xref: /openbmc/qemu/hw/i386/acpi-build.c (revision d3b5a15f)
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,
475             aml_call2("DVNT", aml_name("PCIU"), aml_int(1) /* Device Check */)
476         );
477         aml_append(method,
478             aml_call2("DVNT", aml_name("PCID"), aml_int(3)/* Eject Request */)
479         );
480     }
481 
482     /* Notify about child bus events in any case */
483     if (pcihp_bridge_en) {
484         QLIST_FOREACH(sec, &bus->child, sibling) {
485             int32_t devfn = sec->parent_dev->devfn;
486 
487             if (pci_bus_is_root(sec) || pci_bus_is_express(sec)) {
488                 continue;
489             }
490 
491             aml_append(method, aml_name("^S%.02X.PCNT", devfn));
492         }
493     }
494 
495     if (bsel || pcihp_bridge_en) {
496         aml_append(parent_scope, method);
497     }
498     qobject_unref(bsel);
499 }
500 
501 /**
502  * build_prt_entry:
503  * @link_name: link name for PCI route entry
504  *
505  * build AML package containing a PCI route entry for @link_name
506  */
507 static Aml *build_prt_entry(const char *link_name)
508 {
509     Aml *a_zero = aml_int(0);
510     Aml *pkg = aml_package(4);
511     aml_append(pkg, a_zero);
512     aml_append(pkg, a_zero);
513     aml_append(pkg, aml_name("%s", link_name));
514     aml_append(pkg, a_zero);
515     return pkg;
516 }
517 
518 /*
519  * initialize_route - Initialize the interrupt routing rule
520  * through a specific LINK:
521  *  if (lnk_idx == idx)
522  *      route using link 'link_name'
523  */
524 static Aml *initialize_route(Aml *route, const char *link_name,
525                              Aml *lnk_idx, int idx)
526 {
527     Aml *if_ctx = aml_if(aml_equal(lnk_idx, aml_int(idx)));
528     Aml *pkg = build_prt_entry(link_name);
529 
530     aml_append(if_ctx, aml_store(pkg, route));
531 
532     return if_ctx;
533 }
534 
535 /*
536  * build_prt - Define interrupt rounting rules
537  *
538  * Returns an array of 128 routes, one for each device,
539  * based on device location.
540  * The main goal is to equaly distribute the interrupts
541  * over the 4 existing ACPI links (works only for i440fx).
542  * The hash function is  (slot + pin) & 3 -> "LNK[D|A|B|C]".
543  *
544  */
545 static Aml *build_prt(bool is_pci0_prt)
546 {
547     Aml *method, *while_ctx, *pin, *res;
548 
549     method = aml_method("_PRT", 0, AML_NOTSERIALIZED);
550     res = aml_local(0);
551     pin = aml_local(1);
552     aml_append(method, aml_store(aml_package(128), res));
553     aml_append(method, aml_store(aml_int(0), pin));
554 
555     /* while (pin < 128) */
556     while_ctx = aml_while(aml_lless(pin, aml_int(128)));
557     {
558         Aml *slot = aml_local(2);
559         Aml *lnk_idx = aml_local(3);
560         Aml *route = aml_local(4);
561 
562         /* slot = pin >> 2 */
563         aml_append(while_ctx,
564                    aml_store(aml_shiftright(pin, aml_int(2), NULL), slot));
565         /* lnk_idx = (slot + pin) & 3 */
566         aml_append(while_ctx,
567             aml_store(aml_and(aml_add(pin, slot, NULL), aml_int(3), NULL),
568                       lnk_idx));
569 
570         /* route[2] = "LNK[D|A|B|C]", selection based on pin % 3  */
571         aml_append(while_ctx, initialize_route(route, "LNKD", lnk_idx, 0));
572         if (is_pci0_prt) {
573             Aml *if_device_1, *if_pin_4, *else_pin_4;
574 
575             /* device 1 is the power-management device, needs SCI */
576             if_device_1 = aml_if(aml_equal(lnk_idx, aml_int(1)));
577             {
578                 if_pin_4 = aml_if(aml_equal(pin, aml_int(4)));
579                 {
580                     aml_append(if_pin_4,
581                         aml_store(build_prt_entry("LNKS"), route));
582                 }
583                 aml_append(if_device_1, if_pin_4);
584                 else_pin_4 = aml_else();
585                 {
586                     aml_append(else_pin_4,
587                         aml_store(build_prt_entry("LNKA"), route));
588                 }
589                 aml_append(if_device_1, else_pin_4);
590             }
591             aml_append(while_ctx, if_device_1);
592         } else {
593             aml_append(while_ctx, initialize_route(route, "LNKA", lnk_idx, 1));
594         }
595         aml_append(while_ctx, initialize_route(route, "LNKB", lnk_idx, 2));
596         aml_append(while_ctx, initialize_route(route, "LNKC", lnk_idx, 3));
597 
598         /* route[0] = 0x[slot]FFFF */
599         aml_append(while_ctx,
600             aml_store(aml_or(aml_shiftleft(slot, aml_int(16)), aml_int(0xFFFF),
601                              NULL),
602                       aml_index(route, aml_int(0))));
603         /* route[1] = pin & 3 */
604         aml_append(while_ctx,
605             aml_store(aml_and(pin, aml_int(3), NULL),
606                       aml_index(route, aml_int(1))));
607         /* res[pin] = route */
608         aml_append(while_ctx, aml_store(route, aml_index(res, pin)));
609         /* pin++ */
610         aml_append(while_ctx, aml_increment(pin));
611     }
612     aml_append(method, while_ctx);
613     /* return res*/
614     aml_append(method, aml_return(res));
615 
616     return method;
617 }
618 
619 typedef struct CrsRangeEntry {
620     uint64_t base;
621     uint64_t limit;
622 } CrsRangeEntry;
623 
624 static void crs_range_insert(GPtrArray *ranges, uint64_t base, uint64_t limit)
625 {
626     CrsRangeEntry *entry;
627 
628     entry = g_malloc(sizeof(*entry));
629     entry->base = base;
630     entry->limit = limit;
631 
632     g_ptr_array_add(ranges, entry);
633 }
634 
635 static void crs_range_free(gpointer data)
636 {
637     CrsRangeEntry *entry = (CrsRangeEntry *)data;
638     g_free(entry);
639 }
640 
641 typedef struct CrsRangeSet {
642     GPtrArray *io_ranges;
643     GPtrArray *mem_ranges;
644     GPtrArray *mem_64bit_ranges;
645  } CrsRangeSet;
646 
647 static void crs_range_set_init(CrsRangeSet *range_set)
648 {
649     range_set->io_ranges = g_ptr_array_new_with_free_func(crs_range_free);
650     range_set->mem_ranges = g_ptr_array_new_with_free_func(crs_range_free);
651     range_set->mem_64bit_ranges =
652             g_ptr_array_new_with_free_func(crs_range_free);
653 }
654 
655 static void crs_range_set_free(CrsRangeSet *range_set)
656 {
657     g_ptr_array_free(range_set->io_ranges, true);
658     g_ptr_array_free(range_set->mem_ranges, true);
659     g_ptr_array_free(range_set->mem_64bit_ranges, true);
660 }
661 
662 static gint crs_range_compare(gconstpointer a, gconstpointer b)
663 {
664     CrsRangeEntry *entry_a = *(CrsRangeEntry **)a;
665     CrsRangeEntry *entry_b = *(CrsRangeEntry **)b;
666 
667     if (entry_a->base < entry_b->base) {
668         return -1;
669     } else if (entry_a->base > entry_b->base) {
670         return 1;
671     } else {
672         return 0;
673     }
674 }
675 
676 /*
677  * crs_replace_with_free_ranges - given the 'used' ranges within [start - end]
678  * interval, computes the 'free' ranges from the same interval.
679  * Example: If the input array is { [a1 - a2],[b1 - b2] }, the function
680  * will return { [base - a1], [a2 - b1], [b2 - limit] }.
681  */
682 static void crs_replace_with_free_ranges(GPtrArray *ranges,
683                                          uint64_t start, uint64_t end)
684 {
685     GPtrArray *free_ranges = g_ptr_array_new();
686     uint64_t free_base = start;
687     int i;
688 
689     g_ptr_array_sort(ranges, crs_range_compare);
690     for (i = 0; i < ranges->len; i++) {
691         CrsRangeEntry *used = g_ptr_array_index(ranges, i);
692 
693         if (free_base < used->base) {
694             crs_range_insert(free_ranges, free_base, used->base - 1);
695         }
696 
697         free_base = used->limit + 1;
698     }
699 
700     if (free_base < end) {
701         crs_range_insert(free_ranges, free_base, end);
702     }
703 
704     g_ptr_array_set_size(ranges, 0);
705     for (i = 0; i < free_ranges->len; i++) {
706         g_ptr_array_add(ranges, g_ptr_array_index(free_ranges, i));
707     }
708 
709     g_ptr_array_free(free_ranges, true);
710 }
711 
712 /*
713  * crs_range_merge - merges adjacent ranges in the given array.
714  * Array elements are deleted and replaced with the merged ranges.
715  */
716 static void crs_range_merge(GPtrArray *range)
717 {
718     GPtrArray *tmp =  g_ptr_array_new_with_free_func(crs_range_free);
719     CrsRangeEntry *entry;
720     uint64_t range_base, range_limit;
721     int i;
722 
723     if (!range->len) {
724         return;
725     }
726 
727     g_ptr_array_sort(range, crs_range_compare);
728 
729     entry = g_ptr_array_index(range, 0);
730     range_base = entry->base;
731     range_limit = entry->limit;
732     for (i = 1; i < range->len; i++) {
733         entry = g_ptr_array_index(range, i);
734         if (entry->base - 1 == range_limit) {
735             range_limit = entry->limit;
736         } else {
737             crs_range_insert(tmp, range_base, range_limit);
738             range_base = entry->base;
739             range_limit = entry->limit;
740         }
741     }
742     crs_range_insert(tmp, range_base, range_limit);
743 
744     g_ptr_array_set_size(range, 0);
745     for (i = 0; i < tmp->len; i++) {
746         entry = g_ptr_array_index(tmp, i);
747         crs_range_insert(range, entry->base, entry->limit);
748     }
749     g_ptr_array_free(tmp, true);
750 }
751 
752 static Aml *build_crs(PCIHostState *host, CrsRangeSet *range_set)
753 {
754     Aml *crs = aml_resource_template();
755     CrsRangeSet temp_range_set;
756     CrsRangeEntry *entry;
757     uint8_t max_bus = pci_bus_num(host->bus);
758     uint8_t type;
759     int devfn;
760     int i;
761 
762     crs_range_set_init(&temp_range_set);
763     for (devfn = 0; devfn < ARRAY_SIZE(host->bus->devices); devfn++) {
764         uint64_t range_base, range_limit;
765         PCIDevice *dev = host->bus->devices[devfn];
766 
767         if (!dev) {
768             continue;
769         }
770 
771         for (i = 0; i < PCI_NUM_REGIONS; i++) {
772             PCIIORegion *r = &dev->io_regions[i];
773 
774             range_base = r->addr;
775             range_limit = r->addr + r->size - 1;
776 
777             /*
778              * Work-around for old bioses
779              * that do not support multiple root buses
780              */
781             if (!range_base || range_base > range_limit) {
782                 continue;
783             }
784 
785             if (r->type & PCI_BASE_ADDRESS_SPACE_IO) {
786                 crs_range_insert(temp_range_set.io_ranges,
787                                  range_base, range_limit);
788             } else { /* "memory" */
789                 crs_range_insert(temp_range_set.mem_ranges,
790                                  range_base, range_limit);
791             }
792         }
793 
794         type = dev->config[PCI_HEADER_TYPE] & ~PCI_HEADER_TYPE_MULTI_FUNCTION;
795         if (type == PCI_HEADER_TYPE_BRIDGE) {
796             uint8_t subordinate = dev->config[PCI_SUBORDINATE_BUS];
797             if (subordinate > max_bus) {
798                 max_bus = subordinate;
799             }
800 
801             range_base = pci_bridge_get_base(dev, PCI_BASE_ADDRESS_SPACE_IO);
802             range_limit = pci_bridge_get_limit(dev, PCI_BASE_ADDRESS_SPACE_IO);
803 
804             /*
805              * Work-around for old bioses
806              * that do not support multiple root buses
807              */
808             if (range_base && range_base <= range_limit) {
809                 crs_range_insert(temp_range_set.io_ranges,
810                                  range_base, range_limit);
811             }
812 
813             range_base =
814                 pci_bridge_get_base(dev, PCI_BASE_ADDRESS_SPACE_MEMORY);
815             range_limit =
816                 pci_bridge_get_limit(dev, PCI_BASE_ADDRESS_SPACE_MEMORY);
817 
818             /*
819              * Work-around for old bioses
820              * that do not support multiple root buses
821              */
822             if (range_base && range_base <= range_limit) {
823                 uint64_t length = range_limit - range_base + 1;
824                 if (range_limit <= UINT32_MAX && length <= UINT32_MAX) {
825                     crs_range_insert(temp_range_set.mem_ranges,
826                                      range_base, range_limit);
827                 } else {
828                     crs_range_insert(temp_range_set.mem_64bit_ranges,
829                                      range_base, range_limit);
830                 }
831             }
832 
833             range_base =
834                 pci_bridge_get_base(dev, PCI_BASE_ADDRESS_MEM_PREFETCH);
835             range_limit =
836                 pci_bridge_get_limit(dev, PCI_BASE_ADDRESS_MEM_PREFETCH);
837 
838             /*
839              * Work-around for old bioses
840              * that do not support multiple root buses
841              */
842             if (range_base && range_base <= range_limit) {
843                 uint64_t length = range_limit - range_base + 1;
844                 if (range_limit <= UINT32_MAX && length <= UINT32_MAX) {
845                     crs_range_insert(temp_range_set.mem_ranges,
846                                      range_base, range_limit);
847                 } else {
848                     crs_range_insert(temp_range_set.mem_64bit_ranges,
849                                      range_base, range_limit);
850                 }
851             }
852         }
853     }
854 
855     crs_range_merge(temp_range_set.io_ranges);
856     for (i = 0; i < temp_range_set.io_ranges->len; i++) {
857         entry = g_ptr_array_index(temp_range_set.io_ranges, i);
858         aml_append(crs,
859                    aml_word_io(AML_MIN_FIXED, AML_MAX_FIXED,
860                                AML_POS_DECODE, AML_ENTIRE_RANGE,
861                                0, entry->base, entry->limit, 0,
862                                entry->limit - entry->base + 1));
863         crs_range_insert(range_set->io_ranges, entry->base, entry->limit);
864     }
865 
866     crs_range_merge(temp_range_set.mem_ranges);
867     for (i = 0; i < temp_range_set.mem_ranges->len; i++) {
868         entry = g_ptr_array_index(temp_range_set.mem_ranges, i);
869         aml_append(crs,
870                    aml_dword_memory(AML_POS_DECODE, AML_MIN_FIXED,
871                                     AML_MAX_FIXED, AML_NON_CACHEABLE,
872                                     AML_READ_WRITE,
873                                     0, entry->base, entry->limit, 0,
874                                     entry->limit - entry->base + 1));
875         crs_range_insert(range_set->mem_ranges, entry->base, entry->limit);
876     }
877 
878     crs_range_merge(temp_range_set.mem_64bit_ranges);
879     for (i = 0; i < temp_range_set.mem_64bit_ranges->len; i++) {
880         entry = g_ptr_array_index(temp_range_set.mem_64bit_ranges, i);
881         aml_append(crs,
882                    aml_qword_memory(AML_POS_DECODE, AML_MIN_FIXED,
883                                     AML_MAX_FIXED, AML_NON_CACHEABLE,
884                                     AML_READ_WRITE,
885                                     0, entry->base, entry->limit, 0,
886                                     entry->limit - entry->base + 1));
887         crs_range_insert(range_set->mem_64bit_ranges,
888                          entry->base, entry->limit);
889     }
890 
891     crs_range_set_free(&temp_range_set);
892 
893     aml_append(crs,
894         aml_word_bus_number(AML_MIN_FIXED, AML_MAX_FIXED, AML_POS_DECODE,
895                             0,
896                             pci_bus_num(host->bus),
897                             max_bus,
898                             0,
899                             max_bus - pci_bus_num(host->bus) + 1));
900 
901     return crs;
902 }
903 
904 static void build_hpet_aml(Aml *table)
905 {
906     Aml *crs;
907     Aml *field;
908     Aml *method;
909     Aml *if_ctx;
910     Aml *scope = aml_scope("_SB");
911     Aml *dev = aml_device("HPET");
912     Aml *zero = aml_int(0);
913     Aml *id = aml_local(0);
914     Aml *period = aml_local(1);
915 
916     aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0103")));
917     aml_append(dev, aml_name_decl("_UID", zero));
918 
919     aml_append(dev,
920         aml_operation_region("HPTM", AML_SYSTEM_MEMORY, aml_int(HPET_BASE),
921                              HPET_LEN));
922     field = aml_field("HPTM", AML_DWORD_ACC, AML_LOCK, AML_PRESERVE);
923     aml_append(field, aml_named_field("VEND", 32));
924     aml_append(field, aml_named_field("PRD", 32));
925     aml_append(dev, field);
926 
927     method = aml_method("_STA", 0, AML_NOTSERIALIZED);
928     aml_append(method, aml_store(aml_name("VEND"), id));
929     aml_append(method, aml_store(aml_name("PRD"), period));
930     aml_append(method, aml_shiftright(id, aml_int(16), id));
931     if_ctx = aml_if(aml_lor(aml_equal(id, zero),
932                             aml_equal(id, aml_int(0xffff))));
933     {
934         aml_append(if_ctx, aml_return(zero));
935     }
936     aml_append(method, if_ctx);
937 
938     if_ctx = aml_if(aml_lor(aml_equal(period, zero),
939                             aml_lgreater(period, aml_int(100000000))));
940     {
941         aml_append(if_ctx, aml_return(zero));
942     }
943     aml_append(method, if_ctx);
944 
945     aml_append(method, aml_return(aml_int(0x0F)));
946     aml_append(dev, method);
947 
948     crs = aml_resource_template();
949     aml_append(crs, aml_memory32_fixed(HPET_BASE, HPET_LEN, AML_READ_ONLY));
950     aml_append(dev, aml_name_decl("_CRS", crs));
951 
952     aml_append(scope, dev);
953     aml_append(table, scope);
954 }
955 
956 static Aml *build_vmbus_device_aml(VMBusBridge *vmbus_bridge)
957 {
958     Aml *dev;
959     Aml *method;
960     Aml *crs;
961 
962     dev = aml_device("VMBS");
963     aml_append(dev, aml_name_decl("STA", aml_int(0xF)));
964     aml_append(dev, aml_name_decl("_HID", aml_string("VMBus")));
965     aml_append(dev, aml_name_decl("_UID", aml_int(0x0)));
966     aml_append(dev, aml_name_decl("_DDN", aml_string("VMBUS")));
967 
968     method = aml_method("_DIS", 0, AML_NOTSERIALIZED);
969     aml_append(method, aml_store(aml_and(aml_name("STA"), aml_int(0xD), NULL),
970                                      aml_name("STA")));
971     aml_append(dev, method);
972 
973     method = aml_method("_PS0", 0, AML_NOTSERIALIZED);
974     aml_append(method, aml_store(aml_or(aml_name("STA"), aml_int(0xF), NULL),
975                                      aml_name("STA")));
976     aml_append(dev, method);
977 
978     method = aml_method("_STA", 0, AML_NOTSERIALIZED);
979     aml_append(method, aml_return(aml_name("STA")));
980     aml_append(dev, method);
981 
982     aml_append(dev, aml_name_decl("_PS3", aml_int(0x0)));
983 
984     crs = aml_resource_template();
985     aml_append(crs, aml_irq_no_flags(vmbus_bridge->irq));
986     aml_append(dev, aml_name_decl("_CRS", crs));
987 
988     return dev;
989 }
990 
991 static void build_isa_devices_aml(Aml *table)
992 {
993     bool ambiguous;
994     Object *obj = object_resolve_path_type("", TYPE_ISA_BUS, &ambiguous);
995     Aml *scope;
996 
997     assert(obj && !ambiguous);
998 
999     scope = aml_scope("_SB.PCI0.ISA");
1000     build_acpi_ipmi_devices(scope, BUS(obj), "\\_SB.PCI0.ISA");
1001     isa_build_aml(ISA_BUS(obj), scope);
1002 
1003     aml_append(table, scope);
1004 }
1005 
1006 static void build_dbg_aml(Aml *table)
1007 {
1008     Aml *field;
1009     Aml *method;
1010     Aml *while_ctx;
1011     Aml *scope = aml_scope("\\");
1012     Aml *buf = aml_local(0);
1013     Aml *len = aml_local(1);
1014     Aml *idx = aml_local(2);
1015 
1016     aml_append(scope,
1017        aml_operation_region("DBG", AML_SYSTEM_IO, aml_int(0x0402), 0x01));
1018     field = aml_field("DBG", AML_BYTE_ACC, AML_NOLOCK, AML_PRESERVE);
1019     aml_append(field, aml_named_field("DBGB", 8));
1020     aml_append(scope, field);
1021 
1022     method = aml_method("DBUG", 1, AML_NOTSERIALIZED);
1023 
1024     aml_append(method, aml_to_hexstring(aml_arg(0), buf));
1025     aml_append(method, aml_to_buffer(buf, buf));
1026     aml_append(method, aml_subtract(aml_sizeof(buf), aml_int(1), len));
1027     aml_append(method, aml_store(aml_int(0), idx));
1028 
1029     while_ctx = aml_while(aml_lless(idx, len));
1030     aml_append(while_ctx,
1031         aml_store(aml_derefof(aml_index(buf, idx)), aml_name("DBGB")));
1032     aml_append(while_ctx, aml_increment(idx));
1033     aml_append(method, while_ctx);
1034 
1035     aml_append(method, aml_store(aml_int(0x0A), aml_name("DBGB")));
1036     aml_append(scope, method);
1037 
1038     aml_append(table, scope);
1039 }
1040 
1041 static Aml *build_link_dev(const char *name, uint8_t uid, Aml *reg)
1042 {
1043     Aml *dev;
1044     Aml *crs;
1045     Aml *method;
1046     uint32_t irqs[] = {5, 10, 11};
1047 
1048     dev = aml_device("%s", name);
1049     aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0C0F")));
1050     aml_append(dev, aml_name_decl("_UID", aml_int(uid)));
1051 
1052     crs = aml_resource_template();
1053     aml_append(crs, aml_interrupt(AML_CONSUMER, AML_LEVEL, AML_ACTIVE_HIGH,
1054                                   AML_SHARED, irqs, ARRAY_SIZE(irqs)));
1055     aml_append(dev, aml_name_decl("_PRS", crs));
1056 
1057     method = aml_method("_STA", 0, AML_NOTSERIALIZED);
1058     aml_append(method, aml_return(aml_call1("IQST", reg)));
1059     aml_append(dev, method);
1060 
1061     method = aml_method("_DIS", 0, AML_NOTSERIALIZED);
1062     aml_append(method, aml_or(reg, aml_int(0x80), reg));
1063     aml_append(dev, method);
1064 
1065     method = aml_method("_CRS", 0, AML_NOTSERIALIZED);
1066     aml_append(method, aml_return(aml_call1("IQCR", reg)));
1067     aml_append(dev, method);
1068 
1069     method = aml_method("_SRS", 1, AML_NOTSERIALIZED);
1070     aml_append(method, aml_create_dword_field(aml_arg(0), aml_int(5), "PRRI"));
1071     aml_append(method, aml_store(aml_name("PRRI"), reg));
1072     aml_append(dev, method);
1073 
1074     return dev;
1075  }
1076 
1077 static Aml *build_gsi_link_dev(const char *name, uint8_t uid, uint8_t gsi)
1078 {
1079     Aml *dev;
1080     Aml *crs;
1081     Aml *method;
1082     uint32_t irqs;
1083 
1084     dev = aml_device("%s", name);
1085     aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0C0F")));
1086     aml_append(dev, aml_name_decl("_UID", aml_int(uid)));
1087 
1088     crs = aml_resource_template();
1089     irqs = gsi;
1090     aml_append(crs, aml_interrupt(AML_CONSUMER, AML_LEVEL, AML_ACTIVE_HIGH,
1091                                   AML_SHARED, &irqs, 1));
1092     aml_append(dev, aml_name_decl("_PRS", crs));
1093 
1094     aml_append(dev, aml_name_decl("_CRS", crs));
1095 
1096     /*
1097      * _DIS can be no-op because the interrupt cannot be disabled.
1098      */
1099     method = aml_method("_DIS", 0, AML_NOTSERIALIZED);
1100     aml_append(dev, method);
1101 
1102     method = aml_method("_SRS", 1, AML_NOTSERIALIZED);
1103     aml_append(dev, method);
1104 
1105     return dev;
1106 }
1107 
1108 /* _CRS method - get current settings */
1109 static Aml *build_iqcr_method(bool is_piix4)
1110 {
1111     Aml *if_ctx;
1112     uint32_t irqs;
1113     Aml *method = aml_method("IQCR", 1, AML_SERIALIZED);
1114     Aml *crs = aml_resource_template();
1115 
1116     irqs = 0;
1117     aml_append(crs, aml_interrupt(AML_CONSUMER, AML_LEVEL,
1118                                   AML_ACTIVE_HIGH, AML_SHARED, &irqs, 1));
1119     aml_append(method, aml_name_decl("PRR0", crs));
1120 
1121     aml_append(method,
1122         aml_create_dword_field(aml_name("PRR0"), aml_int(5), "PRRI"));
1123 
1124     if (is_piix4) {
1125         if_ctx = aml_if(aml_lless(aml_arg(0), aml_int(0x80)));
1126         aml_append(if_ctx, aml_store(aml_arg(0), aml_name("PRRI")));
1127         aml_append(method, if_ctx);
1128     } else {
1129         aml_append(method,
1130             aml_store(aml_and(aml_arg(0), aml_int(0xF), NULL),
1131                       aml_name("PRRI")));
1132     }
1133 
1134     aml_append(method, aml_return(aml_name("PRR0")));
1135     return method;
1136 }
1137 
1138 /* _STA method - get status */
1139 static Aml *build_irq_status_method(void)
1140 {
1141     Aml *if_ctx;
1142     Aml *method = aml_method("IQST", 1, AML_NOTSERIALIZED);
1143 
1144     if_ctx = aml_if(aml_and(aml_int(0x80), aml_arg(0), NULL));
1145     aml_append(if_ctx, aml_return(aml_int(0x09)));
1146     aml_append(method, if_ctx);
1147     aml_append(method, aml_return(aml_int(0x0B)));
1148     return method;
1149 }
1150 
1151 static void build_piix4_pci0_int(Aml *table)
1152 {
1153     Aml *dev;
1154     Aml *crs;
1155     Aml *field;
1156     Aml *method;
1157     uint32_t irqs;
1158     Aml *sb_scope = aml_scope("_SB");
1159     Aml *pci0_scope = aml_scope("PCI0");
1160 
1161     aml_append(pci0_scope, build_prt(true));
1162     aml_append(sb_scope, pci0_scope);
1163 
1164     field = aml_field("PCI0.ISA.P40C", AML_BYTE_ACC, AML_NOLOCK, AML_PRESERVE);
1165     aml_append(field, aml_named_field("PRQ0", 8));
1166     aml_append(field, aml_named_field("PRQ1", 8));
1167     aml_append(field, aml_named_field("PRQ2", 8));
1168     aml_append(field, aml_named_field("PRQ3", 8));
1169     aml_append(sb_scope, field);
1170 
1171     aml_append(sb_scope, build_irq_status_method());
1172     aml_append(sb_scope, build_iqcr_method(true));
1173 
1174     aml_append(sb_scope, build_link_dev("LNKA", 0, aml_name("PRQ0")));
1175     aml_append(sb_scope, build_link_dev("LNKB", 1, aml_name("PRQ1")));
1176     aml_append(sb_scope, build_link_dev("LNKC", 2, aml_name("PRQ2")));
1177     aml_append(sb_scope, build_link_dev("LNKD", 3, aml_name("PRQ3")));
1178 
1179     dev = aml_device("LNKS");
1180     {
1181         aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0C0F")));
1182         aml_append(dev, aml_name_decl("_UID", aml_int(4)));
1183 
1184         crs = aml_resource_template();
1185         irqs = 9;
1186         aml_append(crs, aml_interrupt(AML_CONSUMER, AML_LEVEL,
1187                                       AML_ACTIVE_HIGH, AML_SHARED,
1188                                       &irqs, 1));
1189         aml_append(dev, aml_name_decl("_PRS", crs));
1190 
1191         /* The SCI cannot be disabled and is always attached to GSI 9,
1192          * so these are no-ops.  We only need this link to override the
1193          * polarity to active high and match the content of the MADT.
1194          */
1195         method = aml_method("_STA", 0, AML_NOTSERIALIZED);
1196         aml_append(method, aml_return(aml_int(0x0b)));
1197         aml_append(dev, method);
1198 
1199         method = aml_method("_DIS", 0, AML_NOTSERIALIZED);
1200         aml_append(dev, method);
1201 
1202         method = aml_method("_CRS", 0, AML_NOTSERIALIZED);
1203         aml_append(method, aml_return(aml_name("_PRS")));
1204         aml_append(dev, method);
1205 
1206         method = aml_method("_SRS", 1, AML_NOTSERIALIZED);
1207         aml_append(dev, method);
1208     }
1209     aml_append(sb_scope, dev);
1210 
1211     aml_append(table, sb_scope);
1212 }
1213 
1214 static void append_q35_prt_entry(Aml *ctx, uint32_t nr, const char *name)
1215 {
1216     int i;
1217     int head;
1218     Aml *pkg;
1219     char base = name[3] < 'E' ? 'A' : 'E';
1220     char *s = g_strdup(name);
1221     Aml *a_nr = aml_int((nr << 16) | 0xffff);
1222 
1223     assert(strlen(s) == 4);
1224 
1225     head = name[3] - base;
1226     for (i = 0; i < 4; i++) {
1227         if (head + i > 3) {
1228             head = i * -1;
1229         }
1230         s[3] = base + head + i;
1231         pkg = aml_package(4);
1232         aml_append(pkg, a_nr);
1233         aml_append(pkg, aml_int(i));
1234         aml_append(pkg, aml_name("%s", s));
1235         aml_append(pkg, aml_int(0));
1236         aml_append(ctx, pkg);
1237     }
1238     g_free(s);
1239 }
1240 
1241 static Aml *build_q35_routing_table(const char *str)
1242 {
1243     int i;
1244     Aml *pkg;
1245     char *name = g_strdup_printf("%s ", str);
1246 
1247     pkg = aml_package(128);
1248     for (i = 0; i < 0x18; i++) {
1249             name[3] = 'E' + (i & 0x3);
1250             append_q35_prt_entry(pkg, i, name);
1251     }
1252 
1253     name[3] = 'E';
1254     append_q35_prt_entry(pkg, 0x18, name);
1255 
1256     /* INTA -> PIRQA for slot 25 - 31, see the default value of D<N>IR */
1257     for (i = 0x0019; i < 0x1e; i++) {
1258         name[3] = 'A';
1259         append_q35_prt_entry(pkg, i, name);
1260     }
1261 
1262     /* PCIe->PCI bridge. use PIRQ[E-H] */
1263     name[3] = 'E';
1264     append_q35_prt_entry(pkg, 0x1e, name);
1265     name[3] = 'A';
1266     append_q35_prt_entry(pkg, 0x1f, name);
1267 
1268     g_free(name);
1269     return pkg;
1270 }
1271 
1272 static void build_q35_pci0_int(Aml *table)
1273 {
1274     Aml *field;
1275     Aml *method;
1276     Aml *sb_scope = aml_scope("_SB");
1277     Aml *pci0_scope = aml_scope("PCI0");
1278 
1279     /* Zero => PIC mode, One => APIC Mode */
1280     aml_append(table, aml_name_decl("PICF", aml_int(0)));
1281     method = aml_method("_PIC", 1, AML_NOTSERIALIZED);
1282     {
1283         aml_append(method, aml_store(aml_arg(0), aml_name("PICF")));
1284     }
1285     aml_append(table, method);
1286 
1287     aml_append(pci0_scope,
1288         aml_name_decl("PRTP", build_q35_routing_table("LNK")));
1289     aml_append(pci0_scope,
1290         aml_name_decl("PRTA", build_q35_routing_table("GSI")));
1291 
1292     method = aml_method("_PRT", 0, AML_NOTSERIALIZED);
1293     {
1294         Aml *if_ctx;
1295         Aml *else_ctx;
1296 
1297         /* PCI IRQ routing table, example from ACPI 2.0a specification,
1298            section 6.2.8.1 */
1299         /* Note: we provide the same info as the PCI routing
1300            table of the Bochs BIOS */
1301         if_ctx = aml_if(aml_equal(aml_name("PICF"), aml_int(0)));
1302         aml_append(if_ctx, aml_return(aml_name("PRTP")));
1303         aml_append(method, if_ctx);
1304         else_ctx = aml_else();
1305         aml_append(else_ctx, aml_return(aml_name("PRTA")));
1306         aml_append(method, else_ctx);
1307     }
1308     aml_append(pci0_scope, method);
1309     aml_append(sb_scope, pci0_scope);
1310 
1311     field = aml_field("PCI0.ISA.PIRQ", AML_BYTE_ACC, AML_NOLOCK, AML_PRESERVE);
1312     aml_append(field, aml_named_field("PRQA", 8));
1313     aml_append(field, aml_named_field("PRQB", 8));
1314     aml_append(field, aml_named_field("PRQC", 8));
1315     aml_append(field, aml_named_field("PRQD", 8));
1316     aml_append(field, aml_reserved_field(0x20));
1317     aml_append(field, aml_named_field("PRQE", 8));
1318     aml_append(field, aml_named_field("PRQF", 8));
1319     aml_append(field, aml_named_field("PRQG", 8));
1320     aml_append(field, aml_named_field("PRQH", 8));
1321     aml_append(sb_scope, field);
1322 
1323     aml_append(sb_scope, build_irq_status_method());
1324     aml_append(sb_scope, build_iqcr_method(false));
1325 
1326     aml_append(sb_scope, build_link_dev("LNKA", 0, aml_name("PRQA")));
1327     aml_append(sb_scope, build_link_dev("LNKB", 1, aml_name("PRQB")));
1328     aml_append(sb_scope, build_link_dev("LNKC", 2, aml_name("PRQC")));
1329     aml_append(sb_scope, build_link_dev("LNKD", 3, aml_name("PRQD")));
1330     aml_append(sb_scope, build_link_dev("LNKE", 4, aml_name("PRQE")));
1331     aml_append(sb_scope, build_link_dev("LNKF", 5, aml_name("PRQF")));
1332     aml_append(sb_scope, build_link_dev("LNKG", 6, aml_name("PRQG")));
1333     aml_append(sb_scope, build_link_dev("LNKH", 7, aml_name("PRQH")));
1334 
1335     aml_append(sb_scope, build_gsi_link_dev("GSIA", 0x10, 0x10));
1336     aml_append(sb_scope, build_gsi_link_dev("GSIB", 0x11, 0x11));
1337     aml_append(sb_scope, build_gsi_link_dev("GSIC", 0x12, 0x12));
1338     aml_append(sb_scope, build_gsi_link_dev("GSID", 0x13, 0x13));
1339     aml_append(sb_scope, build_gsi_link_dev("GSIE", 0x14, 0x14));
1340     aml_append(sb_scope, build_gsi_link_dev("GSIF", 0x15, 0x15));
1341     aml_append(sb_scope, build_gsi_link_dev("GSIG", 0x16, 0x16));
1342     aml_append(sb_scope, build_gsi_link_dev("GSIH", 0x17, 0x17));
1343 
1344     aml_append(table, sb_scope);
1345 }
1346 
1347 static void build_q35_isa_bridge(Aml *table)
1348 {
1349     Aml *dev;
1350     Aml *scope;
1351 
1352     scope =  aml_scope("_SB.PCI0");
1353     dev = aml_device("ISA");
1354     aml_append(dev, aml_name_decl("_ADR", aml_int(0x001F0000)));
1355 
1356     /* ICH9 PCI to ISA irq remapping */
1357     aml_append(dev, aml_operation_region("PIRQ", AML_PCI_CONFIG,
1358                                          aml_int(0x60), 0x0C));
1359 
1360     aml_append(scope, dev);
1361     aml_append(table, scope);
1362 }
1363 
1364 static void build_piix4_isa_bridge(Aml *table)
1365 {
1366     Aml *dev;
1367     Aml *scope;
1368 
1369     scope =  aml_scope("_SB.PCI0");
1370     dev = aml_device("ISA");
1371     aml_append(dev, aml_name_decl("_ADR", aml_int(0x00010000)));
1372 
1373     /* PIIX PCI to ISA irq remapping */
1374     aml_append(dev, aml_operation_region("P40C", AML_PCI_CONFIG,
1375                                          aml_int(0x60), 0x04));
1376 
1377     aml_append(scope, dev);
1378     aml_append(table, scope);
1379 }
1380 
1381 static void build_piix4_pci_hotplug(Aml *table)
1382 {
1383     Aml *scope;
1384     Aml *field;
1385     Aml *method;
1386 
1387     scope =  aml_scope("_SB.PCI0");
1388 
1389     aml_append(scope,
1390         aml_operation_region("PCST", AML_SYSTEM_IO, aml_int(0xae00), 0x08));
1391     field = aml_field("PCST", AML_DWORD_ACC, AML_NOLOCK, AML_WRITE_AS_ZEROS);
1392     aml_append(field, aml_named_field("PCIU", 32));
1393     aml_append(field, aml_named_field("PCID", 32));
1394     aml_append(scope, field);
1395 
1396     aml_append(scope,
1397         aml_operation_region("SEJ", AML_SYSTEM_IO, aml_int(0xae08), 0x04));
1398     field = aml_field("SEJ", AML_DWORD_ACC, AML_NOLOCK, AML_WRITE_AS_ZEROS);
1399     aml_append(field, aml_named_field("B0EJ", 32));
1400     aml_append(scope, field);
1401 
1402     aml_append(scope,
1403         aml_operation_region("BNMR", AML_SYSTEM_IO, aml_int(0xae10), 0x04));
1404     field = aml_field("BNMR", AML_DWORD_ACC, AML_NOLOCK, AML_WRITE_AS_ZEROS);
1405     aml_append(field, aml_named_field("BNUM", 32));
1406     aml_append(scope, field);
1407 
1408     aml_append(scope, aml_mutex("BLCK", 0));
1409 
1410     method = aml_method("PCEJ", 2, AML_NOTSERIALIZED);
1411     aml_append(method, aml_acquire(aml_name("BLCK"), 0xFFFF));
1412     aml_append(method, aml_store(aml_arg(0), aml_name("BNUM")));
1413     aml_append(method,
1414         aml_store(aml_shiftleft(aml_int(1), aml_arg(1)), aml_name("B0EJ")));
1415     aml_append(method, aml_release(aml_name("BLCK")));
1416     aml_append(method, aml_return(aml_int(0)));
1417     aml_append(scope, method);
1418 
1419     aml_append(table, scope);
1420 }
1421 
1422 static Aml *build_q35_osc_method(void)
1423 {
1424     Aml *if_ctx;
1425     Aml *if_ctx2;
1426     Aml *else_ctx;
1427     Aml *method;
1428     Aml *a_cwd1 = aml_name("CDW1");
1429     Aml *a_ctrl = aml_local(0);
1430 
1431     method = aml_method("_OSC", 4, AML_NOTSERIALIZED);
1432     aml_append(method, aml_create_dword_field(aml_arg(3), aml_int(0), "CDW1"));
1433 
1434     if_ctx = aml_if(aml_equal(
1435         aml_arg(0), aml_touuid("33DB4D5B-1FF7-401C-9657-7441C03DD766")));
1436     aml_append(if_ctx, aml_create_dword_field(aml_arg(3), aml_int(4), "CDW2"));
1437     aml_append(if_ctx, aml_create_dword_field(aml_arg(3), aml_int(8), "CDW3"));
1438 
1439     aml_append(if_ctx, aml_store(aml_name("CDW3"), a_ctrl));
1440 
1441     /*
1442      * Always allow native PME, AER (no dependencies)
1443      * Allow SHPC (PCI bridges can have SHPC controller)
1444      */
1445     aml_append(if_ctx, aml_and(a_ctrl, aml_int(0x1F), a_ctrl));
1446 
1447     if_ctx2 = aml_if(aml_lnot(aml_equal(aml_arg(1), aml_int(1))));
1448     /* Unknown revision */
1449     aml_append(if_ctx2, aml_or(a_cwd1, aml_int(0x08), a_cwd1));
1450     aml_append(if_ctx, if_ctx2);
1451 
1452     if_ctx2 = aml_if(aml_lnot(aml_equal(aml_name("CDW3"), a_ctrl)));
1453     /* Capabilities bits were masked */
1454     aml_append(if_ctx2, aml_or(a_cwd1, aml_int(0x10), a_cwd1));
1455     aml_append(if_ctx, if_ctx2);
1456 
1457     /* Update DWORD3 in the buffer */
1458     aml_append(if_ctx, aml_store(a_ctrl, aml_name("CDW3")));
1459     aml_append(method, if_ctx);
1460 
1461     else_ctx = aml_else();
1462     /* Unrecognized UUID */
1463     aml_append(else_ctx, aml_or(a_cwd1, aml_int(4), a_cwd1));
1464     aml_append(method, else_ctx);
1465 
1466     aml_append(method, aml_return(aml_arg(3)));
1467     return method;
1468 }
1469 
1470 static void build_smb0(Aml *table, I2CBus *smbus, int devnr, int func)
1471 {
1472     Aml *scope = aml_scope("_SB.PCI0");
1473     Aml *dev = aml_device("SMB0");
1474 
1475     aml_append(dev, aml_name_decl("_ADR", aml_int(devnr << 16 | func)));
1476     build_acpi_ipmi_devices(dev, BUS(smbus), "\\_SB.PCI0.SMB0");
1477     aml_append(scope, dev);
1478     aml_append(table, scope);
1479 }
1480 
1481 static void
1482 build_dsdt(GArray *table_data, BIOSLinker *linker,
1483            AcpiPmInfo *pm, AcpiMiscInfo *misc,
1484            Range *pci_hole, Range *pci_hole64, MachineState *machine)
1485 {
1486     CrsRangeEntry *entry;
1487     Aml *dsdt, *sb_scope, *scope, *dev, *method, *field, *pkg, *crs;
1488     CrsRangeSet crs_range_set;
1489     PCMachineState *pcms = PC_MACHINE(machine);
1490     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(machine);
1491     X86MachineState *x86ms = X86_MACHINE(machine);
1492     AcpiMcfgInfo mcfg;
1493     uint32_t nr_mem = machine->ram_slots;
1494     int root_bus_limit = 0xFF;
1495     PCIBus *bus = NULL;
1496     TPMIf *tpm = tpm_find();
1497     int i;
1498     VMBusBridge *vmbus_bridge = vmbus_bridge_find();
1499 
1500     dsdt = init_aml_allocator();
1501 
1502     /* Reserve space for header */
1503     acpi_data_push(dsdt->buf, sizeof(AcpiTableHeader));
1504 
1505     build_dbg_aml(dsdt);
1506     if (misc->is_piix4) {
1507         sb_scope = aml_scope("_SB");
1508         dev = aml_device("PCI0");
1509         aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0A03")));
1510         aml_append(dev, aml_name_decl("_ADR", aml_int(0)));
1511         aml_append(dev, aml_name_decl("_UID", aml_int(0)));
1512         aml_append(sb_scope, dev);
1513         aml_append(dsdt, sb_scope);
1514 
1515         build_hpet_aml(dsdt);
1516         build_piix4_isa_bridge(dsdt);
1517         build_isa_devices_aml(dsdt);
1518         if (pm->pcihp_bridge_en || pm->pcihp_root_en) {
1519             build_piix4_pci_hotplug(dsdt);
1520         }
1521         build_piix4_pci0_int(dsdt);
1522     } else {
1523         sb_scope = aml_scope("_SB");
1524         dev = aml_device("PCI0");
1525         aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0A08")));
1526         aml_append(dev, aml_name_decl("_CID", aml_eisaid("PNP0A03")));
1527         aml_append(dev, aml_name_decl("_ADR", aml_int(0)));
1528         aml_append(dev, aml_name_decl("_UID", aml_int(0)));
1529         aml_append(dev, build_q35_osc_method());
1530         aml_append(sb_scope, dev);
1531 
1532         if (pm->smi_on_cpuhp) {
1533             /* reserve SMI block resources, IO ports 0xB2, 0xB3 */
1534             dev = aml_device("PCI0.SMI0");
1535             aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0A06")));
1536             aml_append(dev, aml_name_decl("_UID", aml_string("SMI resources")));
1537             crs = aml_resource_template();
1538             aml_append(crs,
1539                 aml_io(
1540                        AML_DECODE16,
1541                        ACPI_PORT_SMI_CMD,
1542                        ACPI_PORT_SMI_CMD,
1543                        1,
1544                        2)
1545             );
1546             aml_append(dev, aml_name_decl("_CRS", crs));
1547             aml_append(dev, aml_operation_region("SMIR", AML_SYSTEM_IO,
1548                 aml_int(ACPI_PORT_SMI_CMD), 2));
1549             field = aml_field("SMIR", AML_BYTE_ACC, AML_NOLOCK,
1550                               AML_WRITE_AS_ZEROS);
1551             aml_append(field, aml_named_field("SMIC", 8));
1552             aml_append(field, aml_reserved_field(8));
1553             aml_append(dev, field);
1554             aml_append(sb_scope, dev);
1555         }
1556 
1557         aml_append(dsdt, sb_scope);
1558 
1559         build_hpet_aml(dsdt);
1560         build_q35_isa_bridge(dsdt);
1561         build_isa_devices_aml(dsdt);
1562         build_q35_pci0_int(dsdt);
1563         if (pcms->smbus && !pcmc->do_not_add_smb_acpi) {
1564             build_smb0(dsdt, pcms->smbus, ICH9_SMB_DEV, ICH9_SMB_FUNC);
1565         }
1566     }
1567 
1568     if (vmbus_bridge) {
1569         sb_scope = aml_scope("_SB");
1570         aml_append(sb_scope, build_vmbus_device_aml(vmbus_bridge));
1571         aml_append(dsdt, sb_scope);
1572     }
1573 
1574     if (pcmc->legacy_cpu_hotplug) {
1575         build_legacy_cpu_hotplug_aml(dsdt, machine, pm->cpu_hp_io_base);
1576     } else {
1577         CPUHotplugFeatures opts = {
1578             .acpi_1_compatible = true, .has_legacy_cphp = true,
1579             .smi_path = pm->smi_on_cpuhp ? "\\_SB.PCI0.SMI0.SMIC" : NULL,
1580         };
1581         build_cpus_aml(dsdt, machine, opts, pm->cpu_hp_io_base,
1582                        "\\_SB.PCI0", "\\_GPE._E02");
1583     }
1584 
1585     if (pcms->memhp_io_base && nr_mem) {
1586         build_memory_hotplug_aml(dsdt, nr_mem, "\\_SB.PCI0",
1587                                  "\\_GPE._E03", AML_SYSTEM_IO,
1588                                  pcms->memhp_io_base);
1589     }
1590 
1591     scope =  aml_scope("_GPE");
1592     {
1593         aml_append(scope, aml_name_decl("_HID", aml_string("ACPI0006")));
1594 
1595         if (misc->is_piix4 && (pm->pcihp_bridge_en || pm->pcihp_root_en)) {
1596             method = aml_method("_E01", 0, AML_NOTSERIALIZED);
1597             aml_append(method,
1598                 aml_acquire(aml_name("\\_SB.PCI0.BLCK"), 0xFFFF));
1599             aml_append(method, aml_call0("\\_SB.PCI0.PCNT"));
1600             aml_append(method, aml_release(aml_name("\\_SB.PCI0.BLCK")));
1601             aml_append(scope, method);
1602         }
1603 
1604         if (machine->nvdimms_state->is_enabled) {
1605             method = aml_method("_E04", 0, AML_NOTSERIALIZED);
1606             aml_append(method, aml_notify(aml_name("\\_SB.NVDR"),
1607                                           aml_int(0x80)));
1608             aml_append(scope, method);
1609         }
1610     }
1611     aml_append(dsdt, scope);
1612 
1613     crs_range_set_init(&crs_range_set);
1614     bus = PC_MACHINE(machine)->bus;
1615     if (bus) {
1616         QLIST_FOREACH(bus, &bus->child, sibling) {
1617             uint8_t bus_num = pci_bus_num(bus);
1618             uint8_t numa_node = pci_bus_numa_node(bus);
1619 
1620             /* look only for expander root buses */
1621             if (!pci_bus_is_root(bus)) {
1622                 continue;
1623             }
1624 
1625             if (bus_num < root_bus_limit) {
1626                 root_bus_limit = bus_num - 1;
1627             }
1628 
1629             scope = aml_scope("\\_SB");
1630             dev = aml_device("PC%.02X", bus_num);
1631             aml_append(dev, aml_name_decl("_UID", aml_int(bus_num)));
1632             aml_append(dev, aml_name_decl("_BBN", aml_int(bus_num)));
1633             if (pci_bus_is_express(bus)) {
1634                 aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0A08")));
1635                 aml_append(dev, aml_name_decl("_CID", aml_eisaid("PNP0A03")));
1636                 aml_append(dev, build_q35_osc_method());
1637             } else {
1638                 aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0A03")));
1639             }
1640 
1641             if (numa_node != NUMA_NODE_UNASSIGNED) {
1642                 aml_append(dev, aml_name_decl("_PXM", aml_int(numa_node)));
1643             }
1644 
1645             aml_append(dev, build_prt(false));
1646             crs = build_crs(PCI_HOST_BRIDGE(BUS(bus)->parent), &crs_range_set);
1647             aml_append(dev, aml_name_decl("_CRS", crs));
1648             aml_append(scope, dev);
1649             aml_append(dsdt, scope);
1650         }
1651     }
1652 
1653     /*
1654      * At this point crs_range_set has all the ranges used by pci
1655      * busses *other* than PCI0.  These ranges will be excluded from
1656      * the PCI0._CRS.  Add mmconfig to the set so it will be excluded
1657      * too.
1658      */
1659     if (acpi_get_mcfg(&mcfg)) {
1660         crs_range_insert(crs_range_set.mem_ranges,
1661                          mcfg.base, mcfg.base + mcfg.size - 1);
1662     }
1663 
1664     scope = aml_scope("\\_SB.PCI0");
1665     /* build PCI0._CRS */
1666     crs = aml_resource_template();
1667     aml_append(crs,
1668         aml_word_bus_number(AML_MIN_FIXED, AML_MAX_FIXED, AML_POS_DECODE,
1669                             0x0000, 0x0, root_bus_limit,
1670                             0x0000, root_bus_limit + 1));
1671     aml_append(crs, aml_io(AML_DECODE16, 0x0CF8, 0x0CF8, 0x01, 0x08));
1672 
1673     aml_append(crs,
1674         aml_word_io(AML_MIN_FIXED, AML_MAX_FIXED,
1675                     AML_POS_DECODE, AML_ENTIRE_RANGE,
1676                     0x0000, 0x0000, 0x0CF7, 0x0000, 0x0CF8));
1677 
1678     crs_replace_with_free_ranges(crs_range_set.io_ranges, 0x0D00, 0xFFFF);
1679     for (i = 0; i < crs_range_set.io_ranges->len; i++) {
1680         entry = g_ptr_array_index(crs_range_set.io_ranges, i);
1681         aml_append(crs,
1682             aml_word_io(AML_MIN_FIXED, AML_MAX_FIXED,
1683                         AML_POS_DECODE, AML_ENTIRE_RANGE,
1684                         0x0000, entry->base, entry->limit,
1685                         0x0000, entry->limit - entry->base + 1));
1686     }
1687 
1688     aml_append(crs,
1689         aml_dword_memory(AML_POS_DECODE, AML_MIN_FIXED, AML_MAX_FIXED,
1690                          AML_CACHEABLE, AML_READ_WRITE,
1691                          0, 0x000A0000, 0x000BFFFF, 0, 0x00020000));
1692 
1693     crs_replace_with_free_ranges(crs_range_set.mem_ranges,
1694                                  range_lob(pci_hole),
1695                                  range_upb(pci_hole));
1696     for (i = 0; i < crs_range_set.mem_ranges->len; i++) {
1697         entry = g_ptr_array_index(crs_range_set.mem_ranges, i);
1698         aml_append(crs,
1699             aml_dword_memory(AML_POS_DECODE, AML_MIN_FIXED, AML_MAX_FIXED,
1700                              AML_NON_CACHEABLE, AML_READ_WRITE,
1701                              0, entry->base, entry->limit,
1702                              0, entry->limit - entry->base + 1));
1703     }
1704 
1705     if (!range_is_empty(pci_hole64)) {
1706         crs_replace_with_free_ranges(crs_range_set.mem_64bit_ranges,
1707                                      range_lob(pci_hole64),
1708                                      range_upb(pci_hole64));
1709         for (i = 0; i < crs_range_set.mem_64bit_ranges->len; i++) {
1710             entry = g_ptr_array_index(crs_range_set.mem_64bit_ranges, i);
1711             aml_append(crs,
1712                        aml_qword_memory(AML_POS_DECODE, AML_MIN_FIXED,
1713                                         AML_MAX_FIXED,
1714                                         AML_CACHEABLE, AML_READ_WRITE,
1715                                         0, entry->base, entry->limit,
1716                                         0, entry->limit - entry->base + 1));
1717         }
1718     }
1719 
1720     if (TPM_IS_TIS_ISA(tpm_find())) {
1721         aml_append(crs, aml_memory32_fixed(TPM_TIS_ADDR_BASE,
1722                    TPM_TIS_ADDR_SIZE, AML_READ_WRITE));
1723     }
1724     aml_append(scope, aml_name_decl("_CRS", crs));
1725 
1726     /* reserve GPE0 block resources */
1727     dev = aml_device("GPE0");
1728     aml_append(dev, aml_name_decl("_HID", aml_string("PNP0A06")));
1729     aml_append(dev, aml_name_decl("_UID", aml_string("GPE0 resources")));
1730     /* device present, functioning, decoding, not shown in UI */
1731     aml_append(dev, aml_name_decl("_STA", aml_int(0xB)));
1732     crs = aml_resource_template();
1733     aml_append(crs,
1734         aml_io(
1735                AML_DECODE16,
1736                pm->fadt.gpe0_blk.address,
1737                pm->fadt.gpe0_blk.address,
1738                1,
1739                pm->fadt.gpe0_blk.bit_width / 8)
1740     );
1741     aml_append(dev, aml_name_decl("_CRS", crs));
1742     aml_append(scope, dev);
1743 
1744     crs_range_set_free(&crs_range_set);
1745 
1746     /* reserve PCIHP resources */
1747     if (pm->pcihp_io_len && (pm->pcihp_bridge_en || pm->pcihp_root_en)) {
1748         dev = aml_device("PHPR");
1749         aml_append(dev, aml_name_decl("_HID", aml_string("PNP0A06")));
1750         aml_append(dev,
1751             aml_name_decl("_UID", aml_string("PCI Hotplug resources")));
1752         /* device present, functioning, decoding, not shown in UI */
1753         aml_append(dev, aml_name_decl("_STA", aml_int(0xB)));
1754         crs = aml_resource_template();
1755         aml_append(crs,
1756             aml_io(AML_DECODE16, pm->pcihp_io_base, pm->pcihp_io_base, 1,
1757                    pm->pcihp_io_len)
1758         );
1759         aml_append(dev, aml_name_decl("_CRS", crs));
1760         aml_append(scope, dev);
1761     }
1762     aml_append(dsdt, scope);
1763 
1764     /*  create S3_ / S4_ / S5_ packages if necessary */
1765     scope = aml_scope("\\");
1766     if (!pm->s3_disabled) {
1767         pkg = aml_package(4);
1768         aml_append(pkg, aml_int(1)); /* PM1a_CNT.SLP_TYP */
1769         aml_append(pkg, aml_int(1)); /* PM1b_CNT.SLP_TYP, FIXME: not impl. */
1770         aml_append(pkg, aml_int(0)); /* reserved */
1771         aml_append(pkg, aml_int(0)); /* reserved */
1772         aml_append(scope, aml_name_decl("_S3", pkg));
1773     }
1774 
1775     if (!pm->s4_disabled) {
1776         pkg = aml_package(4);
1777         aml_append(pkg, aml_int(pm->s4_val)); /* PM1a_CNT.SLP_TYP */
1778         /* PM1b_CNT.SLP_TYP, FIXME: not impl. */
1779         aml_append(pkg, aml_int(pm->s4_val));
1780         aml_append(pkg, aml_int(0)); /* reserved */
1781         aml_append(pkg, aml_int(0)); /* reserved */
1782         aml_append(scope, aml_name_decl("_S4", pkg));
1783     }
1784 
1785     pkg = aml_package(4);
1786     aml_append(pkg, aml_int(0)); /* PM1a_CNT.SLP_TYP */
1787     aml_append(pkg, aml_int(0)); /* PM1b_CNT.SLP_TYP not impl. */
1788     aml_append(pkg, aml_int(0)); /* reserved */
1789     aml_append(pkg, aml_int(0)); /* reserved */
1790     aml_append(scope, aml_name_decl("_S5", pkg));
1791     aml_append(dsdt, scope);
1792 
1793     /* create fw_cfg node, unconditionally */
1794     {
1795         scope = aml_scope("\\_SB.PCI0");
1796         fw_cfg_add_acpi_dsdt(scope, x86ms->fw_cfg);
1797         aml_append(dsdt, scope);
1798     }
1799 
1800     if (misc->applesmc_io_base) {
1801         scope = aml_scope("\\_SB.PCI0.ISA");
1802         dev = aml_device("SMC");
1803 
1804         aml_append(dev, aml_name_decl("_HID", aml_eisaid("APP0001")));
1805         /* device present, functioning, decoding, not shown in UI */
1806         aml_append(dev, aml_name_decl("_STA", aml_int(0xB)));
1807 
1808         crs = aml_resource_template();
1809         aml_append(crs,
1810             aml_io(AML_DECODE16, misc->applesmc_io_base, misc->applesmc_io_base,
1811                    0x01, APPLESMC_MAX_DATA_LENGTH)
1812         );
1813         aml_append(crs, aml_irq_no_flags(6));
1814         aml_append(dev, aml_name_decl("_CRS", crs));
1815 
1816         aml_append(scope, dev);
1817         aml_append(dsdt, scope);
1818     }
1819 
1820     if (misc->pvpanic_port) {
1821         scope = aml_scope("\\_SB.PCI0.ISA");
1822 
1823         dev = aml_device("PEVT");
1824         aml_append(dev, aml_name_decl("_HID", aml_string("QEMU0001")));
1825 
1826         crs = aml_resource_template();
1827         aml_append(crs,
1828             aml_io(AML_DECODE16, misc->pvpanic_port, misc->pvpanic_port, 1, 1)
1829         );
1830         aml_append(dev, aml_name_decl("_CRS", crs));
1831 
1832         aml_append(dev, aml_operation_region("PEOR", AML_SYSTEM_IO,
1833                                               aml_int(misc->pvpanic_port), 1));
1834         field = aml_field("PEOR", AML_BYTE_ACC, AML_NOLOCK, AML_PRESERVE);
1835         aml_append(field, aml_named_field("PEPT", 8));
1836         aml_append(dev, field);
1837 
1838         /* device present, functioning, decoding, shown in UI */
1839         aml_append(dev, aml_name_decl("_STA", aml_int(0xF)));
1840 
1841         method = aml_method("RDPT", 0, AML_NOTSERIALIZED);
1842         aml_append(method, aml_store(aml_name("PEPT"), aml_local(0)));
1843         aml_append(method, aml_return(aml_local(0)));
1844         aml_append(dev, method);
1845 
1846         method = aml_method("WRPT", 1, AML_NOTSERIALIZED);
1847         aml_append(method, aml_store(aml_arg(0), aml_name("PEPT")));
1848         aml_append(dev, method);
1849 
1850         aml_append(scope, dev);
1851         aml_append(dsdt, scope);
1852     }
1853 
1854     sb_scope = aml_scope("\\_SB");
1855     {
1856         Object *pci_host;
1857         PCIBus *bus = NULL;
1858 
1859         pci_host = acpi_get_i386_pci_host();
1860         if (pci_host) {
1861             bus = PCI_HOST_BRIDGE(pci_host)->bus;
1862         }
1863 
1864         if (bus) {
1865             Aml *scope = aml_scope("PCI0");
1866             /* Scan all PCI buses. Generate tables to support hotplug. */
1867             build_append_pci_bus_devices(scope, bus, pm->pcihp_bridge_en);
1868 
1869             if (TPM_IS_TIS_ISA(tpm)) {
1870                 if (misc->tpm_version == TPM_VERSION_2_0) {
1871                     dev = aml_device("TPM");
1872                     aml_append(dev, aml_name_decl("_HID",
1873                                                   aml_string("MSFT0101")));
1874                 } else {
1875                     dev = aml_device("ISA.TPM");
1876                     aml_append(dev, aml_name_decl("_HID",
1877                                                   aml_eisaid("PNP0C31")));
1878                 }
1879 
1880                 aml_append(dev, aml_name_decl("_STA", aml_int(0xF)));
1881                 crs = aml_resource_template();
1882                 aml_append(crs, aml_memory32_fixed(TPM_TIS_ADDR_BASE,
1883                            TPM_TIS_ADDR_SIZE, AML_READ_WRITE));
1884                 /*
1885                     FIXME: TPM_TIS_IRQ=5 conflicts with PNP0C0F irqs,
1886                     Rewrite to take IRQ from TPM device model and
1887                     fix default IRQ value there to use some unused IRQ
1888                  */
1889                 /* aml_append(crs, aml_irq_no_flags(TPM_TIS_IRQ)); */
1890                 aml_append(dev, aml_name_decl("_CRS", crs));
1891 
1892                 tpm_build_ppi_acpi(tpm, dev);
1893 
1894                 aml_append(scope, dev);
1895             }
1896 
1897             aml_append(sb_scope, scope);
1898         }
1899     }
1900 
1901     if (TPM_IS_CRB(tpm)) {
1902         dev = aml_device("TPM");
1903         aml_append(dev, aml_name_decl("_HID", aml_string("MSFT0101")));
1904         crs = aml_resource_template();
1905         aml_append(crs, aml_memory32_fixed(TPM_CRB_ADDR_BASE,
1906                                            TPM_CRB_ADDR_SIZE, AML_READ_WRITE));
1907         aml_append(dev, aml_name_decl("_CRS", crs));
1908 
1909         aml_append(dev, aml_name_decl("_STA", aml_int(0xf)));
1910 
1911         tpm_build_ppi_acpi(tpm, dev);
1912 
1913         aml_append(sb_scope, dev);
1914     }
1915 
1916     aml_append(dsdt, sb_scope);
1917 
1918     /* copy AML table into ACPI tables blob and patch header there */
1919     g_array_append_vals(table_data, dsdt->buf->data, dsdt->buf->len);
1920     build_header(linker, table_data,
1921         (void *)(table_data->data + table_data->len - dsdt->buf->len),
1922         "DSDT", dsdt->buf->len, 1, NULL, NULL);
1923     free_aml_allocator();
1924 }
1925 
1926 static void
1927 build_hpet(GArray *table_data, BIOSLinker *linker)
1928 {
1929     Acpi20Hpet *hpet;
1930 
1931     hpet = acpi_data_push(table_data, sizeof(*hpet));
1932     /* Note timer_block_id value must be kept in sync with value advertised by
1933      * emulated hpet
1934      */
1935     hpet->timer_block_id = cpu_to_le32(0x8086a201);
1936     hpet->addr.address = cpu_to_le64(HPET_BASE);
1937     build_header(linker, table_data,
1938                  (void *)hpet, "HPET", sizeof(*hpet), 1, NULL, NULL);
1939 }
1940 
1941 static void
1942 build_tpm_tcpa(GArray *table_data, BIOSLinker *linker, GArray *tcpalog)
1943 {
1944     Acpi20Tcpa *tcpa = acpi_data_push(table_data, sizeof *tcpa);
1945     unsigned log_addr_size = sizeof(tcpa->log_area_start_address);
1946     unsigned log_addr_offset =
1947         (char *)&tcpa->log_area_start_address - table_data->data;
1948 
1949     tcpa->platform_class = cpu_to_le16(TPM_TCPA_ACPI_CLASS_CLIENT);
1950     tcpa->log_area_minimum_length = cpu_to_le32(TPM_LOG_AREA_MINIMUM_SIZE);
1951     acpi_data_push(tcpalog, le32_to_cpu(tcpa->log_area_minimum_length));
1952 
1953     bios_linker_loader_alloc(linker, ACPI_BUILD_TPMLOG_FILE, tcpalog, 1,
1954                              false /* high memory */);
1955 
1956     /* log area start address to be filled by Guest linker */
1957     bios_linker_loader_add_pointer(linker,
1958         ACPI_BUILD_TABLE_FILE, log_addr_offset, log_addr_size,
1959         ACPI_BUILD_TPMLOG_FILE, 0);
1960 
1961     build_header(linker, table_data,
1962                  (void *)tcpa, "TCPA", sizeof(*tcpa), 2, NULL, NULL);
1963 }
1964 
1965 #define HOLE_640K_START  (640 * KiB)
1966 #define HOLE_640K_END   (1 * MiB)
1967 
1968 static void
1969 build_srat(GArray *table_data, BIOSLinker *linker, MachineState *machine)
1970 {
1971     AcpiSystemResourceAffinityTable *srat;
1972     AcpiSratMemoryAffinity *numamem;
1973 
1974     int i;
1975     int srat_start, numa_start, slots;
1976     uint64_t mem_len, mem_base, next_base;
1977     MachineClass *mc = MACHINE_GET_CLASS(machine);
1978     X86MachineState *x86ms = X86_MACHINE(machine);
1979     const CPUArchIdList *apic_ids = mc->possible_cpu_arch_ids(machine);
1980     PCMachineState *pcms = PC_MACHINE(machine);
1981     ram_addr_t hotplugabble_address_space_size =
1982         object_property_get_int(OBJECT(pcms), PC_MACHINE_DEVMEM_REGION_SIZE,
1983                                 NULL);
1984 
1985     srat_start = table_data->len;
1986 
1987     srat = acpi_data_push(table_data, sizeof *srat);
1988     srat->reserved1 = cpu_to_le32(1);
1989 
1990     for (i = 0; i < apic_ids->len; i++) {
1991         int node_id = apic_ids->cpus[i].props.node_id;
1992         uint32_t apic_id = apic_ids->cpus[i].arch_id;
1993 
1994         if (apic_id < 255) {
1995             AcpiSratProcessorAffinity *core;
1996 
1997             core = acpi_data_push(table_data, sizeof *core);
1998             core->type = ACPI_SRAT_PROCESSOR_APIC;
1999             core->length = sizeof(*core);
2000             core->local_apic_id = apic_id;
2001             core->proximity_lo = node_id;
2002             memset(core->proximity_hi, 0, 3);
2003             core->local_sapic_eid = 0;
2004             core->flags = cpu_to_le32(1);
2005         } else {
2006             AcpiSratProcessorX2ApicAffinity *core;
2007 
2008             core = acpi_data_push(table_data, sizeof *core);
2009             core->type = ACPI_SRAT_PROCESSOR_x2APIC;
2010             core->length = sizeof(*core);
2011             core->x2apic_id = cpu_to_le32(apic_id);
2012             core->proximity_domain = cpu_to_le32(node_id);
2013             core->flags = cpu_to_le32(1);
2014         }
2015     }
2016 
2017 
2018     /* the memory map is a bit tricky, it contains at least one hole
2019      * from 640k-1M and possibly another one from 3.5G-4G.
2020      */
2021     next_base = 0;
2022     numa_start = table_data->len;
2023 
2024     for (i = 1; i < pcms->numa_nodes + 1; ++i) {
2025         mem_base = next_base;
2026         mem_len = pcms->node_mem[i - 1];
2027         next_base = mem_base + mem_len;
2028 
2029         /* Cut out the 640K hole */
2030         if (mem_base <= HOLE_640K_START &&
2031             next_base > HOLE_640K_START) {
2032             mem_len -= next_base - HOLE_640K_START;
2033             if (mem_len > 0) {
2034                 numamem = acpi_data_push(table_data, sizeof *numamem);
2035                 build_srat_memory(numamem, mem_base, mem_len, i - 1,
2036                                   MEM_AFFINITY_ENABLED);
2037             }
2038 
2039             /* Check for the rare case: 640K < RAM < 1M */
2040             if (next_base <= HOLE_640K_END) {
2041                 next_base = HOLE_640K_END;
2042                 continue;
2043             }
2044             mem_base = HOLE_640K_END;
2045             mem_len = next_base - HOLE_640K_END;
2046         }
2047 
2048         /* Cut out the ACPI_PCI hole */
2049         if (mem_base <= x86ms->below_4g_mem_size &&
2050             next_base > x86ms->below_4g_mem_size) {
2051             mem_len -= next_base - x86ms->below_4g_mem_size;
2052             if (mem_len > 0) {
2053                 numamem = acpi_data_push(table_data, sizeof *numamem);
2054                 build_srat_memory(numamem, mem_base, mem_len, i - 1,
2055                                   MEM_AFFINITY_ENABLED);
2056             }
2057             mem_base = 1ULL << 32;
2058             mem_len = next_base - x86ms->below_4g_mem_size;
2059             next_base = mem_base + mem_len;
2060         }
2061 
2062         if (mem_len > 0) {
2063             numamem = acpi_data_push(table_data, sizeof *numamem);
2064             build_srat_memory(numamem, mem_base, mem_len, i - 1,
2065                               MEM_AFFINITY_ENABLED);
2066         }
2067     }
2068 
2069     if (machine->nvdimms_state->is_enabled) {
2070         nvdimm_build_srat(table_data);
2071     }
2072 
2073     slots = (table_data->len - numa_start) / sizeof *numamem;
2074     for (; slots < pcms->numa_nodes + 2; slots++) {
2075         numamem = acpi_data_push(table_data, sizeof *numamem);
2076         build_srat_memory(numamem, 0, 0, 0, MEM_AFFINITY_NOFLAGS);
2077     }
2078 
2079     /*
2080      * Entry is required for Windows to enable memory hotplug in OS
2081      * and for Linux to enable SWIOTLB when booted with less than
2082      * 4G of RAM. Windows works better if the entry sets proximity
2083      * to the highest NUMA node in the machine.
2084      * Memory devices may override proximity set by this entry,
2085      * providing _PXM method if necessary.
2086      */
2087     if (hotplugabble_address_space_size) {
2088         numamem = acpi_data_push(table_data, sizeof *numamem);
2089         build_srat_memory(numamem, machine->device_memory->base,
2090                           hotplugabble_address_space_size, pcms->numa_nodes - 1,
2091                           MEM_AFFINITY_HOTPLUGGABLE | MEM_AFFINITY_ENABLED);
2092     }
2093 
2094     build_header(linker, table_data,
2095                  (void *)(table_data->data + srat_start),
2096                  "SRAT",
2097                  table_data->len - srat_start, 1, NULL, NULL);
2098 }
2099 
2100 /*
2101  * VT-d spec 8.1 DMA Remapping Reporting Structure
2102  * (version Oct. 2014 or later)
2103  */
2104 static void
2105 build_dmar_q35(GArray *table_data, BIOSLinker *linker)
2106 {
2107     int dmar_start = table_data->len;
2108 
2109     AcpiTableDmar *dmar;
2110     AcpiDmarHardwareUnit *drhd;
2111     AcpiDmarRootPortATS *atsr;
2112     uint8_t dmar_flags = 0;
2113     X86IOMMUState *iommu = x86_iommu_get_default();
2114     AcpiDmarDeviceScope *scope = NULL;
2115     /* Root complex IOAPIC use one path[0] only */
2116     size_t ioapic_scope_size = sizeof(*scope) + sizeof(scope->path[0]);
2117     IntelIOMMUState *intel_iommu = INTEL_IOMMU_DEVICE(iommu);
2118 
2119     assert(iommu);
2120     if (x86_iommu_ir_supported(iommu)) {
2121         dmar_flags |= 0x1;      /* Flags: 0x1: INT_REMAP */
2122     }
2123 
2124     dmar = acpi_data_push(table_data, sizeof(*dmar));
2125     dmar->host_address_width = intel_iommu->aw_bits - 1;
2126     dmar->flags = dmar_flags;
2127 
2128     /* DMAR Remapping Hardware Unit Definition structure */
2129     drhd = acpi_data_push(table_data, sizeof(*drhd) + ioapic_scope_size);
2130     drhd->type = cpu_to_le16(ACPI_DMAR_TYPE_HARDWARE_UNIT);
2131     drhd->length = cpu_to_le16(sizeof(*drhd) + ioapic_scope_size);
2132     drhd->flags = ACPI_DMAR_INCLUDE_PCI_ALL;
2133     drhd->pci_segment = cpu_to_le16(0);
2134     drhd->address = cpu_to_le64(Q35_HOST_BRIDGE_IOMMU_ADDR);
2135 
2136     /* Scope definition for the root-complex IOAPIC. See VT-d spec
2137      * 8.3.1 (version Oct. 2014 or later). */
2138     scope = &drhd->scope[0];
2139     scope->entry_type = 0x03;   /* Type: 0x03 for IOAPIC */
2140     scope->length = ioapic_scope_size;
2141     scope->enumeration_id = ACPI_BUILD_IOAPIC_ID;
2142     scope->bus = Q35_PSEUDO_BUS_PLATFORM;
2143     scope->path[0].device = PCI_SLOT(Q35_PSEUDO_DEVFN_IOAPIC);
2144     scope->path[0].function = PCI_FUNC(Q35_PSEUDO_DEVFN_IOAPIC);
2145 
2146     if (iommu->dt_supported) {
2147         atsr = acpi_data_push(table_data, sizeof(*atsr));
2148         atsr->type = cpu_to_le16(ACPI_DMAR_TYPE_ATSR);
2149         atsr->length = cpu_to_le16(sizeof(*atsr));
2150         atsr->flags = ACPI_DMAR_ATSR_ALL_PORTS;
2151         atsr->pci_segment = cpu_to_le16(0);
2152     }
2153 
2154     build_header(linker, table_data, (void *)(table_data->data + dmar_start),
2155                  "DMAR", table_data->len - dmar_start, 1, NULL, NULL);
2156 }
2157 
2158 /*
2159  * Windows ACPI Emulated Devices Table
2160  * (Version 1.0 - April 6, 2009)
2161  * Spec: http://download.microsoft.com/download/7/E/7/7E7662CF-CBEA-470B-A97E-CE7CE0D98DC2/WAET.docx
2162  *
2163  * Helpful to speedup Windows guests and ignored by others.
2164  */
2165 static void
2166 build_waet(GArray *table_data, BIOSLinker *linker)
2167 {
2168     int waet_start = table_data->len;
2169 
2170     /* WAET header */
2171     acpi_data_push(table_data, sizeof(AcpiTableHeader));
2172     /*
2173      * Set "ACPI PM timer good" flag.
2174      *
2175      * Tells Windows guests that our ACPI PM timer is reliable in the
2176      * sense that guest can read it only once to obtain a reliable value.
2177      * Which avoids costly VMExits caused by guest re-reading it unnecessarily.
2178      */
2179     build_append_int_noprefix(table_data, 1 << 1 /* ACPI PM timer good */, 4);
2180 
2181     build_header(linker, table_data, (void *)(table_data->data + waet_start),
2182                  "WAET", table_data->len - waet_start, 1, NULL, NULL);
2183 }
2184 
2185 /*
2186  *   IVRS table as specified in AMD IOMMU Specification v2.62, Section 5.2
2187  *   accessible here http://support.amd.com/TechDocs/48882_IOMMU.pdf
2188  */
2189 #define IOAPIC_SB_DEVID   (uint64_t)PCI_BUILD_BDF(0, PCI_DEVFN(0x14, 0))
2190 
2191 /*
2192  * Insert IVHD entry for device and recurse, insert alias, or insert range as
2193  * necessary for the PCI topology.
2194  */
2195 static void
2196 insert_ivhd(PCIBus *bus, PCIDevice *dev, void *opaque)
2197 {
2198     GArray *table_data = opaque;
2199     uint32_t entry;
2200 
2201     /* "Select" IVHD entry, type 0x2 */
2202     entry = PCI_BUILD_BDF(pci_bus_num(bus), dev->devfn) << 8 | 0x2;
2203     build_append_int_noprefix(table_data, entry, 4);
2204 
2205     if (object_dynamic_cast(OBJECT(dev), TYPE_PCI_BRIDGE)) {
2206         PCIBus *sec_bus = pci_bridge_get_sec_bus(PCI_BRIDGE(dev));
2207         uint8_t sec = pci_bus_num(sec_bus);
2208         uint8_t sub = dev->config[PCI_SUBORDINATE_BUS];
2209 
2210         if (pci_bus_is_express(sec_bus)) {
2211             /*
2212              * Walk the bus if there are subordinates, otherwise use a range
2213              * to cover an entire leaf bus.  We could potentially also use a
2214              * range for traversed buses, but we'd need to take care not to
2215              * create both Select and Range entries covering the same device.
2216              * This is easier and potentially more compact.
2217              *
2218              * An example bare metal system seems to use Select entries for
2219              * root ports without a slot (ie. built-ins) and Range entries
2220              * when there is a slot.  The same system also only hard-codes
2221              * the alias range for an onboard PCIe-to-PCI bridge, apparently
2222              * making no effort to support nested bridges.  We attempt to
2223              * be more thorough here.
2224              */
2225             if (sec == sub) { /* leaf bus */
2226                 /* "Start of Range" IVHD entry, type 0x3 */
2227                 entry = PCI_BUILD_BDF(sec, PCI_DEVFN(0, 0)) << 8 | 0x3;
2228                 build_append_int_noprefix(table_data, entry, 4);
2229                 /* "End of Range" IVHD entry, type 0x4 */
2230                 entry = PCI_BUILD_BDF(sub, PCI_DEVFN(31, 7)) << 8 | 0x4;
2231                 build_append_int_noprefix(table_data, entry, 4);
2232             } else {
2233                 pci_for_each_device(sec_bus, sec, insert_ivhd, table_data);
2234             }
2235         } else {
2236             /*
2237              * If the secondary bus is conventional, then we need to create an
2238              * Alias range for everything downstream.  The range covers the
2239              * first devfn on the secondary bus to the last devfn on the
2240              * subordinate bus.  The alias target depends on legacy versus
2241              * express bridges, just as in pci_device_iommu_address_space().
2242              * DeviceIDa vs DeviceIDb as per the AMD IOMMU spec.
2243              */
2244             uint16_t dev_id_a, dev_id_b;
2245 
2246             dev_id_a = PCI_BUILD_BDF(sec, PCI_DEVFN(0, 0));
2247 
2248             if (pci_is_express(dev) &&
2249                 pcie_cap_get_type(dev) == PCI_EXP_TYPE_PCI_BRIDGE) {
2250                 dev_id_b = dev_id_a;
2251             } else {
2252                 dev_id_b = PCI_BUILD_BDF(pci_bus_num(bus), dev->devfn);
2253             }
2254 
2255             /* "Alias Start of Range" IVHD entry, type 0x43, 8 bytes */
2256             build_append_int_noprefix(table_data, dev_id_a << 8 | 0x43, 4);
2257             build_append_int_noprefix(table_data, dev_id_b << 8 | 0x0, 4);
2258 
2259             /* "End of Range" IVHD entry, type 0x4 */
2260             entry = PCI_BUILD_BDF(sub, PCI_DEVFN(31, 7)) << 8 | 0x4;
2261             build_append_int_noprefix(table_data, entry, 4);
2262         }
2263     }
2264 }
2265 
2266 /* For all PCI host bridges, walk and insert IVHD entries */
2267 static int
2268 ivrs_host_bridges(Object *obj, void *opaque)
2269 {
2270     GArray *ivhd_blob = opaque;
2271 
2272     if (object_dynamic_cast(obj, TYPE_PCI_HOST_BRIDGE)) {
2273         PCIBus *bus = PCI_HOST_BRIDGE(obj)->bus;
2274 
2275         if (bus) {
2276             pci_for_each_device(bus, pci_bus_num(bus), insert_ivhd, ivhd_blob);
2277         }
2278     }
2279 
2280     return 0;
2281 }
2282 
2283 static void
2284 build_amd_iommu(GArray *table_data, BIOSLinker *linker)
2285 {
2286     int ivhd_table_len = 24;
2287     int iommu_start = table_data->len;
2288     AMDVIState *s = AMD_IOMMU_DEVICE(x86_iommu_get_default());
2289     GArray *ivhd_blob = g_array_new(false, true, 1);
2290 
2291     /* IVRS header */
2292     acpi_data_push(table_data, sizeof(AcpiTableHeader));
2293     /* IVinfo - IO virtualization information common to all
2294      * IOMMU units in a system
2295      */
2296     build_append_int_noprefix(table_data, 40UL << 8/* PASize */, 4);
2297     /* reserved */
2298     build_append_int_noprefix(table_data, 0, 8);
2299 
2300     /* IVHD definition - type 10h */
2301     build_append_int_noprefix(table_data, 0x10, 1);
2302     /* virtualization flags */
2303     build_append_int_noprefix(table_data,
2304                              (1UL << 0) | /* HtTunEn      */
2305                              (1UL << 4) | /* iotblSup     */
2306                              (1UL << 6) | /* PrefSup      */
2307                              (1UL << 7),  /* PPRSup       */
2308                              1);
2309 
2310     /*
2311      * A PCI bus walk, for each PCI host bridge, is necessary to create a
2312      * complete set of IVHD entries.  Do this into a separate blob so that we
2313      * can calculate the total IVRS table length here and then append the new
2314      * blob further below.  Fall back to an entry covering all devices, which
2315      * is sufficient when no aliases are present.
2316      */
2317     object_child_foreach_recursive(object_get_root(),
2318                                    ivrs_host_bridges, ivhd_blob);
2319 
2320     if (!ivhd_blob->len) {
2321         /*
2322          *   Type 1 device entry reporting all devices
2323          *   These are 4-byte device entries currently reporting the range of
2324          *   Refer to Spec - Table 95:IVHD Device Entry Type Codes(4-byte)
2325          */
2326         build_append_int_noprefix(ivhd_blob, 0x0000001, 4);
2327     }
2328 
2329     ivhd_table_len += ivhd_blob->len;
2330 
2331     /*
2332      * When interrupt remapping is supported, we add a special IVHD device
2333      * for type IO-APIC.
2334      */
2335     if (x86_iommu_ir_supported(x86_iommu_get_default())) {
2336         ivhd_table_len += 8;
2337     }
2338 
2339     /* IVHD length */
2340     build_append_int_noprefix(table_data, ivhd_table_len, 2);
2341     /* DeviceID */
2342     build_append_int_noprefix(table_data, s->devid, 2);
2343     /* Capability offset */
2344     build_append_int_noprefix(table_data, s->capab_offset, 2);
2345     /* IOMMU base address */
2346     build_append_int_noprefix(table_data, s->mmio.addr, 8);
2347     /* PCI Segment Group */
2348     build_append_int_noprefix(table_data, 0, 2);
2349     /* IOMMU info */
2350     build_append_int_noprefix(table_data, 0, 2);
2351     /* IOMMU Feature Reporting */
2352     build_append_int_noprefix(table_data,
2353                              (48UL << 30) | /* HATS   */
2354                              (48UL << 28) | /* GATS   */
2355                              (1UL << 2)   | /* GTSup  */
2356                              (1UL << 6),    /* GASup  */
2357                              4);
2358 
2359     /* IVHD entries as found above */
2360     g_array_append_vals(table_data, ivhd_blob->data, ivhd_blob->len);
2361     g_array_free(ivhd_blob, TRUE);
2362 
2363     /*
2364      * Add a special IVHD device type.
2365      * Refer to spec - Table 95: IVHD device entry type codes
2366      *
2367      * Linux IOMMU driver checks for the special IVHD device (type IO-APIC).
2368      * See Linux kernel commit 'c2ff5cf5294bcbd7fa50f7d860e90a66db7e5059'
2369      */
2370     if (x86_iommu_ir_supported(x86_iommu_get_default())) {
2371         build_append_int_noprefix(table_data,
2372                                  (0x1ull << 56) |           /* type IOAPIC */
2373                                  (IOAPIC_SB_DEVID << 40) |  /* IOAPIC devid */
2374                                  0x48,                      /* special device */
2375                                  8);
2376     }
2377 
2378     build_header(linker, table_data, (void *)(table_data->data + iommu_start),
2379                  "IVRS", table_data->len - iommu_start, 1, NULL, NULL);
2380 }
2381 
2382 typedef
2383 struct AcpiBuildState {
2384     /* Copy of table in RAM (for patching). */
2385     MemoryRegion *table_mr;
2386     /* Is table patched? */
2387     uint8_t patched;
2388     void *rsdp;
2389     MemoryRegion *rsdp_mr;
2390     MemoryRegion *linker_mr;
2391 } AcpiBuildState;
2392 
2393 static bool acpi_get_mcfg(AcpiMcfgInfo *mcfg)
2394 {
2395     Object *pci_host;
2396     QObject *o;
2397 
2398     pci_host = acpi_get_i386_pci_host();
2399     g_assert(pci_host);
2400 
2401     o = object_property_get_qobject(pci_host, PCIE_HOST_MCFG_BASE, NULL);
2402     if (!o) {
2403         return false;
2404     }
2405     mcfg->base = qnum_get_uint(qobject_to(QNum, o));
2406     qobject_unref(o);
2407     if (mcfg->base == PCIE_BASE_ADDR_UNMAPPED) {
2408         return false;
2409     }
2410 
2411     o = object_property_get_qobject(pci_host, PCIE_HOST_MCFG_SIZE, NULL);
2412     assert(o);
2413     mcfg->size = qnum_get_uint(qobject_to(QNum, o));
2414     qobject_unref(o);
2415     return true;
2416 }
2417 
2418 static
2419 void acpi_build(AcpiBuildTables *tables, MachineState *machine)
2420 {
2421     PCMachineState *pcms = PC_MACHINE(machine);
2422     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
2423     X86MachineState *x86ms = X86_MACHINE(machine);
2424     GArray *table_offsets;
2425     unsigned facs, dsdt, rsdt, fadt;
2426     AcpiPmInfo pm;
2427     AcpiMiscInfo misc;
2428     AcpiMcfgInfo mcfg;
2429     Range pci_hole, pci_hole64;
2430     uint8_t *u;
2431     size_t aml_len = 0;
2432     GArray *tables_blob = tables->table_data;
2433     AcpiSlicOem slic_oem = { .id = NULL, .table_id = NULL };
2434     Object *vmgenid_dev;
2435 
2436     acpi_get_pm_info(machine, &pm);
2437     acpi_get_misc_info(&misc);
2438     acpi_get_pci_holes(&pci_hole, &pci_hole64);
2439     acpi_get_slic_oem(&slic_oem);
2440 
2441     table_offsets = g_array_new(false, true /* clear */,
2442                                         sizeof(uint32_t));
2443     ACPI_BUILD_DPRINTF("init ACPI tables\n");
2444 
2445     bios_linker_loader_alloc(tables->linker,
2446                              ACPI_BUILD_TABLE_FILE, tables_blob,
2447                              64 /* Ensure FACS is aligned */,
2448                              false /* high memory */);
2449 
2450     /*
2451      * FACS is pointed to by FADT.
2452      * We place it first since it's the only table that has alignment
2453      * requirements.
2454      */
2455     facs = tables_blob->len;
2456     build_facs(tables_blob);
2457 
2458     /* DSDT is pointed to by FADT */
2459     dsdt = tables_blob->len;
2460     build_dsdt(tables_blob, tables->linker, &pm, &misc,
2461                &pci_hole, &pci_hole64, machine);
2462 
2463     /* Count the size of the DSDT and SSDT, we will need it for legacy
2464      * sizing of ACPI tables.
2465      */
2466     aml_len += tables_blob->len - dsdt;
2467 
2468     /* ACPI tables pointed to by RSDT */
2469     fadt = tables_blob->len;
2470     acpi_add_table(table_offsets, tables_blob);
2471     pm.fadt.facs_tbl_offset = &facs;
2472     pm.fadt.dsdt_tbl_offset = &dsdt;
2473     pm.fadt.xdsdt_tbl_offset = &dsdt;
2474     build_fadt(tables_blob, tables->linker, &pm.fadt,
2475                slic_oem.id, slic_oem.table_id);
2476     aml_len += tables_blob->len - fadt;
2477 
2478     acpi_add_table(table_offsets, tables_blob);
2479     acpi_build_madt(tables_blob, tables->linker, x86ms,
2480                     ACPI_DEVICE_IF(x86ms->acpi_dev));
2481 
2482     vmgenid_dev = find_vmgenid_dev();
2483     if (vmgenid_dev) {
2484         acpi_add_table(table_offsets, tables_blob);
2485         vmgenid_build_acpi(VMGENID(vmgenid_dev), tables_blob,
2486                            tables->vmgenid, tables->linker);
2487     }
2488 
2489     if (misc.has_hpet) {
2490         acpi_add_table(table_offsets, tables_blob);
2491         build_hpet(tables_blob, tables->linker);
2492     }
2493     if (misc.tpm_version != TPM_VERSION_UNSPEC) {
2494         if (misc.tpm_version == TPM_VERSION_1_2) {
2495             acpi_add_table(table_offsets, tables_blob);
2496             build_tpm_tcpa(tables_blob, tables->linker, tables->tcpalog);
2497         } else { /* TPM_VERSION_2_0 */
2498             acpi_add_table(table_offsets, tables_blob);
2499             build_tpm2(tables_blob, tables->linker, tables->tcpalog);
2500         }
2501     }
2502     if (pcms->numa_nodes) {
2503         acpi_add_table(table_offsets, tables_blob);
2504         build_srat(tables_blob, tables->linker, machine);
2505         if (machine->numa_state->have_numa_distance) {
2506             acpi_add_table(table_offsets, tables_blob);
2507             build_slit(tables_blob, tables->linker, machine);
2508         }
2509         if (machine->numa_state->hmat_enabled) {
2510             acpi_add_table(table_offsets, tables_blob);
2511             build_hmat(tables_blob, tables->linker, machine->numa_state);
2512         }
2513     }
2514     if (acpi_get_mcfg(&mcfg)) {
2515         acpi_add_table(table_offsets, tables_blob);
2516         build_mcfg(tables_blob, tables->linker, &mcfg);
2517     }
2518     if (x86_iommu_get_default()) {
2519         IommuType IOMMUType = x86_iommu_get_type();
2520         if (IOMMUType == TYPE_AMD) {
2521             acpi_add_table(table_offsets, tables_blob);
2522             build_amd_iommu(tables_blob, tables->linker);
2523         } else if (IOMMUType == TYPE_INTEL) {
2524             acpi_add_table(table_offsets, tables_blob);
2525             build_dmar_q35(tables_blob, tables->linker);
2526         }
2527     }
2528     if (machine->nvdimms_state->is_enabled) {
2529         nvdimm_build_acpi(table_offsets, tables_blob, tables->linker,
2530                           machine->nvdimms_state, machine->ram_slots);
2531     }
2532 
2533     acpi_add_table(table_offsets, tables_blob);
2534     build_waet(tables_blob, tables->linker);
2535 
2536     /* Add tables supplied by user (if any) */
2537     for (u = acpi_table_first(); u; u = acpi_table_next(u)) {
2538         unsigned len = acpi_table_len(u);
2539 
2540         acpi_add_table(table_offsets, tables_blob);
2541         g_array_append_vals(tables_blob, u, len);
2542     }
2543 
2544     /* RSDT is pointed to by RSDP */
2545     rsdt = tables_blob->len;
2546     build_rsdt(tables_blob, tables->linker, table_offsets,
2547                slic_oem.id, slic_oem.table_id);
2548 
2549     /* RSDP is in FSEG memory, so allocate it separately */
2550     {
2551         AcpiRsdpData rsdp_data = {
2552             .revision = 0,
2553             .oem_id = ACPI_BUILD_APPNAME6,
2554             .xsdt_tbl_offset = NULL,
2555             .rsdt_tbl_offset = &rsdt,
2556         };
2557         build_rsdp(tables->rsdp, tables->linker, &rsdp_data);
2558         if (!pcmc->rsdp_in_ram) {
2559             /* We used to allocate some extra space for RSDP revision 2 but
2560              * only used the RSDP revision 0 space. The extra bytes were
2561              * zeroed out and not used.
2562              * Here we continue wasting those extra 16 bytes to make sure we
2563              * don't break migration for machine types 2.2 and older due to
2564              * RSDP blob size mismatch.
2565              */
2566             build_append_int_noprefix(tables->rsdp, 0, 16);
2567         }
2568     }
2569 
2570     /* We'll expose it all to Guest so we want to reduce
2571      * chance of size changes.
2572      *
2573      * We used to align the tables to 4k, but of course this would
2574      * too simple to be enough.  4k turned out to be too small an
2575      * alignment very soon, and in fact it is almost impossible to
2576      * keep the table size stable for all (max_cpus, max_memory_slots)
2577      * combinations.  So the table size is always 64k for pc-i440fx-2.1
2578      * and we give an error if the table grows beyond that limit.
2579      *
2580      * We still have the problem of migrating from "-M pc-i440fx-2.0".  For
2581      * that, we exploit the fact that QEMU 2.1 generates _smaller_ tables
2582      * than 2.0 and we can always pad the smaller tables with zeros.  We can
2583      * then use the exact size of the 2.0 tables.
2584      *
2585      * All this is for PIIX4, since QEMU 2.0 didn't support Q35 migration.
2586      */
2587     if (pcmc->legacy_acpi_table_size) {
2588         /* Subtracting aml_len gives the size of fixed tables.  Then add the
2589          * size of the PIIX4 DSDT/SSDT in QEMU 2.0.
2590          */
2591         int legacy_aml_len =
2592             pcmc->legacy_acpi_table_size +
2593             ACPI_BUILD_LEGACY_CPU_AML_SIZE * x86ms->apic_id_limit;
2594         int legacy_table_size =
2595             ROUND_UP(tables_blob->len - aml_len + legacy_aml_len,
2596                      ACPI_BUILD_ALIGN_SIZE);
2597         if (tables_blob->len > legacy_table_size) {
2598             /* Should happen only with PCI bridges and -M pc-i440fx-2.0.  */
2599             warn_report("ACPI table size %u exceeds %d bytes,"
2600                         " migration may not work",
2601                         tables_blob->len, legacy_table_size);
2602             error_printf("Try removing CPUs, NUMA nodes, memory slots"
2603                          " or PCI bridges.");
2604         }
2605         g_array_set_size(tables_blob, legacy_table_size);
2606     } else {
2607         /* Make sure we have a buffer in case we need to resize the tables. */
2608         if (tables_blob->len > ACPI_BUILD_TABLE_SIZE / 2) {
2609             /* As of QEMU 2.1, this fires with 160 VCPUs and 255 memory slots.  */
2610             warn_report("ACPI table size %u exceeds %d bytes,"
2611                         " migration may not work",
2612                         tables_blob->len, ACPI_BUILD_TABLE_SIZE / 2);
2613             error_printf("Try removing CPUs, NUMA nodes, memory slots"
2614                          " or PCI bridges.");
2615         }
2616         acpi_align_size(tables_blob, ACPI_BUILD_TABLE_SIZE);
2617     }
2618 
2619     acpi_align_size(tables->linker->cmd_blob, ACPI_BUILD_ALIGN_SIZE);
2620 
2621     /* Cleanup memory that's no longer used. */
2622     g_array_free(table_offsets, true);
2623 }
2624 
2625 static void acpi_ram_update(MemoryRegion *mr, GArray *data)
2626 {
2627     uint32_t size = acpi_data_len(data);
2628 
2629     /* Make sure RAM size is correct - in case it got changed e.g. by migration */
2630     memory_region_ram_resize(mr, size, &error_abort);
2631 
2632     memcpy(memory_region_get_ram_ptr(mr), data->data, size);
2633     memory_region_set_dirty(mr, 0, size);
2634 }
2635 
2636 static void acpi_build_update(void *build_opaque)
2637 {
2638     AcpiBuildState *build_state = build_opaque;
2639     AcpiBuildTables tables;
2640 
2641     /* No state to update or already patched? Nothing to do. */
2642     if (!build_state || build_state->patched) {
2643         return;
2644     }
2645     build_state->patched = 1;
2646 
2647     acpi_build_tables_init(&tables);
2648 
2649     acpi_build(&tables, MACHINE(qdev_get_machine()));
2650 
2651     acpi_ram_update(build_state->table_mr, tables.table_data);
2652 
2653     if (build_state->rsdp) {
2654         memcpy(build_state->rsdp, tables.rsdp->data, acpi_data_len(tables.rsdp));
2655     } else {
2656         acpi_ram_update(build_state->rsdp_mr, tables.rsdp);
2657     }
2658 
2659     acpi_ram_update(build_state->linker_mr, tables.linker->cmd_blob);
2660     acpi_build_tables_cleanup(&tables, true);
2661 }
2662 
2663 static void acpi_build_reset(void *build_opaque)
2664 {
2665     AcpiBuildState *build_state = build_opaque;
2666     build_state->patched = 0;
2667 }
2668 
2669 static const VMStateDescription vmstate_acpi_build = {
2670     .name = "acpi_build",
2671     .version_id = 1,
2672     .minimum_version_id = 1,
2673     .fields = (VMStateField[]) {
2674         VMSTATE_UINT8(patched, AcpiBuildState),
2675         VMSTATE_END_OF_LIST()
2676     },
2677 };
2678 
2679 void acpi_setup(void)
2680 {
2681     PCMachineState *pcms = PC_MACHINE(qdev_get_machine());
2682     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
2683     X86MachineState *x86ms = X86_MACHINE(pcms);
2684     AcpiBuildTables tables;
2685     AcpiBuildState *build_state;
2686     Object *vmgenid_dev;
2687     TPMIf *tpm;
2688     static FwCfgTPMConfig tpm_config;
2689 
2690     if (!x86ms->fw_cfg) {
2691         ACPI_BUILD_DPRINTF("No fw cfg. Bailing out.\n");
2692         return;
2693     }
2694 
2695     if (!pcms->acpi_build_enabled) {
2696         ACPI_BUILD_DPRINTF("ACPI build disabled. Bailing out.\n");
2697         return;
2698     }
2699 
2700     if (!x86_machine_is_acpi_enabled(X86_MACHINE(pcms))) {
2701         ACPI_BUILD_DPRINTF("ACPI disabled. Bailing out.\n");
2702         return;
2703     }
2704 
2705     build_state = g_malloc0(sizeof *build_state);
2706 
2707     acpi_build_tables_init(&tables);
2708     acpi_build(&tables, MACHINE(pcms));
2709 
2710     /* Now expose it all to Guest */
2711     build_state->table_mr = acpi_add_rom_blob(acpi_build_update,
2712                                               build_state, tables.table_data,
2713                                               ACPI_BUILD_TABLE_FILE,
2714                                               ACPI_BUILD_TABLE_MAX_SIZE);
2715     assert(build_state->table_mr != NULL);
2716 
2717     build_state->linker_mr =
2718         acpi_add_rom_blob(acpi_build_update, build_state,
2719                           tables.linker->cmd_blob, ACPI_BUILD_LOADER_FILE, 0);
2720 
2721     fw_cfg_add_file(x86ms->fw_cfg, ACPI_BUILD_TPMLOG_FILE,
2722                     tables.tcpalog->data, acpi_data_len(tables.tcpalog));
2723 
2724     tpm = tpm_find();
2725     if (tpm && object_property_get_bool(OBJECT(tpm), "ppi", &error_abort)) {
2726         tpm_config = (FwCfgTPMConfig) {
2727             .tpmppi_address = cpu_to_le32(TPM_PPI_ADDR_BASE),
2728             .tpm_version = tpm_get_version(tpm),
2729             .tpmppi_version = TPM_PPI_VERSION_1_30
2730         };
2731         fw_cfg_add_file(x86ms->fw_cfg, "etc/tpm/config",
2732                         &tpm_config, sizeof tpm_config);
2733     }
2734 
2735     vmgenid_dev = find_vmgenid_dev();
2736     if (vmgenid_dev) {
2737         vmgenid_add_fw_cfg(VMGENID(vmgenid_dev), x86ms->fw_cfg,
2738                            tables.vmgenid);
2739     }
2740 
2741     if (!pcmc->rsdp_in_ram) {
2742         /*
2743          * Keep for compatibility with old machine types.
2744          * Though RSDP is small, its contents isn't immutable, so
2745          * we'll update it along with the rest of tables on guest access.
2746          */
2747         uint32_t rsdp_size = acpi_data_len(tables.rsdp);
2748 
2749         build_state->rsdp = g_memdup(tables.rsdp->data, rsdp_size);
2750         fw_cfg_add_file_callback(x86ms->fw_cfg, ACPI_BUILD_RSDP_FILE,
2751                                  acpi_build_update, NULL, build_state,
2752                                  build_state->rsdp, rsdp_size, true);
2753         build_state->rsdp_mr = NULL;
2754     } else {
2755         build_state->rsdp = NULL;
2756         build_state->rsdp_mr = acpi_add_rom_blob(acpi_build_update,
2757                                                  build_state, tables.rsdp,
2758                                                  ACPI_BUILD_RSDP_FILE, 0);
2759     }
2760 
2761     qemu_register_reset(acpi_build_reset, build_state);
2762     acpi_build_reset(build_state);
2763     vmstate_register(NULL, 0, &vmstate_acpi_build, build_state);
2764 
2765     /* Cleanup tables but don't free the memory: we track it
2766      * in build_state.
2767      */
2768     acpi_build_tables_cleanup(&tables, false);
2769 }
2770