xref: /openbmc/qemu/hw/i386/acpi-build.c (revision 6e2e2e8a)
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     uint8_t s4_val;
100     AcpiFadtData fadt;
101     uint16_t cpu_hp_io_base;
102     uint16_t pcihp_io_base;
103     uint16_t pcihp_io_len;
104 } AcpiPmInfo;
105 
106 typedef struct AcpiMiscInfo {
107     bool is_piix4;
108     bool has_hpet;
109     TPMVersion tpm_version;
110     const unsigned char *dsdt_code;
111     unsigned dsdt_size;
112     uint16_t pvpanic_port;
113     uint16_t applesmc_io_base;
114 } AcpiMiscInfo;
115 
116 typedef struct AcpiBuildPciBusHotplugState {
117     GArray *device_table;
118     GArray *notify_table;
119     struct AcpiBuildPciBusHotplugState *parent;
120     bool pcihp_bridge_en;
121 } AcpiBuildPciBusHotplugState;
122 
123 typedef struct FwCfgTPMConfig {
124     uint32_t tpmppi_address;
125     uint8_t tpm_version;
126     uint8_t tpmppi_version;
127 } QEMU_PACKED FwCfgTPMConfig;
128 
129 static bool acpi_get_mcfg(AcpiMcfgInfo *mcfg);
130 
131 const struct AcpiGenericAddress x86_nvdimm_acpi_dsmio = {
132     .space_id = AML_AS_SYSTEM_IO,
133     .address = NVDIMM_ACPI_IO_BASE,
134     .bit_width = NVDIMM_ACPI_IO_LEN << 3
135 };
136 
137 static void init_common_fadt_data(MachineState *ms, Object *o,
138                                   AcpiFadtData *data)
139 {
140     uint32_t io = object_property_get_uint(o, ACPI_PM_PROP_PM_IO_BASE, NULL);
141     AmlAddressSpace as = AML_AS_SYSTEM_IO;
142     AcpiFadtData fadt = {
143         .rev = 3,
144         .flags =
145             (1 << ACPI_FADT_F_WBINVD) |
146             (1 << ACPI_FADT_F_PROC_C1) |
147             (1 << ACPI_FADT_F_SLP_BUTTON) |
148             (1 << ACPI_FADT_F_RTC_S4) |
149             (1 << ACPI_FADT_F_USE_PLATFORM_CLOCK) |
150             /* APIC destination mode ("Flat Logical") has an upper limit of 8
151              * CPUs for more than 8 CPUs, "Clustered Logical" mode has to be
152              * used
153              */
154             ((ms->smp.max_cpus > 8) ?
155                         (1 << ACPI_FADT_F_FORCE_APIC_CLUSTER_MODEL) : 0),
156         .int_model = 1 /* Multiple APIC */,
157         .rtc_century = RTC_CENTURY,
158         .plvl2_lat = 0xfff /* C2 state not supported */,
159         .plvl3_lat = 0xfff /* C3 state not supported */,
160         .smi_cmd = ACPI_PORT_SMI_CMD,
161         .sci_int = object_property_get_uint(o, ACPI_PM_PROP_SCI_INT, NULL),
162         .acpi_enable_cmd =
163             object_property_get_uint(o, ACPI_PM_PROP_ACPI_ENABLE_CMD, NULL),
164         .acpi_disable_cmd =
165             object_property_get_uint(o, ACPI_PM_PROP_ACPI_DISABLE_CMD, NULL),
166         .pm1a_evt = { .space_id = as, .bit_width = 4 * 8, .address = io },
167         .pm1a_cnt = { .space_id = as, .bit_width = 2 * 8,
168                       .address = io + 0x04 },
169         .pm_tmr = { .space_id = as, .bit_width = 4 * 8, .address = io + 0x08 },
170         .gpe0_blk = { .space_id = as, .bit_width =
171             object_property_get_uint(o, ACPI_PM_PROP_GPE0_BLK_LEN, NULL) * 8,
172             .address = object_property_get_uint(o, ACPI_PM_PROP_GPE0_BLK, NULL)
173         },
174     };
175     *data = fadt;
176 }
177 
178 static Object *object_resolve_type_unambiguous(const char *typename)
179 {
180     bool ambig;
181     Object *o = object_resolve_path_type("", typename, &ambig);
182 
183     if (ambig || !o) {
184         return NULL;
185     }
186     return o;
187 }
188 
189 static void acpi_get_pm_info(MachineState *machine, AcpiPmInfo *pm)
190 {
191     Object *piix = object_resolve_type_unambiguous(TYPE_PIIX4_PM);
192     Object *lpc = object_resolve_type_unambiguous(TYPE_ICH9_LPC_DEVICE);
193     Object *obj = piix ? piix : lpc;
194     QObject *o;
195     pm->cpu_hp_io_base = 0;
196     pm->pcihp_io_base = 0;
197     pm->pcihp_io_len = 0;
198     pm->smi_on_cpuhp = false;
199 
200     assert(obj);
201     init_common_fadt_data(machine, obj, &pm->fadt);
202     if (piix) {
203         /* w2k requires FADT(rev1) or it won't boot, keep PC compatible */
204         pm->fadt.rev = 1;
205         pm->cpu_hp_io_base = PIIX4_CPU_HOTPLUG_IO_BASE;
206         pm->pcihp_io_base =
207             object_property_get_uint(obj, ACPI_PCIHP_IO_BASE_PROP, NULL);
208         pm->pcihp_io_len =
209             object_property_get_uint(obj, ACPI_PCIHP_IO_LEN_PROP, NULL);
210     }
211     if (lpc) {
212         uint64_t smi_features = object_property_get_uint(lpc,
213             ICH9_LPC_SMI_NEGOTIATED_FEAT_PROP, NULL);
214         struct AcpiGenericAddress r = { .space_id = AML_AS_SYSTEM_IO,
215             .bit_width = 8, .address = ICH9_RST_CNT_IOPORT };
216         pm->fadt.reset_reg = r;
217         pm->fadt.reset_val = 0xf;
218         pm->fadt.flags |= 1 << ACPI_FADT_F_RESET_REG_SUP;
219         pm->cpu_hp_io_base = ICH9_CPU_HOTPLUG_IO_BASE;
220         pm->smi_on_cpuhp =
221             !!(smi_features & BIT_ULL(ICH9_LPC_SMI_F_CPU_HOTPLUG_BIT));
222     }
223 
224     /* The above need not be conditional on machine type because the reset port
225      * happens to be the same on PIIX (pc) and ICH9 (q35). */
226     QEMU_BUILD_BUG_ON(ICH9_RST_CNT_IOPORT != PIIX_RCR_IOPORT);
227 
228     /* Fill in optional s3/s4 related properties */
229     o = object_property_get_qobject(obj, ACPI_PM_PROP_S3_DISABLED, NULL);
230     if (o) {
231         pm->s3_disabled = qnum_get_uint(qobject_to(QNum, o));
232     } else {
233         pm->s3_disabled = false;
234     }
235     qobject_unref(o);
236     o = object_property_get_qobject(obj, ACPI_PM_PROP_S4_DISABLED, NULL);
237     if (o) {
238         pm->s4_disabled = qnum_get_uint(qobject_to(QNum, o));
239     } else {
240         pm->s4_disabled = false;
241     }
242     qobject_unref(o);
243     o = object_property_get_qobject(obj, ACPI_PM_PROP_S4_VAL, NULL);
244     if (o) {
245         pm->s4_val = qnum_get_uint(qobject_to(QNum, o));
246     } else {
247         pm->s4_val = false;
248     }
249     qobject_unref(o);
250 
251     pm->pcihp_bridge_en =
252         object_property_get_bool(obj, "acpi-pci-hotplug-with-bridge-support",
253                                  NULL);
254 }
255 
256 static void acpi_get_misc_info(AcpiMiscInfo *info)
257 {
258     Object *piix = object_resolve_type_unambiguous(TYPE_PIIX4_PM);
259     Object *lpc = object_resolve_type_unambiguous(TYPE_ICH9_LPC_DEVICE);
260     assert(!!piix != !!lpc);
261 
262     if (piix) {
263         info->is_piix4 = true;
264     }
265     if (lpc) {
266         info->is_piix4 = false;
267     }
268 
269     info->has_hpet = hpet_find();
270     info->tpm_version = tpm_get_version(tpm_find());
271     info->pvpanic_port = pvpanic_port();
272     info->applesmc_io_base = applesmc_port();
273 }
274 
275 /*
276  * Because of the PXB hosts we cannot simply query TYPE_PCI_HOST_BRIDGE.
277  * On i386 arch we only have two pci hosts, so we can look only for them.
278  */
279 static Object *acpi_get_i386_pci_host(void)
280 {
281     PCIHostState *host;
282 
283     host = OBJECT_CHECK(PCIHostState,
284                         object_resolve_path("/machine/i440fx", NULL),
285                         TYPE_PCI_HOST_BRIDGE);
286     if (!host) {
287         host = OBJECT_CHECK(PCIHostState,
288                             object_resolve_path("/machine/q35", NULL),
289                             TYPE_PCI_HOST_BRIDGE);
290     }
291 
292     return OBJECT(host);
293 }
294 
295 static void acpi_get_pci_holes(Range *hole, Range *hole64)
296 {
297     Object *pci_host;
298 
299     pci_host = acpi_get_i386_pci_host();
300     g_assert(pci_host);
301 
302     range_set_bounds1(hole,
303                       object_property_get_uint(pci_host,
304                                                PCI_HOST_PROP_PCI_HOLE_START,
305                                                NULL),
306                       object_property_get_uint(pci_host,
307                                                PCI_HOST_PROP_PCI_HOLE_END,
308                                                NULL));
309     range_set_bounds1(hole64,
310                       object_property_get_uint(pci_host,
311                                                PCI_HOST_PROP_PCI_HOLE64_START,
312                                                NULL),
313                       object_property_get_uint(pci_host,
314                                                PCI_HOST_PROP_PCI_HOLE64_END,
315                                                NULL));
316 }
317 
318 static void acpi_align_size(GArray *blob, unsigned align)
319 {
320     /* Align size to multiple of given size. This reduces the chance
321      * we need to change size in the future (breaking cross version migration).
322      */
323     g_array_set_size(blob, ROUND_UP(acpi_data_len(blob), align));
324 }
325 
326 /* FACS */
327 static void
328 build_facs(GArray *table_data)
329 {
330     AcpiFacsDescriptorRev1 *facs = acpi_data_push(table_data, sizeof *facs);
331     memcpy(&facs->signature, "FACS", 4);
332     facs->length = cpu_to_le32(sizeof(*facs));
333 }
334 
335 static void build_append_pcihp_notify_entry(Aml *method, int slot)
336 {
337     Aml *if_ctx;
338     int32_t devfn = PCI_DEVFN(slot, 0);
339 
340     if_ctx = aml_if(aml_and(aml_arg(0), aml_int(0x1U << slot), NULL));
341     aml_append(if_ctx, aml_notify(aml_name("S%.02X", devfn), aml_arg(1)));
342     aml_append(method, if_ctx);
343 }
344 
345 static void build_append_pci_bus_devices(Aml *parent_scope, PCIBus *bus,
346                                          bool pcihp_bridge_en)
347 {
348     Aml *dev, *notify_method = NULL, *method;
349     QObject *bsel;
350     PCIBus *sec;
351     int i;
352 
353     bsel = object_property_get_qobject(OBJECT(bus), ACPI_PCIHP_PROP_BSEL, NULL);
354     if (bsel) {
355         uint64_t bsel_val = qnum_get_uint(qobject_to(QNum, bsel));
356 
357         aml_append(parent_scope, aml_name_decl("BSEL", aml_int(bsel_val)));
358         notify_method = aml_method("DVNT", 2, AML_NOTSERIALIZED);
359     }
360 
361     for (i = 0; i < ARRAY_SIZE(bus->devices); i += PCI_FUNC_MAX) {
362         DeviceClass *dc;
363         PCIDeviceClass *pc;
364         PCIDevice *pdev = bus->devices[i];
365         int slot = PCI_SLOT(i);
366         bool hotplug_enabled_dev;
367         bool bridge_in_acpi;
368 
369         if (!pdev) {
370             if (bsel) { /* add hotplug slots for non present devices */
371                 dev = aml_device("S%.02X", PCI_DEVFN(slot, 0));
372                 aml_append(dev, aml_name_decl("_SUN", aml_int(slot)));
373                 aml_append(dev, aml_name_decl("_ADR", aml_int(slot << 16)));
374                 method = aml_method("_EJ0", 1, AML_NOTSERIALIZED);
375                 aml_append(method,
376                     aml_call2("PCEJ", aml_name("BSEL"), aml_name("_SUN"))
377                 );
378                 aml_append(dev, method);
379                 aml_append(parent_scope, dev);
380 
381                 build_append_pcihp_notify_entry(notify_method, slot);
382             }
383             continue;
384         }
385 
386         pc = PCI_DEVICE_GET_CLASS(pdev);
387         dc = DEVICE_GET_CLASS(pdev);
388 
389         /* When hotplug for bridges is enabled, bridges are
390          * described in ACPI separately (see build_pci_bus_end).
391          * In this case they aren't themselves hot-pluggable.
392          * Hotplugged bridges *are* hot-pluggable.
393          */
394         bridge_in_acpi = pc->is_bridge && pcihp_bridge_en &&
395             !DEVICE(pdev)->hotplugged;
396 
397         hotplug_enabled_dev = bsel && dc->hotpluggable && !bridge_in_acpi;
398 
399         if (pc->class_id == PCI_CLASS_BRIDGE_ISA) {
400             continue;
401         }
402 
403         /* start to compose PCI slot descriptor */
404         dev = aml_device("S%.02X", PCI_DEVFN(slot, 0));
405         aml_append(dev, aml_name_decl("_ADR", aml_int(slot << 16)));
406 
407         if (pc->class_id == PCI_CLASS_DISPLAY_VGA) {
408             /* add VGA specific AML methods */
409             int s3d;
410 
411             if (object_dynamic_cast(OBJECT(pdev), "qxl-vga")) {
412                 s3d = 3;
413             } else {
414                 s3d = 0;
415             }
416 
417             method = aml_method("_S1D", 0, AML_NOTSERIALIZED);
418             aml_append(method, aml_return(aml_int(0)));
419             aml_append(dev, method);
420 
421             method = aml_method("_S2D", 0, AML_NOTSERIALIZED);
422             aml_append(method, aml_return(aml_int(0)));
423             aml_append(dev, method);
424 
425             method = aml_method("_S3D", 0, AML_NOTSERIALIZED);
426             aml_append(method, aml_return(aml_int(s3d)));
427             aml_append(dev, method);
428         } else if (hotplug_enabled_dev) {
429             /* add _SUN/_EJ0 to make slot hotpluggable  */
430             aml_append(dev, aml_name_decl("_SUN", aml_int(slot)));
431 
432             method = aml_method("_EJ0", 1, AML_NOTSERIALIZED);
433             aml_append(method,
434                 aml_call2("PCEJ", aml_name("BSEL"), aml_name("_SUN"))
435             );
436             aml_append(dev, method);
437 
438             if (bsel) {
439                 build_append_pcihp_notify_entry(notify_method, slot);
440             }
441         } else if (bridge_in_acpi) {
442             /*
443              * device is coldplugged bridge,
444              * add child device descriptions into its scope
445              */
446             PCIBus *sec_bus = pci_bridge_get_sec_bus(PCI_BRIDGE(pdev));
447 
448             build_append_pci_bus_devices(dev, sec_bus, pcihp_bridge_en);
449         }
450         /* slot descriptor has been composed, add it into parent context */
451         aml_append(parent_scope, dev);
452     }
453 
454     if (bsel) {
455         aml_append(parent_scope, notify_method);
456     }
457 
458     /* Append PCNT method to notify about events on local and child buses.
459      * Add unconditionally for root since DSDT expects it.
460      */
461     method = aml_method("PCNT", 0, AML_NOTSERIALIZED);
462 
463     /* If bus supports hotplug select it and notify about local events */
464     if (bsel) {
465         uint64_t bsel_val = qnum_get_uint(qobject_to(QNum, bsel));
466 
467         aml_append(method, aml_store(aml_int(bsel_val), aml_name("BNUM")));
468         aml_append(method,
469             aml_call2("DVNT", aml_name("PCIU"), aml_int(1) /* Device Check */)
470         );
471         aml_append(method,
472             aml_call2("DVNT", aml_name("PCID"), aml_int(3)/* Eject Request */)
473         );
474     }
475 
476     /* Notify about child bus events in any case */
477     if (pcihp_bridge_en) {
478         QLIST_FOREACH(sec, &bus->child, sibling) {
479             int32_t devfn = sec->parent_dev->devfn;
480 
481             if (pci_bus_is_root(sec) || pci_bus_is_express(sec)) {
482                 continue;
483             }
484 
485             aml_append(method, aml_name("^S%.02X.PCNT", devfn));
486         }
487     }
488     aml_append(parent_scope, method);
489     qobject_unref(bsel);
490 }
491 
492 /**
493  * build_prt_entry:
494  * @link_name: link name for PCI route entry
495  *
496  * build AML package containing a PCI route entry for @link_name
497  */
498 static Aml *build_prt_entry(const char *link_name)
499 {
500     Aml *a_zero = aml_int(0);
501     Aml *pkg = aml_package(4);
502     aml_append(pkg, a_zero);
503     aml_append(pkg, a_zero);
504     aml_append(pkg, aml_name("%s", link_name));
505     aml_append(pkg, a_zero);
506     return pkg;
507 }
508 
509 /*
510  * initialize_route - Initialize the interrupt routing rule
511  * through a specific LINK:
512  *  if (lnk_idx == idx)
513  *      route using link 'link_name'
514  */
515 static Aml *initialize_route(Aml *route, const char *link_name,
516                              Aml *lnk_idx, int idx)
517 {
518     Aml *if_ctx = aml_if(aml_equal(lnk_idx, aml_int(idx)));
519     Aml *pkg = build_prt_entry(link_name);
520 
521     aml_append(if_ctx, aml_store(pkg, route));
522 
523     return if_ctx;
524 }
525 
526 /*
527  * build_prt - Define interrupt rounting rules
528  *
529  * Returns an array of 128 routes, one for each device,
530  * based on device location.
531  * The main goal is to equaly distribute the interrupts
532  * over the 4 existing ACPI links (works only for i440fx).
533  * The hash function is  (slot + pin) & 3 -> "LNK[D|A|B|C]".
534  *
535  */
536 static Aml *build_prt(bool is_pci0_prt)
537 {
538     Aml *method, *while_ctx, *pin, *res;
539 
540     method = aml_method("_PRT", 0, AML_NOTSERIALIZED);
541     res = aml_local(0);
542     pin = aml_local(1);
543     aml_append(method, aml_store(aml_package(128), res));
544     aml_append(method, aml_store(aml_int(0), pin));
545 
546     /* while (pin < 128) */
547     while_ctx = aml_while(aml_lless(pin, aml_int(128)));
548     {
549         Aml *slot = aml_local(2);
550         Aml *lnk_idx = aml_local(3);
551         Aml *route = aml_local(4);
552 
553         /* slot = pin >> 2 */
554         aml_append(while_ctx,
555                    aml_store(aml_shiftright(pin, aml_int(2), NULL), slot));
556         /* lnk_idx = (slot + pin) & 3 */
557         aml_append(while_ctx,
558             aml_store(aml_and(aml_add(pin, slot, NULL), aml_int(3), NULL),
559                       lnk_idx));
560 
561         /* route[2] = "LNK[D|A|B|C]", selection based on pin % 3  */
562         aml_append(while_ctx, initialize_route(route, "LNKD", lnk_idx, 0));
563         if (is_pci0_prt) {
564             Aml *if_device_1, *if_pin_4, *else_pin_4;
565 
566             /* device 1 is the power-management device, needs SCI */
567             if_device_1 = aml_if(aml_equal(lnk_idx, aml_int(1)));
568             {
569                 if_pin_4 = aml_if(aml_equal(pin, aml_int(4)));
570                 {
571                     aml_append(if_pin_4,
572                         aml_store(build_prt_entry("LNKS"), route));
573                 }
574                 aml_append(if_device_1, if_pin_4);
575                 else_pin_4 = aml_else();
576                 {
577                     aml_append(else_pin_4,
578                         aml_store(build_prt_entry("LNKA"), route));
579                 }
580                 aml_append(if_device_1, else_pin_4);
581             }
582             aml_append(while_ctx, if_device_1);
583         } else {
584             aml_append(while_ctx, initialize_route(route, "LNKA", lnk_idx, 1));
585         }
586         aml_append(while_ctx, initialize_route(route, "LNKB", lnk_idx, 2));
587         aml_append(while_ctx, initialize_route(route, "LNKC", lnk_idx, 3));
588 
589         /* route[0] = 0x[slot]FFFF */
590         aml_append(while_ctx,
591             aml_store(aml_or(aml_shiftleft(slot, aml_int(16)), aml_int(0xFFFF),
592                              NULL),
593                       aml_index(route, aml_int(0))));
594         /* route[1] = pin & 3 */
595         aml_append(while_ctx,
596             aml_store(aml_and(pin, aml_int(3), NULL),
597                       aml_index(route, aml_int(1))));
598         /* res[pin] = route */
599         aml_append(while_ctx, aml_store(route, aml_index(res, pin)));
600         /* pin++ */
601         aml_append(while_ctx, aml_increment(pin));
602     }
603     aml_append(method, while_ctx);
604     /* return res*/
605     aml_append(method, aml_return(res));
606 
607     return method;
608 }
609 
610 typedef struct CrsRangeEntry {
611     uint64_t base;
612     uint64_t limit;
613 } CrsRangeEntry;
614 
615 static void crs_range_insert(GPtrArray *ranges, uint64_t base, uint64_t limit)
616 {
617     CrsRangeEntry *entry;
618 
619     entry = g_malloc(sizeof(*entry));
620     entry->base = base;
621     entry->limit = limit;
622 
623     g_ptr_array_add(ranges, entry);
624 }
625 
626 static void crs_range_free(gpointer data)
627 {
628     CrsRangeEntry *entry = (CrsRangeEntry *)data;
629     g_free(entry);
630 }
631 
632 typedef struct CrsRangeSet {
633     GPtrArray *io_ranges;
634     GPtrArray *mem_ranges;
635     GPtrArray *mem_64bit_ranges;
636  } CrsRangeSet;
637 
638 static void crs_range_set_init(CrsRangeSet *range_set)
639 {
640     range_set->io_ranges = g_ptr_array_new_with_free_func(crs_range_free);
641     range_set->mem_ranges = g_ptr_array_new_with_free_func(crs_range_free);
642     range_set->mem_64bit_ranges =
643             g_ptr_array_new_with_free_func(crs_range_free);
644 }
645 
646 static void crs_range_set_free(CrsRangeSet *range_set)
647 {
648     g_ptr_array_free(range_set->io_ranges, true);
649     g_ptr_array_free(range_set->mem_ranges, true);
650     g_ptr_array_free(range_set->mem_64bit_ranges, true);
651 }
652 
653 static gint crs_range_compare(gconstpointer a, gconstpointer b)
654 {
655     CrsRangeEntry *entry_a = *(CrsRangeEntry **)a;
656     CrsRangeEntry *entry_b = *(CrsRangeEntry **)b;
657 
658     if (entry_a->base < entry_b->base) {
659         return -1;
660     } else if (entry_a->base > entry_b->base) {
661         return 1;
662     } else {
663         return 0;
664     }
665 }
666 
667 /*
668  * crs_replace_with_free_ranges - given the 'used' ranges within [start - end]
669  * interval, computes the 'free' ranges from the same interval.
670  * Example: If the input array is { [a1 - a2],[b1 - b2] }, the function
671  * will return { [base - a1], [a2 - b1], [b2 - limit] }.
672  */
673 static void crs_replace_with_free_ranges(GPtrArray *ranges,
674                                          uint64_t start, uint64_t end)
675 {
676     GPtrArray *free_ranges = g_ptr_array_new();
677     uint64_t free_base = start;
678     int i;
679 
680     g_ptr_array_sort(ranges, crs_range_compare);
681     for (i = 0; i < ranges->len; i++) {
682         CrsRangeEntry *used = g_ptr_array_index(ranges, i);
683 
684         if (free_base < used->base) {
685             crs_range_insert(free_ranges, free_base, used->base - 1);
686         }
687 
688         free_base = used->limit + 1;
689     }
690 
691     if (free_base < end) {
692         crs_range_insert(free_ranges, free_base, end);
693     }
694 
695     g_ptr_array_set_size(ranges, 0);
696     for (i = 0; i < free_ranges->len; i++) {
697         g_ptr_array_add(ranges, g_ptr_array_index(free_ranges, i));
698     }
699 
700     g_ptr_array_free(free_ranges, true);
701 }
702 
703 /*
704  * crs_range_merge - merges adjacent ranges in the given array.
705  * Array elements are deleted and replaced with the merged ranges.
706  */
707 static void crs_range_merge(GPtrArray *range)
708 {
709     GPtrArray *tmp =  g_ptr_array_new_with_free_func(crs_range_free);
710     CrsRangeEntry *entry;
711     uint64_t range_base, range_limit;
712     int i;
713 
714     if (!range->len) {
715         return;
716     }
717 
718     g_ptr_array_sort(range, crs_range_compare);
719 
720     entry = g_ptr_array_index(range, 0);
721     range_base = entry->base;
722     range_limit = entry->limit;
723     for (i = 1; i < range->len; i++) {
724         entry = g_ptr_array_index(range, i);
725         if (entry->base - 1 == range_limit) {
726             range_limit = entry->limit;
727         } else {
728             crs_range_insert(tmp, range_base, range_limit);
729             range_base = entry->base;
730             range_limit = entry->limit;
731         }
732     }
733     crs_range_insert(tmp, range_base, range_limit);
734 
735     g_ptr_array_set_size(range, 0);
736     for (i = 0; i < tmp->len; i++) {
737         entry = g_ptr_array_index(tmp, i);
738         crs_range_insert(range, entry->base, entry->limit);
739     }
740     g_ptr_array_free(tmp, true);
741 }
742 
743 static Aml *build_crs(PCIHostState *host, CrsRangeSet *range_set)
744 {
745     Aml *crs = aml_resource_template();
746     CrsRangeSet temp_range_set;
747     CrsRangeEntry *entry;
748     uint8_t max_bus = pci_bus_num(host->bus);
749     uint8_t type;
750     int devfn;
751     int i;
752 
753     crs_range_set_init(&temp_range_set);
754     for (devfn = 0; devfn < ARRAY_SIZE(host->bus->devices); devfn++) {
755         uint64_t range_base, range_limit;
756         PCIDevice *dev = host->bus->devices[devfn];
757 
758         if (!dev) {
759             continue;
760         }
761 
762         for (i = 0; i < PCI_NUM_REGIONS; i++) {
763             PCIIORegion *r = &dev->io_regions[i];
764 
765             range_base = r->addr;
766             range_limit = r->addr + r->size - 1;
767 
768             /*
769              * Work-around for old bioses
770              * that do not support multiple root buses
771              */
772             if (!range_base || range_base > range_limit) {
773                 continue;
774             }
775 
776             if (r->type & PCI_BASE_ADDRESS_SPACE_IO) {
777                 crs_range_insert(temp_range_set.io_ranges,
778                                  range_base, range_limit);
779             } else { /* "memory" */
780                 crs_range_insert(temp_range_set.mem_ranges,
781                                  range_base, range_limit);
782             }
783         }
784 
785         type = dev->config[PCI_HEADER_TYPE] & ~PCI_HEADER_TYPE_MULTI_FUNCTION;
786         if (type == PCI_HEADER_TYPE_BRIDGE) {
787             uint8_t subordinate = dev->config[PCI_SUBORDINATE_BUS];
788             if (subordinate > max_bus) {
789                 max_bus = subordinate;
790             }
791 
792             range_base = pci_bridge_get_base(dev, PCI_BASE_ADDRESS_SPACE_IO);
793             range_limit = pci_bridge_get_limit(dev, PCI_BASE_ADDRESS_SPACE_IO);
794 
795             /*
796              * Work-around for old bioses
797              * that do not support multiple root buses
798              */
799             if (range_base && range_base <= range_limit) {
800                 crs_range_insert(temp_range_set.io_ranges,
801                                  range_base, range_limit);
802             }
803 
804             range_base =
805                 pci_bridge_get_base(dev, PCI_BASE_ADDRESS_SPACE_MEMORY);
806             range_limit =
807                 pci_bridge_get_limit(dev, PCI_BASE_ADDRESS_SPACE_MEMORY);
808 
809             /*
810              * Work-around for old bioses
811              * that do not support multiple root buses
812              */
813             if (range_base && range_base <= range_limit) {
814                 uint64_t length = range_limit - range_base + 1;
815                 if (range_limit <= UINT32_MAX && length <= UINT32_MAX) {
816                     crs_range_insert(temp_range_set.mem_ranges,
817                                      range_base, range_limit);
818                 } else {
819                     crs_range_insert(temp_range_set.mem_64bit_ranges,
820                                      range_base, range_limit);
821                 }
822             }
823 
824             range_base =
825                 pci_bridge_get_base(dev, PCI_BASE_ADDRESS_MEM_PREFETCH);
826             range_limit =
827                 pci_bridge_get_limit(dev, PCI_BASE_ADDRESS_MEM_PREFETCH);
828 
829             /*
830              * Work-around for old bioses
831              * that do not support multiple root buses
832              */
833             if (range_base && range_base <= range_limit) {
834                 uint64_t length = range_limit - range_base + 1;
835                 if (range_limit <= UINT32_MAX && length <= UINT32_MAX) {
836                     crs_range_insert(temp_range_set.mem_ranges,
837                                      range_base, range_limit);
838                 } else {
839                     crs_range_insert(temp_range_set.mem_64bit_ranges,
840                                      range_base, range_limit);
841                 }
842             }
843         }
844     }
845 
846     crs_range_merge(temp_range_set.io_ranges);
847     for (i = 0; i < temp_range_set.io_ranges->len; i++) {
848         entry = g_ptr_array_index(temp_range_set.io_ranges, i);
849         aml_append(crs,
850                    aml_word_io(AML_MIN_FIXED, AML_MAX_FIXED,
851                                AML_POS_DECODE, AML_ENTIRE_RANGE,
852                                0, entry->base, entry->limit, 0,
853                                entry->limit - entry->base + 1));
854         crs_range_insert(range_set->io_ranges, entry->base, entry->limit);
855     }
856 
857     crs_range_merge(temp_range_set.mem_ranges);
858     for (i = 0; i < temp_range_set.mem_ranges->len; i++) {
859         entry = g_ptr_array_index(temp_range_set.mem_ranges, i);
860         aml_append(crs,
861                    aml_dword_memory(AML_POS_DECODE, AML_MIN_FIXED,
862                                     AML_MAX_FIXED, AML_NON_CACHEABLE,
863                                     AML_READ_WRITE,
864                                     0, entry->base, entry->limit, 0,
865                                     entry->limit - entry->base + 1));
866         crs_range_insert(range_set->mem_ranges, entry->base, entry->limit);
867     }
868 
869     crs_range_merge(temp_range_set.mem_64bit_ranges);
870     for (i = 0; i < temp_range_set.mem_64bit_ranges->len; i++) {
871         entry = g_ptr_array_index(temp_range_set.mem_64bit_ranges, i);
872         aml_append(crs,
873                    aml_qword_memory(AML_POS_DECODE, AML_MIN_FIXED,
874                                     AML_MAX_FIXED, AML_NON_CACHEABLE,
875                                     AML_READ_WRITE,
876                                     0, entry->base, entry->limit, 0,
877                                     entry->limit - entry->base + 1));
878         crs_range_insert(range_set->mem_64bit_ranges,
879                          entry->base, entry->limit);
880     }
881 
882     crs_range_set_free(&temp_range_set);
883 
884     aml_append(crs,
885         aml_word_bus_number(AML_MIN_FIXED, AML_MAX_FIXED, AML_POS_DECODE,
886                             0,
887                             pci_bus_num(host->bus),
888                             max_bus,
889                             0,
890                             max_bus - pci_bus_num(host->bus) + 1));
891 
892     return crs;
893 }
894 
895 static void build_hpet_aml(Aml *table)
896 {
897     Aml *crs;
898     Aml *field;
899     Aml *method;
900     Aml *if_ctx;
901     Aml *scope = aml_scope("_SB");
902     Aml *dev = aml_device("HPET");
903     Aml *zero = aml_int(0);
904     Aml *id = aml_local(0);
905     Aml *period = aml_local(1);
906 
907     aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0103")));
908     aml_append(dev, aml_name_decl("_UID", zero));
909 
910     aml_append(dev,
911         aml_operation_region("HPTM", AML_SYSTEM_MEMORY, aml_int(HPET_BASE),
912                              HPET_LEN));
913     field = aml_field("HPTM", AML_DWORD_ACC, AML_LOCK, AML_PRESERVE);
914     aml_append(field, aml_named_field("VEND", 32));
915     aml_append(field, aml_named_field("PRD", 32));
916     aml_append(dev, field);
917 
918     method = aml_method("_STA", 0, AML_NOTSERIALIZED);
919     aml_append(method, aml_store(aml_name("VEND"), id));
920     aml_append(method, aml_store(aml_name("PRD"), period));
921     aml_append(method, aml_shiftright(id, aml_int(16), id));
922     if_ctx = aml_if(aml_lor(aml_equal(id, zero),
923                             aml_equal(id, aml_int(0xffff))));
924     {
925         aml_append(if_ctx, aml_return(zero));
926     }
927     aml_append(method, if_ctx);
928 
929     if_ctx = aml_if(aml_lor(aml_equal(period, zero),
930                             aml_lgreater(period, aml_int(100000000))));
931     {
932         aml_append(if_ctx, aml_return(zero));
933     }
934     aml_append(method, if_ctx);
935 
936     aml_append(method, aml_return(aml_int(0x0F)));
937     aml_append(dev, method);
938 
939     crs = aml_resource_template();
940     aml_append(crs, aml_memory32_fixed(HPET_BASE, HPET_LEN, AML_READ_ONLY));
941     aml_append(dev, aml_name_decl("_CRS", crs));
942 
943     aml_append(scope, dev);
944     aml_append(table, scope);
945 }
946 
947 static Aml *build_vmbus_device_aml(VMBusBridge *vmbus_bridge)
948 {
949     Aml *dev;
950     Aml *method;
951     Aml *crs;
952 
953     dev = aml_device("VMBS");
954     aml_append(dev, aml_name_decl("STA", aml_int(0xF)));
955     aml_append(dev, aml_name_decl("_HID", aml_string("VMBus")));
956     aml_append(dev, aml_name_decl("_UID", aml_int(0x0)));
957     aml_append(dev, aml_name_decl("_DDN", aml_string("VMBUS")));
958 
959     method = aml_method("_DIS", 0, AML_NOTSERIALIZED);
960     aml_append(method, aml_store(aml_and(aml_name("STA"), aml_int(0xD), NULL),
961                                      aml_name("STA")));
962     aml_append(dev, method);
963 
964     method = aml_method("_PS0", 0, AML_NOTSERIALIZED);
965     aml_append(method, aml_store(aml_or(aml_name("STA"), aml_int(0xF), NULL),
966                                      aml_name("STA")));
967     aml_append(dev, method);
968 
969     method = aml_method("_STA", 0, AML_NOTSERIALIZED);
970     aml_append(method, aml_return(aml_name("STA")));
971     aml_append(dev, method);
972 
973     aml_append(dev, aml_name_decl("_PS3", aml_int(0x0)));
974 
975     crs = aml_resource_template();
976     aml_append(crs, aml_irq_no_flags(vmbus_bridge->irq));
977     aml_append(dev, aml_name_decl("_CRS", crs));
978 
979     return dev;
980 }
981 
982 static void build_isa_devices_aml(Aml *table)
983 {
984     VMBusBridge *vmbus_bridge = vmbus_bridge_find();
985     bool ambiguous;
986     Object *obj = object_resolve_path_type("", TYPE_ISA_BUS, &ambiguous);
987     Aml *scope;
988 
989     assert(obj && !ambiguous);
990 
991     scope = aml_scope("_SB.PCI0.ISA");
992     build_acpi_ipmi_devices(scope, BUS(obj), "\\_SB.PCI0.ISA");
993     isa_build_aml(ISA_BUS(obj), scope);
994 
995     if (vmbus_bridge) {
996         aml_append(scope, build_vmbus_device_aml(vmbus_bridge));
997     }
998 
999     aml_append(table, scope);
1000 }
1001 
1002 static void build_dbg_aml(Aml *table)
1003 {
1004     Aml *field;
1005     Aml *method;
1006     Aml *while_ctx;
1007     Aml *scope = aml_scope("\\");
1008     Aml *buf = aml_local(0);
1009     Aml *len = aml_local(1);
1010     Aml *idx = aml_local(2);
1011 
1012     aml_append(scope,
1013        aml_operation_region("DBG", AML_SYSTEM_IO, aml_int(0x0402), 0x01));
1014     field = aml_field("DBG", AML_BYTE_ACC, AML_NOLOCK, AML_PRESERVE);
1015     aml_append(field, aml_named_field("DBGB", 8));
1016     aml_append(scope, field);
1017 
1018     method = aml_method("DBUG", 1, AML_NOTSERIALIZED);
1019 
1020     aml_append(method, aml_to_hexstring(aml_arg(0), buf));
1021     aml_append(method, aml_to_buffer(buf, buf));
1022     aml_append(method, aml_subtract(aml_sizeof(buf), aml_int(1), len));
1023     aml_append(method, aml_store(aml_int(0), idx));
1024 
1025     while_ctx = aml_while(aml_lless(idx, len));
1026     aml_append(while_ctx,
1027         aml_store(aml_derefof(aml_index(buf, idx)), aml_name("DBGB")));
1028     aml_append(while_ctx, aml_increment(idx));
1029     aml_append(method, while_ctx);
1030 
1031     aml_append(method, aml_store(aml_int(0x0A), aml_name("DBGB")));
1032     aml_append(scope, method);
1033 
1034     aml_append(table, scope);
1035 }
1036 
1037 static Aml *build_link_dev(const char *name, uint8_t uid, Aml *reg)
1038 {
1039     Aml *dev;
1040     Aml *crs;
1041     Aml *method;
1042     uint32_t irqs[] = {5, 10, 11};
1043 
1044     dev = aml_device("%s", name);
1045     aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0C0F")));
1046     aml_append(dev, aml_name_decl("_UID", aml_int(uid)));
1047 
1048     crs = aml_resource_template();
1049     aml_append(crs, aml_interrupt(AML_CONSUMER, AML_LEVEL, AML_ACTIVE_HIGH,
1050                                   AML_SHARED, irqs, ARRAY_SIZE(irqs)));
1051     aml_append(dev, aml_name_decl("_PRS", crs));
1052 
1053     method = aml_method("_STA", 0, AML_NOTSERIALIZED);
1054     aml_append(method, aml_return(aml_call1("IQST", reg)));
1055     aml_append(dev, method);
1056 
1057     method = aml_method("_DIS", 0, AML_NOTSERIALIZED);
1058     aml_append(method, aml_or(reg, aml_int(0x80), reg));
1059     aml_append(dev, method);
1060 
1061     method = aml_method("_CRS", 0, AML_NOTSERIALIZED);
1062     aml_append(method, aml_return(aml_call1("IQCR", reg)));
1063     aml_append(dev, method);
1064 
1065     method = aml_method("_SRS", 1, AML_NOTSERIALIZED);
1066     aml_append(method, aml_create_dword_field(aml_arg(0), aml_int(5), "PRRI"));
1067     aml_append(method, aml_store(aml_name("PRRI"), reg));
1068     aml_append(dev, method);
1069 
1070     return dev;
1071  }
1072 
1073 static Aml *build_gsi_link_dev(const char *name, uint8_t uid, uint8_t gsi)
1074 {
1075     Aml *dev;
1076     Aml *crs;
1077     Aml *method;
1078     uint32_t irqs;
1079 
1080     dev = aml_device("%s", name);
1081     aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0C0F")));
1082     aml_append(dev, aml_name_decl("_UID", aml_int(uid)));
1083 
1084     crs = aml_resource_template();
1085     irqs = gsi;
1086     aml_append(crs, aml_interrupt(AML_CONSUMER, AML_LEVEL, AML_ACTIVE_HIGH,
1087                                   AML_SHARED, &irqs, 1));
1088     aml_append(dev, aml_name_decl("_PRS", crs));
1089 
1090     aml_append(dev, aml_name_decl("_CRS", crs));
1091 
1092     /*
1093      * _DIS can be no-op because the interrupt cannot be disabled.
1094      */
1095     method = aml_method("_DIS", 0, AML_NOTSERIALIZED);
1096     aml_append(dev, method);
1097 
1098     method = aml_method("_SRS", 1, AML_NOTSERIALIZED);
1099     aml_append(dev, method);
1100 
1101     return dev;
1102 }
1103 
1104 /* _CRS method - get current settings */
1105 static Aml *build_iqcr_method(bool is_piix4)
1106 {
1107     Aml *if_ctx;
1108     uint32_t irqs;
1109     Aml *method = aml_method("IQCR", 1, AML_SERIALIZED);
1110     Aml *crs = aml_resource_template();
1111 
1112     irqs = 0;
1113     aml_append(crs, aml_interrupt(AML_CONSUMER, AML_LEVEL,
1114                                   AML_ACTIVE_HIGH, AML_SHARED, &irqs, 1));
1115     aml_append(method, aml_name_decl("PRR0", crs));
1116 
1117     aml_append(method,
1118         aml_create_dword_field(aml_name("PRR0"), aml_int(5), "PRRI"));
1119 
1120     if (is_piix4) {
1121         if_ctx = aml_if(aml_lless(aml_arg(0), aml_int(0x80)));
1122         aml_append(if_ctx, aml_store(aml_arg(0), aml_name("PRRI")));
1123         aml_append(method, if_ctx);
1124     } else {
1125         aml_append(method,
1126             aml_store(aml_and(aml_arg(0), aml_int(0xF), NULL),
1127                       aml_name("PRRI")));
1128     }
1129 
1130     aml_append(method, aml_return(aml_name("PRR0")));
1131     return method;
1132 }
1133 
1134 /* _STA method - get status */
1135 static Aml *build_irq_status_method(void)
1136 {
1137     Aml *if_ctx;
1138     Aml *method = aml_method("IQST", 1, AML_NOTSERIALIZED);
1139 
1140     if_ctx = aml_if(aml_and(aml_int(0x80), aml_arg(0), NULL));
1141     aml_append(if_ctx, aml_return(aml_int(0x09)));
1142     aml_append(method, if_ctx);
1143     aml_append(method, aml_return(aml_int(0x0B)));
1144     return method;
1145 }
1146 
1147 static void build_piix4_pci0_int(Aml *table)
1148 {
1149     Aml *dev;
1150     Aml *crs;
1151     Aml *field;
1152     Aml *method;
1153     uint32_t irqs;
1154     Aml *sb_scope = aml_scope("_SB");
1155     Aml *pci0_scope = aml_scope("PCI0");
1156 
1157     aml_append(pci0_scope, build_prt(true));
1158     aml_append(sb_scope, pci0_scope);
1159 
1160     field = aml_field("PCI0.ISA.P40C", AML_BYTE_ACC, AML_NOLOCK, AML_PRESERVE);
1161     aml_append(field, aml_named_field("PRQ0", 8));
1162     aml_append(field, aml_named_field("PRQ1", 8));
1163     aml_append(field, aml_named_field("PRQ2", 8));
1164     aml_append(field, aml_named_field("PRQ3", 8));
1165     aml_append(sb_scope, field);
1166 
1167     aml_append(sb_scope, build_irq_status_method());
1168     aml_append(sb_scope, build_iqcr_method(true));
1169 
1170     aml_append(sb_scope, build_link_dev("LNKA", 0, aml_name("PRQ0")));
1171     aml_append(sb_scope, build_link_dev("LNKB", 1, aml_name("PRQ1")));
1172     aml_append(sb_scope, build_link_dev("LNKC", 2, aml_name("PRQ2")));
1173     aml_append(sb_scope, build_link_dev("LNKD", 3, aml_name("PRQ3")));
1174 
1175     dev = aml_device("LNKS");
1176     {
1177         aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0C0F")));
1178         aml_append(dev, aml_name_decl("_UID", aml_int(4)));
1179 
1180         crs = aml_resource_template();
1181         irqs = 9;
1182         aml_append(crs, aml_interrupt(AML_CONSUMER, AML_LEVEL,
1183                                       AML_ACTIVE_HIGH, AML_SHARED,
1184                                       &irqs, 1));
1185         aml_append(dev, aml_name_decl("_PRS", crs));
1186 
1187         /* The SCI cannot be disabled and is always attached to GSI 9,
1188          * so these are no-ops.  We only need this link to override the
1189          * polarity to active high and match the content of the MADT.
1190          */
1191         method = aml_method("_STA", 0, AML_NOTSERIALIZED);
1192         aml_append(method, aml_return(aml_int(0x0b)));
1193         aml_append(dev, method);
1194 
1195         method = aml_method("_DIS", 0, AML_NOTSERIALIZED);
1196         aml_append(dev, method);
1197 
1198         method = aml_method("_CRS", 0, AML_NOTSERIALIZED);
1199         aml_append(method, aml_return(aml_name("_PRS")));
1200         aml_append(dev, method);
1201 
1202         method = aml_method("_SRS", 1, AML_NOTSERIALIZED);
1203         aml_append(dev, method);
1204     }
1205     aml_append(sb_scope, dev);
1206 
1207     aml_append(table, sb_scope);
1208 }
1209 
1210 static void append_q35_prt_entry(Aml *ctx, uint32_t nr, const char *name)
1211 {
1212     int i;
1213     int head;
1214     Aml *pkg;
1215     char base = name[3] < 'E' ? 'A' : 'E';
1216     char *s = g_strdup(name);
1217     Aml *a_nr = aml_int((nr << 16) | 0xffff);
1218 
1219     assert(strlen(s) == 4);
1220 
1221     head = name[3] - base;
1222     for (i = 0; i < 4; i++) {
1223         if (head + i > 3) {
1224             head = i * -1;
1225         }
1226         s[3] = base + head + i;
1227         pkg = aml_package(4);
1228         aml_append(pkg, a_nr);
1229         aml_append(pkg, aml_int(i));
1230         aml_append(pkg, aml_name("%s", s));
1231         aml_append(pkg, aml_int(0));
1232         aml_append(ctx, pkg);
1233     }
1234     g_free(s);
1235 }
1236 
1237 static Aml *build_q35_routing_table(const char *str)
1238 {
1239     int i;
1240     Aml *pkg;
1241     char *name = g_strdup_printf("%s ", str);
1242 
1243     pkg = aml_package(128);
1244     for (i = 0; i < 0x18; i++) {
1245             name[3] = 'E' + (i & 0x3);
1246             append_q35_prt_entry(pkg, i, name);
1247     }
1248 
1249     name[3] = 'E';
1250     append_q35_prt_entry(pkg, 0x18, name);
1251 
1252     /* INTA -> PIRQA for slot 25 - 31, see the default value of D<N>IR */
1253     for (i = 0x0019; i < 0x1e; i++) {
1254         name[3] = 'A';
1255         append_q35_prt_entry(pkg, i, name);
1256     }
1257 
1258     /* PCIe->PCI bridge. use PIRQ[E-H] */
1259     name[3] = 'E';
1260     append_q35_prt_entry(pkg, 0x1e, name);
1261     name[3] = 'A';
1262     append_q35_prt_entry(pkg, 0x1f, name);
1263 
1264     g_free(name);
1265     return pkg;
1266 }
1267 
1268 static void build_q35_pci0_int(Aml *table)
1269 {
1270     Aml *field;
1271     Aml *method;
1272     Aml *sb_scope = aml_scope("_SB");
1273     Aml *pci0_scope = aml_scope("PCI0");
1274 
1275     /* Zero => PIC mode, One => APIC Mode */
1276     aml_append(table, aml_name_decl("PICF", aml_int(0)));
1277     method = aml_method("_PIC", 1, AML_NOTSERIALIZED);
1278     {
1279         aml_append(method, aml_store(aml_arg(0), aml_name("PICF")));
1280     }
1281     aml_append(table, method);
1282 
1283     aml_append(pci0_scope,
1284         aml_name_decl("PRTP", build_q35_routing_table("LNK")));
1285     aml_append(pci0_scope,
1286         aml_name_decl("PRTA", build_q35_routing_table("GSI")));
1287 
1288     method = aml_method("_PRT", 0, AML_NOTSERIALIZED);
1289     {
1290         Aml *if_ctx;
1291         Aml *else_ctx;
1292 
1293         /* PCI IRQ routing table, example from ACPI 2.0a specification,
1294            section 6.2.8.1 */
1295         /* Note: we provide the same info as the PCI routing
1296            table of the Bochs BIOS */
1297         if_ctx = aml_if(aml_equal(aml_name("PICF"), aml_int(0)));
1298         aml_append(if_ctx, aml_return(aml_name("PRTP")));
1299         aml_append(method, if_ctx);
1300         else_ctx = aml_else();
1301         aml_append(else_ctx, aml_return(aml_name("PRTA")));
1302         aml_append(method, else_ctx);
1303     }
1304     aml_append(pci0_scope, method);
1305     aml_append(sb_scope, pci0_scope);
1306 
1307     field = aml_field("PCI0.ISA.PIRQ", AML_BYTE_ACC, AML_NOLOCK, AML_PRESERVE);
1308     aml_append(field, aml_named_field("PRQA", 8));
1309     aml_append(field, aml_named_field("PRQB", 8));
1310     aml_append(field, aml_named_field("PRQC", 8));
1311     aml_append(field, aml_named_field("PRQD", 8));
1312     aml_append(field, aml_reserved_field(0x20));
1313     aml_append(field, aml_named_field("PRQE", 8));
1314     aml_append(field, aml_named_field("PRQF", 8));
1315     aml_append(field, aml_named_field("PRQG", 8));
1316     aml_append(field, aml_named_field("PRQH", 8));
1317     aml_append(sb_scope, field);
1318 
1319     aml_append(sb_scope, build_irq_status_method());
1320     aml_append(sb_scope, build_iqcr_method(false));
1321 
1322     aml_append(sb_scope, build_link_dev("LNKA", 0, aml_name("PRQA")));
1323     aml_append(sb_scope, build_link_dev("LNKB", 1, aml_name("PRQB")));
1324     aml_append(sb_scope, build_link_dev("LNKC", 2, aml_name("PRQC")));
1325     aml_append(sb_scope, build_link_dev("LNKD", 3, aml_name("PRQD")));
1326     aml_append(sb_scope, build_link_dev("LNKE", 4, aml_name("PRQE")));
1327     aml_append(sb_scope, build_link_dev("LNKF", 5, aml_name("PRQF")));
1328     aml_append(sb_scope, build_link_dev("LNKG", 6, aml_name("PRQG")));
1329     aml_append(sb_scope, build_link_dev("LNKH", 7, aml_name("PRQH")));
1330 
1331     aml_append(sb_scope, build_gsi_link_dev("GSIA", 0x10, 0x10));
1332     aml_append(sb_scope, build_gsi_link_dev("GSIB", 0x11, 0x11));
1333     aml_append(sb_scope, build_gsi_link_dev("GSIC", 0x12, 0x12));
1334     aml_append(sb_scope, build_gsi_link_dev("GSID", 0x13, 0x13));
1335     aml_append(sb_scope, build_gsi_link_dev("GSIE", 0x14, 0x14));
1336     aml_append(sb_scope, build_gsi_link_dev("GSIF", 0x15, 0x15));
1337     aml_append(sb_scope, build_gsi_link_dev("GSIG", 0x16, 0x16));
1338     aml_append(sb_scope, build_gsi_link_dev("GSIH", 0x17, 0x17));
1339 
1340     aml_append(table, sb_scope);
1341 }
1342 
1343 static void build_q35_isa_bridge(Aml *table)
1344 {
1345     Aml *dev;
1346     Aml *scope;
1347 
1348     scope =  aml_scope("_SB.PCI0");
1349     dev = aml_device("ISA");
1350     aml_append(dev, aml_name_decl("_ADR", aml_int(0x001F0000)));
1351 
1352     /* ICH9 PCI to ISA irq remapping */
1353     aml_append(dev, aml_operation_region("PIRQ", AML_PCI_CONFIG,
1354                                          aml_int(0x60), 0x0C));
1355 
1356     aml_append(scope, dev);
1357     aml_append(table, scope);
1358 }
1359 
1360 static void build_piix4_isa_bridge(Aml *table)
1361 {
1362     Aml *dev;
1363     Aml *scope;
1364 
1365     scope =  aml_scope("_SB.PCI0");
1366     dev = aml_device("ISA");
1367     aml_append(dev, aml_name_decl("_ADR", aml_int(0x00010000)));
1368 
1369     /* PIIX PCI to ISA irq remapping */
1370     aml_append(dev, aml_operation_region("P40C", AML_PCI_CONFIG,
1371                                          aml_int(0x60), 0x04));
1372 
1373     aml_append(scope, dev);
1374     aml_append(table, scope);
1375 }
1376 
1377 static void build_piix4_pci_hotplug(Aml *table)
1378 {
1379     Aml *scope;
1380     Aml *field;
1381     Aml *method;
1382 
1383     scope =  aml_scope("_SB.PCI0");
1384 
1385     aml_append(scope,
1386         aml_operation_region("PCST", AML_SYSTEM_IO, aml_int(0xae00), 0x08));
1387     field = aml_field("PCST", AML_DWORD_ACC, AML_NOLOCK, AML_WRITE_AS_ZEROS);
1388     aml_append(field, aml_named_field("PCIU", 32));
1389     aml_append(field, aml_named_field("PCID", 32));
1390     aml_append(scope, field);
1391 
1392     aml_append(scope,
1393         aml_operation_region("SEJ", AML_SYSTEM_IO, aml_int(0xae08), 0x04));
1394     field = aml_field("SEJ", AML_DWORD_ACC, AML_NOLOCK, AML_WRITE_AS_ZEROS);
1395     aml_append(field, aml_named_field("B0EJ", 32));
1396     aml_append(scope, field);
1397 
1398     aml_append(scope,
1399         aml_operation_region("BNMR", AML_SYSTEM_IO, aml_int(0xae10), 0x04));
1400     field = aml_field("BNMR", AML_DWORD_ACC, AML_NOLOCK, AML_WRITE_AS_ZEROS);
1401     aml_append(field, aml_named_field("BNUM", 32));
1402     aml_append(scope, field);
1403 
1404     aml_append(scope, aml_mutex("BLCK", 0));
1405 
1406     method = aml_method("PCEJ", 2, AML_NOTSERIALIZED);
1407     aml_append(method, aml_acquire(aml_name("BLCK"), 0xFFFF));
1408     aml_append(method, aml_store(aml_arg(0), aml_name("BNUM")));
1409     aml_append(method,
1410         aml_store(aml_shiftleft(aml_int(1), aml_arg(1)), aml_name("B0EJ")));
1411     aml_append(method, aml_release(aml_name("BLCK")));
1412     aml_append(method, aml_return(aml_int(0)));
1413     aml_append(scope, method);
1414 
1415     aml_append(table, scope);
1416 }
1417 
1418 static Aml *build_q35_osc_method(void)
1419 {
1420     Aml *if_ctx;
1421     Aml *if_ctx2;
1422     Aml *else_ctx;
1423     Aml *method;
1424     Aml *a_cwd1 = aml_name("CDW1");
1425     Aml *a_ctrl = aml_local(0);
1426 
1427     method = aml_method("_OSC", 4, AML_NOTSERIALIZED);
1428     aml_append(method, aml_create_dword_field(aml_arg(3), aml_int(0), "CDW1"));
1429 
1430     if_ctx = aml_if(aml_equal(
1431         aml_arg(0), aml_touuid("33DB4D5B-1FF7-401C-9657-7441C03DD766")));
1432     aml_append(if_ctx, aml_create_dword_field(aml_arg(3), aml_int(4), "CDW2"));
1433     aml_append(if_ctx, aml_create_dword_field(aml_arg(3), aml_int(8), "CDW3"));
1434 
1435     aml_append(if_ctx, aml_store(aml_name("CDW3"), a_ctrl));
1436 
1437     /*
1438      * Always allow native PME, AER (no dependencies)
1439      * Allow SHPC (PCI bridges can have SHPC controller)
1440      */
1441     aml_append(if_ctx, aml_and(a_ctrl, aml_int(0x1F), a_ctrl));
1442 
1443     if_ctx2 = aml_if(aml_lnot(aml_equal(aml_arg(1), aml_int(1))));
1444     /* Unknown revision */
1445     aml_append(if_ctx2, aml_or(a_cwd1, aml_int(0x08), a_cwd1));
1446     aml_append(if_ctx, if_ctx2);
1447 
1448     if_ctx2 = aml_if(aml_lnot(aml_equal(aml_name("CDW3"), a_ctrl)));
1449     /* Capabilities bits were masked */
1450     aml_append(if_ctx2, aml_or(a_cwd1, aml_int(0x10), a_cwd1));
1451     aml_append(if_ctx, if_ctx2);
1452 
1453     /* Update DWORD3 in the buffer */
1454     aml_append(if_ctx, aml_store(a_ctrl, aml_name("CDW3")));
1455     aml_append(method, if_ctx);
1456 
1457     else_ctx = aml_else();
1458     /* Unrecognized UUID */
1459     aml_append(else_ctx, aml_or(a_cwd1, aml_int(4), a_cwd1));
1460     aml_append(method, else_ctx);
1461 
1462     aml_append(method, aml_return(aml_arg(3)));
1463     return method;
1464 }
1465 
1466 static void build_smb0(Aml *table, I2CBus *smbus, int devnr, int func)
1467 {
1468     Aml *scope = aml_scope("_SB.PCI0");
1469     Aml *dev = aml_device("SMB0");
1470 
1471     aml_append(dev, aml_name_decl("_ADR", aml_int(devnr << 16 | func)));
1472     build_acpi_ipmi_devices(dev, BUS(smbus), "\\_SB.PCI0.SMB0");
1473     aml_append(scope, dev);
1474     aml_append(table, scope);
1475 }
1476 
1477 static void
1478 build_dsdt(GArray *table_data, BIOSLinker *linker,
1479            AcpiPmInfo *pm, AcpiMiscInfo *misc,
1480            Range *pci_hole, Range *pci_hole64, MachineState *machine)
1481 {
1482     CrsRangeEntry *entry;
1483     Aml *dsdt, *sb_scope, *scope, *dev, *method, *field, *pkg, *crs;
1484     CrsRangeSet crs_range_set;
1485     PCMachineState *pcms = PC_MACHINE(machine);
1486     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(machine);
1487     X86MachineState *x86ms = X86_MACHINE(machine);
1488     AcpiMcfgInfo mcfg;
1489     uint32_t nr_mem = machine->ram_slots;
1490     int root_bus_limit = 0xFF;
1491     PCIBus *bus = NULL;
1492     TPMIf *tpm = tpm_find();
1493     int i;
1494 
1495     dsdt = init_aml_allocator();
1496 
1497     /* Reserve space for header */
1498     acpi_data_push(dsdt->buf, sizeof(AcpiTableHeader));
1499 
1500     build_dbg_aml(dsdt);
1501     if (misc->is_piix4) {
1502         sb_scope = aml_scope("_SB");
1503         dev = aml_device("PCI0");
1504         aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0A03")));
1505         aml_append(dev, aml_name_decl("_ADR", aml_int(0)));
1506         aml_append(dev, aml_name_decl("_UID", aml_int(0)));
1507         aml_append(sb_scope, dev);
1508         aml_append(dsdt, sb_scope);
1509 
1510         build_hpet_aml(dsdt);
1511         build_piix4_isa_bridge(dsdt);
1512         build_isa_devices_aml(dsdt);
1513         build_piix4_pci_hotplug(dsdt);
1514         build_piix4_pci0_int(dsdt);
1515     } else {
1516         sb_scope = aml_scope("_SB");
1517         dev = aml_device("PCI0");
1518         aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0A08")));
1519         aml_append(dev, aml_name_decl("_CID", aml_eisaid("PNP0A03")));
1520         aml_append(dev, aml_name_decl("_ADR", aml_int(0)));
1521         aml_append(dev, aml_name_decl("_UID", aml_int(0)));
1522         aml_append(dev, build_q35_osc_method());
1523         aml_append(sb_scope, dev);
1524 
1525         if (pm->smi_on_cpuhp) {
1526             /* reserve SMI block resources, IO ports 0xB2, 0xB3 */
1527             dev = aml_device("PCI0.SMI0");
1528             aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0A06")));
1529             aml_append(dev, aml_name_decl("_UID", aml_string("SMI resources")));
1530             crs = aml_resource_template();
1531             aml_append(crs,
1532                 aml_io(
1533                        AML_DECODE16,
1534                        ACPI_PORT_SMI_CMD,
1535                        ACPI_PORT_SMI_CMD,
1536                        1,
1537                        2)
1538             );
1539             aml_append(dev, aml_name_decl("_CRS", crs));
1540             aml_append(dev, aml_operation_region("SMIR", AML_SYSTEM_IO,
1541                 aml_int(ACPI_PORT_SMI_CMD), 2));
1542             field = aml_field("SMIR", AML_BYTE_ACC, AML_NOLOCK,
1543                               AML_WRITE_AS_ZEROS);
1544             aml_append(field, aml_named_field("SMIC", 8));
1545             aml_append(field, aml_reserved_field(8));
1546             aml_append(dev, field);
1547             aml_append(sb_scope, dev);
1548         }
1549 
1550         aml_append(dsdt, sb_scope);
1551 
1552         build_hpet_aml(dsdt);
1553         build_q35_isa_bridge(dsdt);
1554         build_isa_devices_aml(dsdt);
1555         build_q35_pci0_int(dsdt);
1556         if (pcms->smbus && !pcmc->do_not_add_smb_acpi) {
1557             build_smb0(dsdt, pcms->smbus, ICH9_SMB_DEV, ICH9_SMB_FUNC);
1558         }
1559     }
1560 
1561     if (pcmc->legacy_cpu_hotplug) {
1562         build_legacy_cpu_hotplug_aml(dsdt, machine, pm->cpu_hp_io_base);
1563     } else {
1564         CPUHotplugFeatures opts = {
1565             .acpi_1_compatible = true, .has_legacy_cphp = true,
1566             .smi_path = pm->smi_on_cpuhp ? "\\_SB.PCI0.SMI0.SMIC" : NULL,
1567         };
1568         build_cpus_aml(dsdt, machine, opts, pm->cpu_hp_io_base,
1569                        "\\_SB.PCI0", "\\_GPE._E02");
1570     }
1571 
1572     if (pcms->memhp_io_base && nr_mem) {
1573         build_memory_hotplug_aml(dsdt, nr_mem, "\\_SB.PCI0",
1574                                  "\\_GPE._E03", AML_SYSTEM_IO,
1575                                  pcms->memhp_io_base);
1576     }
1577 
1578     scope =  aml_scope("_GPE");
1579     {
1580         aml_append(scope, aml_name_decl("_HID", aml_string("ACPI0006")));
1581 
1582         if (misc->is_piix4) {
1583             method = aml_method("_E01", 0, AML_NOTSERIALIZED);
1584             aml_append(method,
1585                 aml_acquire(aml_name("\\_SB.PCI0.BLCK"), 0xFFFF));
1586             aml_append(method, aml_call0("\\_SB.PCI0.PCNT"));
1587             aml_append(method, aml_release(aml_name("\\_SB.PCI0.BLCK")));
1588             aml_append(scope, method);
1589         }
1590 
1591         if (machine->nvdimms_state->is_enabled) {
1592             method = aml_method("_E04", 0, AML_NOTSERIALIZED);
1593             aml_append(method, aml_notify(aml_name("\\_SB.NVDR"),
1594                                           aml_int(0x80)));
1595             aml_append(scope, method);
1596         }
1597     }
1598     aml_append(dsdt, scope);
1599 
1600     crs_range_set_init(&crs_range_set);
1601     bus = PC_MACHINE(machine)->bus;
1602     if (bus) {
1603         QLIST_FOREACH(bus, &bus->child, sibling) {
1604             uint8_t bus_num = pci_bus_num(bus);
1605             uint8_t numa_node = pci_bus_numa_node(bus);
1606 
1607             /* look only for expander root buses */
1608             if (!pci_bus_is_root(bus)) {
1609                 continue;
1610             }
1611 
1612             if (bus_num < root_bus_limit) {
1613                 root_bus_limit = bus_num - 1;
1614             }
1615 
1616             scope = aml_scope("\\_SB");
1617             dev = aml_device("PC%.02X", bus_num);
1618             aml_append(dev, aml_name_decl("_UID", aml_int(bus_num)));
1619             aml_append(dev, aml_name_decl("_BBN", aml_int(bus_num)));
1620             if (pci_bus_is_express(bus)) {
1621                 aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0A08")));
1622                 aml_append(dev, aml_name_decl("_CID", aml_eisaid("PNP0A03")));
1623                 aml_append(dev, build_q35_osc_method());
1624             } else {
1625                 aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0A03")));
1626             }
1627 
1628             if (numa_node != NUMA_NODE_UNASSIGNED) {
1629                 aml_append(dev, aml_name_decl("_PXM", aml_int(numa_node)));
1630             }
1631 
1632             aml_append(dev, build_prt(false));
1633             crs = build_crs(PCI_HOST_BRIDGE(BUS(bus)->parent), &crs_range_set);
1634             aml_append(dev, aml_name_decl("_CRS", crs));
1635             aml_append(scope, dev);
1636             aml_append(dsdt, scope);
1637         }
1638     }
1639 
1640     /*
1641      * At this point crs_range_set has all the ranges used by pci
1642      * busses *other* than PCI0.  These ranges will be excluded from
1643      * the PCI0._CRS.  Add mmconfig to the set so it will be excluded
1644      * too.
1645      */
1646     if (acpi_get_mcfg(&mcfg)) {
1647         crs_range_insert(crs_range_set.mem_ranges,
1648                          mcfg.base, mcfg.base + mcfg.size - 1);
1649     }
1650 
1651     scope = aml_scope("\\_SB.PCI0");
1652     /* build PCI0._CRS */
1653     crs = aml_resource_template();
1654     aml_append(crs,
1655         aml_word_bus_number(AML_MIN_FIXED, AML_MAX_FIXED, AML_POS_DECODE,
1656                             0x0000, 0x0, root_bus_limit,
1657                             0x0000, root_bus_limit + 1));
1658     aml_append(crs, aml_io(AML_DECODE16, 0x0CF8, 0x0CF8, 0x01, 0x08));
1659 
1660     aml_append(crs,
1661         aml_word_io(AML_MIN_FIXED, AML_MAX_FIXED,
1662                     AML_POS_DECODE, AML_ENTIRE_RANGE,
1663                     0x0000, 0x0000, 0x0CF7, 0x0000, 0x0CF8));
1664 
1665     crs_replace_with_free_ranges(crs_range_set.io_ranges, 0x0D00, 0xFFFF);
1666     for (i = 0; i < crs_range_set.io_ranges->len; i++) {
1667         entry = g_ptr_array_index(crs_range_set.io_ranges, i);
1668         aml_append(crs,
1669             aml_word_io(AML_MIN_FIXED, AML_MAX_FIXED,
1670                         AML_POS_DECODE, AML_ENTIRE_RANGE,
1671                         0x0000, entry->base, entry->limit,
1672                         0x0000, entry->limit - entry->base + 1));
1673     }
1674 
1675     aml_append(crs,
1676         aml_dword_memory(AML_POS_DECODE, AML_MIN_FIXED, AML_MAX_FIXED,
1677                          AML_CACHEABLE, AML_READ_WRITE,
1678                          0, 0x000A0000, 0x000BFFFF, 0, 0x00020000));
1679 
1680     crs_replace_with_free_ranges(crs_range_set.mem_ranges,
1681                                  range_lob(pci_hole),
1682                                  range_upb(pci_hole));
1683     for (i = 0; i < crs_range_set.mem_ranges->len; i++) {
1684         entry = g_ptr_array_index(crs_range_set.mem_ranges, i);
1685         aml_append(crs,
1686             aml_dword_memory(AML_POS_DECODE, AML_MIN_FIXED, AML_MAX_FIXED,
1687                              AML_NON_CACHEABLE, AML_READ_WRITE,
1688                              0, entry->base, entry->limit,
1689                              0, entry->limit - entry->base + 1));
1690     }
1691 
1692     if (!range_is_empty(pci_hole64)) {
1693         crs_replace_with_free_ranges(crs_range_set.mem_64bit_ranges,
1694                                      range_lob(pci_hole64),
1695                                      range_upb(pci_hole64));
1696         for (i = 0; i < crs_range_set.mem_64bit_ranges->len; i++) {
1697             entry = g_ptr_array_index(crs_range_set.mem_64bit_ranges, i);
1698             aml_append(crs,
1699                        aml_qword_memory(AML_POS_DECODE, AML_MIN_FIXED,
1700                                         AML_MAX_FIXED,
1701                                         AML_CACHEABLE, AML_READ_WRITE,
1702                                         0, entry->base, entry->limit,
1703                                         0, entry->limit - entry->base + 1));
1704         }
1705     }
1706 
1707     if (TPM_IS_TIS_ISA(tpm_find())) {
1708         aml_append(crs, aml_memory32_fixed(TPM_TIS_ADDR_BASE,
1709                    TPM_TIS_ADDR_SIZE, AML_READ_WRITE));
1710     }
1711     aml_append(scope, aml_name_decl("_CRS", crs));
1712 
1713     /* reserve GPE0 block resources */
1714     dev = aml_device("GPE0");
1715     aml_append(dev, aml_name_decl("_HID", aml_string("PNP0A06")));
1716     aml_append(dev, aml_name_decl("_UID", aml_string("GPE0 resources")));
1717     /* device present, functioning, decoding, not shown in UI */
1718     aml_append(dev, aml_name_decl("_STA", aml_int(0xB)));
1719     crs = aml_resource_template();
1720     aml_append(crs,
1721         aml_io(
1722                AML_DECODE16,
1723                pm->fadt.gpe0_blk.address,
1724                pm->fadt.gpe0_blk.address,
1725                1,
1726                pm->fadt.gpe0_blk.bit_width / 8)
1727     );
1728     aml_append(dev, aml_name_decl("_CRS", crs));
1729     aml_append(scope, dev);
1730 
1731     crs_range_set_free(&crs_range_set);
1732 
1733     /* reserve PCIHP resources */
1734     if (pm->pcihp_io_len) {
1735         dev = aml_device("PHPR");
1736         aml_append(dev, aml_name_decl("_HID", aml_string("PNP0A06")));
1737         aml_append(dev,
1738             aml_name_decl("_UID", aml_string("PCI Hotplug resources")));
1739         /* device present, functioning, decoding, not shown in UI */
1740         aml_append(dev, aml_name_decl("_STA", aml_int(0xB)));
1741         crs = aml_resource_template();
1742         aml_append(crs,
1743             aml_io(AML_DECODE16, pm->pcihp_io_base, pm->pcihp_io_base, 1,
1744                    pm->pcihp_io_len)
1745         );
1746         aml_append(dev, aml_name_decl("_CRS", crs));
1747         aml_append(scope, dev);
1748     }
1749     aml_append(dsdt, scope);
1750 
1751     /*  create S3_ / S4_ / S5_ packages if necessary */
1752     scope = aml_scope("\\");
1753     if (!pm->s3_disabled) {
1754         pkg = aml_package(4);
1755         aml_append(pkg, aml_int(1)); /* PM1a_CNT.SLP_TYP */
1756         aml_append(pkg, aml_int(1)); /* PM1b_CNT.SLP_TYP, FIXME: not impl. */
1757         aml_append(pkg, aml_int(0)); /* reserved */
1758         aml_append(pkg, aml_int(0)); /* reserved */
1759         aml_append(scope, aml_name_decl("_S3", pkg));
1760     }
1761 
1762     if (!pm->s4_disabled) {
1763         pkg = aml_package(4);
1764         aml_append(pkg, aml_int(pm->s4_val)); /* PM1a_CNT.SLP_TYP */
1765         /* PM1b_CNT.SLP_TYP, FIXME: not impl. */
1766         aml_append(pkg, aml_int(pm->s4_val));
1767         aml_append(pkg, aml_int(0)); /* reserved */
1768         aml_append(pkg, aml_int(0)); /* reserved */
1769         aml_append(scope, aml_name_decl("_S4", pkg));
1770     }
1771 
1772     pkg = aml_package(4);
1773     aml_append(pkg, aml_int(0)); /* PM1a_CNT.SLP_TYP */
1774     aml_append(pkg, aml_int(0)); /* PM1b_CNT.SLP_TYP not impl. */
1775     aml_append(pkg, aml_int(0)); /* reserved */
1776     aml_append(pkg, aml_int(0)); /* reserved */
1777     aml_append(scope, aml_name_decl("_S5", pkg));
1778     aml_append(dsdt, scope);
1779 
1780     /* create fw_cfg node, unconditionally */
1781     {
1782         scope = aml_scope("\\_SB.PCI0");
1783         fw_cfg_add_acpi_dsdt(scope, x86ms->fw_cfg);
1784         aml_append(dsdt, scope);
1785     }
1786 
1787     if (misc->applesmc_io_base) {
1788         scope = aml_scope("\\_SB.PCI0.ISA");
1789         dev = aml_device("SMC");
1790 
1791         aml_append(dev, aml_name_decl("_HID", aml_eisaid("APP0001")));
1792         /* device present, functioning, decoding, not shown in UI */
1793         aml_append(dev, aml_name_decl("_STA", aml_int(0xB)));
1794 
1795         crs = aml_resource_template();
1796         aml_append(crs,
1797             aml_io(AML_DECODE16, misc->applesmc_io_base, misc->applesmc_io_base,
1798                    0x01, APPLESMC_MAX_DATA_LENGTH)
1799         );
1800         aml_append(crs, aml_irq_no_flags(6));
1801         aml_append(dev, aml_name_decl("_CRS", crs));
1802 
1803         aml_append(scope, dev);
1804         aml_append(dsdt, scope);
1805     }
1806 
1807     if (misc->pvpanic_port) {
1808         scope = aml_scope("\\_SB.PCI0.ISA");
1809 
1810         dev = aml_device("PEVT");
1811         aml_append(dev, aml_name_decl("_HID", aml_string("QEMU0001")));
1812 
1813         crs = aml_resource_template();
1814         aml_append(crs,
1815             aml_io(AML_DECODE16, misc->pvpanic_port, misc->pvpanic_port, 1, 1)
1816         );
1817         aml_append(dev, aml_name_decl("_CRS", crs));
1818 
1819         aml_append(dev, aml_operation_region("PEOR", AML_SYSTEM_IO,
1820                                               aml_int(misc->pvpanic_port), 1));
1821         field = aml_field("PEOR", AML_BYTE_ACC, AML_NOLOCK, AML_PRESERVE);
1822         aml_append(field, aml_named_field("PEPT", 8));
1823         aml_append(dev, field);
1824 
1825         /* device present, functioning, decoding, shown in UI */
1826         aml_append(dev, aml_name_decl("_STA", aml_int(0xF)));
1827 
1828         method = aml_method("RDPT", 0, AML_NOTSERIALIZED);
1829         aml_append(method, aml_store(aml_name("PEPT"), aml_local(0)));
1830         aml_append(method, aml_return(aml_local(0)));
1831         aml_append(dev, method);
1832 
1833         method = aml_method("WRPT", 1, AML_NOTSERIALIZED);
1834         aml_append(method, aml_store(aml_arg(0), aml_name("PEPT")));
1835         aml_append(dev, method);
1836 
1837         aml_append(scope, dev);
1838         aml_append(dsdt, scope);
1839     }
1840 
1841     sb_scope = aml_scope("\\_SB");
1842     {
1843         Object *pci_host;
1844         PCIBus *bus = NULL;
1845 
1846         pci_host = acpi_get_i386_pci_host();
1847         if (pci_host) {
1848             bus = PCI_HOST_BRIDGE(pci_host)->bus;
1849         }
1850 
1851         if (bus) {
1852             Aml *scope = aml_scope("PCI0");
1853             /* Scan all PCI buses. Generate tables to support hotplug. */
1854             build_append_pci_bus_devices(scope, bus, pm->pcihp_bridge_en);
1855 
1856             if (TPM_IS_TIS_ISA(tpm)) {
1857                 if (misc->tpm_version == TPM_VERSION_2_0) {
1858                     dev = aml_device("TPM");
1859                     aml_append(dev, aml_name_decl("_HID",
1860                                                   aml_string("MSFT0101")));
1861                 } else {
1862                     dev = aml_device("ISA.TPM");
1863                     aml_append(dev, aml_name_decl("_HID",
1864                                                   aml_eisaid("PNP0C31")));
1865                 }
1866 
1867                 aml_append(dev, aml_name_decl("_STA", aml_int(0xF)));
1868                 crs = aml_resource_template();
1869                 aml_append(crs, aml_memory32_fixed(TPM_TIS_ADDR_BASE,
1870                            TPM_TIS_ADDR_SIZE, AML_READ_WRITE));
1871                 /*
1872                     FIXME: TPM_TIS_IRQ=5 conflicts with PNP0C0F irqs,
1873                     Rewrite to take IRQ from TPM device model and
1874                     fix default IRQ value there to use some unused IRQ
1875                  */
1876                 /* aml_append(crs, aml_irq_no_flags(TPM_TIS_IRQ)); */
1877                 aml_append(dev, aml_name_decl("_CRS", crs));
1878 
1879                 tpm_build_ppi_acpi(tpm, dev);
1880 
1881                 aml_append(scope, dev);
1882             }
1883 
1884             aml_append(sb_scope, scope);
1885         }
1886     }
1887 
1888     if (TPM_IS_CRB(tpm)) {
1889         dev = aml_device("TPM");
1890         aml_append(dev, aml_name_decl("_HID", aml_string("MSFT0101")));
1891         crs = aml_resource_template();
1892         aml_append(crs, aml_memory32_fixed(TPM_CRB_ADDR_BASE,
1893                                            TPM_CRB_ADDR_SIZE, AML_READ_WRITE));
1894         aml_append(dev, aml_name_decl("_CRS", crs));
1895 
1896         aml_append(dev, aml_name_decl("_STA", aml_int(0xf)));
1897 
1898         tpm_build_ppi_acpi(tpm, dev);
1899 
1900         aml_append(sb_scope, dev);
1901     }
1902 
1903     aml_append(dsdt, sb_scope);
1904 
1905     /* copy AML table into ACPI tables blob and patch header there */
1906     g_array_append_vals(table_data, dsdt->buf->data, dsdt->buf->len);
1907     build_header(linker, table_data,
1908         (void *)(table_data->data + table_data->len - dsdt->buf->len),
1909         "DSDT", dsdt->buf->len, 1, NULL, NULL);
1910     free_aml_allocator();
1911 }
1912 
1913 static void
1914 build_hpet(GArray *table_data, BIOSLinker *linker)
1915 {
1916     Acpi20Hpet *hpet;
1917 
1918     hpet = acpi_data_push(table_data, sizeof(*hpet));
1919     /* Note timer_block_id value must be kept in sync with value advertised by
1920      * emulated hpet
1921      */
1922     hpet->timer_block_id = cpu_to_le32(0x8086a201);
1923     hpet->addr.address = cpu_to_le64(HPET_BASE);
1924     build_header(linker, table_data,
1925                  (void *)hpet, "HPET", sizeof(*hpet), 1, NULL, NULL);
1926 }
1927 
1928 static void
1929 build_tpm_tcpa(GArray *table_data, BIOSLinker *linker, GArray *tcpalog)
1930 {
1931     Acpi20Tcpa *tcpa = acpi_data_push(table_data, sizeof *tcpa);
1932     unsigned log_addr_size = sizeof(tcpa->log_area_start_address);
1933     unsigned log_addr_offset =
1934         (char *)&tcpa->log_area_start_address - table_data->data;
1935 
1936     tcpa->platform_class = cpu_to_le16(TPM_TCPA_ACPI_CLASS_CLIENT);
1937     tcpa->log_area_minimum_length = cpu_to_le32(TPM_LOG_AREA_MINIMUM_SIZE);
1938     acpi_data_push(tcpalog, le32_to_cpu(tcpa->log_area_minimum_length));
1939 
1940     bios_linker_loader_alloc(linker, ACPI_BUILD_TPMLOG_FILE, tcpalog, 1,
1941                              false /* high memory */);
1942 
1943     /* log area start address to be filled by Guest linker */
1944     bios_linker_loader_add_pointer(linker,
1945         ACPI_BUILD_TABLE_FILE, log_addr_offset, log_addr_size,
1946         ACPI_BUILD_TPMLOG_FILE, 0);
1947 
1948     build_header(linker, table_data,
1949                  (void *)tcpa, "TCPA", sizeof(*tcpa), 2, NULL, NULL);
1950 }
1951 
1952 #define HOLE_640K_START  (640 * KiB)
1953 #define HOLE_640K_END   (1 * MiB)
1954 
1955 static void
1956 build_srat(GArray *table_data, BIOSLinker *linker, MachineState *machine)
1957 {
1958     AcpiSystemResourceAffinityTable *srat;
1959     AcpiSratMemoryAffinity *numamem;
1960 
1961     int i;
1962     int srat_start, numa_start, slots;
1963     uint64_t mem_len, mem_base, next_base;
1964     MachineClass *mc = MACHINE_GET_CLASS(machine);
1965     X86MachineState *x86ms = X86_MACHINE(machine);
1966     const CPUArchIdList *apic_ids = mc->possible_cpu_arch_ids(machine);
1967     PCMachineState *pcms = PC_MACHINE(machine);
1968     ram_addr_t hotplugabble_address_space_size =
1969         object_property_get_int(OBJECT(pcms), PC_MACHINE_DEVMEM_REGION_SIZE,
1970                                 NULL);
1971 
1972     srat_start = table_data->len;
1973 
1974     srat = acpi_data_push(table_data, sizeof *srat);
1975     srat->reserved1 = cpu_to_le32(1);
1976 
1977     for (i = 0; i < apic_ids->len; i++) {
1978         int node_id = apic_ids->cpus[i].props.node_id;
1979         uint32_t apic_id = apic_ids->cpus[i].arch_id;
1980 
1981         if (apic_id < 255) {
1982             AcpiSratProcessorAffinity *core;
1983 
1984             core = acpi_data_push(table_data, sizeof *core);
1985             core->type = ACPI_SRAT_PROCESSOR_APIC;
1986             core->length = sizeof(*core);
1987             core->local_apic_id = apic_id;
1988             core->proximity_lo = node_id;
1989             memset(core->proximity_hi, 0, 3);
1990             core->local_sapic_eid = 0;
1991             core->flags = cpu_to_le32(1);
1992         } else {
1993             AcpiSratProcessorX2ApicAffinity *core;
1994 
1995             core = acpi_data_push(table_data, sizeof *core);
1996             core->type = ACPI_SRAT_PROCESSOR_x2APIC;
1997             core->length = sizeof(*core);
1998             core->x2apic_id = cpu_to_le32(apic_id);
1999             core->proximity_domain = cpu_to_le32(node_id);
2000             core->flags = cpu_to_le32(1);
2001         }
2002     }
2003 
2004 
2005     /* the memory map is a bit tricky, it contains at least one hole
2006      * from 640k-1M and possibly another one from 3.5G-4G.
2007      */
2008     next_base = 0;
2009     numa_start = table_data->len;
2010 
2011     for (i = 1; i < pcms->numa_nodes + 1; ++i) {
2012         mem_base = next_base;
2013         mem_len = pcms->node_mem[i - 1];
2014         next_base = mem_base + mem_len;
2015 
2016         /* Cut out the 640K hole */
2017         if (mem_base <= HOLE_640K_START &&
2018             next_base > HOLE_640K_START) {
2019             mem_len -= next_base - HOLE_640K_START;
2020             if (mem_len > 0) {
2021                 numamem = acpi_data_push(table_data, sizeof *numamem);
2022                 build_srat_memory(numamem, mem_base, mem_len, i - 1,
2023                                   MEM_AFFINITY_ENABLED);
2024             }
2025 
2026             /* Check for the rare case: 640K < RAM < 1M */
2027             if (next_base <= HOLE_640K_END) {
2028                 next_base = HOLE_640K_END;
2029                 continue;
2030             }
2031             mem_base = HOLE_640K_END;
2032             mem_len = next_base - HOLE_640K_END;
2033         }
2034 
2035         /* Cut out the ACPI_PCI hole */
2036         if (mem_base <= x86ms->below_4g_mem_size &&
2037             next_base > x86ms->below_4g_mem_size) {
2038             mem_len -= next_base - x86ms->below_4g_mem_size;
2039             if (mem_len > 0) {
2040                 numamem = acpi_data_push(table_data, sizeof *numamem);
2041                 build_srat_memory(numamem, mem_base, mem_len, i - 1,
2042                                   MEM_AFFINITY_ENABLED);
2043             }
2044             mem_base = 1ULL << 32;
2045             mem_len = next_base - x86ms->below_4g_mem_size;
2046             next_base = mem_base + mem_len;
2047         }
2048 
2049         if (mem_len > 0) {
2050             numamem = acpi_data_push(table_data, sizeof *numamem);
2051             build_srat_memory(numamem, mem_base, mem_len, i - 1,
2052                               MEM_AFFINITY_ENABLED);
2053         }
2054     }
2055 
2056     if (machine->nvdimms_state->is_enabled) {
2057         nvdimm_build_srat(table_data);
2058     }
2059 
2060     slots = (table_data->len - numa_start) / sizeof *numamem;
2061     for (; slots < pcms->numa_nodes + 2; slots++) {
2062         numamem = acpi_data_push(table_data, sizeof *numamem);
2063         build_srat_memory(numamem, 0, 0, 0, MEM_AFFINITY_NOFLAGS);
2064     }
2065 
2066     /*
2067      * Entry is required for Windows to enable memory hotplug in OS
2068      * and for Linux to enable SWIOTLB when booted with less than
2069      * 4G of RAM. Windows works better if the entry sets proximity
2070      * to the highest NUMA node in the machine.
2071      * Memory devices may override proximity set by this entry,
2072      * providing _PXM method if necessary.
2073      */
2074     if (hotplugabble_address_space_size) {
2075         numamem = acpi_data_push(table_data, sizeof *numamem);
2076         build_srat_memory(numamem, machine->device_memory->base,
2077                           hotplugabble_address_space_size, pcms->numa_nodes - 1,
2078                           MEM_AFFINITY_HOTPLUGGABLE | MEM_AFFINITY_ENABLED);
2079     }
2080 
2081     build_header(linker, table_data,
2082                  (void *)(table_data->data + srat_start),
2083                  "SRAT",
2084                  table_data->len - srat_start, 1, NULL, NULL);
2085 }
2086 
2087 /*
2088  * VT-d spec 8.1 DMA Remapping Reporting Structure
2089  * (version Oct. 2014 or later)
2090  */
2091 static void
2092 build_dmar_q35(GArray *table_data, BIOSLinker *linker)
2093 {
2094     int dmar_start = table_data->len;
2095 
2096     AcpiTableDmar *dmar;
2097     AcpiDmarHardwareUnit *drhd;
2098     AcpiDmarRootPortATS *atsr;
2099     uint8_t dmar_flags = 0;
2100     X86IOMMUState *iommu = x86_iommu_get_default();
2101     AcpiDmarDeviceScope *scope = NULL;
2102     /* Root complex IOAPIC use one path[0] only */
2103     size_t ioapic_scope_size = sizeof(*scope) + sizeof(scope->path[0]);
2104     IntelIOMMUState *intel_iommu = INTEL_IOMMU_DEVICE(iommu);
2105 
2106     assert(iommu);
2107     if (x86_iommu_ir_supported(iommu)) {
2108         dmar_flags |= 0x1;      /* Flags: 0x1: INT_REMAP */
2109     }
2110 
2111     dmar = acpi_data_push(table_data, sizeof(*dmar));
2112     dmar->host_address_width = intel_iommu->aw_bits - 1;
2113     dmar->flags = dmar_flags;
2114 
2115     /* DMAR Remapping Hardware Unit Definition structure */
2116     drhd = acpi_data_push(table_data, sizeof(*drhd) + ioapic_scope_size);
2117     drhd->type = cpu_to_le16(ACPI_DMAR_TYPE_HARDWARE_UNIT);
2118     drhd->length = cpu_to_le16(sizeof(*drhd) + ioapic_scope_size);
2119     drhd->flags = ACPI_DMAR_INCLUDE_PCI_ALL;
2120     drhd->pci_segment = cpu_to_le16(0);
2121     drhd->address = cpu_to_le64(Q35_HOST_BRIDGE_IOMMU_ADDR);
2122 
2123     /* Scope definition for the root-complex IOAPIC. See VT-d spec
2124      * 8.3.1 (version Oct. 2014 or later). */
2125     scope = &drhd->scope[0];
2126     scope->entry_type = 0x03;   /* Type: 0x03 for IOAPIC */
2127     scope->length = ioapic_scope_size;
2128     scope->enumeration_id = ACPI_BUILD_IOAPIC_ID;
2129     scope->bus = Q35_PSEUDO_BUS_PLATFORM;
2130     scope->path[0].device = PCI_SLOT(Q35_PSEUDO_DEVFN_IOAPIC);
2131     scope->path[0].function = PCI_FUNC(Q35_PSEUDO_DEVFN_IOAPIC);
2132 
2133     if (iommu->dt_supported) {
2134         atsr = acpi_data_push(table_data, sizeof(*atsr));
2135         atsr->type = cpu_to_le16(ACPI_DMAR_TYPE_ATSR);
2136         atsr->length = cpu_to_le16(sizeof(*atsr));
2137         atsr->flags = ACPI_DMAR_ATSR_ALL_PORTS;
2138         atsr->pci_segment = cpu_to_le16(0);
2139     }
2140 
2141     build_header(linker, table_data, (void *)(table_data->data + dmar_start),
2142                  "DMAR", table_data->len - dmar_start, 1, NULL, NULL);
2143 }
2144 
2145 /*
2146  * Windows ACPI Emulated Devices Table
2147  * (Version 1.0 - April 6, 2009)
2148  * Spec: http://download.microsoft.com/download/7/E/7/7E7662CF-CBEA-470B-A97E-CE7CE0D98DC2/WAET.docx
2149  *
2150  * Helpful to speedup Windows guests and ignored by others.
2151  */
2152 static void
2153 build_waet(GArray *table_data, BIOSLinker *linker)
2154 {
2155     int waet_start = table_data->len;
2156 
2157     /* WAET header */
2158     acpi_data_push(table_data, sizeof(AcpiTableHeader));
2159     /*
2160      * Set "ACPI PM timer good" flag.
2161      *
2162      * Tells Windows guests that our ACPI PM timer is reliable in the
2163      * sense that guest can read it only once to obtain a reliable value.
2164      * Which avoids costly VMExits caused by guest re-reading it unnecessarily.
2165      */
2166     build_append_int_noprefix(table_data, 1 << 1 /* ACPI PM timer good */, 4);
2167 
2168     build_header(linker, table_data, (void *)(table_data->data + waet_start),
2169                  "WAET", table_data->len - waet_start, 1, NULL, NULL);
2170 }
2171 
2172 /*
2173  *   IVRS table as specified in AMD IOMMU Specification v2.62, Section 5.2
2174  *   accessible here http://support.amd.com/TechDocs/48882_IOMMU.pdf
2175  */
2176 #define IOAPIC_SB_DEVID   (uint64_t)PCI_BUILD_BDF(0, PCI_DEVFN(0x14, 0))
2177 
2178 /*
2179  * Insert IVHD entry for device and recurse, insert alias, or insert range as
2180  * necessary for the PCI topology.
2181  */
2182 static void
2183 insert_ivhd(PCIBus *bus, PCIDevice *dev, void *opaque)
2184 {
2185     GArray *table_data = opaque;
2186     uint32_t entry;
2187 
2188     /* "Select" IVHD entry, type 0x2 */
2189     entry = PCI_BUILD_BDF(pci_bus_num(bus), dev->devfn) << 8 | 0x2;
2190     build_append_int_noprefix(table_data, entry, 4);
2191 
2192     if (object_dynamic_cast(OBJECT(dev), TYPE_PCI_BRIDGE)) {
2193         PCIBus *sec_bus = pci_bridge_get_sec_bus(PCI_BRIDGE(dev));
2194         uint8_t sec = pci_bus_num(sec_bus);
2195         uint8_t sub = dev->config[PCI_SUBORDINATE_BUS];
2196 
2197         if (pci_bus_is_express(sec_bus)) {
2198             /*
2199              * Walk the bus if there are subordinates, otherwise use a range
2200              * to cover an entire leaf bus.  We could potentially also use a
2201              * range for traversed buses, but we'd need to take care not to
2202              * create both Select and Range entries covering the same device.
2203              * This is easier and potentially more compact.
2204              *
2205              * An example bare metal system seems to use Select entries for
2206              * root ports without a slot (ie. built-ins) and Range entries
2207              * when there is a slot.  The same system also only hard-codes
2208              * the alias range for an onboard PCIe-to-PCI bridge, apparently
2209              * making no effort to support nested bridges.  We attempt to
2210              * be more thorough here.
2211              */
2212             if (sec == sub) { /* leaf bus */
2213                 /* "Start of Range" IVHD entry, type 0x3 */
2214                 entry = PCI_BUILD_BDF(sec, PCI_DEVFN(0, 0)) << 8 | 0x3;
2215                 build_append_int_noprefix(table_data, entry, 4);
2216                 /* "End of Range" IVHD entry, type 0x4 */
2217                 entry = PCI_BUILD_BDF(sub, PCI_DEVFN(31, 7)) << 8 | 0x4;
2218                 build_append_int_noprefix(table_data, entry, 4);
2219             } else {
2220                 pci_for_each_device(sec_bus, sec, insert_ivhd, table_data);
2221             }
2222         } else {
2223             /*
2224              * If the secondary bus is conventional, then we need to create an
2225              * Alias range for everything downstream.  The range covers the
2226              * first devfn on the secondary bus to the last devfn on the
2227              * subordinate bus.  The alias target depends on legacy versus
2228              * express bridges, just as in pci_device_iommu_address_space().
2229              * DeviceIDa vs DeviceIDb as per the AMD IOMMU spec.
2230              */
2231             uint16_t dev_id_a, dev_id_b;
2232 
2233             dev_id_a = PCI_BUILD_BDF(sec, PCI_DEVFN(0, 0));
2234 
2235             if (pci_is_express(dev) &&
2236                 pcie_cap_get_type(dev) == PCI_EXP_TYPE_PCI_BRIDGE) {
2237                 dev_id_b = dev_id_a;
2238             } else {
2239                 dev_id_b = PCI_BUILD_BDF(pci_bus_num(bus), dev->devfn);
2240             }
2241 
2242             /* "Alias Start of Range" IVHD entry, type 0x43, 8 bytes */
2243             build_append_int_noprefix(table_data, dev_id_a << 8 | 0x43, 4);
2244             build_append_int_noprefix(table_data, dev_id_b << 8 | 0x0, 4);
2245 
2246             /* "End of Range" IVHD entry, type 0x4 */
2247             entry = PCI_BUILD_BDF(sub, PCI_DEVFN(31, 7)) << 8 | 0x4;
2248             build_append_int_noprefix(table_data, entry, 4);
2249         }
2250     }
2251 }
2252 
2253 /* For all PCI host bridges, walk and insert IVHD entries */
2254 static int
2255 ivrs_host_bridges(Object *obj, void *opaque)
2256 {
2257     GArray *ivhd_blob = opaque;
2258 
2259     if (object_dynamic_cast(obj, TYPE_PCI_HOST_BRIDGE)) {
2260         PCIBus *bus = PCI_HOST_BRIDGE(obj)->bus;
2261 
2262         if (bus) {
2263             pci_for_each_device(bus, pci_bus_num(bus), insert_ivhd, ivhd_blob);
2264         }
2265     }
2266 
2267     return 0;
2268 }
2269 
2270 static void
2271 build_amd_iommu(GArray *table_data, BIOSLinker *linker)
2272 {
2273     int ivhd_table_len = 24;
2274     int iommu_start = table_data->len;
2275     AMDVIState *s = AMD_IOMMU_DEVICE(x86_iommu_get_default());
2276     GArray *ivhd_blob = g_array_new(false, true, 1);
2277 
2278     /* IVRS header */
2279     acpi_data_push(table_data, sizeof(AcpiTableHeader));
2280     /* IVinfo - IO virtualization information common to all
2281      * IOMMU units in a system
2282      */
2283     build_append_int_noprefix(table_data, 40UL << 8/* PASize */, 4);
2284     /* reserved */
2285     build_append_int_noprefix(table_data, 0, 8);
2286 
2287     /* IVHD definition - type 10h */
2288     build_append_int_noprefix(table_data, 0x10, 1);
2289     /* virtualization flags */
2290     build_append_int_noprefix(table_data,
2291                              (1UL << 0) | /* HtTunEn      */
2292                              (1UL << 4) | /* iotblSup     */
2293                              (1UL << 6) | /* PrefSup      */
2294                              (1UL << 7),  /* PPRSup       */
2295                              1);
2296 
2297     /*
2298      * A PCI bus walk, for each PCI host bridge, is necessary to create a
2299      * complete set of IVHD entries.  Do this into a separate blob so that we
2300      * can calculate the total IVRS table length here and then append the new
2301      * blob further below.  Fall back to an entry covering all devices, which
2302      * is sufficient when no aliases are present.
2303      */
2304     object_child_foreach_recursive(object_get_root(),
2305                                    ivrs_host_bridges, ivhd_blob);
2306 
2307     if (!ivhd_blob->len) {
2308         /*
2309          *   Type 1 device entry reporting all devices
2310          *   These are 4-byte device entries currently reporting the range of
2311          *   Refer to Spec - Table 95:IVHD Device Entry Type Codes(4-byte)
2312          */
2313         build_append_int_noprefix(ivhd_blob, 0x0000001, 4);
2314     }
2315 
2316     ivhd_table_len += ivhd_blob->len;
2317 
2318     /*
2319      * When interrupt remapping is supported, we add a special IVHD device
2320      * for type IO-APIC.
2321      */
2322     if (x86_iommu_ir_supported(x86_iommu_get_default())) {
2323         ivhd_table_len += 8;
2324     }
2325 
2326     /* IVHD length */
2327     build_append_int_noprefix(table_data, ivhd_table_len, 2);
2328     /* DeviceID */
2329     build_append_int_noprefix(table_data, s->devid, 2);
2330     /* Capability offset */
2331     build_append_int_noprefix(table_data, s->capab_offset, 2);
2332     /* IOMMU base address */
2333     build_append_int_noprefix(table_data, s->mmio.addr, 8);
2334     /* PCI Segment Group */
2335     build_append_int_noprefix(table_data, 0, 2);
2336     /* IOMMU info */
2337     build_append_int_noprefix(table_data, 0, 2);
2338     /* IOMMU Feature Reporting */
2339     build_append_int_noprefix(table_data,
2340                              (48UL << 30) | /* HATS   */
2341                              (48UL << 28) | /* GATS   */
2342                              (1UL << 2)   | /* GTSup  */
2343                              (1UL << 6),    /* GASup  */
2344                              4);
2345 
2346     /* IVHD entries as found above */
2347     g_array_append_vals(table_data, ivhd_blob->data, ivhd_blob->len);
2348     g_array_free(ivhd_blob, TRUE);
2349 
2350     /*
2351      * Add a special IVHD device type.
2352      * Refer to spec - Table 95: IVHD device entry type codes
2353      *
2354      * Linux IOMMU driver checks for the special IVHD device (type IO-APIC).
2355      * See Linux kernel commit 'c2ff5cf5294bcbd7fa50f7d860e90a66db7e5059'
2356      */
2357     if (x86_iommu_ir_supported(x86_iommu_get_default())) {
2358         build_append_int_noprefix(table_data,
2359                                  (0x1ull << 56) |           /* type IOAPIC */
2360                                  (IOAPIC_SB_DEVID << 40) |  /* IOAPIC devid */
2361                                  0x48,                      /* special device */
2362                                  8);
2363     }
2364 
2365     build_header(linker, table_data, (void *)(table_data->data + iommu_start),
2366                  "IVRS", table_data->len - iommu_start, 1, NULL, NULL);
2367 }
2368 
2369 typedef
2370 struct AcpiBuildState {
2371     /* Copy of table in RAM (for patching). */
2372     MemoryRegion *table_mr;
2373     /* Is table patched? */
2374     uint8_t patched;
2375     void *rsdp;
2376     MemoryRegion *rsdp_mr;
2377     MemoryRegion *linker_mr;
2378 } AcpiBuildState;
2379 
2380 static bool acpi_get_mcfg(AcpiMcfgInfo *mcfg)
2381 {
2382     Object *pci_host;
2383     QObject *o;
2384 
2385     pci_host = acpi_get_i386_pci_host();
2386     g_assert(pci_host);
2387 
2388     o = object_property_get_qobject(pci_host, PCIE_HOST_MCFG_BASE, NULL);
2389     if (!o) {
2390         return false;
2391     }
2392     mcfg->base = qnum_get_uint(qobject_to(QNum, o));
2393     qobject_unref(o);
2394     if (mcfg->base == PCIE_BASE_ADDR_UNMAPPED) {
2395         return false;
2396     }
2397 
2398     o = object_property_get_qobject(pci_host, PCIE_HOST_MCFG_SIZE, NULL);
2399     assert(o);
2400     mcfg->size = qnum_get_uint(qobject_to(QNum, o));
2401     qobject_unref(o);
2402     return true;
2403 }
2404 
2405 static
2406 void acpi_build(AcpiBuildTables *tables, MachineState *machine)
2407 {
2408     PCMachineState *pcms = PC_MACHINE(machine);
2409     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
2410     X86MachineState *x86ms = X86_MACHINE(machine);
2411     GArray *table_offsets;
2412     unsigned facs, dsdt, rsdt, fadt;
2413     AcpiPmInfo pm;
2414     AcpiMiscInfo misc;
2415     AcpiMcfgInfo mcfg;
2416     Range pci_hole, pci_hole64;
2417     uint8_t *u;
2418     size_t aml_len = 0;
2419     GArray *tables_blob = tables->table_data;
2420     AcpiSlicOem slic_oem = { .id = NULL, .table_id = NULL };
2421     Object *vmgenid_dev;
2422 
2423     acpi_get_pm_info(machine, &pm);
2424     acpi_get_misc_info(&misc);
2425     acpi_get_pci_holes(&pci_hole, &pci_hole64);
2426     acpi_get_slic_oem(&slic_oem);
2427 
2428     table_offsets = g_array_new(false, true /* clear */,
2429                                         sizeof(uint32_t));
2430     ACPI_BUILD_DPRINTF("init ACPI tables\n");
2431 
2432     bios_linker_loader_alloc(tables->linker,
2433                              ACPI_BUILD_TABLE_FILE, tables_blob,
2434                              64 /* Ensure FACS is aligned */,
2435                              false /* high memory */);
2436 
2437     /*
2438      * FACS is pointed to by FADT.
2439      * We place it first since it's the only table that has alignment
2440      * requirements.
2441      */
2442     facs = tables_blob->len;
2443     build_facs(tables_blob);
2444 
2445     /* DSDT is pointed to by FADT */
2446     dsdt = tables_blob->len;
2447     build_dsdt(tables_blob, tables->linker, &pm, &misc,
2448                &pci_hole, &pci_hole64, machine);
2449 
2450     /* Count the size of the DSDT and SSDT, we will need it for legacy
2451      * sizing of ACPI tables.
2452      */
2453     aml_len += tables_blob->len - dsdt;
2454 
2455     /* ACPI tables pointed to by RSDT */
2456     fadt = tables_blob->len;
2457     acpi_add_table(table_offsets, tables_blob);
2458     pm.fadt.facs_tbl_offset = &facs;
2459     pm.fadt.dsdt_tbl_offset = &dsdt;
2460     pm.fadt.xdsdt_tbl_offset = &dsdt;
2461     build_fadt(tables_blob, tables->linker, &pm.fadt,
2462                slic_oem.id, slic_oem.table_id);
2463     aml_len += tables_blob->len - fadt;
2464 
2465     acpi_add_table(table_offsets, tables_blob);
2466     acpi_build_madt(tables_blob, tables->linker, x86ms,
2467                     ACPI_DEVICE_IF(x86ms->acpi_dev), true);
2468 
2469     vmgenid_dev = find_vmgenid_dev();
2470     if (vmgenid_dev) {
2471         acpi_add_table(table_offsets, tables_blob);
2472         vmgenid_build_acpi(VMGENID(vmgenid_dev), tables_blob,
2473                            tables->vmgenid, tables->linker);
2474     }
2475 
2476     if (misc.has_hpet) {
2477         acpi_add_table(table_offsets, tables_blob);
2478         build_hpet(tables_blob, tables->linker);
2479     }
2480     if (misc.tpm_version != TPM_VERSION_UNSPEC) {
2481         if (misc.tpm_version == TPM_VERSION_1_2) {
2482             acpi_add_table(table_offsets, tables_blob);
2483             build_tpm_tcpa(tables_blob, tables->linker, tables->tcpalog);
2484         } else { /* TPM_VERSION_2_0 */
2485             acpi_add_table(table_offsets, tables_blob);
2486             build_tpm2(tables_blob, tables->linker, tables->tcpalog);
2487         }
2488     }
2489     if (pcms->numa_nodes) {
2490         acpi_add_table(table_offsets, tables_blob);
2491         build_srat(tables_blob, tables->linker, machine);
2492         if (machine->numa_state->have_numa_distance) {
2493             acpi_add_table(table_offsets, tables_blob);
2494             build_slit(tables_blob, tables->linker, machine);
2495         }
2496         if (machine->numa_state->hmat_enabled) {
2497             acpi_add_table(table_offsets, tables_blob);
2498             build_hmat(tables_blob, tables->linker, machine->numa_state);
2499         }
2500     }
2501     if (acpi_get_mcfg(&mcfg)) {
2502         acpi_add_table(table_offsets, tables_blob);
2503         build_mcfg(tables_blob, tables->linker, &mcfg);
2504     }
2505     if (x86_iommu_get_default()) {
2506         IommuType IOMMUType = x86_iommu_get_type();
2507         if (IOMMUType == TYPE_AMD) {
2508             acpi_add_table(table_offsets, tables_blob);
2509             build_amd_iommu(tables_blob, tables->linker);
2510         } else if (IOMMUType == TYPE_INTEL) {
2511             acpi_add_table(table_offsets, tables_blob);
2512             build_dmar_q35(tables_blob, tables->linker);
2513         }
2514     }
2515     if (machine->nvdimms_state->is_enabled) {
2516         nvdimm_build_acpi(table_offsets, tables_blob, tables->linker,
2517                           machine->nvdimms_state, machine->ram_slots);
2518     }
2519 
2520     acpi_add_table(table_offsets, tables_blob);
2521     build_waet(tables_blob, tables->linker);
2522 
2523     /* Add tables supplied by user (if any) */
2524     for (u = acpi_table_first(); u; u = acpi_table_next(u)) {
2525         unsigned len = acpi_table_len(u);
2526 
2527         acpi_add_table(table_offsets, tables_blob);
2528         g_array_append_vals(tables_blob, u, len);
2529     }
2530 
2531     /* RSDT is pointed to by RSDP */
2532     rsdt = tables_blob->len;
2533     build_rsdt(tables_blob, tables->linker, table_offsets,
2534                slic_oem.id, slic_oem.table_id);
2535 
2536     /* RSDP is in FSEG memory, so allocate it separately */
2537     {
2538         AcpiRsdpData rsdp_data = {
2539             .revision = 0,
2540             .oem_id = ACPI_BUILD_APPNAME6,
2541             .xsdt_tbl_offset = NULL,
2542             .rsdt_tbl_offset = &rsdt,
2543         };
2544         build_rsdp(tables->rsdp, tables->linker, &rsdp_data);
2545         if (!pcmc->rsdp_in_ram) {
2546             /* We used to allocate some extra space for RSDP revision 2 but
2547              * only used the RSDP revision 0 space. The extra bytes were
2548              * zeroed out and not used.
2549              * Here we continue wasting those extra 16 bytes to make sure we
2550              * don't break migration for machine types 2.2 and older due to
2551              * RSDP blob size mismatch.
2552              */
2553             build_append_int_noprefix(tables->rsdp, 0, 16);
2554         }
2555     }
2556 
2557     /* We'll expose it all to Guest so we want to reduce
2558      * chance of size changes.
2559      *
2560      * We used to align the tables to 4k, but of course this would
2561      * too simple to be enough.  4k turned out to be too small an
2562      * alignment very soon, and in fact it is almost impossible to
2563      * keep the table size stable for all (max_cpus, max_memory_slots)
2564      * combinations.  So the table size is always 64k for pc-i440fx-2.1
2565      * and we give an error if the table grows beyond that limit.
2566      *
2567      * We still have the problem of migrating from "-M pc-i440fx-2.0".  For
2568      * that, we exploit the fact that QEMU 2.1 generates _smaller_ tables
2569      * than 2.0 and we can always pad the smaller tables with zeros.  We can
2570      * then use the exact size of the 2.0 tables.
2571      *
2572      * All this is for PIIX4, since QEMU 2.0 didn't support Q35 migration.
2573      */
2574     if (pcmc->legacy_acpi_table_size) {
2575         /* Subtracting aml_len gives the size of fixed tables.  Then add the
2576          * size of the PIIX4 DSDT/SSDT in QEMU 2.0.
2577          */
2578         int legacy_aml_len =
2579             pcmc->legacy_acpi_table_size +
2580             ACPI_BUILD_LEGACY_CPU_AML_SIZE * x86ms->apic_id_limit;
2581         int legacy_table_size =
2582             ROUND_UP(tables_blob->len - aml_len + legacy_aml_len,
2583                      ACPI_BUILD_ALIGN_SIZE);
2584         if (tables_blob->len > legacy_table_size) {
2585             /* Should happen only with PCI bridges and -M pc-i440fx-2.0.  */
2586             warn_report("ACPI table size %u exceeds %d bytes,"
2587                         " migration may not work",
2588                         tables_blob->len, legacy_table_size);
2589             error_printf("Try removing CPUs, NUMA nodes, memory slots"
2590                          " or PCI bridges.");
2591         }
2592         g_array_set_size(tables_blob, legacy_table_size);
2593     } else {
2594         /* Make sure we have a buffer in case we need to resize the tables. */
2595         if (tables_blob->len > ACPI_BUILD_TABLE_SIZE / 2) {
2596             /* As of QEMU 2.1, this fires with 160 VCPUs and 255 memory slots.  */
2597             warn_report("ACPI table size %u exceeds %d bytes,"
2598                         " migration may not work",
2599                         tables_blob->len, ACPI_BUILD_TABLE_SIZE / 2);
2600             error_printf("Try removing CPUs, NUMA nodes, memory slots"
2601                          " or PCI bridges.");
2602         }
2603         acpi_align_size(tables_blob, ACPI_BUILD_TABLE_SIZE);
2604     }
2605 
2606     acpi_align_size(tables->linker->cmd_blob, ACPI_BUILD_ALIGN_SIZE);
2607 
2608     /* Cleanup memory that's no longer used. */
2609     g_array_free(table_offsets, true);
2610 }
2611 
2612 static void acpi_ram_update(MemoryRegion *mr, GArray *data)
2613 {
2614     uint32_t size = acpi_data_len(data);
2615 
2616     /* Make sure RAM size is correct - in case it got changed e.g. by migration */
2617     memory_region_ram_resize(mr, size, &error_abort);
2618 
2619     memcpy(memory_region_get_ram_ptr(mr), data->data, size);
2620     memory_region_set_dirty(mr, 0, size);
2621 }
2622 
2623 static void acpi_build_update(void *build_opaque)
2624 {
2625     AcpiBuildState *build_state = build_opaque;
2626     AcpiBuildTables tables;
2627 
2628     /* No state to update or already patched? Nothing to do. */
2629     if (!build_state || build_state->patched) {
2630         return;
2631     }
2632     build_state->patched = 1;
2633 
2634     acpi_build_tables_init(&tables);
2635 
2636     acpi_build(&tables, MACHINE(qdev_get_machine()));
2637 
2638     acpi_ram_update(build_state->table_mr, tables.table_data);
2639 
2640     if (build_state->rsdp) {
2641         memcpy(build_state->rsdp, tables.rsdp->data, acpi_data_len(tables.rsdp));
2642     } else {
2643         acpi_ram_update(build_state->rsdp_mr, tables.rsdp);
2644     }
2645 
2646     acpi_ram_update(build_state->linker_mr, tables.linker->cmd_blob);
2647     acpi_build_tables_cleanup(&tables, true);
2648 }
2649 
2650 static void acpi_build_reset(void *build_opaque)
2651 {
2652     AcpiBuildState *build_state = build_opaque;
2653     build_state->patched = 0;
2654 }
2655 
2656 static const VMStateDescription vmstate_acpi_build = {
2657     .name = "acpi_build",
2658     .version_id = 1,
2659     .minimum_version_id = 1,
2660     .fields = (VMStateField[]) {
2661         VMSTATE_UINT8(patched, AcpiBuildState),
2662         VMSTATE_END_OF_LIST()
2663     },
2664 };
2665 
2666 void acpi_setup(void)
2667 {
2668     PCMachineState *pcms = PC_MACHINE(qdev_get_machine());
2669     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
2670     X86MachineState *x86ms = X86_MACHINE(pcms);
2671     AcpiBuildTables tables;
2672     AcpiBuildState *build_state;
2673     Object *vmgenid_dev;
2674     TPMIf *tpm;
2675     static FwCfgTPMConfig tpm_config;
2676 
2677     if (!x86ms->fw_cfg) {
2678         ACPI_BUILD_DPRINTF("No fw cfg. Bailing out.\n");
2679         return;
2680     }
2681 
2682     if (!pcms->acpi_build_enabled) {
2683         ACPI_BUILD_DPRINTF("ACPI build disabled. Bailing out.\n");
2684         return;
2685     }
2686 
2687     if (!x86_machine_is_acpi_enabled(X86_MACHINE(pcms))) {
2688         ACPI_BUILD_DPRINTF("ACPI disabled. Bailing out.\n");
2689         return;
2690     }
2691 
2692     build_state = g_malloc0(sizeof *build_state);
2693 
2694     acpi_build_tables_init(&tables);
2695     acpi_build(&tables, MACHINE(pcms));
2696 
2697     /* Now expose it all to Guest */
2698     build_state->table_mr = acpi_add_rom_blob(acpi_build_update,
2699                                               build_state, tables.table_data,
2700                                               ACPI_BUILD_TABLE_FILE,
2701                                               ACPI_BUILD_TABLE_MAX_SIZE);
2702     assert(build_state->table_mr != NULL);
2703 
2704     build_state->linker_mr =
2705         acpi_add_rom_blob(acpi_build_update, build_state,
2706                           tables.linker->cmd_blob, ACPI_BUILD_LOADER_FILE, 0);
2707 
2708     fw_cfg_add_file(x86ms->fw_cfg, ACPI_BUILD_TPMLOG_FILE,
2709                     tables.tcpalog->data, acpi_data_len(tables.tcpalog));
2710 
2711     tpm = tpm_find();
2712     if (tpm && object_property_get_bool(OBJECT(tpm), "ppi", &error_abort)) {
2713         tpm_config = (FwCfgTPMConfig) {
2714             .tpmppi_address = cpu_to_le32(TPM_PPI_ADDR_BASE),
2715             .tpm_version = tpm_get_version(tpm),
2716             .tpmppi_version = TPM_PPI_VERSION_1_30
2717         };
2718         fw_cfg_add_file(x86ms->fw_cfg, "etc/tpm/config",
2719                         &tpm_config, sizeof tpm_config);
2720     }
2721 
2722     vmgenid_dev = find_vmgenid_dev();
2723     if (vmgenid_dev) {
2724         vmgenid_add_fw_cfg(VMGENID(vmgenid_dev), x86ms->fw_cfg,
2725                            tables.vmgenid);
2726     }
2727 
2728     if (!pcmc->rsdp_in_ram) {
2729         /*
2730          * Keep for compatibility with old machine types.
2731          * Though RSDP is small, its contents isn't immutable, so
2732          * we'll update it along with the rest of tables on guest access.
2733          */
2734         uint32_t rsdp_size = acpi_data_len(tables.rsdp);
2735 
2736         build_state->rsdp = g_memdup(tables.rsdp->data, rsdp_size);
2737         fw_cfg_add_file_callback(x86ms->fw_cfg, ACPI_BUILD_RSDP_FILE,
2738                                  acpi_build_update, NULL, build_state,
2739                                  build_state->rsdp, rsdp_size, true);
2740         build_state->rsdp_mr = NULL;
2741     } else {
2742         build_state->rsdp = NULL;
2743         build_state->rsdp_mr = acpi_add_rom_blob(acpi_build_update,
2744                                                  build_state, tables.rsdp,
2745                                                  ACPI_BUILD_RSDP_FILE, 0);
2746     }
2747 
2748     qemu_register_reset(acpi_build_reset, build_state);
2749     acpi_build_reset(build_state);
2750     vmstate_register(NULL, 0, &vmstate_acpi_build, build_state);
2751 
2752     /* Cleanup tables but don't free the memory: we track it
2753      * in build_state.
2754      */
2755     acpi_build_tables_cleanup(&tables, false);
2756 }
2757