xref: /openbmc/qemu/hw/i386/acpi-build.c (revision d962f199c7e8625c91e8d1cc54a0323180e6f3e1)
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 /*
567  * build_prt - Define interrupt routing rules
568  *
569  * Returns an array of 128 routes, one for each device,
570  * based on device location.
571  * The main goal is to equally distribute the interrupts
572  * over the 4 existing ACPI links (works only for i440fx).
573  * The hash function is: (slot + pin) & 3 -> "LNK[D|A|B|C]".
574  *
575  */
576 static Aml *build_prt(bool is_pci0_prt)
577 {
578     const int nroutes = 128;
579     Aml *rt_pkg, *method;
580     int pin;
581 
582     method = aml_method("_PRT", 0, AML_NOTSERIALIZED);
583     assert(nroutes < 256);
584     rt_pkg = aml_package(nroutes);
585 
586     for (pin = 0; pin < nroutes; pin++) {
587         Aml *pkg = aml_package(4);
588         int slot = pin >> 2;
589 
590         aml_append(pkg, aml_int((slot << 16) | 0xFFFF));
591         aml_append(pkg, aml_int(pin & 3));
592         /* device 1 is the power-management device, needs SCI */
593         if (is_pci0_prt && pin == 4) {
594             aml_append(pkg, aml_name("%s", "LNKS"));
595         } else {
596             static const char link_name[][5] = {"LNKD", "LNKA", "LNKB", "LNKC"};
597             int hash = (slot + pin) & 3;
598             aml_append(pkg, aml_name("%s", link_name[hash]));
599         }
600         aml_append(pkg, aml_int(0));
601         aml_append(rt_pkg, pkg);
602     }
603 
604     aml_append(method, aml_return(rt_pkg));
605 
606     return method;
607 }
608 
609 static void build_hpet_aml(Aml *table)
610 {
611     Aml *crs;
612     Aml *field;
613     Aml *method;
614     Aml *if_ctx;
615     Aml *scope = aml_scope("_SB");
616     Aml *dev = aml_device("HPET");
617     Aml *zero = aml_int(0);
618     Aml *id = aml_local(0);
619     Aml *period = aml_local(1);
620 
621     aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0103")));
622     aml_append(dev, aml_name_decl("_UID", zero));
623 
624     aml_append(dev,
625         aml_operation_region("HPTM", AML_SYSTEM_MEMORY, aml_int(HPET_BASE),
626                              HPET_LEN));
627     field = aml_field("HPTM", AML_DWORD_ACC, AML_LOCK, AML_PRESERVE);
628     aml_append(field, aml_named_field("VEND", 32));
629     aml_append(field, aml_named_field("PRD", 32));
630     aml_append(dev, field);
631 
632     method = aml_method("_STA", 0, AML_NOTSERIALIZED);
633     aml_append(method, aml_store(aml_name("VEND"), id));
634     aml_append(method, aml_store(aml_name("PRD"), period));
635     aml_append(method, aml_shiftright(id, aml_int(16), id));
636     if_ctx = aml_if(aml_lor(aml_equal(id, zero),
637                             aml_equal(id, aml_int(0xffff))));
638     {
639         aml_append(if_ctx, aml_return(zero));
640     }
641     aml_append(method, if_ctx);
642 
643     if_ctx = aml_if(aml_lor(aml_equal(period, zero),
644                             aml_lgreater(period, aml_int(100000000))));
645     {
646         aml_append(if_ctx, aml_return(zero));
647     }
648     aml_append(method, if_ctx);
649 
650     aml_append(method, aml_return(aml_int(0x0F)));
651     aml_append(dev, method);
652 
653     crs = aml_resource_template();
654     aml_append(crs, aml_memory32_fixed(HPET_BASE, HPET_LEN, AML_READ_ONLY));
655     aml_append(dev, aml_name_decl("_CRS", crs));
656 
657     aml_append(scope, dev);
658     aml_append(table, scope);
659 }
660 
661 static Aml *build_vmbus_device_aml(VMBusBridge *vmbus_bridge)
662 {
663     Aml *dev;
664     Aml *method;
665     Aml *crs;
666 
667     dev = aml_device("VMBS");
668     aml_append(dev, aml_name_decl("STA", aml_int(0xF)));
669     aml_append(dev, aml_name_decl("_HID", aml_string("VMBus")));
670     aml_append(dev, aml_name_decl("_UID", aml_int(0x0)));
671     aml_append(dev, aml_name_decl("_DDN", aml_string("VMBUS")));
672 
673     method = aml_method("_DIS", 0, AML_NOTSERIALIZED);
674     aml_append(method, aml_store(aml_and(aml_name("STA"), aml_int(0xD), NULL),
675                                      aml_name("STA")));
676     aml_append(dev, method);
677 
678     method = aml_method("_PS0", 0, AML_NOTSERIALIZED);
679     aml_append(method, aml_store(aml_or(aml_name("STA"), aml_int(0xF), NULL),
680                                      aml_name("STA")));
681     aml_append(dev, method);
682 
683     method = aml_method("_STA", 0, AML_NOTSERIALIZED);
684     aml_append(method, aml_return(aml_name("STA")));
685     aml_append(dev, method);
686 
687     aml_append(dev, aml_name_decl("_PS3", aml_int(0x0)));
688 
689     crs = aml_resource_template();
690     aml_append(crs, aml_irq_no_flags(vmbus_bridge->irq));
691     aml_append(dev, aml_name_decl("_CRS", crs));
692 
693     return dev;
694 }
695 
696 static void build_dbg_aml(Aml *table)
697 {
698     Aml *field;
699     Aml *method;
700     Aml *while_ctx;
701     Aml *scope = aml_scope("\\");
702     Aml *buf = aml_local(0);
703     Aml *len = aml_local(1);
704     Aml *idx = aml_local(2);
705 
706     aml_append(scope,
707        aml_operation_region("DBG", AML_SYSTEM_IO, aml_int(0x0402), 0x01));
708     field = aml_field("DBG", AML_BYTE_ACC, AML_NOLOCK, AML_PRESERVE);
709     aml_append(field, aml_named_field("DBGB", 8));
710     aml_append(scope, field);
711 
712     method = aml_method("DBUG", 1, AML_NOTSERIALIZED);
713 
714     aml_append(method, aml_to_hexstring(aml_arg(0), buf));
715     aml_append(method, aml_to_buffer(buf, buf));
716     aml_append(method, aml_subtract(aml_sizeof(buf), aml_int(1), len));
717     aml_append(method, aml_store(aml_int(0), idx));
718 
719     while_ctx = aml_while(aml_lless(idx, len));
720     aml_append(while_ctx,
721         aml_store(aml_derefof(aml_index(buf, idx)), aml_name("DBGB")));
722     aml_append(while_ctx, aml_increment(idx));
723     aml_append(method, while_ctx);
724 
725     aml_append(method, aml_store(aml_int(0x0A), aml_name("DBGB")));
726     aml_append(scope, method);
727 
728     aml_append(table, scope);
729 }
730 
731 static Aml *build_link_dev(const char *name, uint8_t uid, Aml *reg)
732 {
733     Aml *dev;
734     Aml *crs;
735     Aml *method;
736     uint32_t irqs[] = {5, 10, 11};
737 
738     dev = aml_device("%s", name);
739     aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0C0F")));
740     aml_append(dev, aml_name_decl("_UID", aml_int(uid)));
741 
742     crs = aml_resource_template();
743     aml_append(crs, aml_interrupt(AML_CONSUMER, AML_LEVEL, AML_ACTIVE_HIGH,
744                                   AML_SHARED, irqs, ARRAY_SIZE(irqs)));
745     aml_append(dev, aml_name_decl("_PRS", crs));
746 
747     method = aml_method("_STA", 0, AML_NOTSERIALIZED);
748     aml_append(method, aml_return(aml_call1("IQST", reg)));
749     aml_append(dev, method);
750 
751     method = aml_method("_DIS", 0, AML_NOTSERIALIZED);
752     aml_append(method, aml_or(reg, aml_int(0x80), reg));
753     aml_append(dev, method);
754 
755     method = aml_method("_CRS", 0, AML_NOTSERIALIZED);
756     aml_append(method, aml_return(aml_call1("IQCR", reg)));
757     aml_append(dev, method);
758 
759     method = aml_method("_SRS", 1, AML_NOTSERIALIZED);
760     aml_append(method, aml_create_dword_field(aml_arg(0), aml_int(5), "PRRI"));
761     aml_append(method, aml_store(aml_name("PRRI"), reg));
762     aml_append(dev, method);
763 
764     return dev;
765  }
766 
767 static Aml *build_gsi_link_dev(const char *name, uint8_t uid, uint8_t gsi)
768 {
769     Aml *dev;
770     Aml *crs;
771     Aml *method;
772     uint32_t irqs;
773 
774     dev = aml_device("%s", name);
775     aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0C0F")));
776     aml_append(dev, aml_name_decl("_UID", aml_int(uid)));
777 
778     crs = aml_resource_template();
779     irqs = gsi;
780     aml_append(crs, aml_interrupt(AML_CONSUMER, AML_LEVEL, AML_ACTIVE_HIGH,
781                                   AML_SHARED, &irqs, 1));
782     aml_append(dev, aml_name_decl("_PRS", crs));
783 
784     aml_append(dev, aml_name_decl("_CRS", crs));
785 
786     /*
787      * _DIS can be no-op because the interrupt cannot be disabled.
788      */
789     method = aml_method("_DIS", 0, AML_NOTSERIALIZED);
790     aml_append(dev, method);
791 
792     method = aml_method("_SRS", 1, AML_NOTSERIALIZED);
793     aml_append(dev, method);
794 
795     return dev;
796 }
797 
798 /* _CRS method - get current settings */
799 static Aml *build_iqcr_method(bool is_piix4)
800 {
801     Aml *if_ctx;
802     uint32_t irqs;
803     Aml *method = aml_method("IQCR", 1, AML_SERIALIZED);
804     Aml *crs = aml_resource_template();
805 
806     irqs = 0;
807     aml_append(crs, aml_interrupt(AML_CONSUMER, AML_LEVEL,
808                                   AML_ACTIVE_HIGH, AML_SHARED, &irqs, 1));
809     aml_append(method, aml_name_decl("PRR0", crs));
810 
811     aml_append(method,
812         aml_create_dword_field(aml_name("PRR0"), aml_int(5), "PRRI"));
813 
814     if (is_piix4) {
815         if_ctx = aml_if(aml_lless(aml_arg(0), aml_int(0x80)));
816         aml_append(if_ctx, aml_store(aml_arg(0), aml_name("PRRI")));
817         aml_append(method, if_ctx);
818     } else {
819         aml_append(method,
820             aml_store(aml_and(aml_arg(0), aml_int(0xF), NULL),
821                       aml_name("PRRI")));
822     }
823 
824     aml_append(method, aml_return(aml_name("PRR0")));
825     return method;
826 }
827 
828 /* _STA method - get status */
829 static Aml *build_irq_status_method(void)
830 {
831     Aml *if_ctx;
832     Aml *method = aml_method("IQST", 1, AML_NOTSERIALIZED);
833 
834     if_ctx = aml_if(aml_and(aml_int(0x80), aml_arg(0), NULL));
835     aml_append(if_ctx, aml_return(aml_int(0x09)));
836     aml_append(method, if_ctx);
837     aml_append(method, aml_return(aml_int(0x0B)));
838     return method;
839 }
840 
841 static void build_piix4_pci0_int(Aml *table)
842 {
843     Aml *dev;
844     Aml *crs;
845     Aml *method;
846     uint32_t irqs;
847     Aml *sb_scope = aml_scope("_SB");
848     Aml *pci0_scope = aml_scope("PCI0");
849 
850     aml_append(pci0_scope, build_prt(true));
851     aml_append(sb_scope, pci0_scope);
852 
853     aml_append(sb_scope, build_irq_status_method());
854     aml_append(sb_scope, build_iqcr_method(true));
855 
856     aml_append(sb_scope, build_link_dev("LNKA", 0, aml_name("PRQ0")));
857     aml_append(sb_scope, build_link_dev("LNKB", 1, aml_name("PRQ1")));
858     aml_append(sb_scope, build_link_dev("LNKC", 2, aml_name("PRQ2")));
859     aml_append(sb_scope, build_link_dev("LNKD", 3, aml_name("PRQ3")));
860 
861     dev = aml_device("LNKS");
862     {
863         aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0C0F")));
864         aml_append(dev, aml_name_decl("_UID", aml_int(4)));
865 
866         crs = aml_resource_template();
867         irqs = 9;
868         aml_append(crs, aml_interrupt(AML_CONSUMER, AML_LEVEL,
869                                       AML_ACTIVE_HIGH, AML_SHARED,
870                                       &irqs, 1));
871         aml_append(dev, aml_name_decl("_PRS", crs));
872 
873         /* The SCI cannot be disabled and is always attached to GSI 9,
874          * so these are no-ops.  We only need this link to override the
875          * polarity to active high and match the content of the MADT.
876          */
877         method = aml_method("_STA", 0, AML_NOTSERIALIZED);
878         aml_append(method, aml_return(aml_int(0x0b)));
879         aml_append(dev, method);
880 
881         method = aml_method("_DIS", 0, AML_NOTSERIALIZED);
882         aml_append(dev, method);
883 
884         method = aml_method("_CRS", 0, AML_NOTSERIALIZED);
885         aml_append(method, aml_return(aml_name("_PRS")));
886         aml_append(dev, method);
887 
888         method = aml_method("_SRS", 1, AML_NOTSERIALIZED);
889         aml_append(dev, method);
890     }
891     aml_append(sb_scope, dev);
892 
893     aml_append(table, sb_scope);
894 }
895 
896 static void append_q35_prt_entry(Aml *ctx, uint32_t nr, const char *name)
897 {
898     int i;
899     int head;
900     Aml *pkg;
901     char base = name[3] < 'E' ? 'A' : 'E';
902     char *s = g_strdup(name);
903     Aml *a_nr = aml_int((nr << 16) | 0xffff);
904 
905     assert(strlen(s) == 4);
906 
907     head = name[3] - base;
908     for (i = 0; i < 4; i++) {
909         if (head + i > 3) {
910             head = i * -1;
911         }
912         s[3] = base + head + i;
913         pkg = aml_package(4);
914         aml_append(pkg, a_nr);
915         aml_append(pkg, aml_int(i));
916         aml_append(pkg, aml_name("%s", s));
917         aml_append(pkg, aml_int(0));
918         aml_append(ctx, pkg);
919     }
920     g_free(s);
921 }
922 
923 static Aml *build_q35_routing_table(const char *str)
924 {
925     int i;
926     Aml *pkg;
927     char *name = g_strdup_printf("%s ", str);
928 
929     pkg = aml_package(128);
930     for (i = 0; i < 0x18; i++) {
931             name[3] = 'E' + (i & 0x3);
932             append_q35_prt_entry(pkg, i, name);
933     }
934 
935     name[3] = 'E';
936     append_q35_prt_entry(pkg, 0x18, name);
937 
938     /* INTA -> PIRQA for slot 25 - 31, see the default value of D<N>IR */
939     for (i = 0x0019; i < 0x1e; i++) {
940         name[3] = 'A';
941         append_q35_prt_entry(pkg, i, name);
942     }
943 
944     /* PCIe->PCI bridge. use PIRQ[E-H] */
945     name[3] = 'E';
946     append_q35_prt_entry(pkg, 0x1e, name);
947     name[3] = 'A';
948     append_q35_prt_entry(pkg, 0x1f, name);
949 
950     g_free(name);
951     return pkg;
952 }
953 
954 static void build_q35_pci0_int(Aml *table)
955 {
956     Aml *method;
957     Aml *sb_scope = aml_scope("_SB");
958     Aml *pci0_scope = aml_scope("PCI0");
959 
960     /* Zero => PIC mode, One => APIC Mode */
961     aml_append(table, aml_name_decl("PICF", aml_int(0)));
962     method = aml_method("_PIC", 1, AML_NOTSERIALIZED);
963     {
964         aml_append(method, aml_store(aml_arg(0), aml_name("PICF")));
965     }
966     aml_append(table, method);
967 
968     aml_append(pci0_scope,
969         aml_name_decl("PRTP", build_q35_routing_table("LNK")));
970     aml_append(pci0_scope,
971         aml_name_decl("PRTA", build_q35_routing_table("GSI")));
972 
973     method = aml_method("_PRT", 0, AML_NOTSERIALIZED);
974     {
975         Aml *if_ctx;
976         Aml *else_ctx;
977 
978         /* PCI IRQ routing table, example from ACPI 2.0a specification,
979            section 6.2.8.1 */
980         /* Note: we provide the same info as the PCI routing
981            table of the Bochs BIOS */
982         if_ctx = aml_if(aml_equal(aml_name("PICF"), aml_int(0)));
983         aml_append(if_ctx, aml_return(aml_name("PRTP")));
984         aml_append(method, if_ctx);
985         else_ctx = aml_else();
986         aml_append(else_ctx, aml_return(aml_name("PRTA")));
987         aml_append(method, else_ctx);
988     }
989     aml_append(pci0_scope, method);
990     aml_append(sb_scope, pci0_scope);
991 
992     aml_append(sb_scope, build_irq_status_method());
993     aml_append(sb_scope, build_iqcr_method(false));
994 
995     aml_append(sb_scope, build_link_dev("LNKA", 0, aml_name("PRQA")));
996     aml_append(sb_scope, build_link_dev("LNKB", 1, aml_name("PRQB")));
997     aml_append(sb_scope, build_link_dev("LNKC", 2, aml_name("PRQC")));
998     aml_append(sb_scope, build_link_dev("LNKD", 3, aml_name("PRQD")));
999     aml_append(sb_scope, build_link_dev("LNKE", 4, aml_name("PRQE")));
1000     aml_append(sb_scope, build_link_dev("LNKF", 5, aml_name("PRQF")));
1001     aml_append(sb_scope, build_link_dev("LNKG", 6, aml_name("PRQG")));
1002     aml_append(sb_scope, build_link_dev("LNKH", 7, aml_name("PRQH")));
1003 
1004     aml_append(sb_scope, build_gsi_link_dev("GSIA", 0x10, 0x10));
1005     aml_append(sb_scope, build_gsi_link_dev("GSIB", 0x11, 0x11));
1006     aml_append(sb_scope, build_gsi_link_dev("GSIC", 0x12, 0x12));
1007     aml_append(sb_scope, build_gsi_link_dev("GSID", 0x13, 0x13));
1008     aml_append(sb_scope, build_gsi_link_dev("GSIE", 0x14, 0x14));
1009     aml_append(sb_scope, build_gsi_link_dev("GSIF", 0x15, 0x15));
1010     aml_append(sb_scope, build_gsi_link_dev("GSIG", 0x16, 0x16));
1011     aml_append(sb_scope, build_gsi_link_dev("GSIH", 0x17, 0x17));
1012 
1013     aml_append(table, sb_scope);
1014 }
1015 
1016 static Aml *build_q35_dram_controller(const AcpiMcfgInfo *mcfg)
1017 {
1018     Aml *dev;
1019     Aml *resource_template;
1020 
1021     /* DRAM controller */
1022     dev = aml_device("DRAC");
1023     aml_append(dev, aml_name_decl("_HID", aml_string("PNP0C01")));
1024 
1025     resource_template = aml_resource_template();
1026     if (mcfg->base + mcfg->size - 1 >= (1ULL << 32)) {
1027         aml_append(resource_template,
1028                    aml_qword_memory(AML_POS_DECODE,
1029                                     AML_MIN_FIXED,
1030                                     AML_MAX_FIXED,
1031                                     AML_NON_CACHEABLE,
1032                                     AML_READ_WRITE,
1033                                     0x0000000000000000,
1034                                     mcfg->base,
1035                                     mcfg->base + mcfg->size - 1,
1036                                     0x0000000000000000,
1037                                     mcfg->size));
1038     } else {
1039         aml_append(resource_template,
1040                    aml_dword_memory(AML_POS_DECODE,
1041                                     AML_MIN_FIXED,
1042                                     AML_MAX_FIXED,
1043                                     AML_NON_CACHEABLE,
1044                                     AML_READ_WRITE,
1045                                     0x0000000000000000,
1046                                     mcfg->base,
1047                                     mcfg->base + mcfg->size - 1,
1048                                     0x0000000000000000,
1049                                     mcfg->size));
1050     }
1051     aml_append(dev, aml_name_decl("_CRS", resource_template));
1052 
1053     return dev;
1054 }
1055 
1056 static void build_acpi0017(Aml *table)
1057 {
1058     Aml *dev, *scope, *method;
1059 
1060     scope =  aml_scope("_SB");
1061     dev = aml_device("CXLM");
1062     aml_append(dev, aml_name_decl("_HID", aml_string("ACPI0017")));
1063 
1064     method = aml_method("_STA", 0, AML_NOTSERIALIZED);
1065     aml_append(method, aml_return(aml_int(0x0B)));
1066     aml_append(dev, method);
1067     build_cxl_dsm_method(dev);
1068 
1069     aml_append(scope, dev);
1070     aml_append(table, scope);
1071 }
1072 
1073 static void
1074 build_dsdt(GArray *table_data, BIOSLinker *linker,
1075            AcpiPmInfo *pm, AcpiMiscInfo *misc,
1076            Range *pci_hole, Range *pci_hole64, MachineState *machine)
1077 {
1078     Object *i440fx = object_resolve_type_unambiguous(TYPE_I440FX_PCI_HOST_BRIDGE,
1079                                                      NULL);
1080     Object *q35 = object_resolve_type_unambiguous(TYPE_Q35_HOST_DEVICE, NULL);
1081     CrsRangeEntry *entry;
1082     Aml *dsdt, *sb_scope, *scope, *dev, *method, *field, *pkg, *crs;
1083     CrsRangeSet crs_range_set;
1084     PCMachineState *pcms = PC_MACHINE(machine);
1085     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(machine);
1086     X86MachineState *x86ms = X86_MACHINE(machine);
1087     AcpiMcfgInfo mcfg;
1088     bool mcfg_valid = !!acpi_get_mcfg(&mcfg);
1089     uint32_t nr_mem = machine->ram_slots;
1090     int root_bus_limit = 0xFF;
1091     PCIBus *bus = NULL;
1092 #ifdef CONFIG_TPM
1093     TPMIf *tpm = tpm_find();
1094 #endif
1095     bool cxl_present = false;
1096     int i;
1097     VMBusBridge *vmbus_bridge = vmbus_bridge_find();
1098     AcpiTable table = { .sig = "DSDT", .rev = 1, .oem_id = x86ms->oem_id,
1099                         .oem_table_id = x86ms->oem_table_id };
1100 
1101     assert(!!i440fx != !!q35);
1102 
1103     acpi_table_begin(&table, table_data);
1104     dsdt = init_aml_allocator();
1105 
1106     build_dbg_aml(dsdt);
1107     if (i440fx) {
1108         sb_scope = aml_scope("_SB");
1109         dev = aml_device("PCI0");
1110         aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0A03")));
1111         aml_append(dev, aml_name_decl("_UID", aml_int(pcmc->pci_root_uid)));
1112         aml_append(dev, aml_pci_edsm());
1113         aml_append(sb_scope, dev);
1114         aml_append(dsdt, sb_scope);
1115 
1116         if (pm->pcihp_bridge_en || pm->pcihp_root_en) {
1117             build_acpi_pci_hotplug(dsdt, AML_SYSTEM_IO, pm->pcihp_io_base);
1118         }
1119         build_piix4_pci0_int(dsdt);
1120     } else if (q35) {
1121         sb_scope = aml_scope("_SB");
1122         dev = aml_device("PCI0");
1123         aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0A08")));
1124         aml_append(dev, aml_name_decl("_CID", aml_eisaid("PNP0A03")));
1125         aml_append(dev, aml_name_decl("_UID", aml_int(pcmc->pci_root_uid)));
1126         aml_append(dev, build_pci_host_bridge_osc_method(!pm->pcihp_bridge_en));
1127         aml_append(dev, aml_pci_edsm());
1128         aml_append(sb_scope, dev);
1129         if (mcfg_valid) {
1130             aml_append(sb_scope, build_q35_dram_controller(&mcfg));
1131         }
1132 
1133         if (pm->smi_on_cpuhp) {
1134             /* reserve SMI block resources, IO ports 0xB2, 0xB3 */
1135             dev = aml_device("PCI0.SMI0");
1136             aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0A06")));
1137             aml_append(dev, aml_name_decl("_UID", aml_string("SMI resources")));
1138             crs = aml_resource_template();
1139             aml_append(crs,
1140                 aml_io(
1141                        AML_DECODE16,
1142                        pm->fadt.smi_cmd,
1143                        pm->fadt.smi_cmd,
1144                        1,
1145                        2)
1146             );
1147             aml_append(dev, aml_name_decl("_CRS", crs));
1148             aml_append(dev, aml_operation_region("SMIR", AML_SYSTEM_IO,
1149                 aml_int(pm->fadt.smi_cmd), 2));
1150             field = aml_field("SMIR", AML_BYTE_ACC, AML_NOLOCK,
1151                               AML_WRITE_AS_ZEROS);
1152             aml_append(field, aml_named_field("SMIC", 8));
1153             aml_append(field, aml_reserved_field(8));
1154             aml_append(dev, field);
1155             aml_append(sb_scope, dev);
1156         }
1157 
1158         aml_append(dsdt, sb_scope);
1159 
1160         if (pm->pcihp_bridge_en) {
1161             build_acpi_pci_hotplug(dsdt, AML_SYSTEM_IO, pm->pcihp_io_base);
1162         }
1163         build_q35_pci0_int(dsdt);
1164     }
1165 
1166     if (misc->has_hpet) {
1167         build_hpet_aml(dsdt);
1168     }
1169 
1170     if (vmbus_bridge) {
1171         sb_scope = aml_scope("_SB");
1172         aml_append(sb_scope, build_vmbus_device_aml(vmbus_bridge));
1173         aml_append(dsdt, sb_scope);
1174     }
1175 
1176     scope =  aml_scope("_GPE");
1177     {
1178         aml_append(scope, aml_name_decl("_HID", aml_string("ACPI0006")));
1179         if (machine->nvdimms_state->is_enabled) {
1180             method = aml_method("_E04", 0, AML_NOTSERIALIZED);
1181             aml_append(method, aml_notify(aml_name("\\_SB.NVDR"),
1182                                           aml_int(0x80)));
1183             aml_append(scope, method);
1184         }
1185     }
1186     aml_append(dsdt, scope);
1187 
1188     if (pcmc->legacy_cpu_hotplug) {
1189         build_legacy_cpu_hotplug_aml(dsdt, machine, pm->cpu_hp_io_base);
1190     } else {
1191         CPUHotplugFeatures opts = {
1192             .acpi_1_compatible = true, .has_legacy_cphp = true,
1193             .smi_path = pm->smi_on_cpuhp ? "\\_SB.PCI0.SMI0.SMIC" : NULL,
1194             .fw_unplugs_cpu = pm->smi_on_cpu_unplug,
1195         };
1196         build_cpus_aml(dsdt, machine, opts, pc_madt_cpu_entry,
1197                        pm->cpu_hp_io_base, "\\_SB.PCI0", "\\_GPE._E02",
1198                        AML_SYSTEM_IO);
1199     }
1200 
1201     if (pcms->memhp_io_base && nr_mem) {
1202         build_memory_hotplug_aml(dsdt, nr_mem, "\\_SB.PCI0",
1203                                  "\\_GPE._E03", AML_SYSTEM_IO,
1204                                  pcms->memhp_io_base);
1205     }
1206 
1207     crs_range_set_init(&crs_range_set);
1208     bus = PC_MACHINE(machine)->pcibus;
1209     if (bus) {
1210         QLIST_FOREACH(bus, &bus->child, sibling) {
1211             uint8_t bus_num = pci_bus_num(bus);
1212             uint8_t numa_node = pci_bus_numa_node(bus);
1213             uint32_t uid;
1214 
1215             /* look only for expander root buses */
1216             if (!pci_bus_is_root(bus)) {
1217                 continue;
1218             }
1219 
1220             if (bus_num < root_bus_limit) {
1221                 root_bus_limit = bus_num - 1;
1222             }
1223 
1224             uid = object_property_get_uint(OBJECT(bus), "acpi_uid",
1225                                            &error_fatal);
1226             scope = aml_scope("\\_SB");
1227 
1228             if (pci_bus_is_cxl(bus)) {
1229                 dev = aml_device("CL%.02X", bus_num);
1230             } else {
1231                 dev = aml_device("PC%.02X", bus_num);
1232             }
1233             aml_append(dev, aml_name_decl("_UID", aml_int(uid)));
1234             aml_append(dev, aml_name_decl("_BBN", aml_int(bus_num)));
1235             if (pci_bus_is_cxl(bus)) {
1236                 struct Aml *aml_pkg = aml_package(2);
1237 
1238                 aml_append(dev, aml_name_decl("_HID", aml_string("ACPI0016")));
1239                 aml_append(aml_pkg, aml_eisaid("PNP0A08"));
1240                 aml_append(aml_pkg, aml_eisaid("PNP0A03"));
1241                 aml_append(dev, aml_name_decl("_CID", aml_pkg));
1242                 build_cxl_osc_method(dev);
1243             } else if (pci_bus_is_express(bus)) {
1244                 aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0A08")));
1245                 aml_append(dev, aml_name_decl("_CID", aml_eisaid("PNP0A03")));
1246 
1247                 /* Expander bridges do not have ACPI PCI Hot-plug enabled */
1248                 aml_append(dev, build_pci_host_bridge_osc_method(true));
1249             } else {
1250                 aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0A03")));
1251             }
1252 
1253             if (numa_node != NUMA_NODE_UNASSIGNED) {
1254                 aml_append(dev, aml_name_decl("_PXM", aml_int(numa_node)));
1255             }
1256 
1257             aml_append(dev, build_prt(false));
1258             crs = build_crs(PCI_HOST_BRIDGE(BUS(bus)->parent), &crs_range_set,
1259                             0, 0, 0, 0);
1260             aml_append(dev, aml_name_decl("_CRS", crs));
1261             aml_append(scope, dev);
1262             aml_append(dsdt, scope);
1263 
1264             /* Handle the ranges for the PXB expanders */
1265             if (pci_bus_is_cxl(bus)) {
1266                 MemoryRegion *mr = &pcms->cxl_devices_state.host_mr;
1267                 uint64_t base = mr->addr;
1268 
1269                 cxl_present = true;
1270                 crs_range_insert(crs_range_set.mem_ranges, base,
1271                                  base + memory_region_size(mr) - 1);
1272             }
1273         }
1274     }
1275 
1276     if (cxl_present) {
1277         build_acpi0017(dsdt);
1278     }
1279 
1280     /*
1281      * At this point crs_range_set has all the ranges used by pci
1282      * busses *other* than PCI0.  These ranges will be excluded from
1283      * the PCI0._CRS.  Add mmconfig to the set so it will be excluded
1284      * too.
1285      */
1286     if (mcfg_valid) {
1287         crs_range_insert(crs_range_set.mem_ranges,
1288                          mcfg.base, mcfg.base + mcfg.size - 1);
1289     }
1290 
1291     scope = aml_scope("\\_SB.PCI0");
1292     /* build PCI0._CRS */
1293     crs = aml_resource_template();
1294     aml_append(crs,
1295         aml_word_bus_number(AML_MIN_FIXED, AML_MAX_FIXED, AML_POS_DECODE,
1296                             0x0000, 0x0, root_bus_limit,
1297                             0x0000, root_bus_limit + 1));
1298     aml_append(crs, aml_io(AML_DECODE16, 0x0CF8, 0x0CF8, 0x01, 0x08));
1299 
1300     aml_append(crs,
1301         aml_word_io(AML_MIN_FIXED, AML_MAX_FIXED,
1302                     AML_POS_DECODE, AML_ENTIRE_RANGE,
1303                     0x0000, 0x0000, 0x0CF7, 0x0000, 0x0CF8));
1304 
1305     crs_replace_with_free_ranges(crs_range_set.io_ranges, 0x0D00, 0xFFFF);
1306     for (i = 0; i < crs_range_set.io_ranges->len; i++) {
1307         entry = g_ptr_array_index(crs_range_set.io_ranges, i);
1308         aml_append(crs,
1309             aml_word_io(AML_MIN_FIXED, AML_MAX_FIXED,
1310                         AML_POS_DECODE, AML_ENTIRE_RANGE,
1311                         0x0000, entry->base, entry->limit,
1312                         0x0000, entry->limit - entry->base + 1));
1313     }
1314 
1315     aml_append(crs,
1316         aml_dword_memory(AML_POS_DECODE, AML_MIN_FIXED, AML_MAX_FIXED,
1317                          AML_CACHEABLE, AML_READ_WRITE,
1318                          0, 0x000A0000, 0x000BFFFF, 0, 0x00020000));
1319 
1320     crs_replace_with_free_ranges(crs_range_set.mem_ranges,
1321                                  range_lob(pci_hole),
1322                                  range_upb(pci_hole));
1323     for (i = 0; i < crs_range_set.mem_ranges->len; i++) {
1324         entry = g_ptr_array_index(crs_range_set.mem_ranges, i);
1325         aml_append(crs,
1326             aml_dword_memory(AML_POS_DECODE, AML_MIN_FIXED, AML_MAX_FIXED,
1327                              AML_NON_CACHEABLE, AML_READ_WRITE,
1328                              0, entry->base, entry->limit,
1329                              0, entry->limit - entry->base + 1));
1330     }
1331 
1332     if (!range_is_empty(pci_hole64)) {
1333         crs_replace_with_free_ranges(crs_range_set.mem_64bit_ranges,
1334                                      range_lob(pci_hole64),
1335                                      range_upb(pci_hole64));
1336         for (i = 0; i < crs_range_set.mem_64bit_ranges->len; i++) {
1337             entry = g_ptr_array_index(crs_range_set.mem_64bit_ranges, i);
1338             aml_append(crs,
1339                        aml_qword_memory(AML_POS_DECODE, AML_MIN_FIXED,
1340                                         AML_MAX_FIXED,
1341                                         AML_CACHEABLE, AML_READ_WRITE,
1342                                         0, entry->base, entry->limit,
1343                                         0, entry->limit - entry->base + 1));
1344         }
1345     }
1346 
1347 #ifdef CONFIG_TPM
1348     if (TPM_IS_TIS_ISA(tpm_find())) {
1349         aml_append(crs, aml_memory32_fixed(TPM_TIS_ADDR_BASE,
1350                    TPM_TIS_ADDR_SIZE, AML_READ_WRITE));
1351     }
1352 #endif
1353     aml_append(scope, aml_name_decl("_CRS", crs));
1354 
1355     /* reserve GPE0 block resources */
1356     dev = aml_device("GPE0");
1357     aml_append(dev, aml_name_decl("_HID", aml_string("PNP0A06")));
1358     aml_append(dev, aml_name_decl("_UID", aml_string("GPE0 resources")));
1359     /* device present, functioning, decoding, not shown in UI */
1360     aml_append(dev, aml_name_decl("_STA", aml_int(0xB)));
1361     crs = aml_resource_template();
1362     aml_append(crs,
1363         aml_io(
1364                AML_DECODE16,
1365                pm->fadt.gpe0_blk.address,
1366                pm->fadt.gpe0_blk.address,
1367                1,
1368                pm->fadt.gpe0_blk.bit_width / 8)
1369     );
1370     aml_append(dev, aml_name_decl("_CRS", crs));
1371     aml_append(scope, dev);
1372 
1373     crs_range_set_free(&crs_range_set);
1374 
1375     /* reserve PCIHP resources */
1376     if (pm->pcihp_io_len && (pm->pcihp_bridge_en || pm->pcihp_root_en)) {
1377         build_append_pcihp_resources(scope,
1378                                       pm->pcihp_io_base, pm->pcihp_io_len);
1379     }
1380     aml_append(dsdt, scope);
1381 
1382     /*  create S3_ / S4_ / S5_ packages if necessary */
1383     scope = aml_scope("\\");
1384     if (!pm->s3_disabled) {
1385         pkg = aml_package(4);
1386         aml_append(pkg, aml_int(1)); /* PM1a_CNT.SLP_TYP */
1387         aml_append(pkg, aml_int(1)); /* PM1b_CNT.SLP_TYP, FIXME: not impl. */
1388         aml_append(pkg, aml_int(0)); /* reserved */
1389         aml_append(pkg, aml_int(0)); /* reserved */
1390         aml_append(scope, aml_name_decl("_S3", pkg));
1391     }
1392 
1393     if (!pm->s4_disabled) {
1394         pkg = aml_package(4);
1395         aml_append(pkg, aml_int(pm->s4_val)); /* PM1a_CNT.SLP_TYP */
1396         /* PM1b_CNT.SLP_TYP, FIXME: not impl. */
1397         aml_append(pkg, aml_int(pm->s4_val));
1398         aml_append(pkg, aml_int(0)); /* reserved */
1399         aml_append(pkg, aml_int(0)); /* reserved */
1400         aml_append(scope, aml_name_decl("_S4", pkg));
1401     }
1402 
1403     pkg = aml_package(4);
1404     aml_append(pkg, aml_int(0)); /* PM1a_CNT.SLP_TYP */
1405     aml_append(pkg, aml_int(0)); /* PM1b_CNT.SLP_TYP not impl. */
1406     aml_append(pkg, aml_int(0)); /* reserved */
1407     aml_append(pkg, aml_int(0)); /* reserved */
1408     aml_append(scope, aml_name_decl("_S5", pkg));
1409     aml_append(dsdt, scope);
1410 
1411     /* create fw_cfg node, unconditionally */
1412     {
1413         scope = aml_scope("\\_SB.PCI0");
1414         fw_cfg_add_acpi_dsdt(scope, x86ms->fw_cfg);
1415         aml_append(dsdt, scope);
1416     }
1417 
1418     sb_scope = aml_scope("\\_SB");
1419     {
1420         Object *pci_host = acpi_get_i386_pci_host();
1421 
1422         if (pci_host) {
1423             PCIBus *pbus = PCI_HOST_BRIDGE(pci_host)->bus;
1424             Aml *ascope = aml_scope("PCI0");
1425             /* Scan all PCI buses. Generate tables to support hotplug. */
1426             build_append_pci_bus_devices(ascope, pbus);
1427             if (object_property_find(OBJECT(pbus), ACPI_PCIHP_PROP_BSEL)) {
1428                 build_append_pcihp_slots(ascope, pbus);
1429             }
1430             aml_append(sb_scope, ascope);
1431         }
1432     }
1433 
1434 #ifdef CONFIG_TPM
1435     if (TPM_IS_CRB(tpm)) {
1436         dev = aml_device("TPM");
1437         aml_append(dev, aml_name_decl("_HID", aml_string("MSFT0101")));
1438         aml_append(dev, aml_name_decl("_STR",
1439                                       aml_string("TPM 2.0 Device")));
1440         crs = aml_resource_template();
1441         aml_append(crs, aml_memory32_fixed(TPM_CRB_ADDR_BASE,
1442                                            TPM_CRB_ADDR_SIZE, AML_READ_WRITE));
1443         aml_append(dev, aml_name_decl("_CRS", crs));
1444 
1445         aml_append(dev, aml_name_decl("_STA", aml_int(0xf)));
1446         aml_append(dev, aml_name_decl("_UID", aml_int(1)));
1447 
1448         tpm_build_ppi_acpi(tpm, dev);
1449 
1450         aml_append(sb_scope, dev);
1451     }
1452 #endif
1453 
1454     if (pcms->sgx_epc.size != 0) {
1455         uint64_t epc_base = pcms->sgx_epc.base;
1456         uint64_t epc_size = pcms->sgx_epc.size;
1457 
1458         dev = aml_device("EPC");
1459         aml_append(dev, aml_name_decl("_HID", aml_eisaid("INT0E0C")));
1460         aml_append(dev, aml_name_decl("_STR",
1461                                       aml_unicode("Enclave Page Cache 1.0")));
1462         crs = aml_resource_template();
1463         aml_append(crs,
1464                    aml_qword_memory(AML_POS_DECODE, AML_MIN_FIXED,
1465                                     AML_MAX_FIXED, AML_NON_CACHEABLE,
1466                                     AML_READ_WRITE, 0, epc_base,
1467                                     epc_base + epc_size - 1, 0, epc_size));
1468         aml_append(dev, aml_name_decl("_CRS", crs));
1469 
1470         method = aml_method("_STA", 0, AML_NOTSERIALIZED);
1471         aml_append(method, aml_return(aml_int(0x0f)));
1472         aml_append(dev, method);
1473 
1474         aml_append(sb_scope, dev);
1475     }
1476     aml_append(dsdt, sb_scope);
1477 
1478     if (pm->pcihp_bridge_en || pm->pcihp_root_en) {
1479         bool has_pcnt;
1480 
1481         Object *pci_host = acpi_get_i386_pci_host();
1482         PCIBus *b = PCI_HOST_BRIDGE(pci_host)->bus;
1483 
1484         scope = aml_scope("\\_SB.PCI0");
1485         has_pcnt = build_append_notification_callback(scope, b);
1486         if (has_pcnt) {
1487             aml_append(dsdt, scope);
1488         }
1489 
1490         scope =  aml_scope("_GPE");
1491         {
1492             method = aml_method("_E01", 0, AML_NOTSERIALIZED);
1493             if (has_pcnt) {
1494                 aml_append(method,
1495                     aml_acquire(aml_name("\\_SB.PCI0.BLCK"), 0xFFFF));
1496                 aml_append(method, aml_call0("\\_SB.PCI0.PCNT"));
1497                 aml_append(method, aml_release(aml_name("\\_SB.PCI0.BLCK")));
1498             }
1499             aml_append(scope, method);
1500         }
1501         aml_append(dsdt, scope);
1502     }
1503 
1504     /* copy AML table into ACPI tables blob and patch header there */
1505     g_array_append_vals(table_data, dsdt->buf->data, dsdt->buf->len);
1506     acpi_table_end(linker, &table);
1507     free_aml_allocator();
1508 }
1509 
1510 /*
1511  * IA-PC HPET (High Precision Event Timers) Specification (Revision: 1.0a)
1512  * 3.2.4The ACPI 2.0 HPET Description Table (HPET)
1513  */
1514 static void
1515 build_hpet(GArray *table_data, BIOSLinker *linker, const char *oem_id,
1516            const char *oem_table_id)
1517 {
1518     AcpiTable table = { .sig = "HPET", .rev = 1,
1519                         .oem_id = oem_id, .oem_table_id = oem_table_id };
1520 
1521     acpi_table_begin(&table, table_data);
1522     /* Note timer_block_id value must be kept in sync with value advertised by
1523      * emulated hpet
1524      */
1525     /* Event Timer Block ID */
1526     build_append_int_noprefix(table_data, 0x8086a201, 4);
1527     /* BASE_ADDRESS */
1528     build_append_gas(table_data, AML_AS_SYSTEM_MEMORY, 0, 0, 0, HPET_BASE);
1529     /* HPET Number */
1530     build_append_int_noprefix(table_data, 0, 1);
1531     /* Main Counter Minimum Clock_tick in Periodic Mode */
1532     build_append_int_noprefix(table_data, 0, 2);
1533     /* Page Protection And OEM Attribute */
1534     build_append_int_noprefix(table_data, 0, 1);
1535     acpi_table_end(linker, &table);
1536 }
1537 
1538 #ifdef CONFIG_TPM
1539 /*
1540  * TCPA Description Table
1541  *
1542  * Following Level 00, Rev 00.37 of specs:
1543  * http://www.trustedcomputinggroup.org/resources/tcg_acpi_specification
1544  * 7.1.2 ACPI Table Layout
1545  */
1546 static void
1547 build_tpm_tcpa(GArray *table_data, BIOSLinker *linker, GArray *tcpalog,
1548                const char *oem_id, const char *oem_table_id)
1549 {
1550     unsigned log_addr_offset;
1551     AcpiTable table = { .sig = "TCPA", .rev = 2,
1552                         .oem_id = oem_id, .oem_table_id = oem_table_id };
1553 
1554     acpi_table_begin(&table, table_data);
1555     /* Platform Class */
1556     build_append_int_noprefix(table_data, TPM_TCPA_ACPI_CLASS_CLIENT, 2);
1557     /* Log Area Minimum Length (LAML) */
1558     build_append_int_noprefix(table_data, TPM_LOG_AREA_MINIMUM_SIZE, 4);
1559     /* Log Area Start Address (LASA) */
1560     log_addr_offset = table_data->len;
1561     build_append_int_noprefix(table_data, 0, 8);
1562 
1563     /* allocate/reserve space for TPM log area */
1564     acpi_data_push(tcpalog, TPM_LOG_AREA_MINIMUM_SIZE);
1565     bios_linker_loader_alloc(linker, ACPI_BUILD_TPMLOG_FILE, tcpalog, 1,
1566                              false /* high memory */);
1567     /* log area start address to be filled by Guest linker */
1568     bios_linker_loader_add_pointer(linker, ACPI_BUILD_TABLE_FILE,
1569         log_addr_offset, 8, ACPI_BUILD_TPMLOG_FILE, 0);
1570 
1571     acpi_table_end(linker, &table);
1572 }
1573 #endif
1574 
1575 #define HOLE_640K_START  (640 * KiB)
1576 #define HOLE_640K_END   (1 * MiB)
1577 
1578 /*
1579  * ACPI spec, Revision 3.0
1580  * 5.2.15 System Resource Affinity Table (SRAT)
1581  */
1582 static void
1583 build_srat(GArray *table_data, BIOSLinker *linker, MachineState *machine)
1584 {
1585     int i;
1586     int numa_mem_start, slots;
1587     uint64_t mem_len, mem_base, next_base;
1588     MachineClass *mc = MACHINE_GET_CLASS(machine);
1589     X86MachineState *x86ms = X86_MACHINE(machine);
1590     const CPUArchIdList *apic_ids = mc->possible_cpu_arch_ids(machine);
1591     int nb_numa_nodes = machine->numa_state->num_nodes;
1592     NodeInfo *numa_info = machine->numa_state->nodes;
1593     AcpiTable table = { .sig = "SRAT", .rev = 1, .oem_id = x86ms->oem_id,
1594                         .oem_table_id = x86ms->oem_table_id };
1595 
1596     acpi_table_begin(&table, table_data);
1597     build_append_int_noprefix(table_data, 1, 4); /* Reserved */
1598     build_append_int_noprefix(table_data, 0, 8); /* Reserved */
1599 
1600     for (i = 0; i < apic_ids->len; i++) {
1601         int node_id = apic_ids->cpus[i].props.node_id;
1602         uint32_t apic_id = apic_ids->cpus[i].arch_id;
1603 
1604         if (apic_id < 255) {
1605             /* 5.2.15.1 Processor Local APIC/SAPIC Affinity Structure */
1606             build_append_int_noprefix(table_data, 0, 1);  /* Type  */
1607             build_append_int_noprefix(table_data, 16, 1); /* Length */
1608             /* Proximity Domain [7:0] */
1609             build_append_int_noprefix(table_data, node_id, 1);
1610             build_append_int_noprefix(table_data, apic_id, 1); /* APIC ID */
1611             /* Flags, Table 5-36 */
1612             build_append_int_noprefix(table_data, 1, 4);
1613             build_append_int_noprefix(table_data, 0, 1); /* Local SAPIC EID */
1614             /* Proximity Domain [31:8] */
1615             build_append_int_noprefix(table_data, 0, 3);
1616             build_append_int_noprefix(table_data, 0, 4); /* Reserved */
1617         } else {
1618             /*
1619              * ACPI spec, Revision 4.0
1620              * 5.2.16.3 Processor Local x2APIC Affinity Structure
1621              */
1622             build_append_int_noprefix(table_data, 2, 1);  /* Type  */
1623             build_append_int_noprefix(table_data, 24, 1); /* Length */
1624             build_append_int_noprefix(table_data, 0, 2); /* Reserved */
1625             /* Proximity Domain */
1626             build_append_int_noprefix(table_data, node_id, 4);
1627             build_append_int_noprefix(table_data, apic_id, 4); /* X2APIC ID */
1628             /* Flags, Table 5-39 */
1629             build_append_int_noprefix(table_data, 1 /* Enabled */, 4);
1630             build_append_int_noprefix(table_data, 0, 4); /* Clock Domain */
1631             build_append_int_noprefix(table_data, 0, 4); /* Reserved */
1632         }
1633     }
1634 
1635     /* the memory map is a bit tricky, it contains at least one hole
1636      * from 640k-1M and possibly another one from 3.5G-4G.
1637      */
1638     next_base = 0;
1639     numa_mem_start = table_data->len;
1640 
1641     for (i = 1; i < nb_numa_nodes + 1; ++i) {
1642         mem_base = next_base;
1643         mem_len = numa_info[i - 1].node_mem;
1644         next_base = mem_base + mem_len;
1645 
1646         /* Cut out the 640K hole */
1647         if (mem_base <= HOLE_640K_START &&
1648             next_base > HOLE_640K_START) {
1649             mem_len -= next_base - HOLE_640K_START;
1650             if (mem_len > 0) {
1651                 build_srat_memory(table_data, mem_base, mem_len, i - 1,
1652                                   MEM_AFFINITY_ENABLED);
1653             }
1654 
1655             /* Check for the rare case: 640K < RAM < 1M */
1656             if (next_base <= HOLE_640K_END) {
1657                 next_base = HOLE_640K_END;
1658                 continue;
1659             }
1660             mem_base = HOLE_640K_END;
1661             mem_len = next_base - HOLE_640K_END;
1662         }
1663 
1664         /* Cut out the ACPI_PCI hole */
1665         if (mem_base <= x86ms->below_4g_mem_size &&
1666             next_base > x86ms->below_4g_mem_size) {
1667             mem_len -= next_base - x86ms->below_4g_mem_size;
1668             if (mem_len > 0) {
1669                 build_srat_memory(table_data, mem_base, mem_len, i - 1,
1670                                   MEM_AFFINITY_ENABLED);
1671             }
1672             mem_base = x86ms->above_4g_mem_start;
1673             mem_len = next_base - x86ms->below_4g_mem_size;
1674             next_base = mem_base + mem_len;
1675         }
1676 
1677         if (mem_len > 0) {
1678             build_srat_memory(table_data, mem_base, mem_len, i - 1,
1679                               MEM_AFFINITY_ENABLED);
1680         }
1681     }
1682 
1683     if (machine->nvdimms_state->is_enabled) {
1684         nvdimm_build_srat(table_data);
1685     }
1686 
1687     sgx_epc_build_srat(table_data);
1688 
1689     /*
1690      * TODO: this part is not in ACPI spec and current linux kernel boots fine
1691      * without these entries. But I recall there were issues the last time I
1692      * tried to remove it with some ancient guest OS, however I can't remember
1693      * what that was so keep this around for now
1694      */
1695     slots = (table_data->len - numa_mem_start) / 40 /* mem affinity len */;
1696     for (; slots < nb_numa_nodes + 2; slots++) {
1697         build_srat_memory(table_data, 0, 0, 0, MEM_AFFINITY_NOFLAGS);
1698     }
1699 
1700     build_srat_generic_affinity_structures(table_data);
1701 
1702     /*
1703      * Entry is required for Windows to enable memory hotplug in OS
1704      * and for Linux to enable SWIOTLB when booted with less than
1705      * 4G of RAM. Windows works better if the entry sets proximity
1706      * to the highest NUMA node in the machine.
1707      * Memory devices may override proximity set by this entry,
1708      * providing _PXM method if necessary.
1709      */
1710     if (machine->device_memory) {
1711         build_srat_memory(table_data, machine->device_memory->base,
1712                           memory_region_size(&machine->device_memory->mr),
1713                           nb_numa_nodes - 1,
1714                           MEM_AFFINITY_HOTPLUGGABLE | MEM_AFFINITY_ENABLED);
1715     }
1716 
1717     acpi_table_end(linker, &table);
1718 }
1719 
1720 /*
1721  * Insert DMAR scope for PCI bridges and endpoint devices
1722  */
1723 static void
1724 insert_scope(PCIBus *bus, PCIDevice *dev, void *opaque)
1725 {
1726     const size_t device_scope_size = 6 /* device scope structure */ +
1727                                      2 /* 1 path entry */;
1728     GArray *scope_blob = opaque;
1729 
1730     if (object_dynamic_cast(OBJECT(dev), TYPE_PCI_BRIDGE)) {
1731         /* Dmar Scope Type: 0x02 for PCI Bridge */
1732         build_append_int_noprefix(scope_blob, 0x02, 1);
1733     } else {
1734         /* Dmar Scope Type: 0x01 for PCI Endpoint Device */
1735         build_append_int_noprefix(scope_blob, 0x01, 1);
1736     }
1737 
1738     /* length */
1739     build_append_int_noprefix(scope_blob, device_scope_size, 1);
1740     /* reserved */
1741     build_append_int_noprefix(scope_blob, 0, 2);
1742     /* enumeration_id */
1743     build_append_int_noprefix(scope_blob, 0, 1);
1744     /* bus */
1745     build_append_int_noprefix(scope_blob, pci_bus_num(bus), 1);
1746     /* device */
1747     build_append_int_noprefix(scope_blob, PCI_SLOT(dev->devfn), 1);
1748     /* function */
1749     build_append_int_noprefix(scope_blob, PCI_FUNC(dev->devfn), 1);
1750 }
1751 
1752 /* For a given PCI host bridge, walk and insert DMAR scope */
1753 static int
1754 dmar_host_bridges(Object *obj, void *opaque)
1755 {
1756     GArray *scope_blob = opaque;
1757 
1758     if (object_dynamic_cast(obj, TYPE_PCI_HOST_BRIDGE)) {
1759         PCIBus *bus = PCI_HOST_BRIDGE(obj)->bus;
1760 
1761         if (bus && !pci_bus_bypass_iommu(bus)) {
1762             pci_for_each_device_under_bus(bus, insert_scope, scope_blob);
1763         }
1764     }
1765 
1766     return 0;
1767 }
1768 
1769 /*
1770  * Intel ® Virtualization Technology for Directed I/O
1771  * Architecture Specification. Revision 3.3
1772  * 8.1 DMA Remapping Reporting Structure
1773  */
1774 static void
1775 build_dmar_q35(GArray *table_data, BIOSLinker *linker, const char *oem_id,
1776                const char *oem_table_id)
1777 {
1778     uint8_t dmar_flags = 0;
1779     uint8_t rsvd10[10] = {};
1780     /* Root complex IOAPIC uses one path only */
1781     const size_t ioapic_scope_size = 6 /* device scope structure */ +
1782                                      2 /* 1 path entry */;
1783     X86IOMMUState *iommu = x86_iommu_get_default();
1784     IntelIOMMUState *intel_iommu = INTEL_IOMMU_DEVICE(iommu);
1785     GArray *scope_blob = g_array_new(false, true, 1);
1786 
1787     AcpiTable table = { .sig = "DMAR", .rev = 1, .oem_id = oem_id,
1788                         .oem_table_id = oem_table_id };
1789 
1790     /*
1791      * A PCI bus walk, for each PCI host bridge.
1792      * Insert scope for each PCI bridge and endpoint device which
1793      * is attached to a bus with iommu enabled.
1794      */
1795     object_child_foreach_recursive(object_get_root(),
1796                                    dmar_host_bridges, scope_blob);
1797 
1798     assert(iommu);
1799     if (x86_iommu_ir_supported(iommu)) {
1800         dmar_flags |= 0x1;      /* Flags: 0x1: INT_REMAP */
1801     }
1802 
1803     acpi_table_begin(&table, table_data);
1804     /* Host Address Width */
1805     build_append_int_noprefix(table_data, intel_iommu->aw_bits - 1, 1);
1806     build_append_int_noprefix(table_data, dmar_flags, 1); /* Flags */
1807     g_array_append_vals(table_data, rsvd10, sizeof(rsvd10)); /* Reserved */
1808 
1809     /* 8.3 DMAR Remapping Hardware Unit Definition structure */
1810     build_append_int_noprefix(table_data, 0, 2); /* Type */
1811     /* Length */
1812     build_append_int_noprefix(table_data,
1813                               16 + ioapic_scope_size + scope_blob->len, 2);
1814     /* Flags */
1815     build_append_int_noprefix(table_data, 0 /* Don't include all pci device */ ,
1816                               1);
1817     build_append_int_noprefix(table_data, 0 , 1); /* Reserved */
1818     build_append_int_noprefix(table_data, 0 , 2); /* Segment Number */
1819     /* Register Base Address */
1820     build_append_int_noprefix(table_data, Q35_HOST_BRIDGE_IOMMU_ADDR , 8);
1821 
1822     /* Scope definition for the root-complex IOAPIC. See VT-d spec
1823      * 8.3.1 (version Oct. 2014 or later). */
1824     build_append_int_noprefix(table_data, 0x03 /* IOAPIC */, 1); /* Type */
1825     build_append_int_noprefix(table_data, ioapic_scope_size, 1); /* Length */
1826     build_append_int_noprefix(table_data, 0, 2); /* Reserved */
1827     /* Enumeration ID */
1828     build_append_int_noprefix(table_data, ACPI_BUILD_IOAPIC_ID, 1);
1829     /* Start Bus Number */
1830     build_append_int_noprefix(table_data, Q35_PSEUDO_BUS_PLATFORM, 1);
1831     /* Path, {Device, Function} pair */
1832     build_append_int_noprefix(table_data, PCI_SLOT(Q35_PSEUDO_DEVFN_IOAPIC), 1);
1833     build_append_int_noprefix(table_data, PCI_FUNC(Q35_PSEUDO_DEVFN_IOAPIC), 1);
1834 
1835     /* Add scope found above */
1836     g_array_append_vals(table_data, scope_blob->data, scope_blob->len);
1837     g_array_free(scope_blob, true);
1838 
1839     if (iommu->dt_supported) {
1840         /* 8.5 Root Port ATS Capability Reporting Structure */
1841         build_append_int_noprefix(table_data, 2, 2); /* Type */
1842         build_append_int_noprefix(table_data, 8, 2); /* Length */
1843         build_append_int_noprefix(table_data, 1 /* ALL_PORTS */, 1); /* Flags */
1844         build_append_int_noprefix(table_data, 0, 1); /* Reserved */
1845         build_append_int_noprefix(table_data, 0, 2); /* Segment Number */
1846     }
1847 
1848     acpi_table_end(linker, &table);
1849 }
1850 
1851 /*
1852  * Windows ACPI Emulated Devices Table
1853  * (Version 1.0 - April 6, 2009)
1854  * Spec: http://download.microsoft.com/download/7/E/7/7E7662CF-CBEA-470B-A97E-CE7CE0D98DC2/WAET.docx
1855  *
1856  * Helpful to speedup Windows guests and ignored by others.
1857  */
1858 static void
1859 build_waet(GArray *table_data, BIOSLinker *linker, const char *oem_id,
1860            const char *oem_table_id)
1861 {
1862     AcpiTable table = { .sig = "WAET", .rev = 1, .oem_id = oem_id,
1863                         .oem_table_id = oem_table_id };
1864 
1865     acpi_table_begin(&table, table_data);
1866     /*
1867      * Set "ACPI PM timer good" flag.
1868      *
1869      * Tells Windows guests that our ACPI PM timer is reliable in the
1870      * sense that guest can read it only once to obtain a reliable value.
1871      * Which avoids costly VMExits caused by guest re-reading it unnecessarily.
1872      */
1873     build_append_int_noprefix(table_data, 1 << 1 /* ACPI PM timer good */, 4);
1874     acpi_table_end(linker, &table);
1875 }
1876 
1877 /*
1878  *   IVRS table as specified in AMD IOMMU Specification v2.62, Section 5.2
1879  *   accessible here http://support.amd.com/TechDocs/48882_IOMMU.pdf
1880  */
1881 #define IOAPIC_SB_DEVID   (uint64_t)PCI_BUILD_BDF(0, PCI_DEVFN(0x14, 0))
1882 
1883 /*
1884  * Insert IVHD entry for device and recurse, insert alias, or insert range as
1885  * necessary for the PCI topology.
1886  */
1887 static void
1888 insert_ivhd(PCIBus *bus, PCIDevice *dev, void *opaque)
1889 {
1890     GArray *table_data = opaque;
1891     uint32_t entry;
1892 
1893     /* "Select" IVHD entry, type 0x2 */
1894     entry = PCI_BUILD_BDF(pci_bus_num(bus), dev->devfn) << 8 | 0x2;
1895     build_append_int_noprefix(table_data, entry, 4);
1896 
1897     if (object_dynamic_cast(OBJECT(dev), TYPE_PCI_BRIDGE)) {
1898         PCIBus *sec_bus = pci_bridge_get_sec_bus(PCI_BRIDGE(dev));
1899         uint8_t sec = pci_bus_num(sec_bus);
1900         uint8_t sub = dev->config[PCI_SUBORDINATE_BUS];
1901 
1902         if (pci_bus_is_express(sec_bus)) {
1903             /*
1904              * Walk the bus if there are subordinates, otherwise use a range
1905              * to cover an entire leaf bus.  We could potentially also use a
1906              * range for traversed buses, but we'd need to take care not to
1907              * create both Select and Range entries covering the same device.
1908              * This is easier and potentially more compact.
1909              *
1910              * An example bare metal system seems to use Select entries for
1911              * root ports without a slot (ie. built-ins) and Range entries
1912              * when there is a slot.  The same system also only hard-codes
1913              * the alias range for an onboard PCIe-to-PCI bridge, apparently
1914              * making no effort to support nested bridges.  We attempt to
1915              * be more thorough here.
1916              */
1917             if (sec == sub) { /* leaf bus */
1918                 /* "Start of Range" IVHD entry, type 0x3 */
1919                 entry = PCI_BUILD_BDF(sec, PCI_DEVFN(0, 0)) << 8 | 0x3;
1920                 build_append_int_noprefix(table_data, entry, 4);
1921                 /* "End of Range" IVHD entry, type 0x4 */
1922                 entry = PCI_BUILD_BDF(sub, PCI_DEVFN(31, 7)) << 8 | 0x4;
1923                 build_append_int_noprefix(table_data, entry, 4);
1924             } else {
1925                 pci_for_each_device(sec_bus, sec, insert_ivhd, table_data);
1926             }
1927         } else {
1928             /*
1929              * If the secondary bus is conventional, then we need to create an
1930              * Alias range for everything downstream.  The range covers the
1931              * first devfn on the secondary bus to the last devfn on the
1932              * subordinate bus.  The alias target depends on legacy versus
1933              * express bridges, just as in pci_device_iommu_address_space().
1934              * DeviceIDa vs DeviceIDb as per the AMD IOMMU spec.
1935              */
1936             uint16_t dev_id_a, dev_id_b;
1937 
1938             dev_id_a = PCI_BUILD_BDF(sec, PCI_DEVFN(0, 0));
1939 
1940             if (pci_is_express(dev) &&
1941                 pcie_cap_get_type(dev) == PCI_EXP_TYPE_PCI_BRIDGE) {
1942                 dev_id_b = dev_id_a;
1943             } else {
1944                 dev_id_b = PCI_BUILD_BDF(pci_bus_num(bus), dev->devfn);
1945             }
1946 
1947             /* "Alias Start of Range" IVHD entry, type 0x43, 8 bytes */
1948             build_append_int_noprefix(table_data, dev_id_a << 8 | 0x43, 4);
1949             build_append_int_noprefix(table_data, dev_id_b << 8 | 0x0, 4);
1950 
1951             /* "End of Range" IVHD entry, type 0x4 */
1952             entry = PCI_BUILD_BDF(sub, PCI_DEVFN(31, 7)) << 8 | 0x4;
1953             build_append_int_noprefix(table_data, entry, 4);
1954         }
1955     }
1956 }
1957 
1958 /* For all PCI host bridges, walk and insert IVHD entries */
1959 static int
1960 ivrs_host_bridges(Object *obj, void *opaque)
1961 {
1962     GArray *ivhd_blob = opaque;
1963 
1964     if (object_dynamic_cast(obj, TYPE_PCI_HOST_BRIDGE)) {
1965         PCIBus *bus = PCI_HOST_BRIDGE(obj)->bus;
1966 
1967         if (bus && !pci_bus_bypass_iommu(bus)) {
1968             pci_for_each_device_under_bus(bus, insert_ivhd, ivhd_blob);
1969         }
1970     }
1971 
1972     return 0;
1973 }
1974 
1975 static void
1976 build_amd_iommu(GArray *table_data, BIOSLinker *linker, const char *oem_id,
1977                 const char *oem_table_id)
1978 {
1979     AMDVIState *s = AMD_IOMMU_DEVICE(x86_iommu_get_default());
1980     GArray *ivhd_blob = g_array_new(false, true, 1);
1981     AcpiTable table = { .sig = "IVRS", .rev = 1, .oem_id = oem_id,
1982                         .oem_table_id = oem_table_id };
1983     uint64_t feature_report;
1984 
1985     acpi_table_begin(&table, table_data);
1986     /* IVinfo - IO virtualization information common to all
1987      * IOMMU units in a system
1988      */
1989     build_append_int_noprefix(table_data,
1990                              (1UL << 0) | /* EFRSup */
1991                              (40UL << 8), /* PASize */
1992                              4);
1993     /* reserved */
1994     build_append_int_noprefix(table_data, 0, 8);
1995 
1996     /*
1997      * A PCI bus walk, for each PCI host bridge, is necessary to create a
1998      * complete set of IVHD entries.  Do this into a separate blob so that we
1999      * can calculate the total IVRS table length here and then append the new
2000      * blob further below.  Fall back to an entry covering all devices, which
2001      * is sufficient when no aliases are present.
2002      */
2003     object_child_foreach_recursive(object_get_root(),
2004                                    ivrs_host_bridges, ivhd_blob);
2005 
2006     if (!ivhd_blob->len) {
2007         /*
2008          *   Type 1 device entry reporting all devices
2009          *   These are 4-byte device entries currently reporting the range of
2010          *   Refer to Spec - Table 95:IVHD Device Entry Type Codes(4-byte)
2011          */
2012         build_append_int_noprefix(ivhd_blob, 0x0000001, 4);
2013     }
2014 
2015     /*
2016      * When interrupt remapping is supported, we add a special IVHD device
2017      * for type IO-APIC
2018      * Refer to spec - Table 95: IVHD device entry type codes
2019      *
2020      * Linux IOMMU driver checks for the special IVHD device (type IO-APIC).
2021      * See Linux kernel commit 'c2ff5cf5294bcbd7fa50f7d860e90a66db7e5059'
2022      */
2023     if (x86_iommu_ir_supported(x86_iommu_get_default())) {
2024         build_append_int_noprefix(ivhd_blob,
2025                                  (0x1ull << 56) |           /* type IOAPIC */
2026                                  (IOAPIC_SB_DEVID << 40) |  /* IOAPIC devid */
2027                                  0x48,                      /* special device */
2028                                  8);
2029     }
2030 
2031     /* IVHD definition - type 10h */
2032     build_append_int_noprefix(table_data, 0x10, 1);
2033     /* virtualization flags */
2034     build_append_int_noprefix(table_data,
2035                              (1UL << 0) | /* HtTunEn      */
2036                              (1UL << 4) | /* iotblSup     */
2037                              (1UL << 6) | /* PrefSup      */
2038                              (1UL << 7),  /* PPRSup       */
2039                              1);
2040 
2041     /* IVHD length */
2042     build_append_int_noprefix(table_data, ivhd_blob->len + 24, 2);
2043     /* DeviceID */
2044     build_append_int_noprefix(table_data,
2045                               object_property_get_int(OBJECT(s->pci), "addr",
2046                                                       &error_abort), 2);
2047     /* Capability offset */
2048     build_append_int_noprefix(table_data, s->pci->capab_offset, 2);
2049     /* IOMMU base address */
2050     build_append_int_noprefix(table_data, s->mr_mmio.addr, 8);
2051     /* PCI Segment Group */
2052     build_append_int_noprefix(table_data, 0, 2);
2053     /* IOMMU info */
2054     build_append_int_noprefix(table_data, 0, 2);
2055     /* IOMMU Feature Reporting */
2056     feature_report = (48UL << 30) | /* HATS   */
2057                      (48UL << 28) | /* GATS   */
2058                      (1UL << 2)   | /* GTSup  */
2059                      (1UL << 6);    /* GASup  */
2060     if (s->xtsup) {
2061         feature_report |= (1UL << 0); /* XTSup */
2062     }
2063     build_append_int_noprefix(table_data, feature_report, 4);
2064 
2065     /* IVHD entries as found above */
2066     g_array_append_vals(table_data, ivhd_blob->data, ivhd_blob->len);
2067 
2068    /* IVHD definition - type 11h */
2069     build_append_int_noprefix(table_data, 0x11, 1);
2070     /* virtualization flags */
2071     build_append_int_noprefix(table_data,
2072                              (1UL << 0) | /* HtTunEn      */
2073                              (1UL << 4),  /* iotblSup     */
2074                              1);
2075 
2076     /* IVHD length */
2077     build_append_int_noprefix(table_data, ivhd_blob->len + 40, 2);
2078     /* DeviceID */
2079     build_append_int_noprefix(table_data,
2080                               object_property_get_int(OBJECT(s->pci), "addr",
2081                                                       &error_abort), 2);
2082     /* Capability offset */
2083     build_append_int_noprefix(table_data, s->pci->capab_offset, 2);
2084     /* IOMMU base address */
2085     build_append_int_noprefix(table_data, s->mr_mmio.addr, 8);
2086     /* PCI Segment Group */
2087     build_append_int_noprefix(table_data, 0, 2);
2088     /* IOMMU info */
2089     build_append_int_noprefix(table_data, 0, 2);
2090     /* IOMMU Attributes */
2091     build_append_int_noprefix(table_data, 0, 4);
2092     /* EFR Register Image */
2093     build_append_int_noprefix(table_data,
2094                               amdvi_extended_feature_register(s),
2095                               8);
2096     /* EFR Register Image 2 */
2097     build_append_int_noprefix(table_data, 0, 8);
2098 
2099     /* IVHD entries as found above */
2100     g_array_append_vals(table_data, ivhd_blob->data, ivhd_blob->len);
2101 
2102     g_array_free(ivhd_blob, TRUE);
2103     acpi_table_end(linker, &table);
2104 }
2105 
2106 typedef
2107 struct AcpiBuildState {
2108     /* Copy of table in RAM (for patching). */
2109     MemoryRegion *table_mr;
2110     /* Is table patched? */
2111     uint8_t patched;
2112     MemoryRegion *rsdp_mr;
2113     MemoryRegion *linker_mr;
2114 } AcpiBuildState;
2115 
2116 static bool acpi_get_mcfg(AcpiMcfgInfo *mcfg)
2117 {
2118     Object *pci_host;
2119     QObject *o;
2120 
2121     pci_host = acpi_get_i386_pci_host();
2122     if (!pci_host) {
2123         return false;
2124     }
2125 
2126     o = object_property_get_qobject(pci_host, PCIE_HOST_MCFG_BASE, NULL);
2127     if (!o) {
2128         return false;
2129     }
2130     mcfg->base = qnum_get_uint(qobject_to(QNum, o));
2131     qobject_unref(o);
2132     if (mcfg->base == PCIE_BASE_ADDR_UNMAPPED) {
2133         return false;
2134     }
2135 
2136     o = object_property_get_qobject(pci_host, PCIE_HOST_MCFG_SIZE, NULL);
2137     assert(o);
2138     mcfg->size = qnum_get_uint(qobject_to(QNum, o));
2139     qobject_unref(o);
2140     return true;
2141 }
2142 
2143 static
2144 void acpi_build(AcpiBuildTables *tables, MachineState *machine)
2145 {
2146     PCMachineState *pcms = PC_MACHINE(machine);
2147     X86MachineState *x86ms = X86_MACHINE(machine);
2148     DeviceState *iommu = pcms->iommu;
2149     GArray *table_offsets;
2150     unsigned facs, dsdt, rsdt;
2151     AcpiPmInfo pm;
2152     AcpiMiscInfo misc;
2153     AcpiMcfgInfo mcfg;
2154     Range pci_hole = {}, pci_hole64 = {};
2155     uint8_t *u;
2156     GArray *tables_blob = tables->table_data;
2157     AcpiSlicOem slic_oem = { .id = NULL, .table_id = NULL };
2158     Object *vmgenid_dev, *vmclock_dev;
2159     char *oem_id;
2160     char *oem_table_id;
2161 
2162     acpi_get_pm_info(machine, &pm);
2163     acpi_get_misc_info(&misc);
2164     acpi_get_pci_holes(&pci_hole, &pci_hole64);
2165     acpi_get_slic_oem(&slic_oem);
2166 
2167     if (slic_oem.id) {
2168         oem_id = slic_oem.id;
2169     } else {
2170         oem_id = x86ms->oem_id;
2171     }
2172 
2173     if (slic_oem.table_id) {
2174         oem_table_id = slic_oem.table_id;
2175     } else {
2176         oem_table_id = x86ms->oem_table_id;
2177     }
2178 
2179     table_offsets = g_array_new(false, true /* clear */,
2180                                         sizeof(uint32_t));
2181     ACPI_BUILD_DPRINTF("init ACPI tables\n");
2182 
2183     bios_linker_loader_alloc(tables->linker,
2184                              ACPI_BUILD_TABLE_FILE, tables_blob,
2185                              64 /* Ensure FACS is aligned */,
2186                              false /* high memory */);
2187 
2188     /*
2189      * FACS is pointed to by FADT.
2190      * We place it first since it's the only table that has alignment
2191      * requirements.
2192      */
2193     facs = tables_blob->len;
2194     build_facs(tables_blob);
2195 
2196     /* DSDT is pointed to by FADT */
2197     dsdt = tables_blob->len;
2198     build_dsdt(tables_blob, tables->linker, &pm, &misc,
2199                &pci_hole, &pci_hole64, machine);
2200 
2201     /* ACPI tables pointed to by RSDT */
2202     acpi_add_table(table_offsets, tables_blob);
2203     pm.fadt.facs_tbl_offset = &facs;
2204     pm.fadt.dsdt_tbl_offset = &dsdt;
2205     pm.fadt.xdsdt_tbl_offset = &dsdt;
2206     build_fadt(tables_blob, tables->linker, &pm.fadt, oem_id, oem_table_id);
2207 
2208     acpi_add_table(table_offsets, tables_blob);
2209     acpi_build_madt(tables_blob, tables->linker, x86ms,
2210                     x86ms->oem_id, x86ms->oem_table_id);
2211 
2212 #ifdef CONFIG_ACPI_ERST
2213     {
2214         Object *erst_dev;
2215         erst_dev = find_erst_dev();
2216         if (erst_dev) {
2217             acpi_add_table(table_offsets, tables_blob);
2218             build_erst(tables_blob, tables->linker, erst_dev,
2219                        x86ms->oem_id, x86ms->oem_table_id);
2220         }
2221     }
2222 #endif
2223 
2224     vmgenid_dev = find_vmgenid_dev();
2225     if (vmgenid_dev) {
2226         acpi_add_table(table_offsets, tables_blob);
2227         vmgenid_build_acpi(VMGENID(vmgenid_dev), tables_blob,
2228                            tables->vmgenid, tables->linker, x86ms->oem_id);
2229     }
2230 
2231     vmclock_dev = find_vmclock_dev();
2232     if (vmclock_dev) {
2233         acpi_add_table(table_offsets, tables_blob);
2234         vmclock_build_acpi(VMCLOCK(vmclock_dev), tables_blob, tables->linker,
2235                            x86ms->oem_id);
2236     }
2237 
2238     if (misc.has_hpet) {
2239         acpi_add_table(table_offsets, tables_blob);
2240         build_hpet(tables_blob, tables->linker, x86ms->oem_id,
2241                    x86ms->oem_table_id);
2242     }
2243 #ifdef CONFIG_TPM
2244     if (misc.tpm_version != TPM_VERSION_UNSPEC) {
2245         if (misc.tpm_version == TPM_VERSION_1_2) {
2246             acpi_add_table(table_offsets, tables_blob);
2247             build_tpm_tcpa(tables_blob, tables->linker, tables->tcpalog,
2248                            x86ms->oem_id, x86ms->oem_table_id);
2249         } else { /* TPM_VERSION_2_0 */
2250             acpi_add_table(table_offsets, tables_blob);
2251             build_tpm2(tables_blob, tables->linker, tables->tcpalog,
2252                        x86ms->oem_id, x86ms->oem_table_id);
2253         }
2254     }
2255 #endif
2256     if (machine->numa_state->num_nodes) {
2257         acpi_add_table(table_offsets, tables_blob);
2258         build_srat(tables_blob, tables->linker, machine);
2259         if (machine->numa_state->have_numa_distance) {
2260             acpi_add_table(table_offsets, tables_blob);
2261             build_slit(tables_blob, tables->linker, machine, x86ms->oem_id,
2262                        x86ms->oem_table_id);
2263         }
2264         if (machine->numa_state->hmat_enabled) {
2265             acpi_add_table(table_offsets, tables_blob);
2266             build_hmat(tables_blob, tables->linker, machine->numa_state,
2267                        x86ms->oem_id, x86ms->oem_table_id);
2268         }
2269     }
2270     if (acpi_get_mcfg(&mcfg)) {
2271         acpi_add_table(table_offsets, tables_blob);
2272         build_mcfg(tables_blob, tables->linker, &mcfg, x86ms->oem_id,
2273                    x86ms->oem_table_id);
2274     }
2275     if (object_dynamic_cast(OBJECT(iommu), TYPE_AMD_IOMMU_DEVICE)) {
2276         acpi_add_table(table_offsets, tables_blob);
2277         build_amd_iommu(tables_blob, tables->linker, x86ms->oem_id,
2278                         x86ms->oem_table_id);
2279     } else if (object_dynamic_cast(OBJECT(iommu), TYPE_INTEL_IOMMU_DEVICE)) {
2280         acpi_add_table(table_offsets, tables_blob);
2281         build_dmar_q35(tables_blob, tables->linker, x86ms->oem_id,
2282                        x86ms->oem_table_id);
2283     } else if (object_dynamic_cast(OBJECT(iommu), TYPE_VIRTIO_IOMMU_PCI)) {
2284         PCIDevice *pdev = PCI_DEVICE(iommu);
2285 
2286         acpi_add_table(table_offsets, tables_blob);
2287         build_viot(machine, tables_blob, tables->linker, pci_get_bdf(pdev),
2288                    x86ms->oem_id, x86ms->oem_table_id);
2289     }
2290     if (machine->nvdimms_state->is_enabled) {
2291         nvdimm_build_acpi(table_offsets, tables_blob, tables->linker,
2292                           machine->nvdimms_state, machine->ram_slots,
2293                           x86ms->oem_id, x86ms->oem_table_id);
2294     }
2295     if (pcms->cxl_devices_state.is_enabled) {
2296         cxl_build_cedt(table_offsets, tables_blob, tables->linker,
2297                        x86ms->oem_id, x86ms->oem_table_id, &pcms->cxl_devices_state);
2298     }
2299 
2300     acpi_add_table(table_offsets, tables_blob);
2301     build_waet(tables_blob, tables->linker, x86ms->oem_id, x86ms->oem_table_id);
2302 
2303     /* Add tables supplied by user (if any) */
2304     for (u = acpi_table_first(); u; u = acpi_table_next(u)) {
2305         unsigned len = acpi_table_len(u);
2306 
2307         acpi_add_table(table_offsets, tables_blob);
2308         g_array_append_vals(tables_blob, u, len);
2309     }
2310 
2311     /* RSDT is pointed to by RSDP */
2312     rsdt = tables_blob->len;
2313     build_rsdt(tables_blob, tables->linker, table_offsets,
2314                oem_id, oem_table_id);
2315 
2316     /* RSDP is in FSEG memory, so allocate it separately */
2317     {
2318         AcpiRsdpData rsdp_data = {
2319             .revision = 0,
2320             .oem_id = x86ms->oem_id,
2321             .xsdt_tbl_offset = NULL,
2322             .rsdt_tbl_offset = &rsdt,
2323         };
2324         build_rsdp(tables->rsdp, tables->linker, &rsdp_data);
2325     }
2326 
2327     /* We'll expose it all to Guest so we want to reduce
2328      * chance of size changes.
2329      *
2330      * We used to align the tables to 4k, but of course this would
2331      * too simple to be enough.  4k turned out to be too small an
2332      * alignment very soon, and in fact it is almost impossible to
2333      * keep the table size stable for all (max_cpus, max_memory_slots)
2334      * combinations.
2335      */
2336     acpi_align_size(tables_blob, ACPI_BUILD_TABLE_SIZE);
2337 
2338     acpi_align_size(tables->linker->cmd_blob, ACPI_BUILD_ALIGN_SIZE);
2339 
2340     /* Cleanup memory that's no longer used. */
2341     g_array_free(table_offsets, true);
2342     g_free(slic_oem.id);
2343     g_free(slic_oem.table_id);
2344 }
2345 
2346 static void acpi_ram_update(MemoryRegion *mr, GArray *data)
2347 {
2348     uint32_t size = acpi_data_len(data);
2349 
2350     /* Make sure RAM size is correct - in case it got changed e.g. by migration */
2351     memory_region_ram_resize(mr, size, &error_abort);
2352 
2353     memcpy(memory_region_get_ram_ptr(mr), data->data, size);
2354     memory_region_set_dirty(mr, 0, size);
2355 }
2356 
2357 static void acpi_build_update(void *build_opaque)
2358 {
2359     AcpiBuildState *build_state = build_opaque;
2360     AcpiBuildTables tables;
2361 
2362     /* No state to update or already patched? Nothing to do. */
2363     if (!build_state || build_state->patched) {
2364         return;
2365     }
2366     build_state->patched = 1;
2367 
2368     acpi_build_tables_init(&tables);
2369 
2370     acpi_build(&tables, MACHINE(qdev_get_machine()));
2371 
2372     acpi_ram_update(build_state->table_mr, tables.table_data);
2373 
2374     acpi_ram_update(build_state->rsdp_mr, tables.rsdp);
2375 
2376     acpi_ram_update(build_state->linker_mr, tables.linker->cmd_blob);
2377     acpi_build_tables_cleanup(&tables, true);
2378 }
2379 
2380 static void acpi_build_reset(void *build_opaque)
2381 {
2382     AcpiBuildState *build_state = build_opaque;
2383     build_state->patched = 0;
2384 }
2385 
2386 static const VMStateDescription vmstate_acpi_build = {
2387     .name = "acpi_build",
2388     .version_id = 1,
2389     .minimum_version_id = 1,
2390     .fields = (const VMStateField[]) {
2391         VMSTATE_UINT8(patched, AcpiBuildState),
2392         VMSTATE_END_OF_LIST()
2393     },
2394 };
2395 
2396 void acpi_setup(void)
2397 {
2398     PCMachineState *pcms = PC_MACHINE(qdev_get_machine());
2399     X86MachineState *x86ms = X86_MACHINE(pcms);
2400     AcpiBuildTables tables;
2401     AcpiBuildState *build_state;
2402     Object *vmgenid_dev;
2403 #ifdef CONFIG_TPM
2404     TPMIf *tpm;
2405     static FwCfgTPMConfig tpm_config;
2406 #endif
2407 
2408     if (!x86ms->fw_cfg) {
2409         ACPI_BUILD_DPRINTF("No fw cfg. Bailing out.\n");
2410         return;
2411     }
2412 
2413     if (!pcms->acpi_build_enabled) {
2414         ACPI_BUILD_DPRINTF("ACPI build disabled. Bailing out.\n");
2415         return;
2416     }
2417 
2418     if (!x86_machine_is_acpi_enabled(X86_MACHINE(pcms))) {
2419         ACPI_BUILD_DPRINTF("ACPI disabled. Bailing out.\n");
2420         return;
2421     }
2422 
2423     build_state = g_malloc0(sizeof *build_state);
2424 
2425     acpi_build_tables_init(&tables);
2426     acpi_build(&tables, MACHINE(pcms));
2427 
2428     /* Now expose it all to Guest */
2429     build_state->table_mr = acpi_add_rom_blob(acpi_build_update,
2430                                               build_state, tables.table_data,
2431                                               ACPI_BUILD_TABLE_FILE);
2432     assert(build_state->table_mr != NULL);
2433 
2434     build_state->linker_mr =
2435         acpi_add_rom_blob(acpi_build_update, build_state,
2436                           tables.linker->cmd_blob, ACPI_BUILD_LOADER_FILE);
2437 
2438 #ifdef CONFIG_TPM
2439     fw_cfg_add_file(x86ms->fw_cfg, ACPI_BUILD_TPMLOG_FILE,
2440                     tables.tcpalog->data, acpi_data_len(tables.tcpalog));
2441 
2442     tpm = tpm_find();
2443     if (tpm && object_property_get_bool(OBJECT(tpm), "ppi", &error_abort)) {
2444         tpm_config = (FwCfgTPMConfig) {
2445             .tpmppi_address = cpu_to_le32(TPM_PPI_ADDR_BASE),
2446             .tpm_version = tpm_get_version(tpm),
2447             .tpmppi_version = TPM_PPI_VERSION_1_30
2448         };
2449         fw_cfg_add_file(x86ms->fw_cfg, "etc/tpm/config",
2450                         &tpm_config, sizeof tpm_config);
2451     }
2452 #endif
2453 
2454     vmgenid_dev = find_vmgenid_dev();
2455     if (vmgenid_dev) {
2456         vmgenid_add_fw_cfg(VMGENID(vmgenid_dev), x86ms->fw_cfg,
2457                            tables.vmgenid);
2458     }
2459 
2460     build_state->rsdp_mr = acpi_add_rom_blob(acpi_build_update,
2461                                              build_state, tables.rsdp,
2462                                              ACPI_BUILD_RSDP_FILE);
2463 
2464     qemu_register_reset(acpi_build_reset, build_state);
2465     acpi_build_reset(build_state);
2466     vmstate_register(NULL, 0, &vmstate_acpi_build, build_state);
2467 
2468     /* Cleanup tables but don't free the memory: we track it
2469      * in build_state.
2470      */
2471     acpi_build_tables_cleanup(&tables, false);
2472 }
2473