xref: /openbmc/qemu/hw/i386/acpi-build.c (revision 2b37e9f8)
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 "qapi/qmp/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.h"
31 #include "hw/core/cpu.h"
32 #include "target/i386/cpu.h"
33 #include "hw/misc/pvpanic.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/isa/isa.h"
41 #include "hw/block/fdc.h"
42 #include "hw/acpi/memory_hotplug.h"
43 #include "sysemu/tpm.h"
44 #include "hw/acpi/tpm.h"
45 #include "hw/acpi/vmgenid.h"
46 #include "hw/boards.h"
47 #include "sysemu/tpm_backend.h"
48 #include "hw/rtc/mc146818rtc_regs.h"
49 #include "migration/vmstate.h"
50 #include "hw/mem/memory-device.h"
51 #include "hw/mem/nvdimm.h"
52 #include "sysemu/numa.h"
53 #include "sysemu/reset.h"
54 #include "hw/hyperv/vmbus-bridge.h"
55 
56 /* Supported chipsets: */
57 #include "hw/southbridge/piix.h"
58 #include "hw/acpi/pcihp.h"
59 #include "hw/i386/fw_cfg.h"
60 #include "hw/i386/ich9.h"
61 #include "hw/pci/pci_bus.h"
62 #include "hw/pci-host/q35.h"
63 #include "hw/i386/x86-iommu.h"
64 
65 #include "hw/acpi/aml-build.h"
66 #include "hw/acpi/utils.h"
67 #include "hw/acpi/pci.h"
68 
69 #include "qom/qom-qobject.h"
70 #include "hw/i386/amd_iommu.h"
71 #include "hw/i386/intel_iommu.h"
72 
73 #include "hw/acpi/ipmi.h"
74 #include "hw/acpi/hmat.h"
75 
76 /* These are used to size the ACPI tables for -M pc-i440fx-1.7 and
77  * -M pc-i440fx-2.0.  Even if the actual amount of AML generated grows
78  * a little bit, there should be plenty of free space since the DSDT
79  * shrunk by ~1.5k between QEMU 2.0 and QEMU 2.1.
80  */
81 #define ACPI_BUILD_LEGACY_CPU_AML_SIZE    97
82 #define ACPI_BUILD_ALIGN_SIZE             0x1000
83 
84 #define ACPI_BUILD_TABLE_SIZE             0x20000
85 
86 /* #define DEBUG_ACPI_BUILD */
87 #ifdef DEBUG_ACPI_BUILD
88 #define ACPI_BUILD_DPRINTF(fmt, ...)        \
89     do {printf("ACPI_BUILD: " fmt, ## __VA_ARGS__); } while (0)
90 #else
91 #define ACPI_BUILD_DPRINTF(fmt, ...)
92 #endif
93 
94 typedef struct AcpiPmInfo {
95     bool s3_disabled;
96     bool s4_disabled;
97     bool pcihp_bridge_en;
98     bool smi_on_cpuhp;
99     bool pcihp_root_en;
100     uint8_t s4_val;
101     AcpiFadtData fadt;
102     uint16_t cpu_hp_io_base;
103     uint16_t pcihp_io_base;
104     uint16_t pcihp_io_len;
105 } AcpiPmInfo;
106 
107 typedef struct AcpiMiscInfo {
108     bool is_piix4;
109     bool has_hpet;
110     TPMVersion tpm_version;
111     const unsigned char *dsdt_code;
112     unsigned dsdt_size;
113     uint16_t pvpanic_port;
114     uint16_t applesmc_io_base;
115 } AcpiMiscInfo;
116 
117 typedef struct AcpiBuildPciBusHotplugState {
118     GArray *device_table;
119     GArray *notify_table;
120     struct AcpiBuildPciBusHotplugState *parent;
121     bool pcihp_bridge_en;
122 } AcpiBuildPciBusHotplugState;
123 
124 typedef struct FwCfgTPMConfig {
125     uint32_t tpmppi_address;
126     uint8_t tpm_version;
127     uint8_t tpmppi_version;
128 } QEMU_PACKED FwCfgTPMConfig;
129 
130 static bool acpi_get_mcfg(AcpiMcfgInfo *mcfg);
131 
132 const struct AcpiGenericAddress x86_nvdimm_acpi_dsmio = {
133     .space_id = AML_AS_SYSTEM_IO,
134     .address = NVDIMM_ACPI_IO_BASE,
135     .bit_width = NVDIMM_ACPI_IO_LEN << 3
136 };
137 
138 static void init_common_fadt_data(MachineState *ms, Object *o,
139                                   AcpiFadtData *data)
140 {
141     uint32_t io = object_property_get_uint(o, ACPI_PM_PROP_PM_IO_BASE, NULL);
142     AmlAddressSpace as = AML_AS_SYSTEM_IO;
143     AcpiFadtData fadt = {
144         .rev = 3,
145         .flags =
146             (1 << ACPI_FADT_F_WBINVD) |
147             (1 << ACPI_FADT_F_PROC_C1) |
148             (1 << ACPI_FADT_F_SLP_BUTTON) |
149             (1 << ACPI_FADT_F_RTC_S4) |
150             (1 << ACPI_FADT_F_USE_PLATFORM_CLOCK) |
151             /* APIC destination mode ("Flat Logical") has an upper limit of 8
152              * CPUs for more than 8 CPUs, "Clustered Logical" mode has to be
153              * used
154              */
155             ((ms->smp.max_cpus > 8) ?
156                         (1 << ACPI_FADT_F_FORCE_APIC_CLUSTER_MODEL) : 0),
157         .int_model = 1 /* Multiple APIC */,
158         .rtc_century = RTC_CENTURY,
159         .plvl2_lat = 0xfff /* C2 state not supported */,
160         .plvl3_lat = 0xfff /* C3 state not supported */,
161         .smi_cmd = ACPI_PORT_SMI_CMD,
162         .sci_int = object_property_get_uint(o, ACPI_PM_PROP_SCI_INT, NULL),
163         .acpi_enable_cmd =
164             object_property_get_uint(o, ACPI_PM_PROP_ACPI_ENABLE_CMD, NULL),
165         .acpi_disable_cmd =
166             object_property_get_uint(o, ACPI_PM_PROP_ACPI_DISABLE_CMD, NULL),
167         .pm1a_evt = { .space_id = as, .bit_width = 4 * 8, .address = io },
168         .pm1a_cnt = { .space_id = as, .bit_width = 2 * 8,
169                       .address = io + 0x04 },
170         .pm_tmr = { .space_id = as, .bit_width = 4 * 8, .address = io + 0x08 },
171         .gpe0_blk = { .space_id = as, .bit_width =
172             object_property_get_uint(o, ACPI_PM_PROP_GPE0_BLK_LEN, NULL) * 8,
173             .address = object_property_get_uint(o, ACPI_PM_PROP_GPE0_BLK, NULL)
174         },
175     };
176     *data = fadt;
177 }
178 
179 static Object *object_resolve_type_unambiguous(const char *typename)
180 {
181     bool ambig;
182     Object *o = object_resolve_path_type("", typename, &ambig);
183 
184     if (ambig || !o) {
185         return NULL;
186     }
187     return o;
188 }
189 
190 static void acpi_get_pm_info(MachineState *machine, AcpiPmInfo *pm)
191 {
192     Object *piix = object_resolve_type_unambiguous(TYPE_PIIX4_PM);
193     Object *lpc = object_resolve_type_unambiguous(TYPE_ICH9_LPC_DEVICE);
194     Object *obj = piix ? piix : lpc;
195     QObject *o;
196     pm->cpu_hp_io_base = 0;
197     pm->pcihp_io_base = 0;
198     pm->pcihp_io_len = 0;
199     pm->smi_on_cpuhp = false;
200 
201     assert(obj);
202     init_common_fadt_data(machine, obj, &pm->fadt);
203     if (piix) {
204         /* w2k requires FADT(rev1) or it won't boot, keep PC compatible */
205         pm->fadt.rev = 1;
206         pm->cpu_hp_io_base = PIIX4_CPU_HOTPLUG_IO_BASE;
207         pm->pcihp_io_base =
208             object_property_get_uint(obj, ACPI_PCIHP_IO_BASE_PROP, NULL);
209         pm->pcihp_io_len =
210             object_property_get_uint(obj, ACPI_PCIHP_IO_LEN_PROP, NULL);
211     }
212     if (lpc) {
213         uint64_t smi_features = object_property_get_uint(lpc,
214             ICH9_LPC_SMI_NEGOTIATED_FEAT_PROP, NULL);
215         struct AcpiGenericAddress r = { .space_id = AML_AS_SYSTEM_IO,
216             .bit_width = 8, .address = ICH9_RST_CNT_IOPORT };
217         pm->fadt.reset_reg = r;
218         pm->fadt.reset_val = 0xf;
219         pm->fadt.flags |= 1 << ACPI_FADT_F_RESET_REG_SUP;
220         pm->cpu_hp_io_base = ICH9_CPU_HOTPLUG_IO_BASE;
221         pm->smi_on_cpuhp =
222             !!(smi_features & BIT_ULL(ICH9_LPC_SMI_F_CPU_HOTPLUG_BIT));
223     }
224 
225     /* The above need not be conditional on machine type because the reset port
226      * happens to be the same on PIIX (pc) and ICH9 (q35). */
227     QEMU_BUILD_BUG_ON(ICH9_RST_CNT_IOPORT != PIIX_RCR_IOPORT);
228 
229     /* Fill in optional s3/s4 related properties */
230     o = object_property_get_qobject(obj, ACPI_PM_PROP_S3_DISABLED, NULL);
231     if (o) {
232         pm->s3_disabled = qnum_get_uint(qobject_to(QNum, o));
233     } else {
234         pm->s3_disabled = false;
235     }
236     qobject_unref(o);
237     o = object_property_get_qobject(obj, ACPI_PM_PROP_S4_DISABLED, NULL);
238     if (o) {
239         pm->s4_disabled = qnum_get_uint(qobject_to(QNum, o));
240     } else {
241         pm->s4_disabled = false;
242     }
243     qobject_unref(o);
244     o = object_property_get_qobject(obj, ACPI_PM_PROP_S4_VAL, NULL);
245     if (o) {
246         pm->s4_val = qnum_get_uint(qobject_to(QNum, o));
247     } else {
248         pm->s4_val = false;
249     }
250     qobject_unref(o);
251 
252     pm->pcihp_bridge_en =
253         object_property_get_bool(obj, "acpi-pci-hotplug-with-bridge-support",
254                                  NULL);
255     pm->pcihp_root_en =
256         object_property_get_bool(obj, "acpi-root-pci-hotplug",
257                                  NULL);
258 }
259 
260 static void acpi_get_misc_info(AcpiMiscInfo *info)
261 {
262     Object *piix = object_resolve_type_unambiguous(TYPE_PIIX4_PM);
263     Object *lpc = object_resolve_type_unambiguous(TYPE_ICH9_LPC_DEVICE);
264     assert(!!piix != !!lpc);
265 
266     if (piix) {
267         info->is_piix4 = true;
268     }
269     if (lpc) {
270         info->is_piix4 = false;
271     }
272 
273     info->has_hpet = hpet_find();
274     info->tpm_version = tpm_get_version(tpm_find());
275     info->pvpanic_port = pvpanic_port();
276     info->applesmc_io_base = applesmc_port();
277 }
278 
279 /*
280  * Because of the PXB hosts we cannot simply query TYPE_PCI_HOST_BRIDGE.
281  * On i386 arch we only have two pci hosts, so we can look only for them.
282  */
283 static Object *acpi_get_i386_pci_host(void)
284 {
285     PCIHostState *host;
286 
287     host = OBJECT_CHECK(PCIHostState,
288                         object_resolve_path("/machine/i440fx", NULL),
289                         TYPE_PCI_HOST_BRIDGE);
290     if (!host) {
291         host = OBJECT_CHECK(PCIHostState,
292                             object_resolve_path("/machine/q35", NULL),
293                             TYPE_PCI_HOST_BRIDGE);
294     }
295 
296     return OBJECT(host);
297 }
298 
299 static void acpi_get_pci_holes(Range *hole, Range *hole64)
300 {
301     Object *pci_host;
302 
303     pci_host = acpi_get_i386_pci_host();
304     g_assert(pci_host);
305 
306     range_set_bounds1(hole,
307                       object_property_get_uint(pci_host,
308                                                PCI_HOST_PROP_PCI_HOLE_START,
309                                                NULL),
310                       object_property_get_uint(pci_host,
311                                                PCI_HOST_PROP_PCI_HOLE_END,
312                                                NULL));
313     range_set_bounds1(hole64,
314                       object_property_get_uint(pci_host,
315                                                PCI_HOST_PROP_PCI_HOLE64_START,
316                                                NULL),
317                       object_property_get_uint(pci_host,
318                                                PCI_HOST_PROP_PCI_HOLE64_END,
319                                                NULL));
320 }
321 
322 static void acpi_align_size(GArray *blob, unsigned align)
323 {
324     /* Align size to multiple of given size. This reduces the chance
325      * we need to change size in the future (breaking cross version migration).
326      */
327     g_array_set_size(blob, ROUND_UP(acpi_data_len(blob), align));
328 }
329 
330 /* FACS */
331 static void
332 build_facs(GArray *table_data)
333 {
334     AcpiFacsDescriptorRev1 *facs = acpi_data_push(table_data, sizeof *facs);
335     memcpy(&facs->signature, "FACS", 4);
336     facs->length = cpu_to_le32(sizeof(*facs));
337 }
338 
339 static void build_append_pcihp_notify_entry(Aml *method, int slot)
340 {
341     Aml *if_ctx;
342     int32_t devfn = PCI_DEVFN(slot, 0);
343 
344     if_ctx = aml_if(aml_and(aml_arg(0), aml_int(0x1U << slot), NULL));
345     aml_append(if_ctx, aml_notify(aml_name("S%.02X", devfn), aml_arg(1)));
346     aml_append(method, if_ctx);
347 }
348 
349 static void build_append_pci_bus_devices(Aml *parent_scope, PCIBus *bus,
350                                          bool pcihp_bridge_en)
351 {
352     Aml *dev, *notify_method = NULL, *method;
353     QObject *bsel;
354     PCIBus *sec;
355     int i;
356 
357     bsel = object_property_get_qobject(OBJECT(bus), ACPI_PCIHP_PROP_BSEL, NULL);
358     if (bsel) {
359         uint64_t bsel_val = qnum_get_uint(qobject_to(QNum, bsel));
360 
361         aml_append(parent_scope, aml_name_decl("BSEL", aml_int(bsel_val)));
362         notify_method = aml_method("DVNT", 2, AML_NOTSERIALIZED);
363     }
364 
365     for (i = 0; i < ARRAY_SIZE(bus->devices); i += PCI_FUNC_MAX) {
366         DeviceClass *dc;
367         PCIDeviceClass *pc;
368         PCIDevice *pdev = bus->devices[i];
369         int slot = PCI_SLOT(i);
370         bool hotplug_enabled_dev;
371         bool bridge_in_acpi;
372         bool cold_plugged_bridge;
373 
374         if (!pdev) {
375             if (bsel) { /* add hotplug slots for non present devices */
376                 dev = aml_device("S%.02X", PCI_DEVFN(slot, 0));
377                 aml_append(dev, aml_name_decl("_SUN", aml_int(slot)));
378                 aml_append(dev, aml_name_decl("_ADR", aml_int(slot << 16)));
379                 method = aml_method("_EJ0", 1, AML_NOTSERIALIZED);
380                 aml_append(method,
381                     aml_call2("PCEJ", aml_name("BSEL"), aml_name("_SUN"))
382                 );
383                 aml_append(dev, method);
384                 aml_append(parent_scope, dev);
385 
386                 build_append_pcihp_notify_entry(notify_method, slot);
387             }
388             continue;
389         }
390 
391         pc = PCI_DEVICE_GET_CLASS(pdev);
392         dc = DEVICE_GET_CLASS(pdev);
393 
394         /*
395          * Cold plugged bridges aren't themselves hot-pluggable.
396          * Hotplugged bridges *are* hot-pluggable.
397          */
398         cold_plugged_bridge = pc->is_bridge && !DEVICE(pdev)->hotplugged;
399         bridge_in_acpi =  cold_plugged_bridge && pcihp_bridge_en;
400 
401         hotplug_enabled_dev = bsel && dc->hotpluggable && !cold_plugged_bridge;
402 
403         if (pc->class_id == PCI_CLASS_BRIDGE_ISA) {
404             continue;
405         }
406 
407         /* start to compose PCI slot descriptor */
408         dev = aml_device("S%.02X", PCI_DEVFN(slot, 0));
409         aml_append(dev, aml_name_decl("_ADR", aml_int(slot << 16)));
410 
411         if (pc->class_id == PCI_CLASS_DISPLAY_VGA) {
412             /* add VGA specific AML methods */
413             int s3d;
414 
415             if (object_dynamic_cast(OBJECT(pdev), "qxl-vga")) {
416                 s3d = 3;
417             } else {
418                 s3d = 0;
419             }
420 
421             method = aml_method("_S1D", 0, AML_NOTSERIALIZED);
422             aml_append(method, aml_return(aml_int(0)));
423             aml_append(dev, method);
424 
425             method = aml_method("_S2D", 0, AML_NOTSERIALIZED);
426             aml_append(method, aml_return(aml_int(0)));
427             aml_append(dev, method);
428 
429             method = aml_method("_S3D", 0, AML_NOTSERIALIZED);
430             aml_append(method, aml_return(aml_int(s3d)));
431             aml_append(dev, method);
432         } else if (hotplug_enabled_dev) {
433             /* add _SUN/_EJ0 to make slot hotpluggable  */
434             aml_append(dev, aml_name_decl("_SUN", aml_int(slot)));
435 
436             method = aml_method("_EJ0", 1, AML_NOTSERIALIZED);
437             aml_append(method,
438                 aml_call2("PCEJ", aml_name("BSEL"), aml_name("_SUN"))
439             );
440             aml_append(dev, method);
441 
442             if (bsel) {
443                 build_append_pcihp_notify_entry(notify_method, slot);
444             }
445         } else if (bridge_in_acpi) {
446             /*
447              * device is coldplugged bridge,
448              * add child device descriptions into its scope
449              */
450             PCIBus *sec_bus = pci_bridge_get_sec_bus(PCI_BRIDGE(pdev));
451 
452             build_append_pci_bus_devices(dev, sec_bus, pcihp_bridge_en);
453         }
454         /* slot descriptor has been composed, add it into parent context */
455         aml_append(parent_scope, dev);
456     }
457 
458     if (bsel) {
459         aml_append(parent_scope, notify_method);
460     }
461 
462     /* Append PCNT method to notify about events on local and child buses.
463      * Add this method for root bus only when hotplug is enabled since DSDT
464      * expects it.
465      */
466     if (bsel || pcihp_bridge_en) {
467         method = aml_method("PCNT", 0, AML_NOTSERIALIZED);
468 
469         /* If bus supports hotplug select it and notify about local events */
470         if (bsel) {
471             uint64_t bsel_val = qnum_get_uint(qobject_to(QNum, bsel));
472 
473             aml_append(method, aml_store(aml_int(bsel_val), aml_name("BNUM")));
474             aml_append(method, aml_call2("DVNT", aml_name("PCIU"),
475                                          aml_int(1))); /* Device Check */
476             aml_append(method, aml_call2("DVNT", aml_name("PCID"),
477                                          aml_int(3))); /* Eject Request */
478         }
479 
480         /* Notify about child bus events in any case */
481         if (pcihp_bridge_en) {
482             QLIST_FOREACH(sec, &bus->child, sibling) {
483                 int32_t devfn = sec->parent_dev->devfn;
484 
485                 if (pci_bus_is_root(sec) || pci_bus_is_express(sec)) {
486                     continue;
487                 }
488 
489                 aml_append(method, aml_name("^S%.02X.PCNT", devfn));
490             }
491         }
492 
493         aml_append(parent_scope, method);
494     }
495     qobject_unref(bsel);
496 }
497 
498 /**
499  * build_prt_entry:
500  * @link_name: link name for PCI route entry
501  *
502  * build AML package containing a PCI route entry for @link_name
503  */
504 static Aml *build_prt_entry(const char *link_name)
505 {
506     Aml *a_zero = aml_int(0);
507     Aml *pkg = aml_package(4);
508     aml_append(pkg, a_zero);
509     aml_append(pkg, a_zero);
510     aml_append(pkg, aml_name("%s", link_name));
511     aml_append(pkg, a_zero);
512     return pkg;
513 }
514 
515 /*
516  * initialize_route - Initialize the interrupt routing rule
517  * through a specific LINK:
518  *  if (lnk_idx == idx)
519  *      route using link 'link_name'
520  */
521 static Aml *initialize_route(Aml *route, const char *link_name,
522                              Aml *lnk_idx, int idx)
523 {
524     Aml *if_ctx = aml_if(aml_equal(lnk_idx, aml_int(idx)));
525     Aml *pkg = build_prt_entry(link_name);
526 
527     aml_append(if_ctx, aml_store(pkg, route));
528 
529     return if_ctx;
530 }
531 
532 /*
533  * build_prt - Define interrupt rounting rules
534  *
535  * Returns an array of 128 routes, one for each device,
536  * based on device location.
537  * The main goal is to equaly distribute the interrupts
538  * over the 4 existing ACPI links (works only for i440fx).
539  * The hash function is  (slot + pin) & 3 -> "LNK[D|A|B|C]".
540  *
541  */
542 static Aml *build_prt(bool is_pci0_prt)
543 {
544     Aml *method, *while_ctx, *pin, *res;
545 
546     method = aml_method("_PRT", 0, AML_NOTSERIALIZED);
547     res = aml_local(0);
548     pin = aml_local(1);
549     aml_append(method, aml_store(aml_package(128), res));
550     aml_append(method, aml_store(aml_int(0), pin));
551 
552     /* while (pin < 128) */
553     while_ctx = aml_while(aml_lless(pin, aml_int(128)));
554     {
555         Aml *slot = aml_local(2);
556         Aml *lnk_idx = aml_local(3);
557         Aml *route = aml_local(4);
558 
559         /* slot = pin >> 2 */
560         aml_append(while_ctx,
561                    aml_store(aml_shiftright(pin, aml_int(2), NULL), slot));
562         /* lnk_idx = (slot + pin) & 3 */
563         aml_append(while_ctx,
564             aml_store(aml_and(aml_add(pin, slot, NULL), aml_int(3), NULL),
565                       lnk_idx));
566 
567         /* route[2] = "LNK[D|A|B|C]", selection based on pin % 3  */
568         aml_append(while_ctx, initialize_route(route, "LNKD", lnk_idx, 0));
569         if (is_pci0_prt) {
570             Aml *if_device_1, *if_pin_4, *else_pin_4;
571 
572             /* device 1 is the power-management device, needs SCI */
573             if_device_1 = aml_if(aml_equal(lnk_idx, aml_int(1)));
574             {
575                 if_pin_4 = aml_if(aml_equal(pin, aml_int(4)));
576                 {
577                     aml_append(if_pin_4,
578                         aml_store(build_prt_entry("LNKS"), route));
579                 }
580                 aml_append(if_device_1, if_pin_4);
581                 else_pin_4 = aml_else();
582                 {
583                     aml_append(else_pin_4,
584                         aml_store(build_prt_entry("LNKA"), route));
585                 }
586                 aml_append(if_device_1, else_pin_4);
587             }
588             aml_append(while_ctx, if_device_1);
589         } else {
590             aml_append(while_ctx, initialize_route(route, "LNKA", lnk_idx, 1));
591         }
592         aml_append(while_ctx, initialize_route(route, "LNKB", lnk_idx, 2));
593         aml_append(while_ctx, initialize_route(route, "LNKC", lnk_idx, 3));
594 
595         /* route[0] = 0x[slot]FFFF */
596         aml_append(while_ctx,
597             aml_store(aml_or(aml_shiftleft(slot, aml_int(16)), aml_int(0xFFFF),
598                              NULL),
599                       aml_index(route, aml_int(0))));
600         /* route[1] = pin & 3 */
601         aml_append(while_ctx,
602             aml_store(aml_and(pin, aml_int(3), NULL),
603                       aml_index(route, aml_int(1))));
604         /* res[pin] = route */
605         aml_append(while_ctx, aml_store(route, aml_index(res, pin)));
606         /* pin++ */
607         aml_append(while_ctx, aml_increment(pin));
608     }
609     aml_append(method, while_ctx);
610     /* return res*/
611     aml_append(method, aml_return(res));
612 
613     return method;
614 }
615 
616 typedef struct CrsRangeEntry {
617     uint64_t base;
618     uint64_t limit;
619 } CrsRangeEntry;
620 
621 static void crs_range_insert(GPtrArray *ranges, uint64_t base, uint64_t limit)
622 {
623     CrsRangeEntry *entry;
624 
625     entry = g_malloc(sizeof(*entry));
626     entry->base = base;
627     entry->limit = limit;
628 
629     g_ptr_array_add(ranges, entry);
630 }
631 
632 static void crs_range_free(gpointer data)
633 {
634     CrsRangeEntry *entry = (CrsRangeEntry *)data;
635     g_free(entry);
636 }
637 
638 typedef struct CrsRangeSet {
639     GPtrArray *io_ranges;
640     GPtrArray *mem_ranges;
641     GPtrArray *mem_64bit_ranges;
642  } CrsRangeSet;
643 
644 static void crs_range_set_init(CrsRangeSet *range_set)
645 {
646     range_set->io_ranges = g_ptr_array_new_with_free_func(crs_range_free);
647     range_set->mem_ranges = g_ptr_array_new_with_free_func(crs_range_free);
648     range_set->mem_64bit_ranges =
649             g_ptr_array_new_with_free_func(crs_range_free);
650 }
651 
652 static void crs_range_set_free(CrsRangeSet *range_set)
653 {
654     g_ptr_array_free(range_set->io_ranges, true);
655     g_ptr_array_free(range_set->mem_ranges, true);
656     g_ptr_array_free(range_set->mem_64bit_ranges, true);
657 }
658 
659 static gint crs_range_compare(gconstpointer a, gconstpointer b)
660 {
661     CrsRangeEntry *entry_a = *(CrsRangeEntry **)a;
662     CrsRangeEntry *entry_b = *(CrsRangeEntry **)b;
663 
664     if (entry_a->base < entry_b->base) {
665         return -1;
666     } else if (entry_a->base > entry_b->base) {
667         return 1;
668     } else {
669         return 0;
670     }
671 }
672 
673 /*
674  * crs_replace_with_free_ranges - given the 'used' ranges within [start - end]
675  * interval, computes the 'free' ranges from the same interval.
676  * Example: If the input array is { [a1 - a2],[b1 - b2] }, the function
677  * will return { [base - a1], [a2 - b1], [b2 - limit] }.
678  */
679 static void crs_replace_with_free_ranges(GPtrArray *ranges,
680                                          uint64_t start, uint64_t end)
681 {
682     GPtrArray *free_ranges = g_ptr_array_new();
683     uint64_t free_base = start;
684     int i;
685 
686     g_ptr_array_sort(ranges, crs_range_compare);
687     for (i = 0; i < ranges->len; i++) {
688         CrsRangeEntry *used = g_ptr_array_index(ranges, i);
689 
690         if (free_base < used->base) {
691             crs_range_insert(free_ranges, free_base, used->base - 1);
692         }
693 
694         free_base = used->limit + 1;
695     }
696 
697     if (free_base < end) {
698         crs_range_insert(free_ranges, free_base, end);
699     }
700 
701     g_ptr_array_set_size(ranges, 0);
702     for (i = 0; i < free_ranges->len; i++) {
703         g_ptr_array_add(ranges, g_ptr_array_index(free_ranges, i));
704     }
705 
706     g_ptr_array_free(free_ranges, true);
707 }
708 
709 /*
710  * crs_range_merge - merges adjacent ranges in the given array.
711  * Array elements are deleted and replaced with the merged ranges.
712  */
713 static void crs_range_merge(GPtrArray *range)
714 {
715     GPtrArray *tmp =  g_ptr_array_new_with_free_func(crs_range_free);
716     CrsRangeEntry *entry;
717     uint64_t range_base, range_limit;
718     int i;
719 
720     if (!range->len) {
721         return;
722     }
723 
724     g_ptr_array_sort(range, crs_range_compare);
725 
726     entry = g_ptr_array_index(range, 0);
727     range_base = entry->base;
728     range_limit = entry->limit;
729     for (i = 1; i < range->len; i++) {
730         entry = g_ptr_array_index(range, i);
731         if (entry->base - 1 == range_limit) {
732             range_limit = entry->limit;
733         } else {
734             crs_range_insert(tmp, range_base, range_limit);
735             range_base = entry->base;
736             range_limit = entry->limit;
737         }
738     }
739     crs_range_insert(tmp, range_base, range_limit);
740 
741     g_ptr_array_set_size(range, 0);
742     for (i = 0; i < tmp->len; i++) {
743         entry = g_ptr_array_index(tmp, i);
744         crs_range_insert(range, entry->base, entry->limit);
745     }
746     g_ptr_array_free(tmp, true);
747 }
748 
749 static Aml *build_crs(PCIHostState *host, CrsRangeSet *range_set)
750 {
751     Aml *crs = aml_resource_template();
752     CrsRangeSet temp_range_set;
753     CrsRangeEntry *entry;
754     uint8_t max_bus = pci_bus_num(host->bus);
755     uint8_t type;
756     int devfn;
757     int i;
758 
759     crs_range_set_init(&temp_range_set);
760     for (devfn = 0; devfn < ARRAY_SIZE(host->bus->devices); devfn++) {
761         uint64_t range_base, range_limit;
762         PCIDevice *dev = host->bus->devices[devfn];
763 
764         if (!dev) {
765             continue;
766         }
767 
768         for (i = 0; i < PCI_NUM_REGIONS; i++) {
769             PCIIORegion *r = &dev->io_regions[i];
770 
771             range_base = r->addr;
772             range_limit = r->addr + r->size - 1;
773 
774             /*
775              * Work-around for old bioses
776              * that do not support multiple root buses
777              */
778             if (!range_base || range_base > range_limit) {
779                 continue;
780             }
781 
782             if (r->type & PCI_BASE_ADDRESS_SPACE_IO) {
783                 crs_range_insert(temp_range_set.io_ranges,
784                                  range_base, range_limit);
785             } else { /* "memory" */
786                 uint64_t length = range_limit - range_base + 1;
787                 if (range_limit <= UINT32_MAX && length <= UINT32_MAX) {
788                     crs_range_insert(temp_range_set.mem_ranges, range_base,
789                                      range_limit);
790                 } else {
791                     crs_range_insert(temp_range_set.mem_64bit_ranges,
792                                      range_base, range_limit);
793                 }
794             }
795         }
796 
797         type = dev->config[PCI_HEADER_TYPE] & ~PCI_HEADER_TYPE_MULTI_FUNCTION;
798         if (type == PCI_HEADER_TYPE_BRIDGE) {
799             uint8_t subordinate = dev->config[PCI_SUBORDINATE_BUS];
800             if (subordinate > max_bus) {
801                 max_bus = subordinate;
802             }
803 
804             range_base = pci_bridge_get_base(dev, PCI_BASE_ADDRESS_SPACE_IO);
805             range_limit = pci_bridge_get_limit(dev, PCI_BASE_ADDRESS_SPACE_IO);
806 
807             /*
808              * Work-around for old bioses
809              * that do not support multiple root buses
810              */
811             if (range_base && range_base <= range_limit) {
812                 crs_range_insert(temp_range_set.io_ranges,
813                                  range_base, range_limit);
814             }
815 
816             range_base =
817                 pci_bridge_get_base(dev, PCI_BASE_ADDRESS_SPACE_MEMORY);
818             range_limit =
819                 pci_bridge_get_limit(dev, PCI_BASE_ADDRESS_SPACE_MEMORY);
820 
821             /*
822              * Work-around for old bioses
823              * that do not support multiple root buses
824              */
825             if (range_base && range_base <= range_limit) {
826                 uint64_t length = range_limit - range_base + 1;
827                 if (range_limit <= UINT32_MAX && length <= UINT32_MAX) {
828                     crs_range_insert(temp_range_set.mem_ranges,
829                                      range_base, range_limit);
830                 } else {
831                     crs_range_insert(temp_range_set.mem_64bit_ranges,
832                                      range_base, range_limit);
833                 }
834             }
835 
836             range_base =
837                 pci_bridge_get_base(dev, PCI_BASE_ADDRESS_MEM_PREFETCH);
838             range_limit =
839                 pci_bridge_get_limit(dev, PCI_BASE_ADDRESS_MEM_PREFETCH);
840 
841             /*
842              * Work-around for old bioses
843              * that do not support multiple root buses
844              */
845             if (range_base && range_base <= range_limit) {
846                 uint64_t length = range_limit - range_base + 1;
847                 if (range_limit <= UINT32_MAX && length <= UINT32_MAX) {
848                     crs_range_insert(temp_range_set.mem_ranges,
849                                      range_base, range_limit);
850                 } else {
851                     crs_range_insert(temp_range_set.mem_64bit_ranges,
852                                      range_base, range_limit);
853                 }
854             }
855         }
856     }
857 
858     crs_range_merge(temp_range_set.io_ranges);
859     for (i = 0; i < temp_range_set.io_ranges->len; i++) {
860         entry = g_ptr_array_index(temp_range_set.io_ranges, i);
861         aml_append(crs,
862                    aml_word_io(AML_MIN_FIXED, AML_MAX_FIXED,
863                                AML_POS_DECODE, AML_ENTIRE_RANGE,
864                                0, entry->base, entry->limit, 0,
865                                entry->limit - entry->base + 1));
866         crs_range_insert(range_set->io_ranges, entry->base, entry->limit);
867     }
868 
869     crs_range_merge(temp_range_set.mem_ranges);
870     for (i = 0; i < temp_range_set.mem_ranges->len; i++) {
871         entry = g_ptr_array_index(temp_range_set.mem_ranges, i);
872         assert(entry->limit <= UINT32_MAX &&
873                (entry->limit - entry->base + 1) <= UINT32_MAX);
874         aml_append(crs,
875                    aml_dword_memory(AML_POS_DECODE, AML_MIN_FIXED,
876                                     AML_MAX_FIXED, AML_NON_CACHEABLE,
877                                     AML_READ_WRITE,
878                                     0, entry->base, entry->limit, 0,
879                                     entry->limit - entry->base + 1));
880         crs_range_insert(range_set->mem_ranges, entry->base, entry->limit);
881     }
882 
883     crs_range_merge(temp_range_set.mem_64bit_ranges);
884     for (i = 0; i < temp_range_set.mem_64bit_ranges->len; i++) {
885         entry = g_ptr_array_index(temp_range_set.mem_64bit_ranges, i);
886         aml_append(crs,
887                    aml_qword_memory(AML_POS_DECODE, AML_MIN_FIXED,
888                                     AML_MAX_FIXED, AML_NON_CACHEABLE,
889                                     AML_READ_WRITE,
890                                     0, entry->base, entry->limit, 0,
891                                     entry->limit - entry->base + 1));
892         crs_range_insert(range_set->mem_64bit_ranges,
893                          entry->base, entry->limit);
894     }
895 
896     crs_range_set_free(&temp_range_set);
897 
898     aml_append(crs,
899         aml_word_bus_number(AML_MIN_FIXED, AML_MAX_FIXED, AML_POS_DECODE,
900                             0,
901                             pci_bus_num(host->bus),
902                             max_bus,
903                             0,
904                             max_bus - pci_bus_num(host->bus) + 1));
905 
906     return crs;
907 }
908 
909 static void build_hpet_aml(Aml *table)
910 {
911     Aml *crs;
912     Aml *field;
913     Aml *method;
914     Aml *if_ctx;
915     Aml *scope = aml_scope("_SB");
916     Aml *dev = aml_device("HPET");
917     Aml *zero = aml_int(0);
918     Aml *id = aml_local(0);
919     Aml *period = aml_local(1);
920 
921     aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0103")));
922     aml_append(dev, aml_name_decl("_UID", zero));
923 
924     aml_append(dev,
925         aml_operation_region("HPTM", AML_SYSTEM_MEMORY, aml_int(HPET_BASE),
926                              HPET_LEN));
927     field = aml_field("HPTM", AML_DWORD_ACC, AML_LOCK, AML_PRESERVE);
928     aml_append(field, aml_named_field("VEND", 32));
929     aml_append(field, aml_named_field("PRD", 32));
930     aml_append(dev, field);
931 
932     method = aml_method("_STA", 0, AML_NOTSERIALIZED);
933     aml_append(method, aml_store(aml_name("VEND"), id));
934     aml_append(method, aml_store(aml_name("PRD"), period));
935     aml_append(method, aml_shiftright(id, aml_int(16), id));
936     if_ctx = aml_if(aml_lor(aml_equal(id, zero),
937                             aml_equal(id, aml_int(0xffff))));
938     {
939         aml_append(if_ctx, aml_return(zero));
940     }
941     aml_append(method, if_ctx);
942 
943     if_ctx = aml_if(aml_lor(aml_equal(period, zero),
944                             aml_lgreater(period, aml_int(100000000))));
945     {
946         aml_append(if_ctx, aml_return(zero));
947     }
948     aml_append(method, if_ctx);
949 
950     aml_append(method, aml_return(aml_int(0x0F)));
951     aml_append(dev, method);
952 
953     crs = aml_resource_template();
954     aml_append(crs, aml_memory32_fixed(HPET_BASE, HPET_LEN, AML_READ_ONLY));
955     aml_append(dev, aml_name_decl("_CRS", crs));
956 
957     aml_append(scope, dev);
958     aml_append(table, scope);
959 }
960 
961 static Aml *build_vmbus_device_aml(VMBusBridge *vmbus_bridge)
962 {
963     Aml *dev;
964     Aml *method;
965     Aml *crs;
966 
967     dev = aml_device("VMBS");
968     aml_append(dev, aml_name_decl("STA", aml_int(0xF)));
969     aml_append(dev, aml_name_decl("_HID", aml_string("VMBus")));
970     aml_append(dev, aml_name_decl("_UID", aml_int(0x0)));
971     aml_append(dev, aml_name_decl("_DDN", aml_string("VMBUS")));
972 
973     method = aml_method("_DIS", 0, AML_NOTSERIALIZED);
974     aml_append(method, aml_store(aml_and(aml_name("STA"), aml_int(0xD), NULL),
975                                      aml_name("STA")));
976     aml_append(dev, method);
977 
978     method = aml_method("_PS0", 0, AML_NOTSERIALIZED);
979     aml_append(method, aml_store(aml_or(aml_name("STA"), aml_int(0xF), NULL),
980                                      aml_name("STA")));
981     aml_append(dev, method);
982 
983     method = aml_method("_STA", 0, AML_NOTSERIALIZED);
984     aml_append(method, aml_return(aml_name("STA")));
985     aml_append(dev, method);
986 
987     aml_append(dev, aml_name_decl("_PS3", aml_int(0x0)));
988 
989     crs = aml_resource_template();
990     aml_append(crs, aml_irq_no_flags(vmbus_bridge->irq));
991     aml_append(dev, aml_name_decl("_CRS", crs));
992 
993     return dev;
994 }
995 
996 static void build_isa_devices_aml(Aml *table)
997 {
998     bool ambiguous;
999     Object *obj = object_resolve_path_type("", TYPE_ISA_BUS, &ambiguous);
1000     Aml *scope;
1001 
1002     assert(obj && !ambiguous);
1003 
1004     scope = aml_scope("_SB.PCI0.ISA");
1005     build_acpi_ipmi_devices(scope, BUS(obj), "\\_SB.PCI0.ISA");
1006     isa_build_aml(ISA_BUS(obj), scope);
1007 
1008     aml_append(table, scope);
1009 }
1010 
1011 static void build_dbg_aml(Aml *table)
1012 {
1013     Aml *field;
1014     Aml *method;
1015     Aml *while_ctx;
1016     Aml *scope = aml_scope("\\");
1017     Aml *buf = aml_local(0);
1018     Aml *len = aml_local(1);
1019     Aml *idx = aml_local(2);
1020 
1021     aml_append(scope,
1022        aml_operation_region("DBG", AML_SYSTEM_IO, aml_int(0x0402), 0x01));
1023     field = aml_field("DBG", AML_BYTE_ACC, AML_NOLOCK, AML_PRESERVE);
1024     aml_append(field, aml_named_field("DBGB", 8));
1025     aml_append(scope, field);
1026 
1027     method = aml_method("DBUG", 1, AML_NOTSERIALIZED);
1028 
1029     aml_append(method, aml_to_hexstring(aml_arg(0), buf));
1030     aml_append(method, aml_to_buffer(buf, buf));
1031     aml_append(method, aml_subtract(aml_sizeof(buf), aml_int(1), len));
1032     aml_append(method, aml_store(aml_int(0), idx));
1033 
1034     while_ctx = aml_while(aml_lless(idx, len));
1035     aml_append(while_ctx,
1036         aml_store(aml_derefof(aml_index(buf, idx)), aml_name("DBGB")));
1037     aml_append(while_ctx, aml_increment(idx));
1038     aml_append(method, while_ctx);
1039 
1040     aml_append(method, aml_store(aml_int(0x0A), aml_name("DBGB")));
1041     aml_append(scope, method);
1042 
1043     aml_append(table, scope);
1044 }
1045 
1046 static Aml *build_link_dev(const char *name, uint8_t uid, Aml *reg)
1047 {
1048     Aml *dev;
1049     Aml *crs;
1050     Aml *method;
1051     uint32_t irqs[] = {5, 10, 11};
1052 
1053     dev = aml_device("%s", name);
1054     aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0C0F")));
1055     aml_append(dev, aml_name_decl("_UID", aml_int(uid)));
1056 
1057     crs = aml_resource_template();
1058     aml_append(crs, aml_interrupt(AML_CONSUMER, AML_LEVEL, AML_ACTIVE_HIGH,
1059                                   AML_SHARED, irqs, ARRAY_SIZE(irqs)));
1060     aml_append(dev, aml_name_decl("_PRS", crs));
1061 
1062     method = aml_method("_STA", 0, AML_NOTSERIALIZED);
1063     aml_append(method, aml_return(aml_call1("IQST", reg)));
1064     aml_append(dev, method);
1065 
1066     method = aml_method("_DIS", 0, AML_NOTSERIALIZED);
1067     aml_append(method, aml_or(reg, aml_int(0x80), reg));
1068     aml_append(dev, method);
1069 
1070     method = aml_method("_CRS", 0, AML_NOTSERIALIZED);
1071     aml_append(method, aml_return(aml_call1("IQCR", reg)));
1072     aml_append(dev, method);
1073 
1074     method = aml_method("_SRS", 1, AML_NOTSERIALIZED);
1075     aml_append(method, aml_create_dword_field(aml_arg(0), aml_int(5), "PRRI"));
1076     aml_append(method, aml_store(aml_name("PRRI"), reg));
1077     aml_append(dev, method);
1078 
1079     return dev;
1080  }
1081 
1082 static Aml *build_gsi_link_dev(const char *name, uint8_t uid, uint8_t gsi)
1083 {
1084     Aml *dev;
1085     Aml *crs;
1086     Aml *method;
1087     uint32_t irqs;
1088 
1089     dev = aml_device("%s", name);
1090     aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0C0F")));
1091     aml_append(dev, aml_name_decl("_UID", aml_int(uid)));
1092 
1093     crs = aml_resource_template();
1094     irqs = gsi;
1095     aml_append(crs, aml_interrupt(AML_CONSUMER, AML_LEVEL, AML_ACTIVE_HIGH,
1096                                   AML_SHARED, &irqs, 1));
1097     aml_append(dev, aml_name_decl("_PRS", crs));
1098 
1099     aml_append(dev, aml_name_decl("_CRS", crs));
1100 
1101     /*
1102      * _DIS can be no-op because the interrupt cannot be disabled.
1103      */
1104     method = aml_method("_DIS", 0, AML_NOTSERIALIZED);
1105     aml_append(dev, method);
1106 
1107     method = aml_method("_SRS", 1, AML_NOTSERIALIZED);
1108     aml_append(dev, method);
1109 
1110     return dev;
1111 }
1112 
1113 /* _CRS method - get current settings */
1114 static Aml *build_iqcr_method(bool is_piix4)
1115 {
1116     Aml *if_ctx;
1117     uint32_t irqs;
1118     Aml *method = aml_method("IQCR", 1, AML_SERIALIZED);
1119     Aml *crs = aml_resource_template();
1120 
1121     irqs = 0;
1122     aml_append(crs, aml_interrupt(AML_CONSUMER, AML_LEVEL,
1123                                   AML_ACTIVE_HIGH, AML_SHARED, &irqs, 1));
1124     aml_append(method, aml_name_decl("PRR0", crs));
1125 
1126     aml_append(method,
1127         aml_create_dword_field(aml_name("PRR0"), aml_int(5), "PRRI"));
1128 
1129     if (is_piix4) {
1130         if_ctx = aml_if(aml_lless(aml_arg(0), aml_int(0x80)));
1131         aml_append(if_ctx, aml_store(aml_arg(0), aml_name("PRRI")));
1132         aml_append(method, if_ctx);
1133     } else {
1134         aml_append(method,
1135             aml_store(aml_and(aml_arg(0), aml_int(0xF), NULL),
1136                       aml_name("PRRI")));
1137     }
1138 
1139     aml_append(method, aml_return(aml_name("PRR0")));
1140     return method;
1141 }
1142 
1143 /* _STA method - get status */
1144 static Aml *build_irq_status_method(void)
1145 {
1146     Aml *if_ctx;
1147     Aml *method = aml_method("IQST", 1, AML_NOTSERIALIZED);
1148 
1149     if_ctx = aml_if(aml_and(aml_int(0x80), aml_arg(0), NULL));
1150     aml_append(if_ctx, aml_return(aml_int(0x09)));
1151     aml_append(method, if_ctx);
1152     aml_append(method, aml_return(aml_int(0x0B)));
1153     return method;
1154 }
1155 
1156 static void build_piix4_pci0_int(Aml *table)
1157 {
1158     Aml *dev;
1159     Aml *crs;
1160     Aml *field;
1161     Aml *method;
1162     uint32_t irqs;
1163     Aml *sb_scope = aml_scope("_SB");
1164     Aml *pci0_scope = aml_scope("PCI0");
1165 
1166     aml_append(pci0_scope, build_prt(true));
1167     aml_append(sb_scope, pci0_scope);
1168 
1169     field = aml_field("PCI0.ISA.P40C", AML_BYTE_ACC, AML_NOLOCK, AML_PRESERVE);
1170     aml_append(field, aml_named_field("PRQ0", 8));
1171     aml_append(field, aml_named_field("PRQ1", 8));
1172     aml_append(field, aml_named_field("PRQ2", 8));
1173     aml_append(field, aml_named_field("PRQ3", 8));
1174     aml_append(sb_scope, field);
1175 
1176     aml_append(sb_scope, build_irq_status_method());
1177     aml_append(sb_scope, build_iqcr_method(true));
1178 
1179     aml_append(sb_scope, build_link_dev("LNKA", 0, aml_name("PRQ0")));
1180     aml_append(sb_scope, build_link_dev("LNKB", 1, aml_name("PRQ1")));
1181     aml_append(sb_scope, build_link_dev("LNKC", 2, aml_name("PRQ2")));
1182     aml_append(sb_scope, build_link_dev("LNKD", 3, aml_name("PRQ3")));
1183 
1184     dev = aml_device("LNKS");
1185     {
1186         aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0C0F")));
1187         aml_append(dev, aml_name_decl("_UID", aml_int(4)));
1188 
1189         crs = aml_resource_template();
1190         irqs = 9;
1191         aml_append(crs, aml_interrupt(AML_CONSUMER, AML_LEVEL,
1192                                       AML_ACTIVE_HIGH, AML_SHARED,
1193                                       &irqs, 1));
1194         aml_append(dev, aml_name_decl("_PRS", crs));
1195 
1196         /* The SCI cannot be disabled and is always attached to GSI 9,
1197          * so these are no-ops.  We only need this link to override the
1198          * polarity to active high and match the content of the MADT.
1199          */
1200         method = aml_method("_STA", 0, AML_NOTSERIALIZED);
1201         aml_append(method, aml_return(aml_int(0x0b)));
1202         aml_append(dev, method);
1203 
1204         method = aml_method("_DIS", 0, AML_NOTSERIALIZED);
1205         aml_append(dev, method);
1206 
1207         method = aml_method("_CRS", 0, AML_NOTSERIALIZED);
1208         aml_append(method, aml_return(aml_name("_PRS")));
1209         aml_append(dev, method);
1210 
1211         method = aml_method("_SRS", 1, AML_NOTSERIALIZED);
1212         aml_append(dev, method);
1213     }
1214     aml_append(sb_scope, dev);
1215 
1216     aml_append(table, sb_scope);
1217 }
1218 
1219 static void append_q35_prt_entry(Aml *ctx, uint32_t nr, const char *name)
1220 {
1221     int i;
1222     int head;
1223     Aml *pkg;
1224     char base = name[3] < 'E' ? 'A' : 'E';
1225     char *s = g_strdup(name);
1226     Aml *a_nr = aml_int((nr << 16) | 0xffff);
1227 
1228     assert(strlen(s) == 4);
1229 
1230     head = name[3] - base;
1231     for (i = 0; i < 4; i++) {
1232         if (head + i > 3) {
1233             head = i * -1;
1234         }
1235         s[3] = base + head + i;
1236         pkg = aml_package(4);
1237         aml_append(pkg, a_nr);
1238         aml_append(pkg, aml_int(i));
1239         aml_append(pkg, aml_name("%s", s));
1240         aml_append(pkg, aml_int(0));
1241         aml_append(ctx, pkg);
1242     }
1243     g_free(s);
1244 }
1245 
1246 static Aml *build_q35_routing_table(const char *str)
1247 {
1248     int i;
1249     Aml *pkg;
1250     char *name = g_strdup_printf("%s ", str);
1251 
1252     pkg = aml_package(128);
1253     for (i = 0; i < 0x18; i++) {
1254             name[3] = 'E' + (i & 0x3);
1255             append_q35_prt_entry(pkg, i, name);
1256     }
1257 
1258     name[3] = 'E';
1259     append_q35_prt_entry(pkg, 0x18, name);
1260 
1261     /* INTA -> PIRQA for slot 25 - 31, see the default value of D<N>IR */
1262     for (i = 0x0019; i < 0x1e; i++) {
1263         name[3] = 'A';
1264         append_q35_prt_entry(pkg, i, name);
1265     }
1266 
1267     /* PCIe->PCI bridge. use PIRQ[E-H] */
1268     name[3] = 'E';
1269     append_q35_prt_entry(pkg, 0x1e, name);
1270     name[3] = 'A';
1271     append_q35_prt_entry(pkg, 0x1f, name);
1272 
1273     g_free(name);
1274     return pkg;
1275 }
1276 
1277 static void build_q35_pci0_int(Aml *table)
1278 {
1279     Aml *field;
1280     Aml *method;
1281     Aml *sb_scope = aml_scope("_SB");
1282     Aml *pci0_scope = aml_scope("PCI0");
1283 
1284     /* Zero => PIC mode, One => APIC Mode */
1285     aml_append(table, aml_name_decl("PICF", aml_int(0)));
1286     method = aml_method("_PIC", 1, AML_NOTSERIALIZED);
1287     {
1288         aml_append(method, aml_store(aml_arg(0), aml_name("PICF")));
1289     }
1290     aml_append(table, method);
1291 
1292     aml_append(pci0_scope,
1293         aml_name_decl("PRTP", build_q35_routing_table("LNK")));
1294     aml_append(pci0_scope,
1295         aml_name_decl("PRTA", build_q35_routing_table("GSI")));
1296 
1297     method = aml_method("_PRT", 0, AML_NOTSERIALIZED);
1298     {
1299         Aml *if_ctx;
1300         Aml *else_ctx;
1301 
1302         /* PCI IRQ routing table, example from ACPI 2.0a specification,
1303            section 6.2.8.1 */
1304         /* Note: we provide the same info as the PCI routing
1305            table of the Bochs BIOS */
1306         if_ctx = aml_if(aml_equal(aml_name("PICF"), aml_int(0)));
1307         aml_append(if_ctx, aml_return(aml_name("PRTP")));
1308         aml_append(method, if_ctx);
1309         else_ctx = aml_else();
1310         aml_append(else_ctx, aml_return(aml_name("PRTA")));
1311         aml_append(method, else_ctx);
1312     }
1313     aml_append(pci0_scope, method);
1314     aml_append(sb_scope, pci0_scope);
1315 
1316     field = aml_field("PCI0.ISA.PIRQ", AML_BYTE_ACC, AML_NOLOCK, AML_PRESERVE);
1317     aml_append(field, aml_named_field("PRQA", 8));
1318     aml_append(field, aml_named_field("PRQB", 8));
1319     aml_append(field, aml_named_field("PRQC", 8));
1320     aml_append(field, aml_named_field("PRQD", 8));
1321     aml_append(field, aml_reserved_field(0x20));
1322     aml_append(field, aml_named_field("PRQE", 8));
1323     aml_append(field, aml_named_field("PRQF", 8));
1324     aml_append(field, aml_named_field("PRQG", 8));
1325     aml_append(field, aml_named_field("PRQH", 8));
1326     aml_append(sb_scope, field);
1327 
1328     aml_append(sb_scope, build_irq_status_method());
1329     aml_append(sb_scope, build_iqcr_method(false));
1330 
1331     aml_append(sb_scope, build_link_dev("LNKA", 0, aml_name("PRQA")));
1332     aml_append(sb_scope, build_link_dev("LNKB", 1, aml_name("PRQB")));
1333     aml_append(sb_scope, build_link_dev("LNKC", 2, aml_name("PRQC")));
1334     aml_append(sb_scope, build_link_dev("LNKD", 3, aml_name("PRQD")));
1335     aml_append(sb_scope, build_link_dev("LNKE", 4, aml_name("PRQE")));
1336     aml_append(sb_scope, build_link_dev("LNKF", 5, aml_name("PRQF")));
1337     aml_append(sb_scope, build_link_dev("LNKG", 6, aml_name("PRQG")));
1338     aml_append(sb_scope, build_link_dev("LNKH", 7, aml_name("PRQH")));
1339 
1340     aml_append(sb_scope, build_gsi_link_dev("GSIA", 0x10, 0x10));
1341     aml_append(sb_scope, build_gsi_link_dev("GSIB", 0x11, 0x11));
1342     aml_append(sb_scope, build_gsi_link_dev("GSIC", 0x12, 0x12));
1343     aml_append(sb_scope, build_gsi_link_dev("GSID", 0x13, 0x13));
1344     aml_append(sb_scope, build_gsi_link_dev("GSIE", 0x14, 0x14));
1345     aml_append(sb_scope, build_gsi_link_dev("GSIF", 0x15, 0x15));
1346     aml_append(sb_scope, build_gsi_link_dev("GSIG", 0x16, 0x16));
1347     aml_append(sb_scope, build_gsi_link_dev("GSIH", 0x17, 0x17));
1348 
1349     aml_append(table, sb_scope);
1350 }
1351 
1352 static void build_q35_isa_bridge(Aml *table)
1353 {
1354     Aml *dev;
1355     Aml *scope;
1356 
1357     scope =  aml_scope("_SB.PCI0");
1358     dev = aml_device("ISA");
1359     aml_append(dev, aml_name_decl("_ADR", aml_int(0x001F0000)));
1360 
1361     /* ICH9 PCI to ISA irq remapping */
1362     aml_append(dev, aml_operation_region("PIRQ", AML_PCI_CONFIG,
1363                                          aml_int(0x60), 0x0C));
1364 
1365     aml_append(scope, dev);
1366     aml_append(table, scope);
1367 }
1368 
1369 static void build_piix4_isa_bridge(Aml *table)
1370 {
1371     Aml *dev;
1372     Aml *scope;
1373 
1374     scope =  aml_scope("_SB.PCI0");
1375     dev = aml_device("ISA");
1376     aml_append(dev, aml_name_decl("_ADR", aml_int(0x00010000)));
1377 
1378     /* PIIX PCI to ISA irq remapping */
1379     aml_append(dev, aml_operation_region("P40C", AML_PCI_CONFIG,
1380                                          aml_int(0x60), 0x04));
1381 
1382     aml_append(scope, dev);
1383     aml_append(table, scope);
1384 }
1385 
1386 static void build_piix4_pci_hotplug(Aml *table)
1387 {
1388     Aml *scope;
1389     Aml *field;
1390     Aml *method;
1391 
1392     scope =  aml_scope("_SB.PCI0");
1393 
1394     aml_append(scope,
1395         aml_operation_region("PCST", AML_SYSTEM_IO, aml_int(0xae00), 0x08));
1396     field = aml_field("PCST", AML_DWORD_ACC, AML_NOLOCK, AML_WRITE_AS_ZEROS);
1397     aml_append(field, aml_named_field("PCIU", 32));
1398     aml_append(field, aml_named_field("PCID", 32));
1399     aml_append(scope, field);
1400 
1401     aml_append(scope,
1402         aml_operation_region("SEJ", AML_SYSTEM_IO, aml_int(0xae08), 0x04));
1403     field = aml_field("SEJ", AML_DWORD_ACC, AML_NOLOCK, AML_WRITE_AS_ZEROS);
1404     aml_append(field, aml_named_field("B0EJ", 32));
1405     aml_append(scope, field);
1406 
1407     aml_append(scope,
1408         aml_operation_region("BNMR", AML_SYSTEM_IO, aml_int(0xae10), 0x04));
1409     field = aml_field("BNMR", AML_DWORD_ACC, AML_NOLOCK, AML_WRITE_AS_ZEROS);
1410     aml_append(field, aml_named_field("BNUM", 32));
1411     aml_append(scope, field);
1412 
1413     aml_append(scope, aml_mutex("BLCK", 0));
1414 
1415     method = aml_method("PCEJ", 2, AML_NOTSERIALIZED);
1416     aml_append(method, aml_acquire(aml_name("BLCK"), 0xFFFF));
1417     aml_append(method, aml_store(aml_arg(0), aml_name("BNUM")));
1418     aml_append(method,
1419         aml_store(aml_shiftleft(aml_int(1), aml_arg(1)), aml_name("B0EJ")));
1420     aml_append(method, aml_release(aml_name("BLCK")));
1421     aml_append(method, aml_return(aml_int(0)));
1422     aml_append(scope, method);
1423 
1424     aml_append(table, scope);
1425 }
1426 
1427 static Aml *build_q35_osc_method(void)
1428 {
1429     Aml *if_ctx;
1430     Aml *if_ctx2;
1431     Aml *else_ctx;
1432     Aml *method;
1433     Aml *a_cwd1 = aml_name("CDW1");
1434     Aml *a_ctrl = aml_local(0);
1435 
1436     method = aml_method("_OSC", 4, AML_NOTSERIALIZED);
1437     aml_append(method, aml_create_dword_field(aml_arg(3), aml_int(0), "CDW1"));
1438 
1439     if_ctx = aml_if(aml_equal(
1440         aml_arg(0), aml_touuid("33DB4D5B-1FF7-401C-9657-7441C03DD766")));
1441     aml_append(if_ctx, aml_create_dword_field(aml_arg(3), aml_int(4), "CDW2"));
1442     aml_append(if_ctx, aml_create_dword_field(aml_arg(3), aml_int(8), "CDW3"));
1443 
1444     aml_append(if_ctx, aml_store(aml_name("CDW3"), a_ctrl));
1445 
1446     /*
1447      * Always allow native PME, AER (no dependencies)
1448      * Allow SHPC (PCI bridges can have SHPC controller)
1449      */
1450     aml_append(if_ctx, aml_and(a_ctrl, aml_int(0x1F), a_ctrl));
1451 
1452     if_ctx2 = aml_if(aml_lnot(aml_equal(aml_arg(1), aml_int(1))));
1453     /* Unknown revision */
1454     aml_append(if_ctx2, aml_or(a_cwd1, aml_int(0x08), a_cwd1));
1455     aml_append(if_ctx, if_ctx2);
1456 
1457     if_ctx2 = aml_if(aml_lnot(aml_equal(aml_name("CDW3"), a_ctrl)));
1458     /* Capabilities bits were masked */
1459     aml_append(if_ctx2, aml_or(a_cwd1, aml_int(0x10), a_cwd1));
1460     aml_append(if_ctx, if_ctx2);
1461 
1462     /* Update DWORD3 in the buffer */
1463     aml_append(if_ctx, aml_store(a_ctrl, aml_name("CDW3")));
1464     aml_append(method, if_ctx);
1465 
1466     else_ctx = aml_else();
1467     /* Unrecognized UUID */
1468     aml_append(else_ctx, aml_or(a_cwd1, aml_int(4), a_cwd1));
1469     aml_append(method, else_ctx);
1470 
1471     aml_append(method, aml_return(aml_arg(3)));
1472     return method;
1473 }
1474 
1475 static void build_smb0(Aml *table, I2CBus *smbus, int devnr, int func)
1476 {
1477     Aml *scope = aml_scope("_SB.PCI0");
1478     Aml *dev = aml_device("SMB0");
1479 
1480     aml_append(dev, aml_name_decl("_ADR", aml_int(devnr << 16 | func)));
1481     build_acpi_ipmi_devices(dev, BUS(smbus), "\\_SB.PCI0.SMB0");
1482     aml_append(scope, dev);
1483     aml_append(table, scope);
1484 }
1485 
1486 static void
1487 build_dsdt(GArray *table_data, BIOSLinker *linker,
1488            AcpiPmInfo *pm, AcpiMiscInfo *misc,
1489            Range *pci_hole, Range *pci_hole64, MachineState *machine)
1490 {
1491     CrsRangeEntry *entry;
1492     Aml *dsdt, *sb_scope, *scope, *dev, *method, *field, *pkg, *crs;
1493     CrsRangeSet crs_range_set;
1494     PCMachineState *pcms = PC_MACHINE(machine);
1495     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(machine);
1496     X86MachineState *x86ms = X86_MACHINE(machine);
1497     AcpiMcfgInfo mcfg;
1498     uint32_t nr_mem = machine->ram_slots;
1499     int root_bus_limit = 0xFF;
1500     PCIBus *bus = NULL;
1501     TPMIf *tpm = tpm_find();
1502     int i;
1503     VMBusBridge *vmbus_bridge = vmbus_bridge_find();
1504 
1505     dsdt = init_aml_allocator();
1506 
1507     /* Reserve space for header */
1508     acpi_data_push(dsdt->buf, sizeof(AcpiTableHeader));
1509 
1510     build_dbg_aml(dsdt);
1511     if (misc->is_piix4) {
1512         sb_scope = aml_scope("_SB");
1513         dev = aml_device("PCI0");
1514         aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0A03")));
1515         aml_append(dev, aml_name_decl("_ADR", aml_int(0)));
1516         aml_append(dev, aml_name_decl("_UID", aml_int(0)));
1517         aml_append(sb_scope, dev);
1518         aml_append(dsdt, sb_scope);
1519 
1520         build_hpet_aml(dsdt);
1521         build_piix4_isa_bridge(dsdt);
1522         build_isa_devices_aml(dsdt);
1523         if (pm->pcihp_bridge_en || pm->pcihp_root_en) {
1524             build_piix4_pci_hotplug(dsdt);
1525         }
1526         build_piix4_pci0_int(dsdt);
1527     } else {
1528         sb_scope = aml_scope("_SB");
1529         dev = aml_device("PCI0");
1530         aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0A08")));
1531         aml_append(dev, aml_name_decl("_CID", aml_eisaid("PNP0A03")));
1532         aml_append(dev, aml_name_decl("_ADR", aml_int(0)));
1533         aml_append(dev, aml_name_decl("_UID", aml_int(0)));
1534         aml_append(dev, build_q35_osc_method());
1535         aml_append(sb_scope, dev);
1536 
1537         if (pm->smi_on_cpuhp) {
1538             /* reserve SMI block resources, IO ports 0xB2, 0xB3 */
1539             dev = aml_device("PCI0.SMI0");
1540             aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0A06")));
1541             aml_append(dev, aml_name_decl("_UID", aml_string("SMI resources")));
1542             crs = aml_resource_template();
1543             aml_append(crs,
1544                 aml_io(
1545                        AML_DECODE16,
1546                        ACPI_PORT_SMI_CMD,
1547                        ACPI_PORT_SMI_CMD,
1548                        1,
1549                        2)
1550             );
1551             aml_append(dev, aml_name_decl("_CRS", crs));
1552             aml_append(dev, aml_operation_region("SMIR", AML_SYSTEM_IO,
1553                 aml_int(ACPI_PORT_SMI_CMD), 2));
1554             field = aml_field("SMIR", AML_BYTE_ACC, AML_NOLOCK,
1555                               AML_WRITE_AS_ZEROS);
1556             aml_append(field, aml_named_field("SMIC", 8));
1557             aml_append(field, aml_reserved_field(8));
1558             aml_append(dev, field);
1559             aml_append(sb_scope, dev);
1560         }
1561 
1562         aml_append(dsdt, sb_scope);
1563 
1564         build_hpet_aml(dsdt);
1565         build_q35_isa_bridge(dsdt);
1566         build_isa_devices_aml(dsdt);
1567         build_q35_pci0_int(dsdt);
1568         if (pcms->smbus && !pcmc->do_not_add_smb_acpi) {
1569             build_smb0(dsdt, pcms->smbus, ICH9_SMB_DEV, ICH9_SMB_FUNC);
1570         }
1571     }
1572 
1573     if (vmbus_bridge) {
1574         sb_scope = aml_scope("_SB");
1575         aml_append(sb_scope, build_vmbus_device_aml(vmbus_bridge));
1576         aml_append(dsdt, sb_scope);
1577     }
1578 
1579     if (pcmc->legacy_cpu_hotplug) {
1580         build_legacy_cpu_hotplug_aml(dsdt, machine, pm->cpu_hp_io_base);
1581     } else {
1582         CPUHotplugFeatures opts = {
1583             .acpi_1_compatible = true, .has_legacy_cphp = true,
1584             .smi_path = pm->smi_on_cpuhp ? "\\_SB.PCI0.SMI0.SMIC" : NULL,
1585         };
1586         build_cpus_aml(dsdt, machine, opts, pm->cpu_hp_io_base,
1587                        "\\_SB.PCI0", "\\_GPE._E02");
1588     }
1589 
1590     if (pcms->memhp_io_base && nr_mem) {
1591         build_memory_hotplug_aml(dsdt, nr_mem, "\\_SB.PCI0",
1592                                  "\\_GPE._E03", AML_SYSTEM_IO,
1593                                  pcms->memhp_io_base);
1594     }
1595 
1596     scope =  aml_scope("_GPE");
1597     {
1598         aml_append(scope, aml_name_decl("_HID", aml_string("ACPI0006")));
1599 
1600         if (misc->is_piix4 && (pm->pcihp_bridge_en || pm->pcihp_root_en)) {
1601             method = aml_method("_E01", 0, AML_NOTSERIALIZED);
1602             aml_append(method,
1603                 aml_acquire(aml_name("\\_SB.PCI0.BLCK"), 0xFFFF));
1604             aml_append(method, aml_call0("\\_SB.PCI0.PCNT"));
1605             aml_append(method, aml_release(aml_name("\\_SB.PCI0.BLCK")));
1606             aml_append(scope, method);
1607         }
1608 
1609         if (machine->nvdimms_state->is_enabled) {
1610             method = aml_method("_E04", 0, AML_NOTSERIALIZED);
1611             aml_append(method, aml_notify(aml_name("\\_SB.NVDR"),
1612                                           aml_int(0x80)));
1613             aml_append(scope, method);
1614         }
1615     }
1616     aml_append(dsdt, scope);
1617 
1618     crs_range_set_init(&crs_range_set);
1619     bus = PC_MACHINE(machine)->bus;
1620     if (bus) {
1621         QLIST_FOREACH(bus, &bus->child, sibling) {
1622             uint8_t bus_num = pci_bus_num(bus);
1623             uint8_t numa_node = pci_bus_numa_node(bus);
1624 
1625             /* look only for expander root buses */
1626             if (!pci_bus_is_root(bus)) {
1627                 continue;
1628             }
1629 
1630             if (bus_num < root_bus_limit) {
1631                 root_bus_limit = bus_num - 1;
1632             }
1633 
1634             scope = aml_scope("\\_SB");
1635             dev = aml_device("PC%.02X", bus_num);
1636             aml_append(dev, aml_name_decl("_UID", aml_int(bus_num)));
1637             aml_append(dev, aml_name_decl("_BBN", aml_int(bus_num)));
1638             if (pci_bus_is_express(bus)) {
1639                 aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0A08")));
1640                 aml_append(dev, aml_name_decl("_CID", aml_eisaid("PNP0A03")));
1641                 aml_append(dev, build_q35_osc_method());
1642             } else {
1643                 aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0A03")));
1644             }
1645 
1646             if (numa_node != NUMA_NODE_UNASSIGNED) {
1647                 aml_append(dev, aml_name_decl("_PXM", aml_int(numa_node)));
1648             }
1649 
1650             aml_append(dev, build_prt(false));
1651             crs = build_crs(PCI_HOST_BRIDGE(BUS(bus)->parent), &crs_range_set);
1652             aml_append(dev, aml_name_decl("_CRS", crs));
1653             aml_append(scope, dev);
1654             aml_append(dsdt, scope);
1655         }
1656     }
1657 
1658     /*
1659      * At this point crs_range_set has all the ranges used by pci
1660      * busses *other* than PCI0.  These ranges will be excluded from
1661      * the PCI0._CRS.  Add mmconfig to the set so it will be excluded
1662      * too.
1663      */
1664     if (acpi_get_mcfg(&mcfg)) {
1665         crs_range_insert(crs_range_set.mem_ranges,
1666                          mcfg.base, mcfg.base + mcfg.size - 1);
1667     }
1668 
1669     scope = aml_scope("\\_SB.PCI0");
1670     /* build PCI0._CRS */
1671     crs = aml_resource_template();
1672     aml_append(crs,
1673         aml_word_bus_number(AML_MIN_FIXED, AML_MAX_FIXED, AML_POS_DECODE,
1674                             0x0000, 0x0, root_bus_limit,
1675                             0x0000, root_bus_limit + 1));
1676     aml_append(crs, aml_io(AML_DECODE16, 0x0CF8, 0x0CF8, 0x01, 0x08));
1677 
1678     aml_append(crs,
1679         aml_word_io(AML_MIN_FIXED, AML_MAX_FIXED,
1680                     AML_POS_DECODE, AML_ENTIRE_RANGE,
1681                     0x0000, 0x0000, 0x0CF7, 0x0000, 0x0CF8));
1682 
1683     crs_replace_with_free_ranges(crs_range_set.io_ranges, 0x0D00, 0xFFFF);
1684     for (i = 0; i < crs_range_set.io_ranges->len; i++) {
1685         entry = g_ptr_array_index(crs_range_set.io_ranges, i);
1686         aml_append(crs,
1687             aml_word_io(AML_MIN_FIXED, AML_MAX_FIXED,
1688                         AML_POS_DECODE, AML_ENTIRE_RANGE,
1689                         0x0000, entry->base, entry->limit,
1690                         0x0000, entry->limit - entry->base + 1));
1691     }
1692 
1693     aml_append(crs,
1694         aml_dword_memory(AML_POS_DECODE, AML_MIN_FIXED, AML_MAX_FIXED,
1695                          AML_CACHEABLE, AML_READ_WRITE,
1696                          0, 0x000A0000, 0x000BFFFF, 0, 0x00020000));
1697 
1698     crs_replace_with_free_ranges(crs_range_set.mem_ranges,
1699                                  range_lob(pci_hole),
1700                                  range_upb(pci_hole));
1701     for (i = 0; i < crs_range_set.mem_ranges->len; i++) {
1702         entry = g_ptr_array_index(crs_range_set.mem_ranges, i);
1703         aml_append(crs,
1704             aml_dword_memory(AML_POS_DECODE, AML_MIN_FIXED, AML_MAX_FIXED,
1705                              AML_NON_CACHEABLE, AML_READ_WRITE,
1706                              0, entry->base, entry->limit,
1707                              0, entry->limit - entry->base + 1));
1708     }
1709 
1710     if (!range_is_empty(pci_hole64)) {
1711         crs_replace_with_free_ranges(crs_range_set.mem_64bit_ranges,
1712                                      range_lob(pci_hole64),
1713                                      range_upb(pci_hole64));
1714         for (i = 0; i < crs_range_set.mem_64bit_ranges->len; i++) {
1715             entry = g_ptr_array_index(crs_range_set.mem_64bit_ranges, i);
1716             aml_append(crs,
1717                        aml_qword_memory(AML_POS_DECODE, AML_MIN_FIXED,
1718                                         AML_MAX_FIXED,
1719                                         AML_CACHEABLE, AML_READ_WRITE,
1720                                         0, entry->base, entry->limit,
1721                                         0, entry->limit - entry->base + 1));
1722         }
1723     }
1724 
1725     if (TPM_IS_TIS_ISA(tpm_find())) {
1726         aml_append(crs, aml_memory32_fixed(TPM_TIS_ADDR_BASE,
1727                    TPM_TIS_ADDR_SIZE, AML_READ_WRITE));
1728     }
1729     aml_append(scope, aml_name_decl("_CRS", crs));
1730 
1731     /* reserve GPE0 block resources */
1732     dev = aml_device("GPE0");
1733     aml_append(dev, aml_name_decl("_HID", aml_string("PNP0A06")));
1734     aml_append(dev, aml_name_decl("_UID", aml_string("GPE0 resources")));
1735     /* device present, functioning, decoding, not shown in UI */
1736     aml_append(dev, aml_name_decl("_STA", aml_int(0xB)));
1737     crs = aml_resource_template();
1738     aml_append(crs,
1739         aml_io(
1740                AML_DECODE16,
1741                pm->fadt.gpe0_blk.address,
1742                pm->fadt.gpe0_blk.address,
1743                1,
1744                pm->fadt.gpe0_blk.bit_width / 8)
1745     );
1746     aml_append(dev, aml_name_decl("_CRS", crs));
1747     aml_append(scope, dev);
1748 
1749     crs_range_set_free(&crs_range_set);
1750 
1751     /* reserve PCIHP resources */
1752     if (pm->pcihp_io_len && (pm->pcihp_bridge_en || pm->pcihp_root_en)) {
1753         dev = aml_device("PHPR");
1754         aml_append(dev, aml_name_decl("_HID", aml_string("PNP0A06")));
1755         aml_append(dev,
1756             aml_name_decl("_UID", aml_string("PCI Hotplug resources")));
1757         /* device present, functioning, decoding, not shown in UI */
1758         aml_append(dev, aml_name_decl("_STA", aml_int(0xB)));
1759         crs = aml_resource_template();
1760         aml_append(crs,
1761             aml_io(AML_DECODE16, pm->pcihp_io_base, pm->pcihp_io_base, 1,
1762                    pm->pcihp_io_len)
1763         );
1764         aml_append(dev, aml_name_decl("_CRS", crs));
1765         aml_append(scope, dev);
1766     }
1767     aml_append(dsdt, scope);
1768 
1769     /*  create S3_ / S4_ / S5_ packages if necessary */
1770     scope = aml_scope("\\");
1771     if (!pm->s3_disabled) {
1772         pkg = aml_package(4);
1773         aml_append(pkg, aml_int(1)); /* PM1a_CNT.SLP_TYP */
1774         aml_append(pkg, aml_int(1)); /* PM1b_CNT.SLP_TYP, FIXME: not impl. */
1775         aml_append(pkg, aml_int(0)); /* reserved */
1776         aml_append(pkg, aml_int(0)); /* reserved */
1777         aml_append(scope, aml_name_decl("_S3", pkg));
1778     }
1779 
1780     if (!pm->s4_disabled) {
1781         pkg = aml_package(4);
1782         aml_append(pkg, aml_int(pm->s4_val)); /* PM1a_CNT.SLP_TYP */
1783         /* PM1b_CNT.SLP_TYP, FIXME: not impl. */
1784         aml_append(pkg, aml_int(pm->s4_val));
1785         aml_append(pkg, aml_int(0)); /* reserved */
1786         aml_append(pkg, aml_int(0)); /* reserved */
1787         aml_append(scope, aml_name_decl("_S4", pkg));
1788     }
1789 
1790     pkg = aml_package(4);
1791     aml_append(pkg, aml_int(0)); /* PM1a_CNT.SLP_TYP */
1792     aml_append(pkg, aml_int(0)); /* PM1b_CNT.SLP_TYP not impl. */
1793     aml_append(pkg, aml_int(0)); /* reserved */
1794     aml_append(pkg, aml_int(0)); /* reserved */
1795     aml_append(scope, aml_name_decl("_S5", pkg));
1796     aml_append(dsdt, scope);
1797 
1798     /* create fw_cfg node, unconditionally */
1799     {
1800         scope = aml_scope("\\_SB.PCI0");
1801         fw_cfg_add_acpi_dsdt(scope, x86ms->fw_cfg);
1802         aml_append(dsdt, scope);
1803     }
1804 
1805     if (misc->applesmc_io_base) {
1806         scope = aml_scope("\\_SB.PCI0.ISA");
1807         dev = aml_device("SMC");
1808 
1809         aml_append(dev, aml_name_decl("_HID", aml_eisaid("APP0001")));
1810         /* device present, functioning, decoding, not shown in UI */
1811         aml_append(dev, aml_name_decl("_STA", aml_int(0xB)));
1812 
1813         crs = aml_resource_template();
1814         aml_append(crs,
1815             aml_io(AML_DECODE16, misc->applesmc_io_base, misc->applesmc_io_base,
1816                    0x01, APPLESMC_MAX_DATA_LENGTH)
1817         );
1818         aml_append(crs, aml_irq_no_flags(6));
1819         aml_append(dev, aml_name_decl("_CRS", crs));
1820 
1821         aml_append(scope, dev);
1822         aml_append(dsdt, scope);
1823     }
1824 
1825     if (misc->pvpanic_port) {
1826         scope = aml_scope("\\_SB.PCI0.ISA");
1827 
1828         dev = aml_device("PEVT");
1829         aml_append(dev, aml_name_decl("_HID", aml_string("QEMU0001")));
1830 
1831         crs = aml_resource_template();
1832         aml_append(crs,
1833             aml_io(AML_DECODE16, misc->pvpanic_port, misc->pvpanic_port, 1, 1)
1834         );
1835         aml_append(dev, aml_name_decl("_CRS", crs));
1836 
1837         aml_append(dev, aml_operation_region("PEOR", AML_SYSTEM_IO,
1838                                               aml_int(misc->pvpanic_port), 1));
1839         field = aml_field("PEOR", AML_BYTE_ACC, AML_NOLOCK, AML_PRESERVE);
1840         aml_append(field, aml_named_field("PEPT", 8));
1841         aml_append(dev, field);
1842 
1843         /* device present, functioning, decoding, shown in UI */
1844         aml_append(dev, aml_name_decl("_STA", aml_int(0xF)));
1845 
1846         method = aml_method("RDPT", 0, AML_NOTSERIALIZED);
1847         aml_append(method, aml_store(aml_name("PEPT"), aml_local(0)));
1848         aml_append(method, aml_return(aml_local(0)));
1849         aml_append(dev, method);
1850 
1851         method = aml_method("WRPT", 1, AML_NOTSERIALIZED);
1852         aml_append(method, aml_store(aml_arg(0), aml_name("PEPT")));
1853         aml_append(dev, method);
1854 
1855         aml_append(scope, dev);
1856         aml_append(dsdt, scope);
1857     }
1858 
1859     sb_scope = aml_scope("\\_SB");
1860     {
1861         Object *pci_host;
1862         PCIBus *bus = NULL;
1863 
1864         pci_host = acpi_get_i386_pci_host();
1865         if (pci_host) {
1866             bus = PCI_HOST_BRIDGE(pci_host)->bus;
1867         }
1868 
1869         if (bus) {
1870             Aml *scope = aml_scope("PCI0");
1871             /* Scan all PCI buses. Generate tables to support hotplug. */
1872             build_append_pci_bus_devices(scope, bus, pm->pcihp_bridge_en);
1873 
1874             if (TPM_IS_TIS_ISA(tpm)) {
1875                 if (misc->tpm_version == TPM_VERSION_2_0) {
1876                     dev = aml_device("TPM");
1877                     aml_append(dev, aml_name_decl("_HID",
1878                                                   aml_string("MSFT0101")));
1879                 } else {
1880                     dev = aml_device("ISA.TPM");
1881                     aml_append(dev, aml_name_decl("_HID",
1882                                                   aml_eisaid("PNP0C31")));
1883                 }
1884 
1885                 aml_append(dev, aml_name_decl("_STA", aml_int(0xF)));
1886                 crs = aml_resource_template();
1887                 aml_append(crs, aml_memory32_fixed(TPM_TIS_ADDR_BASE,
1888                            TPM_TIS_ADDR_SIZE, AML_READ_WRITE));
1889                 /*
1890                     FIXME: TPM_TIS_IRQ=5 conflicts with PNP0C0F irqs,
1891                     Rewrite to take IRQ from TPM device model and
1892                     fix default IRQ value there to use some unused IRQ
1893                  */
1894                 /* aml_append(crs, aml_irq_no_flags(TPM_TIS_IRQ)); */
1895                 aml_append(dev, aml_name_decl("_CRS", crs));
1896 
1897                 tpm_build_ppi_acpi(tpm, dev);
1898 
1899                 aml_append(scope, dev);
1900             }
1901 
1902             aml_append(sb_scope, scope);
1903         }
1904     }
1905 
1906     if (TPM_IS_CRB(tpm)) {
1907         dev = aml_device("TPM");
1908         aml_append(dev, aml_name_decl("_HID", aml_string("MSFT0101")));
1909         crs = aml_resource_template();
1910         aml_append(crs, aml_memory32_fixed(TPM_CRB_ADDR_BASE,
1911                                            TPM_CRB_ADDR_SIZE, AML_READ_WRITE));
1912         aml_append(dev, aml_name_decl("_CRS", crs));
1913 
1914         aml_append(dev, aml_name_decl("_STA", aml_int(0xf)));
1915 
1916         tpm_build_ppi_acpi(tpm, dev);
1917 
1918         aml_append(sb_scope, dev);
1919     }
1920 
1921     aml_append(dsdt, sb_scope);
1922 
1923     /* copy AML table into ACPI tables blob and patch header there */
1924     g_array_append_vals(table_data, dsdt->buf->data, dsdt->buf->len);
1925     build_header(linker, table_data,
1926         (void *)(table_data->data + table_data->len - dsdt->buf->len),
1927         "DSDT", dsdt->buf->len, 1, NULL, NULL);
1928     free_aml_allocator();
1929 }
1930 
1931 static void
1932 build_hpet(GArray *table_data, BIOSLinker *linker)
1933 {
1934     Acpi20Hpet *hpet;
1935 
1936     hpet = acpi_data_push(table_data, sizeof(*hpet));
1937     /* Note timer_block_id value must be kept in sync with value advertised by
1938      * emulated hpet
1939      */
1940     hpet->timer_block_id = cpu_to_le32(0x8086a201);
1941     hpet->addr.address = cpu_to_le64(HPET_BASE);
1942     build_header(linker, table_data,
1943                  (void *)hpet, "HPET", sizeof(*hpet), 1, NULL, NULL);
1944 }
1945 
1946 static void
1947 build_tpm_tcpa(GArray *table_data, BIOSLinker *linker, GArray *tcpalog)
1948 {
1949     Acpi20Tcpa *tcpa = acpi_data_push(table_data, sizeof *tcpa);
1950     unsigned log_addr_size = sizeof(tcpa->log_area_start_address);
1951     unsigned log_addr_offset =
1952         (char *)&tcpa->log_area_start_address - table_data->data;
1953 
1954     tcpa->platform_class = cpu_to_le16(TPM_TCPA_ACPI_CLASS_CLIENT);
1955     tcpa->log_area_minimum_length = cpu_to_le32(TPM_LOG_AREA_MINIMUM_SIZE);
1956     acpi_data_push(tcpalog, le32_to_cpu(tcpa->log_area_minimum_length));
1957 
1958     bios_linker_loader_alloc(linker, ACPI_BUILD_TPMLOG_FILE, tcpalog, 1,
1959                              false /* high memory */);
1960 
1961     /* log area start address to be filled by Guest linker */
1962     bios_linker_loader_add_pointer(linker,
1963         ACPI_BUILD_TABLE_FILE, log_addr_offset, log_addr_size,
1964         ACPI_BUILD_TPMLOG_FILE, 0);
1965 
1966     build_header(linker, table_data,
1967                  (void *)tcpa, "TCPA", sizeof(*tcpa), 2, NULL, NULL);
1968 }
1969 
1970 #define HOLE_640K_START  (640 * KiB)
1971 #define HOLE_640K_END   (1 * MiB)
1972 
1973 static void
1974 build_srat(GArray *table_data, BIOSLinker *linker, MachineState *machine)
1975 {
1976     AcpiSystemResourceAffinityTable *srat;
1977     AcpiSratMemoryAffinity *numamem;
1978 
1979     int i;
1980     int srat_start, numa_start, slots;
1981     uint64_t mem_len, mem_base, next_base;
1982     MachineClass *mc = MACHINE_GET_CLASS(machine);
1983     X86MachineState *x86ms = X86_MACHINE(machine);
1984     const CPUArchIdList *apic_ids = mc->possible_cpu_arch_ids(machine);
1985     PCMachineState *pcms = PC_MACHINE(machine);
1986     ram_addr_t hotplugabble_address_space_size =
1987         object_property_get_int(OBJECT(pcms), PC_MACHINE_DEVMEM_REGION_SIZE,
1988                                 NULL);
1989 
1990     srat_start = table_data->len;
1991 
1992     srat = acpi_data_push(table_data, sizeof *srat);
1993     srat->reserved1 = cpu_to_le32(1);
1994 
1995     for (i = 0; i < apic_ids->len; i++) {
1996         int node_id = apic_ids->cpus[i].props.node_id;
1997         uint32_t apic_id = apic_ids->cpus[i].arch_id;
1998 
1999         if (apic_id < 255) {
2000             AcpiSratProcessorAffinity *core;
2001 
2002             core = acpi_data_push(table_data, sizeof *core);
2003             core->type = ACPI_SRAT_PROCESSOR_APIC;
2004             core->length = sizeof(*core);
2005             core->local_apic_id = apic_id;
2006             core->proximity_lo = node_id;
2007             memset(core->proximity_hi, 0, 3);
2008             core->local_sapic_eid = 0;
2009             core->flags = cpu_to_le32(1);
2010         } else {
2011             AcpiSratProcessorX2ApicAffinity *core;
2012 
2013             core = acpi_data_push(table_data, sizeof *core);
2014             core->type = ACPI_SRAT_PROCESSOR_x2APIC;
2015             core->length = sizeof(*core);
2016             core->x2apic_id = cpu_to_le32(apic_id);
2017             core->proximity_domain = cpu_to_le32(node_id);
2018             core->flags = cpu_to_le32(1);
2019         }
2020     }
2021 
2022 
2023     /* the memory map is a bit tricky, it contains at least one hole
2024      * from 640k-1M and possibly another one from 3.5G-4G.
2025      */
2026     next_base = 0;
2027     numa_start = table_data->len;
2028 
2029     for (i = 1; i < pcms->numa_nodes + 1; ++i) {
2030         mem_base = next_base;
2031         mem_len = pcms->node_mem[i - 1];
2032         next_base = mem_base + mem_len;
2033 
2034         /* Cut out the 640K hole */
2035         if (mem_base <= HOLE_640K_START &&
2036             next_base > HOLE_640K_START) {
2037             mem_len -= next_base - HOLE_640K_START;
2038             if (mem_len > 0) {
2039                 numamem = acpi_data_push(table_data, sizeof *numamem);
2040                 build_srat_memory(numamem, mem_base, mem_len, i - 1,
2041                                   MEM_AFFINITY_ENABLED);
2042             }
2043 
2044             /* Check for the rare case: 640K < RAM < 1M */
2045             if (next_base <= HOLE_640K_END) {
2046                 next_base = HOLE_640K_END;
2047                 continue;
2048             }
2049             mem_base = HOLE_640K_END;
2050             mem_len = next_base - HOLE_640K_END;
2051         }
2052 
2053         /* Cut out the ACPI_PCI hole */
2054         if (mem_base <= x86ms->below_4g_mem_size &&
2055             next_base > x86ms->below_4g_mem_size) {
2056             mem_len -= next_base - x86ms->below_4g_mem_size;
2057             if (mem_len > 0) {
2058                 numamem = acpi_data_push(table_data, sizeof *numamem);
2059                 build_srat_memory(numamem, mem_base, mem_len, i - 1,
2060                                   MEM_AFFINITY_ENABLED);
2061             }
2062             mem_base = 1ULL << 32;
2063             mem_len = next_base - x86ms->below_4g_mem_size;
2064             next_base = mem_base + mem_len;
2065         }
2066 
2067         if (mem_len > 0) {
2068             numamem = acpi_data_push(table_data, sizeof *numamem);
2069             build_srat_memory(numamem, mem_base, mem_len, i - 1,
2070                               MEM_AFFINITY_ENABLED);
2071         }
2072     }
2073 
2074     if (machine->nvdimms_state->is_enabled) {
2075         nvdimm_build_srat(table_data);
2076     }
2077 
2078     slots = (table_data->len - numa_start) / sizeof *numamem;
2079     for (; slots < pcms->numa_nodes + 2; slots++) {
2080         numamem = acpi_data_push(table_data, sizeof *numamem);
2081         build_srat_memory(numamem, 0, 0, 0, MEM_AFFINITY_NOFLAGS);
2082     }
2083 
2084     /*
2085      * Entry is required for Windows to enable memory hotplug in OS
2086      * and for Linux to enable SWIOTLB when booted with less than
2087      * 4G of RAM. Windows works better if the entry sets proximity
2088      * to the highest NUMA node in the machine.
2089      * Memory devices may override proximity set by this entry,
2090      * providing _PXM method if necessary.
2091      */
2092     if (hotplugabble_address_space_size) {
2093         numamem = acpi_data_push(table_data, sizeof *numamem);
2094         build_srat_memory(numamem, machine->device_memory->base,
2095                           hotplugabble_address_space_size, pcms->numa_nodes - 1,
2096                           MEM_AFFINITY_HOTPLUGGABLE | MEM_AFFINITY_ENABLED);
2097     }
2098 
2099     build_header(linker, table_data,
2100                  (void *)(table_data->data + srat_start),
2101                  "SRAT",
2102                  table_data->len - srat_start, 1, NULL, NULL);
2103 }
2104 
2105 /*
2106  * VT-d spec 8.1 DMA Remapping Reporting Structure
2107  * (version Oct. 2014 or later)
2108  */
2109 static void
2110 build_dmar_q35(GArray *table_data, BIOSLinker *linker)
2111 {
2112     int dmar_start = table_data->len;
2113 
2114     AcpiTableDmar *dmar;
2115     AcpiDmarHardwareUnit *drhd;
2116     AcpiDmarRootPortATS *atsr;
2117     uint8_t dmar_flags = 0;
2118     X86IOMMUState *iommu = x86_iommu_get_default();
2119     AcpiDmarDeviceScope *scope = NULL;
2120     /* Root complex IOAPIC use one path[0] only */
2121     size_t ioapic_scope_size = sizeof(*scope) + sizeof(scope->path[0]);
2122     IntelIOMMUState *intel_iommu = INTEL_IOMMU_DEVICE(iommu);
2123 
2124     assert(iommu);
2125     if (x86_iommu_ir_supported(iommu)) {
2126         dmar_flags |= 0x1;      /* Flags: 0x1: INT_REMAP */
2127     }
2128 
2129     dmar = acpi_data_push(table_data, sizeof(*dmar));
2130     dmar->host_address_width = intel_iommu->aw_bits - 1;
2131     dmar->flags = dmar_flags;
2132 
2133     /* DMAR Remapping Hardware Unit Definition structure */
2134     drhd = acpi_data_push(table_data, sizeof(*drhd) + ioapic_scope_size);
2135     drhd->type = cpu_to_le16(ACPI_DMAR_TYPE_HARDWARE_UNIT);
2136     drhd->length = cpu_to_le16(sizeof(*drhd) + ioapic_scope_size);
2137     drhd->flags = ACPI_DMAR_INCLUDE_PCI_ALL;
2138     drhd->pci_segment = cpu_to_le16(0);
2139     drhd->address = cpu_to_le64(Q35_HOST_BRIDGE_IOMMU_ADDR);
2140 
2141     /* Scope definition for the root-complex IOAPIC. See VT-d spec
2142      * 8.3.1 (version Oct. 2014 or later). */
2143     scope = &drhd->scope[0];
2144     scope->entry_type = 0x03;   /* Type: 0x03 for IOAPIC */
2145     scope->length = ioapic_scope_size;
2146     scope->enumeration_id = ACPI_BUILD_IOAPIC_ID;
2147     scope->bus = Q35_PSEUDO_BUS_PLATFORM;
2148     scope->path[0].device = PCI_SLOT(Q35_PSEUDO_DEVFN_IOAPIC);
2149     scope->path[0].function = PCI_FUNC(Q35_PSEUDO_DEVFN_IOAPIC);
2150 
2151     if (iommu->dt_supported) {
2152         atsr = acpi_data_push(table_data, sizeof(*atsr));
2153         atsr->type = cpu_to_le16(ACPI_DMAR_TYPE_ATSR);
2154         atsr->length = cpu_to_le16(sizeof(*atsr));
2155         atsr->flags = ACPI_DMAR_ATSR_ALL_PORTS;
2156         atsr->pci_segment = cpu_to_le16(0);
2157     }
2158 
2159     build_header(linker, table_data, (void *)(table_data->data + dmar_start),
2160                  "DMAR", table_data->len - dmar_start, 1, NULL, NULL);
2161 }
2162 
2163 /*
2164  * Windows ACPI Emulated Devices Table
2165  * (Version 1.0 - April 6, 2009)
2166  * Spec: http://download.microsoft.com/download/7/E/7/7E7662CF-CBEA-470B-A97E-CE7CE0D98DC2/WAET.docx
2167  *
2168  * Helpful to speedup Windows guests and ignored by others.
2169  */
2170 static void
2171 build_waet(GArray *table_data, BIOSLinker *linker)
2172 {
2173     int waet_start = table_data->len;
2174 
2175     /* WAET header */
2176     acpi_data_push(table_data, sizeof(AcpiTableHeader));
2177     /*
2178      * Set "ACPI PM timer good" flag.
2179      *
2180      * Tells Windows guests that our ACPI PM timer is reliable in the
2181      * sense that guest can read it only once to obtain a reliable value.
2182      * Which avoids costly VMExits caused by guest re-reading it unnecessarily.
2183      */
2184     build_append_int_noprefix(table_data, 1 << 1 /* ACPI PM timer good */, 4);
2185 
2186     build_header(linker, table_data, (void *)(table_data->data + waet_start),
2187                  "WAET", table_data->len - waet_start, 1, NULL, NULL);
2188 }
2189 
2190 /*
2191  *   IVRS table as specified in AMD IOMMU Specification v2.62, Section 5.2
2192  *   accessible here http://support.amd.com/TechDocs/48882_IOMMU.pdf
2193  */
2194 #define IOAPIC_SB_DEVID   (uint64_t)PCI_BUILD_BDF(0, PCI_DEVFN(0x14, 0))
2195 
2196 /*
2197  * Insert IVHD entry for device and recurse, insert alias, or insert range as
2198  * necessary for the PCI topology.
2199  */
2200 static void
2201 insert_ivhd(PCIBus *bus, PCIDevice *dev, void *opaque)
2202 {
2203     GArray *table_data = opaque;
2204     uint32_t entry;
2205 
2206     /* "Select" IVHD entry, type 0x2 */
2207     entry = PCI_BUILD_BDF(pci_bus_num(bus), dev->devfn) << 8 | 0x2;
2208     build_append_int_noprefix(table_data, entry, 4);
2209 
2210     if (object_dynamic_cast(OBJECT(dev), TYPE_PCI_BRIDGE)) {
2211         PCIBus *sec_bus = pci_bridge_get_sec_bus(PCI_BRIDGE(dev));
2212         uint8_t sec = pci_bus_num(sec_bus);
2213         uint8_t sub = dev->config[PCI_SUBORDINATE_BUS];
2214 
2215         if (pci_bus_is_express(sec_bus)) {
2216             /*
2217              * Walk the bus if there are subordinates, otherwise use a range
2218              * to cover an entire leaf bus.  We could potentially also use a
2219              * range for traversed buses, but we'd need to take care not to
2220              * create both Select and Range entries covering the same device.
2221              * This is easier and potentially more compact.
2222              *
2223              * An example bare metal system seems to use Select entries for
2224              * root ports without a slot (ie. built-ins) and Range entries
2225              * when there is a slot.  The same system also only hard-codes
2226              * the alias range for an onboard PCIe-to-PCI bridge, apparently
2227              * making no effort to support nested bridges.  We attempt to
2228              * be more thorough here.
2229              */
2230             if (sec == sub) { /* leaf bus */
2231                 /* "Start of Range" IVHD entry, type 0x3 */
2232                 entry = PCI_BUILD_BDF(sec, PCI_DEVFN(0, 0)) << 8 | 0x3;
2233                 build_append_int_noprefix(table_data, entry, 4);
2234                 /* "End of Range" IVHD entry, type 0x4 */
2235                 entry = PCI_BUILD_BDF(sub, PCI_DEVFN(31, 7)) << 8 | 0x4;
2236                 build_append_int_noprefix(table_data, entry, 4);
2237             } else {
2238                 pci_for_each_device(sec_bus, sec, insert_ivhd, table_data);
2239             }
2240         } else {
2241             /*
2242              * If the secondary bus is conventional, then we need to create an
2243              * Alias range for everything downstream.  The range covers the
2244              * first devfn on the secondary bus to the last devfn on the
2245              * subordinate bus.  The alias target depends on legacy versus
2246              * express bridges, just as in pci_device_iommu_address_space().
2247              * DeviceIDa vs DeviceIDb as per the AMD IOMMU spec.
2248              */
2249             uint16_t dev_id_a, dev_id_b;
2250 
2251             dev_id_a = PCI_BUILD_BDF(sec, PCI_DEVFN(0, 0));
2252 
2253             if (pci_is_express(dev) &&
2254                 pcie_cap_get_type(dev) == PCI_EXP_TYPE_PCI_BRIDGE) {
2255                 dev_id_b = dev_id_a;
2256             } else {
2257                 dev_id_b = PCI_BUILD_BDF(pci_bus_num(bus), dev->devfn);
2258             }
2259 
2260             /* "Alias Start of Range" IVHD entry, type 0x43, 8 bytes */
2261             build_append_int_noprefix(table_data, dev_id_a << 8 | 0x43, 4);
2262             build_append_int_noprefix(table_data, dev_id_b << 8 | 0x0, 4);
2263 
2264             /* "End of Range" IVHD entry, type 0x4 */
2265             entry = PCI_BUILD_BDF(sub, PCI_DEVFN(31, 7)) << 8 | 0x4;
2266             build_append_int_noprefix(table_data, entry, 4);
2267         }
2268     }
2269 }
2270 
2271 /* For all PCI host bridges, walk and insert IVHD entries */
2272 static int
2273 ivrs_host_bridges(Object *obj, void *opaque)
2274 {
2275     GArray *ivhd_blob = opaque;
2276 
2277     if (object_dynamic_cast(obj, TYPE_PCI_HOST_BRIDGE)) {
2278         PCIBus *bus = PCI_HOST_BRIDGE(obj)->bus;
2279 
2280         if (bus) {
2281             pci_for_each_device(bus, pci_bus_num(bus), insert_ivhd, ivhd_blob);
2282         }
2283     }
2284 
2285     return 0;
2286 }
2287 
2288 static void
2289 build_amd_iommu(GArray *table_data, BIOSLinker *linker)
2290 {
2291     int ivhd_table_len = 24;
2292     int iommu_start = table_data->len;
2293     AMDVIState *s = AMD_IOMMU_DEVICE(x86_iommu_get_default());
2294     GArray *ivhd_blob = g_array_new(false, true, 1);
2295 
2296     /* IVRS header */
2297     acpi_data_push(table_data, sizeof(AcpiTableHeader));
2298     /* IVinfo - IO virtualization information common to all
2299      * IOMMU units in a system
2300      */
2301     build_append_int_noprefix(table_data, 40UL << 8/* PASize */, 4);
2302     /* reserved */
2303     build_append_int_noprefix(table_data, 0, 8);
2304 
2305     /* IVHD definition - type 10h */
2306     build_append_int_noprefix(table_data, 0x10, 1);
2307     /* virtualization flags */
2308     build_append_int_noprefix(table_data,
2309                              (1UL << 0) | /* HtTunEn      */
2310                              (1UL << 4) | /* iotblSup     */
2311                              (1UL << 6) | /* PrefSup      */
2312                              (1UL << 7),  /* PPRSup       */
2313                              1);
2314 
2315     /*
2316      * A PCI bus walk, for each PCI host bridge, is necessary to create a
2317      * complete set of IVHD entries.  Do this into a separate blob so that we
2318      * can calculate the total IVRS table length here and then append the new
2319      * blob further below.  Fall back to an entry covering all devices, which
2320      * is sufficient when no aliases are present.
2321      */
2322     object_child_foreach_recursive(object_get_root(),
2323                                    ivrs_host_bridges, ivhd_blob);
2324 
2325     if (!ivhd_blob->len) {
2326         /*
2327          *   Type 1 device entry reporting all devices
2328          *   These are 4-byte device entries currently reporting the range of
2329          *   Refer to Spec - Table 95:IVHD Device Entry Type Codes(4-byte)
2330          */
2331         build_append_int_noprefix(ivhd_blob, 0x0000001, 4);
2332     }
2333 
2334     ivhd_table_len += ivhd_blob->len;
2335 
2336     /*
2337      * When interrupt remapping is supported, we add a special IVHD device
2338      * for type IO-APIC.
2339      */
2340     if (x86_iommu_ir_supported(x86_iommu_get_default())) {
2341         ivhd_table_len += 8;
2342     }
2343 
2344     /* IVHD length */
2345     build_append_int_noprefix(table_data, ivhd_table_len, 2);
2346     /* DeviceID */
2347     build_append_int_noprefix(table_data, s->devid, 2);
2348     /* Capability offset */
2349     build_append_int_noprefix(table_data, s->capab_offset, 2);
2350     /* IOMMU base address */
2351     build_append_int_noprefix(table_data, s->mmio.addr, 8);
2352     /* PCI Segment Group */
2353     build_append_int_noprefix(table_data, 0, 2);
2354     /* IOMMU info */
2355     build_append_int_noprefix(table_data, 0, 2);
2356     /* IOMMU Feature Reporting */
2357     build_append_int_noprefix(table_data,
2358                              (48UL << 30) | /* HATS   */
2359                              (48UL << 28) | /* GATS   */
2360                              (1UL << 2)   | /* GTSup  */
2361                              (1UL << 6),    /* GASup  */
2362                              4);
2363 
2364     /* IVHD entries as found above */
2365     g_array_append_vals(table_data, ivhd_blob->data, ivhd_blob->len);
2366     g_array_free(ivhd_blob, TRUE);
2367 
2368     /*
2369      * Add a special IVHD device type.
2370      * Refer to spec - Table 95: IVHD device entry type codes
2371      *
2372      * Linux IOMMU driver checks for the special IVHD device (type IO-APIC).
2373      * See Linux kernel commit 'c2ff5cf5294bcbd7fa50f7d860e90a66db7e5059'
2374      */
2375     if (x86_iommu_ir_supported(x86_iommu_get_default())) {
2376         build_append_int_noprefix(table_data,
2377                                  (0x1ull << 56) |           /* type IOAPIC */
2378                                  (IOAPIC_SB_DEVID << 40) |  /* IOAPIC devid */
2379                                  0x48,                      /* special device */
2380                                  8);
2381     }
2382 
2383     build_header(linker, table_data, (void *)(table_data->data + iommu_start),
2384                  "IVRS", table_data->len - iommu_start, 1, NULL, NULL);
2385 }
2386 
2387 typedef
2388 struct AcpiBuildState {
2389     /* Copy of table in RAM (for patching). */
2390     MemoryRegion *table_mr;
2391     /* Is table patched? */
2392     uint8_t patched;
2393     void *rsdp;
2394     MemoryRegion *rsdp_mr;
2395     MemoryRegion *linker_mr;
2396 } AcpiBuildState;
2397 
2398 static bool acpi_get_mcfg(AcpiMcfgInfo *mcfg)
2399 {
2400     Object *pci_host;
2401     QObject *o;
2402 
2403     pci_host = acpi_get_i386_pci_host();
2404     g_assert(pci_host);
2405 
2406     o = object_property_get_qobject(pci_host, PCIE_HOST_MCFG_BASE, NULL);
2407     if (!o) {
2408         return false;
2409     }
2410     mcfg->base = qnum_get_uint(qobject_to(QNum, o));
2411     qobject_unref(o);
2412     if (mcfg->base == PCIE_BASE_ADDR_UNMAPPED) {
2413         return false;
2414     }
2415 
2416     o = object_property_get_qobject(pci_host, PCIE_HOST_MCFG_SIZE, NULL);
2417     assert(o);
2418     mcfg->size = qnum_get_uint(qobject_to(QNum, o));
2419     qobject_unref(o);
2420     return true;
2421 }
2422 
2423 static
2424 void acpi_build(AcpiBuildTables *tables, MachineState *machine)
2425 {
2426     PCMachineState *pcms = PC_MACHINE(machine);
2427     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
2428     X86MachineState *x86ms = X86_MACHINE(machine);
2429     GArray *table_offsets;
2430     unsigned facs, dsdt, rsdt, fadt;
2431     AcpiPmInfo pm;
2432     AcpiMiscInfo misc;
2433     AcpiMcfgInfo mcfg;
2434     Range pci_hole, pci_hole64;
2435     uint8_t *u;
2436     size_t aml_len = 0;
2437     GArray *tables_blob = tables->table_data;
2438     AcpiSlicOem slic_oem = { .id = NULL, .table_id = NULL };
2439     Object *vmgenid_dev;
2440 
2441     acpi_get_pm_info(machine, &pm);
2442     acpi_get_misc_info(&misc);
2443     acpi_get_pci_holes(&pci_hole, &pci_hole64);
2444     acpi_get_slic_oem(&slic_oem);
2445 
2446     table_offsets = g_array_new(false, true /* clear */,
2447                                         sizeof(uint32_t));
2448     ACPI_BUILD_DPRINTF("init ACPI tables\n");
2449 
2450     bios_linker_loader_alloc(tables->linker,
2451                              ACPI_BUILD_TABLE_FILE, tables_blob,
2452                              64 /* Ensure FACS is aligned */,
2453                              false /* high memory */);
2454 
2455     /*
2456      * FACS is pointed to by FADT.
2457      * We place it first since it's the only table that has alignment
2458      * requirements.
2459      */
2460     facs = tables_blob->len;
2461     build_facs(tables_blob);
2462 
2463     /* DSDT is pointed to by FADT */
2464     dsdt = tables_blob->len;
2465     build_dsdt(tables_blob, tables->linker, &pm, &misc,
2466                &pci_hole, &pci_hole64, machine);
2467 
2468     /* Count the size of the DSDT and SSDT, we will need it for legacy
2469      * sizing of ACPI tables.
2470      */
2471     aml_len += tables_blob->len - dsdt;
2472 
2473     /* ACPI tables pointed to by RSDT */
2474     fadt = tables_blob->len;
2475     acpi_add_table(table_offsets, tables_blob);
2476     pm.fadt.facs_tbl_offset = &facs;
2477     pm.fadt.dsdt_tbl_offset = &dsdt;
2478     pm.fadt.xdsdt_tbl_offset = &dsdt;
2479     build_fadt(tables_blob, tables->linker, &pm.fadt,
2480                slic_oem.id, slic_oem.table_id);
2481     aml_len += tables_blob->len - fadt;
2482 
2483     acpi_add_table(table_offsets, tables_blob);
2484     acpi_build_madt(tables_blob, tables->linker, x86ms,
2485                     ACPI_DEVICE_IF(x86ms->acpi_dev));
2486 
2487     vmgenid_dev = find_vmgenid_dev();
2488     if (vmgenid_dev) {
2489         acpi_add_table(table_offsets, tables_blob);
2490         vmgenid_build_acpi(VMGENID(vmgenid_dev), tables_blob,
2491                            tables->vmgenid, tables->linker);
2492     }
2493 
2494     if (misc.has_hpet) {
2495         acpi_add_table(table_offsets, tables_blob);
2496         build_hpet(tables_blob, tables->linker);
2497     }
2498     if (misc.tpm_version != TPM_VERSION_UNSPEC) {
2499         if (misc.tpm_version == TPM_VERSION_1_2) {
2500             acpi_add_table(table_offsets, tables_blob);
2501             build_tpm_tcpa(tables_blob, tables->linker, tables->tcpalog);
2502         } else { /* TPM_VERSION_2_0 */
2503             acpi_add_table(table_offsets, tables_blob);
2504             build_tpm2(tables_blob, tables->linker, tables->tcpalog);
2505         }
2506     }
2507     if (pcms->numa_nodes) {
2508         acpi_add_table(table_offsets, tables_blob);
2509         build_srat(tables_blob, tables->linker, machine);
2510         if (machine->numa_state->have_numa_distance) {
2511             acpi_add_table(table_offsets, tables_blob);
2512             build_slit(tables_blob, tables->linker, machine);
2513         }
2514         if (machine->numa_state->hmat_enabled) {
2515             acpi_add_table(table_offsets, tables_blob);
2516             build_hmat(tables_blob, tables->linker, machine->numa_state);
2517         }
2518     }
2519     if (acpi_get_mcfg(&mcfg)) {
2520         acpi_add_table(table_offsets, tables_blob);
2521         build_mcfg(tables_blob, tables->linker, &mcfg);
2522     }
2523     if (x86_iommu_get_default()) {
2524         IommuType IOMMUType = x86_iommu_get_type();
2525         if (IOMMUType == TYPE_AMD) {
2526             acpi_add_table(table_offsets, tables_blob);
2527             build_amd_iommu(tables_blob, tables->linker);
2528         } else if (IOMMUType == TYPE_INTEL) {
2529             acpi_add_table(table_offsets, tables_blob);
2530             build_dmar_q35(tables_blob, tables->linker);
2531         }
2532     }
2533     if (machine->nvdimms_state->is_enabled) {
2534         nvdimm_build_acpi(table_offsets, tables_blob, tables->linker,
2535                           machine->nvdimms_state, machine->ram_slots);
2536     }
2537 
2538     acpi_add_table(table_offsets, tables_blob);
2539     build_waet(tables_blob, tables->linker);
2540 
2541     /* Add tables supplied by user (if any) */
2542     for (u = acpi_table_first(); u; u = acpi_table_next(u)) {
2543         unsigned len = acpi_table_len(u);
2544 
2545         acpi_add_table(table_offsets, tables_blob);
2546         g_array_append_vals(tables_blob, u, len);
2547     }
2548 
2549     /* RSDT is pointed to by RSDP */
2550     rsdt = tables_blob->len;
2551     build_rsdt(tables_blob, tables->linker, table_offsets,
2552                slic_oem.id, slic_oem.table_id);
2553 
2554     /* RSDP is in FSEG memory, so allocate it separately */
2555     {
2556         AcpiRsdpData rsdp_data = {
2557             .revision = 0,
2558             .oem_id = ACPI_BUILD_APPNAME6,
2559             .xsdt_tbl_offset = NULL,
2560             .rsdt_tbl_offset = &rsdt,
2561         };
2562         build_rsdp(tables->rsdp, tables->linker, &rsdp_data);
2563         if (!pcmc->rsdp_in_ram) {
2564             /* We used to allocate some extra space for RSDP revision 2 but
2565              * only used the RSDP revision 0 space. The extra bytes were
2566              * zeroed out and not used.
2567              * Here we continue wasting those extra 16 bytes to make sure we
2568              * don't break migration for machine types 2.2 and older due to
2569              * RSDP blob size mismatch.
2570              */
2571             build_append_int_noprefix(tables->rsdp, 0, 16);
2572         }
2573     }
2574 
2575     /* We'll expose it all to Guest so we want to reduce
2576      * chance of size changes.
2577      *
2578      * We used to align the tables to 4k, but of course this would
2579      * too simple to be enough.  4k turned out to be too small an
2580      * alignment very soon, and in fact it is almost impossible to
2581      * keep the table size stable for all (max_cpus, max_memory_slots)
2582      * combinations.  So the table size is always 64k for pc-i440fx-2.1
2583      * and we give an error if the table grows beyond that limit.
2584      *
2585      * We still have the problem of migrating from "-M pc-i440fx-2.0".  For
2586      * that, we exploit the fact that QEMU 2.1 generates _smaller_ tables
2587      * than 2.0 and we can always pad the smaller tables with zeros.  We can
2588      * then use the exact size of the 2.0 tables.
2589      *
2590      * All this is for PIIX4, since QEMU 2.0 didn't support Q35 migration.
2591      */
2592     if (pcmc->legacy_acpi_table_size) {
2593         /* Subtracting aml_len gives the size of fixed tables.  Then add the
2594          * size of the PIIX4 DSDT/SSDT in QEMU 2.0.
2595          */
2596         int legacy_aml_len =
2597             pcmc->legacy_acpi_table_size +
2598             ACPI_BUILD_LEGACY_CPU_AML_SIZE * x86ms->apic_id_limit;
2599         int legacy_table_size =
2600             ROUND_UP(tables_blob->len - aml_len + legacy_aml_len,
2601                      ACPI_BUILD_ALIGN_SIZE);
2602         if (tables_blob->len > legacy_table_size) {
2603             /* Should happen only with PCI bridges and -M pc-i440fx-2.0.  */
2604             warn_report("ACPI table size %u exceeds %d bytes,"
2605                         " migration may not work",
2606                         tables_blob->len, legacy_table_size);
2607             error_printf("Try removing CPUs, NUMA nodes, memory slots"
2608                          " or PCI bridges.");
2609         }
2610         g_array_set_size(tables_blob, legacy_table_size);
2611     } else {
2612         /* Make sure we have a buffer in case we need to resize the tables. */
2613         if (tables_blob->len > ACPI_BUILD_TABLE_SIZE / 2) {
2614             /* As of QEMU 2.1, this fires with 160 VCPUs and 255 memory slots.  */
2615             warn_report("ACPI table size %u exceeds %d bytes,"
2616                         " migration may not work",
2617                         tables_blob->len, ACPI_BUILD_TABLE_SIZE / 2);
2618             error_printf("Try removing CPUs, NUMA nodes, memory slots"
2619                          " or PCI bridges.");
2620         }
2621         acpi_align_size(tables_blob, ACPI_BUILD_TABLE_SIZE);
2622     }
2623 
2624     acpi_align_size(tables->linker->cmd_blob, ACPI_BUILD_ALIGN_SIZE);
2625 
2626     /* Cleanup memory that's no longer used. */
2627     g_array_free(table_offsets, true);
2628 }
2629 
2630 static void acpi_ram_update(MemoryRegion *mr, GArray *data)
2631 {
2632     uint32_t size = acpi_data_len(data);
2633 
2634     /* Make sure RAM size is correct - in case it got changed e.g. by migration */
2635     memory_region_ram_resize(mr, size, &error_abort);
2636 
2637     memcpy(memory_region_get_ram_ptr(mr), data->data, size);
2638     memory_region_set_dirty(mr, 0, size);
2639 }
2640 
2641 static void acpi_build_update(void *build_opaque)
2642 {
2643     AcpiBuildState *build_state = build_opaque;
2644     AcpiBuildTables tables;
2645 
2646     /* No state to update or already patched? Nothing to do. */
2647     if (!build_state || build_state->patched) {
2648         return;
2649     }
2650     build_state->patched = 1;
2651 
2652     acpi_build_tables_init(&tables);
2653 
2654     acpi_build(&tables, MACHINE(qdev_get_machine()));
2655 
2656     acpi_ram_update(build_state->table_mr, tables.table_data);
2657 
2658     if (build_state->rsdp) {
2659         memcpy(build_state->rsdp, tables.rsdp->data, acpi_data_len(tables.rsdp));
2660     } else {
2661         acpi_ram_update(build_state->rsdp_mr, tables.rsdp);
2662     }
2663 
2664     acpi_ram_update(build_state->linker_mr, tables.linker->cmd_blob);
2665     acpi_build_tables_cleanup(&tables, true);
2666 }
2667 
2668 static void acpi_build_reset(void *build_opaque)
2669 {
2670     AcpiBuildState *build_state = build_opaque;
2671     build_state->patched = 0;
2672 }
2673 
2674 static const VMStateDescription vmstate_acpi_build = {
2675     .name = "acpi_build",
2676     .version_id = 1,
2677     .minimum_version_id = 1,
2678     .fields = (VMStateField[]) {
2679         VMSTATE_UINT8(patched, AcpiBuildState),
2680         VMSTATE_END_OF_LIST()
2681     },
2682 };
2683 
2684 void acpi_setup(void)
2685 {
2686     PCMachineState *pcms = PC_MACHINE(qdev_get_machine());
2687     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
2688     X86MachineState *x86ms = X86_MACHINE(pcms);
2689     AcpiBuildTables tables;
2690     AcpiBuildState *build_state;
2691     Object *vmgenid_dev;
2692     TPMIf *tpm;
2693     static FwCfgTPMConfig tpm_config;
2694 
2695     if (!x86ms->fw_cfg) {
2696         ACPI_BUILD_DPRINTF("No fw cfg. Bailing out.\n");
2697         return;
2698     }
2699 
2700     if (!pcms->acpi_build_enabled) {
2701         ACPI_BUILD_DPRINTF("ACPI build disabled. Bailing out.\n");
2702         return;
2703     }
2704 
2705     if (!x86_machine_is_acpi_enabled(X86_MACHINE(pcms))) {
2706         ACPI_BUILD_DPRINTF("ACPI disabled. Bailing out.\n");
2707         return;
2708     }
2709 
2710     build_state = g_malloc0(sizeof *build_state);
2711 
2712     acpi_build_tables_init(&tables);
2713     acpi_build(&tables, MACHINE(pcms));
2714 
2715     /* Now expose it all to Guest */
2716     build_state->table_mr = acpi_add_rom_blob(acpi_build_update,
2717                                               build_state, tables.table_data,
2718                                               ACPI_BUILD_TABLE_FILE,
2719                                               ACPI_BUILD_TABLE_MAX_SIZE);
2720     assert(build_state->table_mr != NULL);
2721 
2722     build_state->linker_mr =
2723         acpi_add_rom_blob(acpi_build_update, build_state,
2724                           tables.linker->cmd_blob, ACPI_BUILD_LOADER_FILE, 0);
2725 
2726     fw_cfg_add_file(x86ms->fw_cfg, ACPI_BUILD_TPMLOG_FILE,
2727                     tables.tcpalog->data, acpi_data_len(tables.tcpalog));
2728 
2729     tpm = tpm_find();
2730     if (tpm && object_property_get_bool(OBJECT(tpm), "ppi", &error_abort)) {
2731         tpm_config = (FwCfgTPMConfig) {
2732             .tpmppi_address = cpu_to_le32(TPM_PPI_ADDR_BASE),
2733             .tpm_version = tpm_get_version(tpm),
2734             .tpmppi_version = TPM_PPI_VERSION_1_30
2735         };
2736         fw_cfg_add_file(x86ms->fw_cfg, "etc/tpm/config",
2737                         &tpm_config, sizeof tpm_config);
2738     }
2739 
2740     vmgenid_dev = find_vmgenid_dev();
2741     if (vmgenid_dev) {
2742         vmgenid_add_fw_cfg(VMGENID(vmgenid_dev), x86ms->fw_cfg,
2743                            tables.vmgenid);
2744     }
2745 
2746     if (!pcmc->rsdp_in_ram) {
2747         /*
2748          * Keep for compatibility with old machine types.
2749          * Though RSDP is small, its contents isn't immutable, so
2750          * we'll update it along with the rest of tables on guest access.
2751          */
2752         uint32_t rsdp_size = acpi_data_len(tables.rsdp);
2753 
2754         build_state->rsdp = g_memdup(tables.rsdp->data, rsdp_size);
2755         fw_cfg_add_file_callback(x86ms->fw_cfg, ACPI_BUILD_RSDP_FILE,
2756                                  acpi_build_update, NULL, build_state,
2757                                  build_state->rsdp, rsdp_size, true);
2758         build_state->rsdp_mr = NULL;
2759     } else {
2760         build_state->rsdp = NULL;
2761         build_state->rsdp_mr = acpi_add_rom_blob(acpi_build_update,
2762                                                  build_state, tables.rsdp,
2763                                                  ACPI_BUILD_RSDP_FILE, 0);
2764     }
2765 
2766     qemu_register_reset(acpi_build_reset, build_state);
2767     acpi_build_reset(build_state);
2768     vmstate_register(NULL, 0, &vmstate_acpi_build, build_state);
2769 
2770     /* Cleanup tables but don't free the memory: we track it
2771      * in build_state.
2772      */
2773     acpi_build_tables_cleanup(&tables, false);
2774 }
2775