(0), NULL));
893     aml_append(if_ctx, aml_return(aml_int(0x09)));
894     aml_append(method, if_ctx);
895     aml_append(method, aml_return(aml_int(0x0B)));
896     return method;
897 }
898 
899 static void build_piix4_pci0_int(Aml *table)
900 {
901     Aml *dev;
902     Aml *crs;
903     Aml *method;
904     uint32_t irqs;
905     Aml *sb_scope = aml_scope("_SB");
906     Aml *pci0_scope = aml_scope("PCI0");
907 
908     aml_append(pci0_scope, build_prt(true));
909     aml_append(sb_scope, pci0_scope);
910 
911     aml_append(sb_scope, build_irq_status_method());
912     aml_append(sb_scope, build_iqcr_method(true));
913 
914     aml_append(sb_scope, build_link_dev("LNKA", 0, aml_name("PRQ0")));
915     aml_append(sb_scope, build_link_dev("LNKB", 1, aml_name("PRQ1")));
916     aml_append(sb_scope, build_link_dev("LNKC", 2, aml_name("PRQ2")));
917     aml_append(sb_scope, build_link_dev("LNKD", 3, aml_name("PRQ3")));
918 
919     dev = aml_device("LNKS");
920     {
921         aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0C0F")));
922         aml_append(dev, aml_name_decl("_UID", aml_int(4)));
923 
924         crs = aml_resource_template();
925         irqs = 9;
926         aml_append(crs, aml_interrupt(AML_CONSUMER, AML_LEVEL,
927                                       AML_ACTIVE_HIGH, AML_SHARED,
928                                       &irqs, 1));
929         aml_append(dev, aml_name_decl("_PRS", crs));
930 
931         /* The SCI cannot be disabled and is always attached to GSI 9,
932          * so these are no-ops.  We only need this link to override the
933          * polarity to active high and match the content of the MADT.
934          */
935         method = aml_method("_STA", 0, AML_NOTSERIALIZED);
936         aml_append(method, aml_return(aml_int(0x0b)));
937         aml_append(dev, method);
938 
939         method = aml_method("_DIS", 0, AML_NOTSERIALIZED);
940         aml_append(dev, method);
941 
942         method = aml_method("_CRS", 0, AML_NOTSERIALIZED);
943         aml_append(method, aml_return(aml_name("_PRS")));
944         aml_append(dev, method);
945 
946         method = aml_method("_SRS", 1, AML_NOTSERIALIZED);
947         aml_append(dev, method);
948     }
949     aml_append(sb_scope, dev);
950 
951     aml_append(table, sb_scope);
952 }
953 
954 static void append_q35_prt_entry(Aml *ctx, uint32_t nr, const char *name)
955 {
956     int i;
957     int head;
958     Aml *pkg;
959     char base = name[3] < 'E' ? 'A' : 'E';
960     char *s = g_strdup(name);
961     Aml *a_nr = aml_int((nr << 16) | 0xffff);
962 
963     assert(strlen(s) == 4);
964 
965     head = name[3] - base;
966     for (i = 0; i < 4; i++) {
967         if (head + i > 3) {
968             head = i * -1;
969         }
970         s[3] = base + head + i;
971         pkg = aml_package(4);
972         aml_append(pkg, a_nr);
973         aml_append(pkg, aml_int(i));
974         aml_append(pkg, aml_name("%s", s));
975         aml_append(pkg, aml_int(0));
976         aml_append(ctx, pkg);
977     }
978     g_free(s);
979 }
980 
981 static Aml *build_q35_routing_table(const char *str)
982 {
983     int i;
984     Aml *pkg;
985     char *name = g_strdup_printf("%s ", str);
986 
987     pkg = aml_package(128);
988     for (i = 0; i < 0x18; i++) {
989             name[3] = 'E' + (i & 0x3);
990             append_q35_prt_entry(pkg, i, name);
991     }
992 
993     name[3] = 'E';
994     append_q35_prt_entry(pkg, 0x18, name);
995 
996     /* INTA -> PIRQA for slot 25 - 31, see the default value of D<N>IR */
997     for (i = 0x0019; i < 0x1e; i++) {
998         name[3] = 'A';
999         append_q35_prt_entry(pkg, i, name);
1000     }
1001 
1002     /* PCIe->PCI bridge. use PIRQ[E-H] */
1003     name[3] = 'E';
1004     append_q35_prt_entry(pkg, 0x1e, name);
1005     name[3] = 'A';
1006     append_q35_prt_entry(pkg, 0x1f, name);
1007 
1008     g_free(name);
1009     return pkg;
1010 }
1011 
1012 static void build_q35_pci0_int(Aml *table)
1013 {
1014     Aml *method;
1015     Aml *sb_scope = aml_scope("_SB");
1016     Aml *pci0_scope = aml_scope("PCI0");
1017 
1018     /* Zero => PIC mode, One => APIC Mode */
1019     aml_append(table, aml_name_decl("PICF", aml_int(0)));
1020     method = aml_method("_PIC", 1, AML_NOTSERIALIZED);
1021     {
1022         aml_append(method, aml_store(aml_arg(0), aml_name("PICF")));
1023     }
1024     aml_append(table, method);
1025 
1026     aml_append(pci0_scope,
1027         aml_name_decl("PRTP", build_q35_routing_table("LNK")));
1028     aml_append(pci0_scope,
1029         aml_name_decl("PRTA", build_q35_routing_table("GSI")));
1030 
1031     method = aml_method("_PRT", 0, AML_NOTSERIALIZED);
1032     {
1033         Aml *if_ctx;
1034         Aml *else_ctx;
1035 
1036         /* PCI IRQ routing table, example from ACPI 2.0a specification,
1037            section 6.2.8.1 */
1038         /* Note: we provide the same info as the PCI routing
1039            table of the Bochs BIOS */
1040         if_ctx = aml_if(aml_equal(aml_name("PICF"), aml_int(0)));
1041         aml_append(if_ctx, aml_return(aml_name("PRTP")));
1042         aml_append(method, if_ctx);
1043         else_ctx = aml_else();
1044         aml_append(else_ctx, aml_return(aml_name("PRTA")));
1045         aml_append(method, else_ctx);
1046     }
1047     aml_append(pci0_scope, method);
1048     aml_append(sb_scope, pci0_scope);
1049 
1050     aml_append(sb_scope, build_irq_status_method());
1051     aml_append(sb_scope, build_iqcr_method(false));
1052 
1053     aml_append(sb_scope, build_link_dev("LNKA", 0, aml_name("PRQA")));
1054     aml_append(sb_scope, build_link_dev("LNKB", 1, aml_name("PRQB")));
1055     aml_append(sb_scope, build_link_dev("LNKC", 2, aml_name("PRQC")));
1056     aml_append(sb_scope, build_link_dev("LNKD", 3, aml_name("PRQD")));
1057     aml_append(sb_scope, build_link_dev("LNKE", 4, aml_name("PRQE")));
1058     aml_append(sb_scope, build_link_dev("LNKF", 5, aml_name("PRQF")));
1059     aml_append(sb_scope, build_link_dev("LNKG", 6, aml_name("PRQG")));
1060     aml_append(sb_scope, build_link_dev("LNKH", 7, aml_name("PRQH")));
1061 
1062     aml_append(sb_scope, build_gsi_link_dev("GSIA", 0x10, 0x10));
1063     aml_append(sb_scope, build_gsi_link_dev("GSIB", 0x11, 0x11));
1064     aml_append(sb_scope, build_gsi_link_dev("GSIC", 0x12, 0x12));
1065     aml_append(sb_scope, build_gsi_link_dev("GSID", 0x13, 0x13));
1066     aml_append(sb_scope, build_gsi_link_dev("GSIE", 0x14, 0x14));
1067     aml_append(sb_scope, build_gsi_link_dev("GSIF", 0x15, 0x15));
1068     aml_append(sb_scope, build_gsi_link_dev("GSIG", 0x16, 0x16));
1069     aml_append(sb_scope, build_gsi_link_dev("GSIH", 0x17, 0x17));
1070 
1071     aml_append(table, sb_scope);
1072 }
1073 
1074 static Aml *build_q35_dram_controller(const AcpiMcfgInfo *mcfg)
1075 {
1076     Aml *dev;
1077     Aml *resource_template;
1078 
1079     /* DRAM controller */
1080     dev = aml_device("DRAC");
1081     aml_append(dev, aml_name_decl("_HID", aml_string("PNP0C01")));
1082 
1083     resource_template = aml_resource_template();
1084     if (mcfg->base + mcfg->size - 1 >= (1ULL << 32)) {
1085         aml_append(resource_template,
1086                    aml_qword_memory(AML_POS_DECODE,
1087                                     AML_MIN_FIXED,
1088                                     AML_MAX_FIXED,
1089                                     AML_NON_CACHEABLE,
1090                                     AML_READ_WRITE,
1091                                     0x0000000000000000,
1092                                     mcfg->base,
1093                                     mcfg->base + mcfg->size - 1,
1094                                     0x0000000000000000,
1095                                     mcfg->size));
1096     } else {
1097         aml_append(resource_template,
1098                    aml_dword_memory(AML_POS_DECODE,
1099                                     AML_MIN_FIXED,
1100                                     AML_MAX_FIXED,
1101                                     AML_NON_CACHEABLE,
1102                                     AML_READ_WRITE,
1103                                     0x0000000000000000,
1104                                     mcfg->base,
1105                                     mcfg->base + mcfg->size - 1,
1106                                     0x0000000000000000,
1107                                     mcfg->size));
1108     }
1109     aml_append(dev, aml_name_decl("_CRS", resource_template));
1110 
1111     return dev;
1112 }
1113 
1114 static void build_acpi0017(Aml *table)
1115 {
1116     Aml *dev, *scope, *method;
1117 
1118     scope =  aml_scope("_SB");
1119     dev = aml_device("CXLM");
1120     aml_append(dev, aml_name_decl("_HID", aml_string("ACPI0017")));
1121 
1122     method = aml_method("_STA", 0, AML_NOTSERIALIZED);
1123     aml_append(method, aml_return(aml_int(0x0B)));
1124     aml_append(dev, method);
1125     build_cxl_dsm_method(dev);
1126 
1127     aml_append(scope, dev);
1128     aml_append(table, scope);
1129 }
1130 
1131 static void
1132 build_dsdt(GArray *table_data, BIOSLinker *linker,
1133            AcpiPmInfo *pm, AcpiMiscInfo *misc,
1134            Range *pci_hole, Range *pci_hole64, MachineState *machine)
1135 {
1136     Object *i440fx = object_resolve_type_unambiguous(TYPE_I440FX_PCI_HOST_BRIDGE,
1137                                                      NULL);
1138     Object *q35 = object_resolve_type_unambiguous(TYPE_Q35_HOST_DEVICE, NULL);
1139     CrsRangeEntry *entry;
1140     Aml *dsdt, *sb_scope, *scope, *dev, *method, *field, *pkg, *crs;
1141     CrsRangeSet crs_range_set;
1142     PCMachineState *pcms = PC_MACHINE(machine);
1143     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(machine);
1144     X86MachineState *x86ms = X86_MACHINE(machine);
1145     AcpiMcfgInfo mcfg;
1146     bool mcfg_valid = !!acpi_get_mcfg(&mcfg);
1147     uint32_t nr_mem = machine->ram_slots;
1148     int root_bus_limit = 0xFF;
1149     PCIBus *bus = NULL;
1150 #ifdef CONFIG_TPM
1151     TPMIf *tpm = tpm_find();
1152 #endif
1153     bool cxl_present = false;
1154     int i;
1155     VMBusBridge *vmbus_bridge = vmbus_bridge_find();
1156     AcpiTable table = { .sig = "DSDT", .rev = 1, .oem_id = x86ms->oem_id,
1157                         .oem_table_id = x86ms->oem_table_id };
1158 
1159     assert(!!i440fx != !!q35);
1160 
1161     acpi_table_begin(&table, table_data);
1162     dsdt = init_aml_allocator();
1163 
1164     build_dbg_aml(dsdt);
1165     if (i440fx) {
1166         sb_scope = aml_scope("_SB");
1167         dev = aml_device("PCI0");
1168         aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0A03")));
1169         aml_append(dev, aml_name_decl("_UID", aml_int(pcmc->pci_root_uid)));
1170         aml_append(dev, aml_pci_edsm());
1171         aml_append(sb_scope, dev);
1172         aml_append(dsdt, sb_scope);
1173 
1174         if (pm->pcihp_bridge_en || pm->pcihp_root_en) {
1175             build_acpi_pci_hotplug(dsdt, pm->pcihp_io_base);
1176         }
1177         build_piix4_pci0_int(dsdt);
1178     } else if (q35) {
1179         sb_scope = aml_scope("_SB");
1180         dev = aml_device("PCI0");
1181         aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0A08")));
1182         aml_append(dev, aml_name_decl("_CID", aml_eisaid("PNP0A03")));
1183         aml_append(dev, aml_name_decl("_UID", aml_int(pcmc->pci_root_uid)));
1184         aml_append(dev, build_pci_host_bridge_osc_method(!pm->pcihp_bridge_en));
1185         aml_append(dev, aml_pci_edsm());
1186         aml_append(sb_scope, dev);
1187         if (mcfg_valid) {
1188             aml_append(sb_scope, build_q35_dram_controller(&mcfg));
1189         }
1190 
1191         if (pm->smi_on_cpuhp) {
1192             /* reserve SMI block resources, IO ports 0xB2, 0xB3 */
1193             dev = aml_device("PCI0.SMI0");
1194             aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0A06")));
1195             aml_append(dev, aml_name_decl("_UID", aml_string("SMI resources")));
1196             crs = aml_resource_template();
1197             aml_append(crs,
1198                 aml_io(
1199                        AML_DECODE16,
1200                        pm->fadt.smi_cmd,
1201                        pm->fadt.smi_cmd,
1202                        1,
1203                        2)
1204             );
1205             aml_append(dev, aml_name_decl("_CRS", crs));
1206             aml_append(dev, aml_operation_region("SMIR", AML_SYSTEM_IO,
1207                 aml_int(pm->fadt.smi_cmd), 2));
1208             field = aml_field("SMIR", AML_BYTE_ACC, AML_NOLOCK,
1209                               AML_WRITE_AS_ZEROS);
1210             aml_append(field, aml_named_field("SMIC", 8));
1211             aml_append(field, aml_reserved_field(8));
1212             aml_append(dev, field);
1213             aml_append(sb_scope, dev);
1214         }
1215 
1216         aml_append(dsdt, sb_scope);
1217 
1218         if (pm->pcihp_bridge_en) {
1219             build_acpi_pci_hotplug(dsdt, pm->pcihp_io_base);
1220         }
1221         build_q35_pci0_int(dsdt);
1222     }
1223 
1224     if (misc->has_hpet) {
1225         build_hpet_aml(dsdt);
1226     }
1227 
1228     if (vmbus_bridge) {
1229         sb_scope = aml_scope("_SB");
1230         aml_append(sb_scope, build_vmbus_device_aml(vmbus_bridge));
1231         aml_append(dsdt, sb_scope);
1232     }
1233 
1234     scope =  aml_scope("_GPE");
1235     {
1236         aml_append(scope, aml_name_decl("_HID", aml_string("ACPI0006")));
1237         if (machine->nvdimms_state->is_enabled) {
1238             method = aml_method("_E04", 0, AML_NOTSERIALIZED);
1239             aml_append(method, aml_notify(aml_name("\\_SB.NVDR"),
1240                                           aml_int(0x80)));
1241             aml_append(scope, method);
1242         }
1243     }
1244     aml_append(dsdt, scope);
1245 
1246     if (pcmc->legacy_cpu_hotplug) {
1247         build_legacy_cpu_hotplug_aml(dsdt, machine, pm->cpu_hp_io_base);
1248     } else {
1249         CPUHotplugFeatures opts = {
1250             .acpi_1_compatible = true, .has_legacy_cphp = true,
1251             .smi_path = pm->smi_on_cpuhp ? "\\_SB.PCI0.SMI0.SMIC" : NULL,
1252             .fw_unplugs_cpu = pm->smi_on_cpu_unplug,
1253         };
1254         build_cpus_aml(dsdt, machine, opts, pc_madt_cpu_entry,
1255                        pm->cpu_hp_io_base, "\\_SB.PCI0", "\\_GPE._E02",
1256                        AML_SYSTEM_IO);
1257     }
1258 
1259     if (pcms->memhp_io_base && nr_mem) {
1260         build_memory_hotplug_aml(dsdt, nr_mem, "\\_SB.PCI0",
1261                                  "\\_GPE._E03", AML_SYSTEM_IO,
1262                                  pcms->memhp_io_base);
1263     }
1264 
1265     crs_range_set_init(&crs_range_set);
1266     bus = PC_MACHINE(machine)->pcibus;
1267     if (bus) {
1268         QLIST_FOREACH(bus, &bus->child, sibling) {
1269             uint8_t bus_num = pci_bus_num(bus);
1270             uint8_t numa_node = pci_bus_numa_node(bus);
1271             uint32_t uid;
1272 
1273             /* look only for expander root buses */
1274             if (!pci_bus_is_root(bus)) {
1275                 continue;
1276             }
1277 
1278             if (bus_num < root_bus_limit) {
1279                 root_bus_limit = bus_num - 1;
1280             }
1281 
1282             uid = object_property_get_uint(OBJECT(bus), "acpi_uid",
1283                                            &error_fatal);
1284             scope = aml_scope("\\_SB");
1285 
1286             if (pci_bus_is_cxl(bus)) {
1287                 dev = aml_device("CL%.02X", bus_num);
1288             } else {
1289                 dev = aml_device("PC%.02X", bus_num);
1290             }
1291             aml_append(dev, aml_name_decl("_UID", aml_int(uid)));
1292             aml_append(dev, aml_name_decl("_BBN", aml_int(bus_num)));
1293             if (pci_bus_is_cxl(bus)) {
1294                 struct Aml *aml_pkg = aml_package(2);
1295 
1296                 aml_append(dev, aml_name_decl("_HID", aml_string("ACPI0016")));
1297                 aml_append(aml_pkg, aml_eisaid("PNP0A08"));
1298                 aml_append(aml_pkg, aml_eisaid("PNP0A03"));
1299                 aml_append(dev, aml_name_decl("_CID", aml_pkg));
1300                 build_cxl_osc_method(dev);
1301             } else if (pci_bus_is_express(bus)) {
1302                 aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0A08")));
1303                 aml_append(dev, aml_name_decl("_CID", aml_eisaid("PNP0A03")));
1304 
1305                 /* Expander bridges do not have ACPI PCI Hot-plug enabled */
1306                 aml_append(dev, build_pci_host_bridge_osc_method(true));
1307             } else {
1308                 aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0A03")));
1309             }
1310 
1311             if (numa_node != NUMA_NODE_UNASSIGNED) {
1312                 aml_append(dev, aml_name_decl("_PXM", aml_int(numa_node)));
1313             }
1314 
1315             aml_append(dev, build_prt(false));
1316             crs = build_crs(PCI_HOST_BRIDGE(BUS(bus)->parent), &crs_range_set,
1317                             0, 0, 0, 0);
1318             aml_append(dev, aml_name_decl("_CRS", crs));
1319             aml_append(scope, dev);
1320             aml_append(dsdt, scope);
1321 
1322             /* Handle the ranges for the PXB expanders */
1323             if (pci_bus_is_cxl(bus)) {
1324                 MemoryRegion *mr = &pcms->cxl_devices_state.host_mr;
1325                 uint64_t base = mr->addr;
1326 
1327                 cxl_present = true;
1328                 crs_range_insert(crs_range_set.mem_ranges, base,
1329                                  base + memory_region_size(mr) - 1);
1330             }
1331         }
1332     }
1333 
1334     if (cxl_present) {
1335         build_acpi0017(dsdt);
1336     }
1337 
1338     /*
1339      * At this point crs_range_set has all the ranges used by pci
1340      * busses *other* than PCI0.  These ranges will be excluded from
1341      * the PCI0._CRS.  Add mmconfig to the set so it will be excluded
1342      * too.
1343      */
1344     if (mcfg_valid) {
1345         crs_range_insert(crs_range_set.mem_ranges,
1346                          mcfg.base, mcfg.base + mcfg.size - 1);
1347     }
1348 
1349     scope = aml_scope("\\_SB.PCI0");
1350     /* build PCI0._CRS */
1351     crs = aml_resource_template();
1352     aml_append(crs,
1353         aml_word_bus_number(AML_MIN_FIXED, AML_MAX_FIXED, AML_POS_DECODE,
1354                             0x0000, 0x0, root_bus_limit,
1355                             0x0000, root_bus_limit + 1));
1356     aml_append(crs, aml_io(AML_DECODE16, 0x0CF8, 0x0CF8, 0x01, 0x08));
1357 
1358     aml_append(crs,
1359         aml_word_io(AML_MIN_FIXED, AML_MAX_FIXED,
1360                     AML_POS_DECODE, AML_ENTIRE_RANGE,
1361                     0x0000, 0x0000, 0x0CF7, 0x0000, 0x0CF8));
1362 
1363     crs_replace_with_free_ranges(crs_range_set.io_ranges, 0x0D00, 0xFFFF);
1364     for (i = 0; i < crs_range_set.io_ranges->len; i++) {
1365         entry = g_ptr_array_index(crs_range_set.io_ranges, i);
1366         aml_append(crs,
1367             aml_word_io(AML_MIN_FIXED, AML_MAX_FIXED,
1368                         AML_POS_DECODE, AML_ENTIRE_RANGE,
1369                         0x0000, entry->base, entry->limit,
1370                         0x0000, entry->limit - entry->base + 1));
1371     }
1372 
1373     aml_append(crs,
1374         aml_dword_memory(AML_POS_DECODE, AML_MIN_FIXED, AML_MAX_FIXED,
1375                          AML_CACHEABLE, AML_READ_WRITE,
1376                          0, 0x000A0000, 0x000BFFFF, 0, 0x00020000));
1377 
1378     crs_replace_with_free_ranges(crs_range_set.mem_ranges,
1379                                  range_lob(pci_hole),
1380                                  range_upb(pci_hole));
1381     for (i = 0; i < crs_range_set.mem_ranges->len; i++) {
1382         entry = g_ptr_array_index(crs_range_set.mem_ranges, i);
1383         aml_append(crs,
1384             aml_dword_memory(AML_POS_DECODE, AML_MIN_FIXED, AML_MAX_FIXED,
1385                              AML_NON_CACHEABLE, AML_READ_WRITE,
1386                              0, entry->base, entry->limit,
1387                              0, entry->limit - entry->base + 1));
1388     }
1389 
1390     if (!range_is_empty(pci_hole64)) {
1391         crs_replace_with_free_ranges(crs_range_set.mem_64bit_ranges,
1392                                      range_lob(pci_hole64),
1393                                      range_upb(pci_hole64));
1394         for (i = 0; i < crs_range_set.mem_64bit_ranges->len; i++) {
1395             entry = g_ptr_array_index(crs_range_set.mem_64bit_ranges, i);
1396             aml_append(crs,
1397                        aml_qword_memory(AML_POS_DECODE, AML_MIN_FIXED,
1398                                         AML_MAX_FIXED,
1399                                         AML_CACHEABLE, AML_READ_WRITE,
1400                                         0, entry->base, entry->limit,
1401                                         0, entry->limit - entry->base + 1));
1402         }
1403     }
1404 
1405 #ifdef CONFIG_TPM
1406     if (TPM_IS_TIS_ISA(tpm_find())) {
1407         aml_append(crs, aml_memory32_fixed(TPM_TIS_ADDR_BASE,
1408                    TPM_TIS_ADDR_SIZE, AML_READ_WRITE));
1409     }
1410 #endif
1411     aml_append(scope, aml_name_decl("_CRS", crs));
1412 
1413     /* reserve GPE0 block resources */
1414     dev = aml_device("GPE0");
1415     aml_append(dev, aml_name_decl("_HID", aml_string("PNP0A06")));
1416     aml_append(dev, aml_name_decl("_UID", aml_string("GPE0 resources")));
1417     /* device present, functioning, decoding, not shown in UI */
1418     aml_append(dev, aml_name_decl("_STA", aml_int(0xB)));
1419     crs = aml_resource_template();
1420     aml_append(crs,
1421         aml_io(
1422                AML_DECODE16,
1423                pm->fadt.gpe0_blk.address,
1424                pm->fadt.gpe0_blk.address,
1425                1,
1426                pm->fadt.gpe0_blk.bit_width / 8)
1427     );
1428     aml_append(dev, aml_name_decl("_CRS", crs));
1429     aml_append(scope, dev);
1430 
1431     crs_range_set_free(&crs_range_set);
1432 
1433     /* reserve PCIHP resources */
1434     if (pm->pcihp_io_len && (pm->pcihp_bridge_en || pm->pcihp_root_en)) {
1435         build_append_pcihp_resources(scope,
1436                                       pm->pcihp_io_base, pm->pcihp_io_len);
1437     }
1438     aml_append(dsdt, scope);
1439 
1440     /*  create S3_ / S4_ / S5_ packages if necessary */
1441     scope = aml_scope("\\");
1442     if (!pm->s3_disabled) {
1443         pkg = aml_package(4);
1444         aml_append(pkg, aml_int(1)); /* PM1a_CNT.SLP_TYP */
1445         aml_append(pkg, aml_int(1)); /* PM1b_CNT.SLP_TYP, FIXME: not impl. */
1446         aml_append(pkg, aml_int(0)); /* reserved */
1447         aml_append(pkg, aml_int(0)); /* reserved */
1448         aml_append(scope, aml_name_decl("_S3", pkg));
1449     }
1450 
1451     if (!pm->s4_disabled) {
1452         pkg = aml_package(4);
1453         aml_append(pkg, aml_int(pm->s4_val)); /* PM1a_CNT.SLP_TYP */
1454         /* PM1b_CNT.SLP_TYP, FIXME: not impl. */
1455         aml_append(pkg, aml_int(pm->s4_val));
1456         aml_append(pkg, aml_int(0)); /* reserved */
1457         aml_append(pkg, aml_int(0)); /* reserved */
1458         aml_append(scope, aml_name_decl("_S4", pkg));
1459     }
1460 
1461     pkg = aml_package(4);
1462     aml_append(pkg, aml_int(0)); /* PM1a_CNT.SLP_TYP */
1463     aml_append(pkg, aml_int(0)); /* PM1b_CNT.SLP_TYP not impl. */
1464     aml_append(pkg, aml_int(0)); /* reserved */
1465     aml_append(pkg, aml_int(0)); /* reserved */
1466     aml_append(scope, aml_name_decl("_S5", pkg));
1467     aml_append(dsdt, scope);
1468 
1469     /* create fw_cfg node, unconditionally */
1470     {
1471         scope = aml_scope("\\_SB.PCI0");
1472         fw_cfg_add_acpi_dsdt(scope, x86ms->fw_cfg);
1473         aml_append(dsdt, scope);
1474     }
1475 
1476     sb_scope = aml_scope("\\_SB");
1477     {
1478         Object *pci_host = acpi_get_i386_pci_host();
1479 
1480         if (pci_host) {
1481             PCIBus *pbus = PCI_HOST_BRIDGE(pci_host)->bus;
1482             Aml *ascope = aml_scope("PCI0");
1483             /* Scan all PCI buses. Generate tables to support hotplug. */
1484             build_append_pci_bus_devices(ascope, pbus);
1485             if (object_property_find(OBJECT(pbus), ACPI_PCIHP_PROP_BSEL)) {
1486                 build_append_pcihp_slots(ascope, pbus);
1487             }
1488             aml_append(sb_scope, ascope);
1489         }
1490     }
1491 
1492 #ifdef CONFIG_TPM
1493     if (TPM_IS_CRB(tpm)) {
1494         dev = aml_device("TPM");
1495         aml_append(dev, aml_name_decl("_HID", aml_string("MSFT0101")));
1496         aml_append(dev, aml_name_decl("_STR",
1497                                       aml_string("TPM 2.0 Device")));
1498         crs = aml_resource_template();
1499         aml_append(crs, aml_memory32_fixed(TPM_CRB_ADDR_BASE,
1500                                            TPM_CRB_ADDR_SIZE, AML_READ_WRITE));
1501         aml_append(dev, aml_name_decl("_CRS", crs));
1502 
1503         aml_append(dev, aml_name_decl("_STA", aml_int(0xf)));
1504         aml_append(dev, aml_name_decl("_UID", aml_int(1)));
1505 
1506         tpm_build_ppi_acpi(tpm, dev);
1507 
1508         aml_append(sb_scope, dev);
1509     }
1510 #endif
1511 
1512     if (pcms->sgx_epc.size != 0) {
1513         uint64_t epc_base = pcms->sgx_epc.base;
1514         uint64_t epc_size = pcms->sgx_epc.size;
1515 
1516         dev = aml_device("EPC");
1517         aml_append(dev, aml_name_decl("_HID", aml_eisaid("INT0E0C")));
1518         aml_append(dev, aml_name_decl("_STR",
1519                                       aml_unicode("Enclave Page Cache 1.0")));
1520         crs = aml_resource_template();
1521         aml_append(crs,
1522                    aml_qword_memory(AML_POS_DECODE, AML_MIN_FIXED,
1523                                     AML_MAX_FIXED, AML_NON_CACHEABLE,
1524                                     AML_READ_WRITE, 0, epc_base,
1525                                     epc_base + epc_size - 1, 0, epc_size));
1526         aml_append(dev, aml_name_decl("_CRS", crs));
1527 
1528         method = aml_method("_STA", 0, AML_NOTSERIALIZED);
1529         aml_append(method, aml_return(aml_int(0x0f)));
1530         aml_append(dev, method);
1531 
1532         aml_append(sb_scope, dev);
1533     }
1534     aml_append(dsdt, sb_scope);
1535 
1536     if (pm->pcihp_bridge_en || pm->pcihp_root_en) {
1537         bool has_pcnt;
1538 
1539         Object *pci_host = acpi_get_i386_pci_host();
1540         PCIBus *b = PCI_HOST_BRIDGE(pci_host)->bus;
1541 
1542         scope = aml_scope("\\_SB.PCI0");
1543         has_pcnt = build_append_notification_callback(scope, b);
1544         if (has_pcnt) {
1545             aml_append(dsdt, scope);
1546         }
1547 
1548         scope =  aml_scope("_GPE");
1549         {
1550             method = aml_method("_E01", 0, AML_NOTSERIALIZED);
1551             if (has_pcnt) {
1552                 aml_append(method,
1553                     aml_acquire(aml_name("\\_SB.PCI0.BLCK"), 0xFFFF));
1554                 aml_append(method, aml_call0("\\_SB.PCI0.PCNT"));
1555                 aml_append(method, aml_release(aml_name("\\_SB.PCI0.BLCK")));
1556             }
1557             aml_append(scope, method);
1558         }
1559         aml_append(dsdt, scope);
1560     }
1561 
1562     /* copy AML table into ACPI tables blob and patch header there */
1563     g_array_append_vals(table_data, dsdt->buf->data, dsdt->buf->len);
1564     acpi_table_end(linker, &table);
1565     free_aml_allocator();
1566 }
1567 
1568 /*
1569  * IA-PC HPET (High Precision Event Timers) Specification (Revision: 1.0a)
1570  * 3.2.4The ACPI 2.0 HPET Description Table (HPET)
1571  */
1572 static void
1573 build_hpet(GArray *table_data, BIOSLinker *linker, const char *oem_id,
1574            const char *oem_table_id)
1575 {
1576     AcpiTable table = { .sig = "HPET", .rev = 1,
1577                         .oem_id = oem_id, .oem_table_id = oem_table_id };
1578 
1579     acpi_table_begin(&table, table_data);
1580     /* Note timer_block_id value must be kept in sync with value advertised by
1581      * emulated hpet
1582      */
1583     /* Event Timer Block ID */
1584     build_append_int_noprefix(table_data, 0x8086a201, 4);
1585     /* BASE_ADDRESS */
1586     build_append_gas(table_data, AML_AS_SYSTEM_MEMORY, 0, 0, 0, HPET_BASE);
1587     /* HPET Number */
1588     build_append_int_noprefix(table_data, 0, 1);
1589     /* Main Counter Minimum Clock_tick in Periodic Mode */
1590     build_append_int_noprefix(table_data, 0, 2);
1591     /* Page Protection And OEM Attribute */
1592     build_append_int_noprefix(table_data, 0, 1);
1593     acpi_table_end(linker, &table);
1594 }
1595 
1596 #ifdef CONFIG_TPM
1597 /*
1598  * TCPA Description Table
1599  *
1600  * Following Level 00, Rev 00.37 of specs:
1601  * http://www.trustedcomputinggroup.org/resources/tcg_acpi_specification
1602  * 7.1.2 ACPI Table Layout
1603  */
1604 static void
1605 build_tpm_tcpa(GArray *table_data, BIOSLinker *linker, GArray *tcpalog,
1606                const char *oem_id, const char *oem_table_id)
1607 {
1608     unsigned log_addr_offset;
1609     AcpiTable table = { .sig = "TCPA", .rev = 2,
1610                         .oem_id = oem_id, .oem_table_id = oem_table_id };
1611 
1612     acpi_table_begin(&table, table_data);
1613     /* Platform Class */
1614     build_append_int_noprefix(table_data, TPM_TCPA_ACPI_CLASS_CLIENT, 2);
1615     /* Log Area Minimum Length (LAML) */
1616     build_append_int_noprefix(table_data, TPM_LOG_AREA_MINIMUM_SIZE, 4);
1617     /* Log Area Start Address (LASA) */
1618     log_addr_offset = table_data->len;
1619     build_append_int_noprefix(table_data, 0, 8);
1620 
1621     /* allocate/reserve space for TPM log area */
1622     acpi_data_push(tcpalog, TPM_LOG_AREA_MINIMUM_SIZE);
1623     bios_linker_loader_alloc(linker, ACPI_BUILD_TPMLOG_FILE, tcpalog, 1,
1624                              false /* high memory */);
1625     /* log area start address to be filled by Guest linker */
1626     bios_linker_loader_add_pointer(linker, ACPI_BUILD_TABLE_FILE,
1627         log_addr_offset, 8, ACPI_BUILD_TPMLOG_FILE, 0);
1628 
1629     acpi_table_end(linker, &table);
1630 }
1631 #endif
1632 
1633 #define HOLE_640K_START  (640 * KiB)
1634 #define HOLE_640K_END   (1 * MiB)
1635 
1636 /*
1637  * ACPI spec, Revision 3.0
1638  * 5.2.15 System Resource Affinity Table (SRAT)
1639  */
1640 static void
1641 build_srat(GArray *table_data, BIOSLinker *linker, MachineState *machine)
1642 {
1643     int i;
1644     int numa_mem_start, slots;
1645     uint64_t mem_len, mem_base, next_base;
1646     MachineClass *mc = MACHINE_GET_CLASS(machine);
1647     X86MachineState *x86ms = X86_MACHINE(machine);
1648     const CPUArchIdList *apic_ids = mc->possible_cpu_arch_ids(machine);
1649     int nb_numa_nodes = machine->numa_state->num_nodes;
1650     NodeInfo *numa_info = machine->numa_state->nodes;
1651     AcpiTable table = { .sig = "SRAT", .rev = 1, .oem_id = x86ms->oem_id,
1652                         .oem_table_id = x86ms->oem_table_id };
1653 
1654     acpi_table_begin(&table, table_data);
1655     build_append_int_noprefix(table_data, 1, 4); /* Reserved */
1656     build_append_int_noprefix(table_data, 0, 8); /* Reserved */
1657 
1658     for (i = 0; i < apic_ids->len; i++) {
1659         int node_id = apic_ids->cpus[i].props.node_id;
1660         uint32_t apic_id = apic_ids->cpus[i].arch_id;
1661 
1662         if (apic_id < 255) {
1663             /* 5.2.15.1 Processor Local APIC/SAPIC Affinity Structure */
1664             build_append_int_noprefix(table_data, 0, 1);  /* Type  */
1665             build_append_int_noprefix(table_data, 16, 1); /* Length */
1666             /* Proximity Domain [7:0] */
1667             build_append_int_noprefix(table_data, node_id, 1);
1668             build_append_int_noprefix(table_data, apic_id, 1); /* APIC ID */
1669             /* Flags, Table 5-36 */
1670             build_append_int_noprefix(table_data, 1, 4);
1671             build_append_int_noprefix(table_data, 0, 1); /* Local SAPIC EID */
1672             /* Proximity Domain [31:8] */
1673             build_append_int_noprefix(table_data, 0, 3);
1674             build_append_int_noprefix(table_data, 0, 4); /* Reserved */
1675         } else {
1676             /*
1677              * ACPI spec, Revision 4.0
1678              * 5.2.16.3 Processor Local x2APIC Affinity Structure
1679              */
1680             build_append_int_noprefix(table_data, 2, 1);  /* Type  */
1681             build_append_int_noprefix(table_data, 24, 1); /* Length */
1682             build_append_int_noprefix(table_data, 0, 2); /* Reserved */
1683             /* Proximity Domain */
1684             build_append_int_noprefix(table_data, node_id, 4);
1685             build_append_int_noprefix(table_data, apic_id, 4); /* X2APIC ID */
1686             /* Flags, Table 5-39 */
1687             build_append_int_noprefix(table_data, 1 /* Enabled */, 4);
1688             build_append_int_noprefix(table_data, 0, 4); /* Clock Domain */
1689             build_append_int_noprefix(table_data, 0, 4); /* Reserved */
1690         }
1691     }
1692 
1693     /* the memory map is a bit tricky, it contains at least one hole
1694      * from 640k-1M and possibly another one from 3.5G-4G.
1695      */
1696     next_base = 0;
1697     numa_mem_start = table_data->len;
1698 
1699     for (i = 1; i < nb_numa_nodes + 1; ++i) {
1700         mem_base = next_base;
1701         mem_len = numa_info[i - 1].node_mem;
1702         next_base = mem_base + mem_len;
1703 
1704         /* Cut out the 640K hole */
1705         if (mem_base <= HOLE_640K_START &&
1706             next_base > HOLE_640K_START) {
1707             mem_len -= next_base - HOLE_640K_START;
1708             if (mem_len > 0) {
1709                 build_srat_memory(table_data, mem_base, mem_len, i - 1,
1710                                   MEM_AFFINITY_ENABLED);
1711             }
1712 
1713             /* Check for the rare case: 640K < RAM < 1M */
1714             if (next_base <= HOLE_640K_END) {
1715                 next_base = HOLE_640K_END;
1716                 continue;
1717             }
1718             mem_base = HOLE_640K_END;
1719             mem_len = next_base - HOLE_640K_END;
1720         }
1721 
1722         /* Cut out the ACPI_PCI hole */
1723         if (mem_base <= x86ms->below_4g_mem_size &&
1724             next_base > x86ms->below_4g_mem_size) {
1725             mem_len -= next_base - x86ms->below_4g_mem_size;
1726             if (mem_len > 0) {
1727                 build_srat_memory(table_data, mem_base, mem_len, i - 1,
1728                                   MEM_AFFINITY_ENABLED);
1729             }
1730             mem_base = x86ms->above_4g_mem_start;
1731             mem_len = next_base - x86ms->below_4g_mem_size;
1732             next_base = mem_base + mem_len;
1733         }
1734 
1735         if (mem_len > 0) {
1736             build_srat_memory(table_data, mem_base, mem_len, i - 1,
1737                               MEM_AFFINITY_ENABLED);
1738         }
1739     }
1740 
1741     if (machine->nvdimms_state->is_enabled) {
1742         nvdimm_build_srat(table_data);
1743     }
1744 
1745     sgx_epc_build_srat(table_data);
1746 
1747     /*
1748      * TODO: this part is not in ACPI spec and current linux kernel boots fine
1749      * without these entries. But I recall there were issues the last time I
1750      * tried to remove it with some ancient guest OS, however I can't remember
1751      * what that was so keep this around for now
1752      */
1753     slots = (table_data->len - numa_mem_start) / 40 /* mem affinity len */;
1754     for (; slots < nb_numa_nodes + 2; slots++) {
1755         build_srat_memory(table_data, 0, 0, 0, MEM_AFFINITY_NOFLAGS);
1756     }
1757 
1758     build_srat_generic_affinity_structures(table_data);
1759 
1760     /*
1761      * Entry is required for Windows to enable memory hotplug in OS
1762      * and for Linux to enable SWIOTLB when booted with less than
1763      * 4G of RAM. Windows works better if the entry sets proximity
1764      * to the highest NUMA node in the machine.
1765      * Memory devices may override proximity set by this entry,
1766      * providing _PXM method if necessary.
1767      */
1768     if (machine->device_memory) {
1769         build_srat_memory(table_data, machine->device_memory->base,
1770                           memory_region_size(&machine->device_memory->mr),
1771                           nb_numa_nodes - 1,
1772                           MEM_AFFINITY_HOTPLUGGABLE | MEM_AFFINITY_ENABLED);
1773     }
1774 
1775     acpi_table_end(linker, &table);
1776 }
1777 
1778 /*
1779  * Insert DMAR scope for PCI bridges and endpoint devices
1780  */
1781 static void
1782 insert_scope(PCIBus *bus, PCIDevice *dev, void *opaque)
1783 {
1784     const size_t device_scope_size = 6 /* device scope structure */ +
1785                                      2 /* 1 path entry */;
1786     GArray *scope_blob = opaque;
1787 
1788     if (object_dynamic_cast(OBJECT(dev), TYPE_PCI_BRIDGE)) {
1789         /* Dmar Scope Type: 0x02 for PCI Bridge */
1790         build_append_int_noprefix(scope_blob, 0x02, 1);
1791     } else {
1792         /* Dmar Scope Type: 0x01 for PCI Endpoint Device */
1793         build_append_int_noprefix(scope_blob, 0x01, 1);
1794     }
1795 
1796     /* length */
1797     build_append_int_noprefix(scope_blob, device_scope_size, 1);
1798     /* reserved */
1799     build_append_int_noprefix(scope_blob, 0, 2);
1800     /* enumeration_id */
1801     build_append_int_noprefix(scope_blob, 0, 1);
1802     /* bus */
1803     build_append_int_noprefix(scope_blob, pci_bus_num(bus), 1);
1804     /* device */
1805     build_append_int_noprefix(scope_blob, PCI_SLOT(dev->devfn), 1);
1806     /* function */
1807     build_append_int_noprefix(scope_blob, PCI_FUNC(dev->devfn), 1);
1808 }
1809 
1810 /* For a given PCI host bridge, walk and insert DMAR scope */
1811 static int
1812 dmar_host_bridges(Object *obj, void *opaque)
1813 {
1814     GArray *scope_blob = opaque;
1815 
1816     if (object_dynamic_cast(obj, TYPE_PCI_HOST_BRIDGE)) {
1817         PCIBus *bus = PCI_HOST_BRIDGE(obj)->bus;
1818 
1819         if (bus && !pci_bus_bypass_iommu(bus)) {
1820             pci_for_each_device_under_bus(bus, insert_scope, scope_blob);
1821         }
1822     }
1823 
1824     return 0;
1825 }
1826 
1827 /*
1828  * Intel ® Virtualization Technology for Directed I/O
1829  * Architecture Specification. Revision 3.3
1830  * 8.1 DMA Remapping Reporting Structure
1831  */
1832 static void
1833 build_dmar_q35(GArray *table_data, BIOSLinker *linker, const char *oem_id,
1834                const char *oem_table_id)
1835 {
1836     uint8_t dmar_flags = 0;
1837     uint8_t rsvd10[10] = {};
1838     /* Root complex IOAPIC uses one path only */
1839     const size_t ioapic_scope_size = 6 /* device scope structure */ +
1840                                      2 /* 1 path entry */;
1841     X86IOMMUState *iommu = x86_iommu_get_default();
1842     IntelIOMMUState *intel_iommu = INTEL_IOMMU_DEVICE(iommu);
1843     GArray *scope_blob = g_array_new(false, true, 1);
1844 
1845     AcpiTable table = { .sig = "DMAR", .rev = 1, .oem_id = oem_id,
1846                         .oem_table_id = oem_table_id };
1847 
1848     /*
1849      * A PCI bus walk, for each PCI host bridge.
1850      * Insert scope for each PCI bridge and endpoint device which
1851      * is attached to a bus with iommu enabled.
1852      */
1853     object_child_foreach_recursive(object_get_root(),
1854                                    dmar_host_bridges, scope_blob);
1855 
1856     assert(iommu);
1857     if (x86_iommu_ir_supported(iommu)) {
1858         dmar_flags |= 0x1;      /* Flags: 0x1: INT_REMAP */
1859     }
1860 
1861     acpi_table_begin(&table, table_data);
1862     /* Host Address Width */
1863     build_append_int_noprefix(table_data, intel_iommu->aw_bits - 1, 1);
1864     build_append_int_noprefix(table_data, dmar_flags, 1); /* Flags */
1865     g_array_append_vals(table_data, rsvd10, sizeof(rsvd10)); /* Reserved */
1866 
1867     /* 8.3 DMAR Remapping Hardware Unit Definition structure */
1868     build_append_int_noprefix(table_data, 0, 2); /* Type */
1869     /* Length */
1870     build_append_int_noprefix(table_data,
1871                               16 + ioapic_scope_size + scope_blob->len, 2);
1872     /* Flags */
1873     build_append_int_noprefix(table_data, 0 /* Don't include all pci device */ ,
1874                               1);
1875     build_append_int_noprefix(table_data, 0 , 1); /* Reserved */
1876     build_append_int_noprefix(table_data, 0 , 2); /* Segment Number */
1877     /* Register Base Address */
1878     build_append_int_noprefix(table_data, Q35_HOST_BRIDGE_IOMMU_ADDR , 8);
1879 
1880     /* Scope definition for the root-complex IOAPIC. See VT-d spec
1881      * 8.3.1 (version Oct. 2014 or later). */
1882     build_append_int_noprefix(table_data, 0x03 /* IOAPIC */, 1); /* Type */
1883     build_append_int_noprefix(table_data, ioapic_scope_size, 1); /* Length */
1884     build_append_int_noprefix(table_data, 0, 2); /* Reserved */
1885     /* Enumeration ID */
1886     build_append_int_noprefix(table_data, ACPI_BUILD_IOAPIC_ID, 1);
1887     /* Start Bus Number */
1888     build_append_int_noprefix(table_data, Q35_PSEUDO_BUS_PLATFORM, 1);
1889     /* Path, {Device, Function} pair */
1890     build_append_int_noprefix(table_data, PCI_SLOT(Q35_PSEUDO_DEVFN_IOAPIC), 1);
1891     build_append_int_noprefix(table_data, PCI_FUNC(Q35_PSEUDO_DEVFN_IOAPIC), 1);
1892 
1893     /* Add scope found above */
1894     g_array_append_vals(table_data, scope_blob->data, scope_blob->len);
1895     g_array_free(scope_blob, true);
1896 
1897     if (iommu->dt_supported) {
1898         /* 8.5 Root Port ATS Capability Reporting Structure */
1899         build_append_int_noprefix(table_data, 2, 2); /* Type */
1900         build_append_int_noprefix(table_data, 8, 2); /* Length */
1901         build_append_int_noprefix(table_data, 1 /* ALL_PORTS */, 1); /* Flags */
1902         build_append_int_noprefix(table_data, 0, 1); /* Reserved */
1903         build_append_int_noprefix(table_data, 0, 2); /* Segment Number */
1904     }
1905 
1906     acpi_table_end(linker, &table);
1907 }
1908 
1909 /*
1910  * Windows ACPI Emulated Devices Table
1911  * (Version 1.0 - April 6, 2009)
1912  * Spec: http://download.microsoft.com/download/7/E/7/7E7662CF-CBEA-470B-A97E-CE7CE0D98DC2/WAET.docx
1913  *
1914  * Helpful to speedup Windows guests and ignored by others.
1915  */
1916 static void
1917 build_waet(GArray *table_data, BIOSLinker *linker, const char *oem_id,
1918            const char *oem_table_id)
1919 {
1920     AcpiTable table = { .sig = "WAET", .rev = 1, .oem_id = oem_id,
1921                         .oem_table_id = oem_table_id };
1922 
1923     acpi_table_begin(&table, table_data);
1924     /*
1925      * Set "ACPI PM timer good" flag.
1926      *
1927      * Tells Windows guests that our ACPI PM timer is reliable in the
1928      * sense that guest can read it only once to obtain a reliable value.
1929      * Which avoids costly VMExits caused by guest re-reading it unnecessarily.
1930      */
1931     build_append_int_noprefix(table_data, 1 << 1 /* ACPI PM timer good */, 4);
1932     acpi_table_end(linker, &table);
1933 }
1934 
1935 /*
1936  *   IVRS table as specified in AMD IOMMU Specification v2.62, Section 5.2
1937  *   accessible here http://support.amd.com/TechDocs/48882_IOMMU.pdf
1938  */
1939 #define IOAPIC_SB_DEVID   (uint64_t)PCI_BUILD_BDF(0, PCI_DEVFN(0x14, 0))
1940 
1941 /*
1942  * Insert IVHD entry for device and recurse, insert alias, or insert range as
1943  * necessary for the PCI topology.
1944  */
1945 static void
1946 insert_ivhd(PCIBus *bus, PCIDevice *dev, void *opaque)
1947 {
1948     GArray *table_data = opaque;
1949     uint32_t entry;
1950 
1951     /* "Select" IVHD entry, type 0x2 */
1952     entry = PCI_BUILD_BDF(pci_bus_num(bus), dev->devfn) << 8 | 0x2;
1953     build_append_int_noprefix(table_data, entry, 4);
1954 
1955     if (object_dynamic_cast(OBJECT(dev), TYPE_PCI_BRIDGE)) {
1956         PCIBus *sec_bus = pci_bridge_get_sec_bus(PCI_BRIDGE(dev));
1957         uint8_t sec = pci_bus_num(sec_bus);
1958         uint8_t sub = dev->config[PCI_SUBORDINATE_BUS];
1959 
1960         if (pci_bus_is_express(sec_bus)) {
1961             /*
1962              * Walk the bus if there are subordinates, otherwise use a range
1963              * to cover an entire leaf bus.  We could potentially also use a
1964              * range for traversed buses, but we'd need to take care not to
1965              * create both Select and Range entries covering the same device.
1966              * This is easier and potentially more compact.
1967              *
1968              * An example bare metal system seems to use Select entries for
1969              * root ports without a slot (ie. built-ins) and Range entries
1970              * when there is a slot.  The same system also only hard-codes
1971              * the alias range for an onboard PCIe-to-PCI bridge, apparently
1972              * making no effort to support nested bridges.  We attempt to
1973              * be more thorough here.
1974              */
1975             if (sec == sub) { /* leaf bus */
1976                 /* "Start of Range" IVHD entry, type 0x3 */
1977                 entry = PCI_BUILD_BDF(sec, PCI_DEVFN(0, 0)) << 8 | 0x3;
1978                 build_append_int_noprefix(table_data, entry, 4);
1979                 /* "End of Range" IVHD entry, type 0x4 */
1980                 entry = PCI_BUILD_BDF(sub, PCI_DEVFN(31, 7)) << 8 | 0x4;
1981                 build_append_int_noprefix(table_data, entry, 4);
1982             } else {
1983                 pci_for_each_device(sec_bus, sec, insert_ivhd, table_data);
1984             }
1985         } else {
1986             /*
1987              * If the secondary bus is conventional, then we need to create an
1988              * Alias range for everything downstream.  The range covers the
1989              * first devfn on the secondary bus to the last devfn on the
1990              * subordinate bus.  The alias target depends on legacy versus
1991              * express bridges, just as in pci_device_iommu_address_space().
1992              * DeviceIDa vs DeviceIDb as per the AMD IOMMU spec.
1993              */
1994             uint16_t dev_id_a, dev_id_b;
1995 
1996             dev_id_a = PCI_BUILD_BDF(sec, PCI_DEVFN(0, 0));
1997 
1998             if (pci_is_express(dev) &&
1999                 pcie_cap_get_type(dev) == PCI_EXP_TYPE_PCI_BRIDGE) {
2000                 dev_id_b = dev_id_a;
2001             } else {
2002                 dev_id_b = PCI_BUILD_BDF(pci_bus_num(bus), dev->devfn);
2003             }
2004 
2005             /* "Alias Start of Range" IVHD entry, type 0x43, 8 bytes */
2006             build_append_int_noprefix(table_data, dev_id_a << 8 | 0x43, 4);
2007             build_append_int_noprefix(table_data, dev_id_b << 8 | 0x0, 4);
2008 
2009             /* "End of Range" IVHD entry, type 0x4 */
2010             entry = PCI_BUILD_BDF(sub, PCI_DEVFN(31, 7)) << 8 | 0x4;
2011             build_append_int_noprefix(table_data, entry, 4);
2012         }
2013     }
2014 }
2015 
2016 /* For all PCI host bridges, walk and insert IVHD entries */
2017 static int
2018 ivrs_host_bridges(Object *obj, void *opaque)
2019 {
2020     GArray *ivhd_blob = opaque;
2021 
2022     if (object_dynamic_cast(obj, TYPE_PCI_HOST_BRIDGE)) {
2023         PCIBus *bus = PCI_HOST_BRIDGE(obj)->bus;
2024 
2025         if (bus && !pci_bus_bypass_iommu(bus)) {
2026             pci_for_each_device_under_bus(bus, insert_ivhd, ivhd_blob);
2027         }
2028     }
2029 
2030     return 0;
2031 }
2032 
2033 static void
2034 build_amd_iommu(GArray *table_data, BIOSLinker *linker, const char *oem_id,
2035                 const char *oem_table_id)
2036 {
2037     AMDVIState *s = AMD_IOMMU_DEVICE(x86_iommu_get_default());
2038     GArray *ivhd_blob = g_array_new(false, true, 1);
2039     AcpiTable table = { .sig = "IVRS", .rev = 1, .oem_id = oem_id,
2040                         .oem_table_id = oem_table_id };
2041     uint64_t feature_report;
2042 
2043     acpi_table_begin(&table, table_data);
2044     /* IVinfo - IO virtualization information common to all
2045      * IOMMU units in a system
2046      */
2047     build_append_int_noprefix(table_data,
2048                              (1UL << 0) | /* EFRSup */
2049                              (40UL << 8), /* PASize */
2050                              4);
2051     /* reserved */
2052     build_append_int_noprefix(table_data, 0, 8);
2053 
2054     /*
2055      * A PCI bus walk, for each PCI host bridge, is necessary to create a
2056      * complete set of IVHD entries.  Do this into a separate blob so that we
2057      * can calculate the total IVRS table length here and then append the new
2058      * blob further below.  Fall back to an entry covering all devices, which
2059      * is sufficient when no aliases are present.
2060      */
2061     object_child_foreach_recursive(object_get_root(),
2062                                    ivrs_host_bridges, ivhd_blob);
2063 
2064     if (!ivhd_blob->len) {
2065         /*
2066          *   Type 1 device entry reporting all devices
2067          *   These are 4-byte device entries currently reporting the range of
2068          *   Refer to Spec - Table 95:IVHD Device Entry Type Codes(4-byte)
2069          */
2070         build_append_int_noprefix(ivhd_blob, 0x0000001, 4);
2071     }
2072 
2073     /*
2074      * When interrupt remapping is supported, we add a special IVHD device
2075      * for type IO-APIC
2076      * Refer to spec - Table 95: IVHD device entry type codes
2077      *
2078      * Linux IOMMU driver checks for the special IVHD device (type IO-APIC).
2079      * See Linux kernel commit 'c2ff5cf5294bcbd7fa50f7d860e90a66db7e5059'
2080      */
2081     if (x86_iommu_ir_supported(x86_iommu_get_default())) {
2082         build_append_int_noprefix(ivhd_blob,
2083                                  (0x1ull << 56) |           /* type IOAPIC */
2084                                  (IOAPIC_SB_DEVID << 40) |  /* IOAPIC devid */
2085                                  0x48,                      /* special device */
2086                                  8);
2087     }
2088 
2089     /* IVHD definition - type 10h */
2090     build_append_int_noprefix(table_data, 0x10, 1);
2091     /* virtualization flags */
2092     build_append_int_noprefix(table_data,
2093                              (1UL << 0) | /* HtTunEn      */
2094                              (1UL << 4) | /* iotblSup     */
2095                              (1UL << 6) | /* PrefSup      */
2096                              (1UL << 7),  /* PPRSup       */
2097                              1);
2098 
2099     /* IVHD length */
2100     build_append_int_noprefix(table_data, ivhd_blob->len + 24, 2);
2101     /* DeviceID */
2102     build_append_int_noprefix(table_data,
2103                               object_property_get_int(OBJECT(s->pci), "addr",
2104                                                       &error_abort), 2);
2105     /* Capability offset */
2106     build_append_int_noprefix(table_data, s->pci->capab_offset, 2);
2107     /* IOMMU base address */
2108     build_append_int_noprefix(table_data, s->mr_mmio.addr, 8);
2109     /* PCI Segment Group */
2110     build_append_int_noprefix(table_data, 0, 2);
2111     /* IOMMU info */
2112     build_append_int_noprefix(table_data, 0, 2);
2113     /* IOMMU Feature Reporting */
2114     feature_report = (48UL << 30) | /* HATS   */
2115                      (48UL << 28) | /* GATS   */
2116                      (1UL << 2)   | /* GTSup  */
2117                      (1UL << 6);    /* GASup  */
2118     if (s->xtsup) {
2119         feature_report |= (1UL << 0); /* XTSup */
2120     }
2121     build_append_int_noprefix(table_data, feature_report, 4);
2122 
2123     /* IVHD entries as found above */
2124     g_array_append_vals(table_data, ivhd_blob->data, ivhd_blob->len);
2125 
2126    /* IVHD definition - type 11h */
2127     build_append_int_noprefix(table_data, 0x11, 1);
2128     /* virtualization flags */
2129     build_append_int_noprefix(table_data,
2130                              (1UL << 0) | /* HtTunEn      */
2131                              (1UL << 4),  /* iotblSup     */
2132                              1);
2133 
2134     /* IVHD length */
2135     build_append_int_noprefix(table_data, ivhd_blob->len + 40, 2);
2136     /* DeviceID */
2137     build_append_int_noprefix(table_data,
2138                               object_property_get_int(OBJECT(s->pci), "addr",
2139                                                       &error_abort), 2);
2140     /* Capability offset */
2141     build_append_int_noprefix(table_data, s->pci->capab_offset, 2);
2142     /* IOMMU base address */
2143     build_append_int_noprefix(table_data, s->mr_mmio.addr, 8);
2144     /* PCI Segment Group */
2145     build_append_int_noprefix(table_data, 0, 2);
2146     /* IOMMU info */
2147     build_append_int_noprefix(table_data, 0, 2);
2148     /* IOMMU Attributes */
2149     build_append_int_noprefix(table_data, 0, 4);
2150     /* EFR Register Image */
2151     build_append_int_noprefix(table_data,
2152                               amdvi_extended_feature_register(s),
2153                               8);
2154     /* EFR Register Image 2 */
2155     build_append_int_noprefix(table_data, 0, 8);
2156 
2157     /* IVHD entries as found above */
2158     g_array_append_vals(table_data, ivhd_blob->data, ivhd_blob->len);
2159 
2160     g_array_free(ivhd_blob, TRUE);
2161     acpi_table_end(linker, &table);
2162 }
2163 
2164 typedef
2165 struct AcpiBuildState {
2166     /* Copy of table in RAM (for patching). */
2167     MemoryRegion *table_mr;
2168     /* Is table patched? */
2169     uint8_t patched;
2170     MemoryRegion *rsdp_mr;
2171     MemoryRegion *linker_mr;
2172 } AcpiBuildState;
2173 
2174 static bool acpi_get_mcfg(AcpiMcfgInfo *mcfg)
2175 {
2176     Object *pci_host;
2177     QObject *o;
2178 
2179     pci_host = acpi_get_i386_pci_host();
2180     if (!pci_host) {
2181         return false;
2182     }
2183 
2184     o = object_property_get_qobject(pci_host, PCIE_HOST_MCFG_BASE, NULL);
2185     if (!o) {
2186         return false;
2187     }
2188     mcfg->base = qnum_get_uint(qobject_to(QNum, o));
2189     qobject_unref(o);
2190     if (mcfg->base == PCIE_BASE_ADDR_UNMAPPED) {
2191         return false;
2192     }
2193 
2194     o = object_property_get_qobject(pci_host, PCIE_HOST_MCFG_SIZE, NULL);
2195     assert(o);
2196     mcfg->size = qnum_get_uint(qobject_to(QNum, o));
2197     qobject_unref(o);
2198     return true;
2199 }
2200 
2201 static
2202 void acpi_build(AcpiBuildTables *tables, MachineState *machine)
2203 {
2204     PCMachineState *pcms = PC_MACHINE(machine);
2205     X86MachineState *x86ms = X86_MACHINE(machine);
2206     DeviceState *iommu = pcms->iommu;
2207     GArray *table_offsets;
2208     unsigned facs, dsdt, rsdt;
2209     AcpiPmInfo pm;
2210     AcpiMiscInfo misc;
2211     AcpiMcfgInfo mcfg;
2212     Range pci_hole = {}, pci_hole64 = {};
2213     uint8_t *u;
2214     GArray *tables_blob = tables->table_data;
2215     AcpiSlicOem slic_oem = { .id = NULL, .table_id = NULL };
2216     Object *vmgenid_dev, *vmclock_dev;
2217     char *oem_id;
2218     char *oem_table_id;
2219 
2220     acpi_get_pm_info(machine, &pm);
2221     acpi_get_misc_info(&misc);
2222     acpi_get_pci_holes(&pci_hole, &pci_hole64);
2223     acpi_get_slic_oem(&slic_oem);
2224 
2225     if (slic_oem.id) {
2226         oem_id = slic_oem.id;
2227     } else {
2228         oem_id = x86ms->oem_id;
2229     }
2230 
2231     if (slic_oem.table_id) {
2232         oem_table_id = slic_oem.table_id;
2233     } else {
2234         oem_table_id = x86ms->oem_table_id;
2235     }
2236 
2237     table_offsets = g_array_new(false, true /* clear */,
2238                                         sizeof(uint32_t));
2239     ACPI_BUILD_DPRINTF("init ACPI tables\n");
2240 
2241     bios_linker_loader_alloc(tables->linker,
2242                              ACPI_BUILD_TABLE_FILE, tables_blob,
2243                              64 /* Ensure FACS is aligned */,
2244                              false /* high memory */);
2245 
2246     /*
2247      * FACS is pointed to by FADT.
2248      * We place it first since it's the only table that has alignment
2249      * requirements.
2250      */
2251     facs = tables_blob->len;
2252     build_facs(tables_blob);
2253 
2254     /* DSDT is pointed to by FADT */
2255     dsdt = tables_blob->len;
2256     build_dsdt(tables_blob, tables->linker, &pm, &misc,
2257                &pci_hole, &pci_hole64, machine);
2258 
2259     /* ACPI tables pointed to by RSDT */
2260     acpi_add_table(table_offsets, tables_blob);
2261     pm.fadt.facs_tbl_offset = &facs;
2262     pm.fadt.dsdt_tbl_offset = &dsdt;
2263     pm.fadt.xdsdt_tbl_offset = &dsdt;
2264     build_fadt(tables_blob, tables->linker, &pm.fadt, oem_id, oem_table_id);
2265 
2266     acpi_add_table(table_offsets, tables_blob);
2267     acpi_build_madt(tables_blob, tables->linker, x86ms,
2268                     x86ms->oem_id, x86ms->oem_table_id);
2269 
2270 #ifdef CONFIG_ACPI_ERST
2271     {
2272         Object *erst_dev;
2273         erst_dev = find_erst_dev();
2274         if (erst_dev) {
2275             acpi_add_table(table_offsets, tables_blob);
2276             build_erst(tables_blob, tables->linker, erst_dev,
2277                        x86ms->oem_id, x86ms->oem_table_id);
2278         }
2279     }
2280 #endif
2281 
2282     vmgenid_dev = find_vmgenid_dev();
2283     if (vmgenid_dev) {
2284         acpi_add_table(table_offsets, tables_blob);
2285         vmgenid_build_acpi(VMGENID(vmgenid_dev), tables_blob,
2286                            tables->vmgenid, tables->linker, x86ms->oem_id);
2287     }
2288 
2289     vmclock_dev = find_vmclock_dev();
2290     if (vmclock_dev) {
2291         acpi_add_table(table_offsets, tables_blob);
2292         vmclock_build_acpi(VMCLOCK(vmclock_dev), tables_blob, tables->linker,
2293                            x86ms->oem_id);
2294     }
2295 
2296     if (misc.has_hpet) {
2297         acpi_add_table(table_offsets, tables_blob);
2298         build_hpet(tables_blob, tables->linker, x86ms->oem_id,
2299                    x86ms->oem_table_id);
2300     }
2301 #ifdef CONFIG_TPM
2302     if (misc.tpm_version != TPM_VERSION_UNSPEC) {
2303         if (misc.tpm_version == TPM_VERSION_1_2) {
2304             acpi_add_table(table_offsets, tables_blob);
2305             build_tpm_tcpa(tables_blob, tables->linker, tables->tcpalog,
2306                            x86ms->oem_id, x86ms->oem_table_id);
2307         } else { /* TPM_VERSION_2_0 */
2308             acpi_add_table(table_offsets, tables_blob);
2309             build_tpm2(tables_blob, tables->linker, tables->tcpalog,
2310                        x86ms->oem_id, x86ms->oem_table_id);
2311         }
2312     }
2313 #endif
2314     if (machine->numa_state->num_nodes) {
2315         acpi_add_table(table_offsets, tables_blob);
2316         build_srat(tables_blob, tables->linker, machine);
2317         if (machine->numa_state->have_numa_distance) {
2318             acpi_add_table(table_offsets, tables_blob);
2319             build_slit(tables_blob, tables->linker, machine, x86ms->oem_id,
2320                        x86ms->oem_table_id);
2321         }
2322         if (machine->numa_state->hmat_enabled) {
2323             acpi_add_table(table_offsets, tables_blob);
2324             build_hmat(tables_blob, tables->linker, machine->numa_state,
2325                        x86ms->oem_id, x86ms->oem_table_id);
2326         }
2327     }
2328     if (acpi_get_mcfg(&mcfg)) {
2329         acpi_add_table(table_offsets, tables_blob);
2330         build_mcfg(tables_blob, tables->linker, &mcfg, x86ms->oem_id,
2331                    x86ms->oem_table_id);
2332     }
2333     if (object_dynamic_cast(OBJECT(iommu), TYPE_AMD_IOMMU_DEVICE)) {
2334         acpi_add_table(table_offsets, tables_blob);
2335         build_amd_iommu(tables_blob, tables->linker, x86ms->oem_id,
2336                         x86ms->oem_table_id);
2337     } else if (object_dynamic_cast(OBJECT(iommu), TYPE_INTEL_IOMMU_DEVICE)) {
2338         acpi_add_table(table_offsets, tables_blob);
2339         build_dmar_q35(tables_blob, tables->linker, x86ms->oem_id,
2340                        x86ms->oem_table_id);
2341     } else if (object_dynamic_cast(OBJECT(iommu), TYPE_VIRTIO_IOMMU_PCI)) {
2342         PCIDevice *pdev = PCI_DEVICE(iommu);
2343 
2344         acpi_add_table(table_offsets, tables_blob);
2345         build_viot(machine, tables_blob, tables->linker, pci_get_bdf(pdev),
2346                    x86ms->oem_id, x86ms->oem_table_id);
2347     }
2348     if (machine->nvdimms_state->is_enabled) {
2349         nvdimm_build_acpi(table_offsets, tables_blob, tables->linker,
2350                           machine->nvdimms_state, machine->ram_slots,
2351                           x86ms->oem_id, x86ms->oem_table_id);
2352     }
2353     if (pcms->cxl_devices_state.is_enabled) {
2354         cxl_build_cedt(table_offsets, tables_blob, tables->linker,
2355                        x86ms->oem_id, x86ms->oem_table_id, &pcms->cxl_devices_state);
2356     }
2357 
2358     acpi_add_table(table_offsets, tables_blob);
2359     build_waet(tables_blob, tables->linker, x86ms->oem_id, x86ms->oem_table_id);
2360 
2361     /* Add tables supplied by user (if any) */
2362     for (u = acpi_table_first(); u; u = acpi_table_next(u)) {
2363         unsigned len = acpi_table_len(u);
2364 
2365         acpi_add_table(table_offsets, tables_blob);
2366         g_array_append_vals(tables_blob, u, len);
2367     }
2368 
2369     /* RSDT is pointed to by RSDP */
2370     rsdt = tables_blob->len;
2371     build_rsdt(tables_blob, tables->linker, table_offsets,
2372                oem_id, oem_table_id);
2373 
2374     /* RSDP is in FSEG memory, so allocate it separately */
2375     {
2376         AcpiRsdpData rsdp_data = {
2377             .revision = 0,
2378             .oem_id = x86ms->oem_id,
2379             .xsdt_tbl_offset = NULL,
2380             .rsdt_tbl_offset = &rsdt,
2381         };
2382         build_rsdp(tables->rsdp, tables->linker, &rsdp_data);
2383     }
2384 
2385     /* We'll expose it all to Guest so we want to reduce
2386      * chance of size changes.
2387      *
2388      * We used to align the tables to 4k, but of course this would
2389      * too simple to be enough.  4k turned out to be too small an
2390      * alignment very soon, and in fact it is almost impossible to
2391      * keep the table size stable for all (max_cpus, max_memory_slots)
2392      * combinations.
2393      */
2394     acpi_align_size(tables_blob, ACPI_BUILD_TABLE_SIZE);
2395 
2396     acpi_align_size(tables->linker->cmd_blob, ACPI_BUILD_ALIGN_SIZE);
2397 
2398     /* Cleanup memory that's no longer used. */
2399     g_array_free(table_offsets, true);
2400     g_free(slic_oem.id);
2401     g_free(slic_oem.table_id);
2402 }
2403 
2404 static void acpi_ram_update(MemoryRegion *mr, GArray *data)
2405 {
2406     uint32_t size = acpi_data_len(data);
2407 
2408     /* Make sure RAM size is correct - in case it got changed e.g. by migration */
2409     memory_region_ram_resize(mr, size, &error_abort);
2410 
2411     memcpy(memory_region_get_ram_ptr(mr), data->data, size);
2412     memory_region_set_dirty(mr, 0, size);
2413 }
2414 
2415 static void acpi_build_update(void *build_opaque)
2416 {
2417     AcpiBuildState *build_state = build_opaque;
2418     AcpiBuildTables tables;
2419 
2420     /* No state to update or already patched? Nothing to do. */
2421     if (!build_state || build_state->patched) {
2422         return;
2423     }
2424     build_state->patched = 1;
2425 
2426     acpi_build_tables_init(&tables);
2427 
2428     acpi_build(&tables, MACHINE(qdev_get_machine()));
2429 
2430     acpi_ram_update(build_state->table_mr, tables.table_data);
2431 
2432     acpi_ram_update(build_state->rsdp_mr, tables.rsdp);
2433 
2434     acpi_ram_update(build_state->linker_mr, tables.linker->cmd_blob);
2435     acpi_build_tables_cleanup(&tables, true);
2436 }
2437 
2438 static void acpi_build_reset(void *build_opaque)
2439 {
2440     AcpiBuildState *build_state = build_opaque;
2441     build_state->patched = 0;
2442 }
2443 
2444 static const VMStateDescription vmstate_acpi_build = {
2445     .name = "acpi_build",
2446     .version_id = 1,
2447     .minimum_version_id = 1,
2448     .fields = (const VMStateField[]) {
2449         VMSTATE_UINT8(patched, AcpiBuildState),
2450         VMSTATE_END_OF_LIST()
2451     },
2452 };
2453 
2454 void acpi_setup(void)
2455 {
2456     PCMachineState *pcms = PC_MACHINE(qdev_get_machine());
2457     X86MachineState *x86ms = X86_MACHINE(pcms);
2458     AcpiBuildTables tables;
2459     AcpiBuildState *build_state;
2460     Object *vmgenid_dev;
2461 #ifdef CONFIG_TPM
2462     TPMIf *tpm;
2463     static FwCfgTPMConfig tpm_config;
2464 #endif
2465 
2466     if (!x86ms->fw_cfg) {
2467         ACPI_BUILD_DPRINTF("No fw cfg. Bailing out.\n");
2468         return;
2469     }
2470 
2471     if (!pcms->acpi_build_enabled) {
2472         ACPI_BUILD_DPRINTF("ACPI build disabled. Bailing out.\n");
2473         return;
2474     }
2475 
2476     if (!x86_machine_is_acpi_enabled(X86_MACHINE(pcms))) {
2477         ACPI_BUILD_DPRINTF("ACPI disabled. Bailing out.\n");
2478         return;
2479     }
2480 
2481     build_state = g_malloc0(sizeof *build_state);
2482 
2483     acpi_build_tables_init(&tables);
2484     acpi_build(&tables, MACHINE(pcms));
2485 
2486     /* Now expose it all to Guest */
2487     build_state->table_mr = acpi_add_rom_blob(acpi_build_update,
2488                                               build_state, tables.table_data,
2489                                               ACPI_BUILD_TABLE_FILE);
2490     assert(build_state->table_mr != NULL);
2491 
2492     build_state->linker_mr =
2493         acpi_add_rom_blob(acpi_build_update, build_state,
2494                           tables.linker->cmd_blob, ACPI_BUILD_LOADER_FILE);
2495 
2496 #ifdef CONFIG_TPM
2497     fw_cfg_add_file(x86ms->fw_cfg, ACPI_BUILD_TPMLOG_FILE,
2498                     tables.tcpalog->data, acpi_data_len(tables.tcpalog));
2499 
2500     tpm = tpm_find();
2501     if (tpm && object_property_get_bool(OBJECT(tpm), "ppi", &error_abort)) {
2502         tpm_config = (FwCfgTPMConfig) {
2503             .tpmppi_address = cpu_to_le32(TPM_PPI_ADDR_BASE),
2504             .tpm_version = tpm_get_version(tpm),
2505             .tpmppi_version = TPM_PPI_VERSION_1_30
2506         };
2507         fw_cfg_add_file(x86ms->fw_cfg, "etc/tpm/config",
2508                         &tpm_config, sizeof tpm_config);
2509     }
2510 #endif
2511 
2512     vmgenid_dev = find_vmgenid_dev();
2513     if (vmgenid_dev) {
2514         vmgenid_add_fw_cfg(VMGENID(vmgenid_dev), x86ms->fw_cfg,
2515                            tables.vmgenid);
2516     }
2517 
2518     build_state->rsdp_mr = acpi_add_rom_blob(acpi_build_update,
2519                                              build_state, tables.rsdp,
2520                                              ACPI_BUILD_RSDP_FILE);
2521 
2522     qemu_register_reset(acpi_build_reset, build_state);
2523     acpi_build_reset(build_state);
2524     vmstate_register(NULL, 0, &vmstate_acpi_build, build_state);
2525 
2526     /* Cleanup tables but don't free the memory: we track it
2527      * in build_state.
2528      */
2529     acpi_build_tables_cleanup(&tables, false);
2530 }
2531