xref: /openbmc/qemu/hw/i386/acpi-build.c (revision 34b36c3b)
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     uint8_t s4_val;
99     AcpiFadtData fadt;
100     uint16_t cpu_hp_io_base;
101     uint16_t pcihp_io_base;
102     uint16_t pcihp_io_len;
103 } AcpiPmInfo;
104 
105 typedef struct AcpiMiscInfo {
106     bool is_piix4;
107     bool has_hpet;
108     TPMVersion tpm_version;
109     const unsigned char *dsdt_code;
110     unsigned dsdt_size;
111     uint16_t pvpanic_port;
112     uint16_t applesmc_io_base;
113 } AcpiMiscInfo;
114 
115 typedef struct AcpiBuildPciBusHotplugState {
116     GArray *device_table;
117     GArray *notify_table;
118     struct AcpiBuildPciBusHotplugState *parent;
119     bool pcihp_bridge_en;
120 } AcpiBuildPciBusHotplugState;
121 
122 typedef struct FwCfgTPMConfig {
123     uint32_t tpmppi_address;
124     uint8_t tpm_version;
125     uint8_t tpmppi_version;
126 } QEMU_PACKED FwCfgTPMConfig;
127 
128 static bool acpi_get_mcfg(AcpiMcfgInfo *mcfg);
129 
130 const struct AcpiGenericAddress x86_nvdimm_acpi_dsmio = {
131     .space_id = AML_AS_SYSTEM_IO,
132     .address = NVDIMM_ACPI_IO_BASE,
133     .bit_width = NVDIMM_ACPI_IO_LEN << 3
134 };
135 
136 static void init_common_fadt_data(MachineState *ms, Object *o,
137                                   AcpiFadtData *data)
138 {
139     uint32_t io = object_property_get_uint(o, ACPI_PM_PROP_PM_IO_BASE, NULL);
140     AmlAddressSpace as = AML_AS_SYSTEM_IO;
141     AcpiFadtData fadt = {
142         .rev = 3,
143         .flags =
144             (1 << ACPI_FADT_F_WBINVD) |
145             (1 << ACPI_FADT_F_PROC_C1) |
146             (1 << ACPI_FADT_F_SLP_BUTTON) |
147             (1 << ACPI_FADT_F_RTC_S4) |
148             (1 << ACPI_FADT_F_USE_PLATFORM_CLOCK) |
149             /* APIC destination mode ("Flat Logical") has an upper limit of 8
150              * CPUs for more than 8 CPUs, "Clustered Logical" mode has to be
151              * used
152              */
153             ((ms->smp.max_cpus > 8) ?
154                         (1 << ACPI_FADT_F_FORCE_APIC_CLUSTER_MODEL) : 0),
155         .int_model = 1 /* Multiple APIC */,
156         .rtc_century = RTC_CENTURY,
157         .plvl2_lat = 0xfff /* C2 state not supported */,
158         .plvl3_lat = 0xfff /* C3 state not supported */,
159         .smi_cmd = ACPI_PORT_SMI_CMD,
160         .sci_int = object_property_get_uint(o, ACPI_PM_PROP_SCI_INT, NULL),
161         .acpi_enable_cmd =
162             object_property_get_uint(o, ACPI_PM_PROP_ACPI_ENABLE_CMD, NULL),
163         .acpi_disable_cmd =
164             object_property_get_uint(o, ACPI_PM_PROP_ACPI_DISABLE_CMD, NULL),
165         .pm1a_evt = { .space_id = as, .bit_width = 4 * 8, .address = io },
166         .pm1a_cnt = { .space_id = as, .bit_width = 2 * 8,
167                       .address = io + 0x04 },
168         .pm_tmr = { .space_id = as, .bit_width = 4 * 8, .address = io + 0x08 },
169         .gpe0_blk = { .space_id = as, .bit_width =
170             object_property_get_uint(o, ACPI_PM_PROP_GPE0_BLK_LEN, NULL) * 8,
171             .address = object_property_get_uint(o, ACPI_PM_PROP_GPE0_BLK, NULL)
172         },
173     };
174     *data = fadt;
175 }
176 
177 static Object *object_resolve_type_unambiguous(const char *typename)
178 {
179     bool ambig;
180     Object *o = object_resolve_path_type("", typename, &ambig);
181 
182     if (ambig || !o) {
183         return NULL;
184     }
185     return o;
186 }
187 
188 static void acpi_get_pm_info(MachineState *machine, AcpiPmInfo *pm)
189 {
190     Object *piix = object_resolve_type_unambiguous(TYPE_PIIX4_PM);
191     Object *lpc = object_resolve_type_unambiguous(TYPE_ICH9_LPC_DEVICE);
192     Object *obj = piix ? piix : lpc;
193     QObject *o;
194     pm->cpu_hp_io_base = 0;
195     pm->pcihp_io_base = 0;
196     pm->pcihp_io_len = 0;
197 
198     assert(obj);
199     init_common_fadt_data(machine, obj, &pm->fadt);
200     if (piix) {
201         /* w2k requires FADT(rev1) or it won't boot, keep PC compatible */
202         pm->fadt.rev = 1;
203         pm->cpu_hp_io_base = PIIX4_CPU_HOTPLUG_IO_BASE;
204         pm->pcihp_io_base =
205             object_property_get_uint(obj, ACPI_PCIHP_IO_BASE_PROP, NULL);
206         pm->pcihp_io_len =
207             object_property_get_uint(obj, ACPI_PCIHP_IO_LEN_PROP, NULL);
208     }
209     if (lpc) {
210         struct AcpiGenericAddress r = { .space_id = AML_AS_SYSTEM_IO,
211             .bit_width = 8, .address = ICH9_RST_CNT_IOPORT };
212         pm->fadt.reset_reg = r;
213         pm->fadt.reset_val = 0xf;
214         pm->fadt.flags |= 1 << ACPI_FADT_F_RESET_REG_SUP;
215         pm->cpu_hp_io_base = ICH9_CPU_HOTPLUG_IO_BASE;
216     }
217 
218     /* The above need not be conditional on machine type because the reset port
219      * happens to be the same on PIIX (pc) and ICH9 (q35). */
220     QEMU_BUILD_BUG_ON(ICH9_RST_CNT_IOPORT != PIIX_RCR_IOPORT);
221 
222     /* Fill in optional s3/s4 related properties */
223     o = object_property_get_qobject(obj, ACPI_PM_PROP_S3_DISABLED, NULL);
224     if (o) {
225         pm->s3_disabled = qnum_get_uint(qobject_to(QNum, o));
226     } else {
227         pm->s3_disabled = false;
228     }
229     qobject_unref(o);
230     o = object_property_get_qobject(obj, ACPI_PM_PROP_S4_DISABLED, NULL);
231     if (o) {
232         pm->s4_disabled = qnum_get_uint(qobject_to(QNum, o));
233     } else {
234         pm->s4_disabled = false;
235     }
236     qobject_unref(o);
237     o = object_property_get_qobject(obj, ACPI_PM_PROP_S4_VAL, NULL);
238     if (o) {
239         pm->s4_val = qnum_get_uint(qobject_to(QNum, o));
240     } else {
241         pm->s4_val = false;
242     }
243     qobject_unref(o);
244 
245     pm->pcihp_bridge_en =
246         object_property_get_bool(obj, "acpi-pci-hotplug-with-bridge-support",
247                                  NULL);
248 }
249 
250 static void acpi_get_misc_info(AcpiMiscInfo *info)
251 {
252     Object *piix = object_resolve_type_unambiguous(TYPE_PIIX4_PM);
253     Object *lpc = object_resolve_type_unambiguous(TYPE_ICH9_LPC_DEVICE);
254     assert(!!piix != !!lpc);
255 
256     if (piix) {
257         info->is_piix4 = true;
258     }
259     if (lpc) {
260         info->is_piix4 = false;
261     }
262 
263     info->has_hpet = hpet_find();
264     info->tpm_version = tpm_get_version(tpm_find());
265     info->pvpanic_port = pvpanic_port();
266     info->applesmc_io_base = applesmc_port();
267 }
268 
269 /*
270  * Because of the PXB hosts we cannot simply query TYPE_PCI_HOST_BRIDGE.
271  * On i386 arch we only have two pci hosts, so we can look only for them.
272  */
273 static Object *acpi_get_i386_pci_host(void)
274 {
275     PCIHostState *host;
276 
277     host = OBJECT_CHECK(PCIHostState,
278                         object_resolve_path("/machine/i440fx", NULL),
279                         TYPE_PCI_HOST_BRIDGE);
280     if (!host) {
281         host = OBJECT_CHECK(PCIHostState,
282                             object_resolve_path("/machine/q35", NULL),
283                             TYPE_PCI_HOST_BRIDGE);
284     }
285 
286     return OBJECT(host);
287 }
288 
289 static void acpi_get_pci_holes(Range *hole, Range *hole64)
290 {
291     Object *pci_host;
292 
293     pci_host = acpi_get_i386_pci_host();
294     g_assert(pci_host);
295 
296     range_set_bounds1(hole,
297                       object_property_get_uint(pci_host,
298                                                PCI_HOST_PROP_PCI_HOLE_START,
299                                                NULL),
300                       object_property_get_uint(pci_host,
301                                                PCI_HOST_PROP_PCI_HOLE_END,
302                                                NULL));
303     range_set_bounds1(hole64,
304                       object_property_get_uint(pci_host,
305                                                PCI_HOST_PROP_PCI_HOLE64_START,
306                                                NULL),
307                       object_property_get_uint(pci_host,
308                                                PCI_HOST_PROP_PCI_HOLE64_END,
309                                                NULL));
310 }
311 
312 static void acpi_align_size(GArray *blob, unsigned align)
313 {
314     /* Align size to multiple of given size. This reduces the chance
315      * we need to change size in the future (breaking cross version migration).
316      */
317     g_array_set_size(blob, ROUND_UP(acpi_data_len(blob), align));
318 }
319 
320 /* FACS */
321 static void
322 build_facs(GArray *table_data)
323 {
324     AcpiFacsDescriptorRev1 *facs = acpi_data_push(table_data, sizeof *facs);
325     memcpy(&facs->signature, "FACS", 4);
326     facs->length = cpu_to_le32(sizeof(*facs));
327 }
328 
329 static void build_append_pcihp_notify_entry(Aml *method, int slot)
330 {
331     Aml *if_ctx;
332     int32_t devfn = PCI_DEVFN(slot, 0);
333 
334     if_ctx = aml_if(aml_and(aml_arg(0), aml_int(0x1U << slot), NULL));
335     aml_append(if_ctx, aml_notify(aml_name("S%.02X", devfn), aml_arg(1)));
336     aml_append(method, if_ctx);
337 }
338 
339 static void build_append_pci_bus_devices(Aml *parent_scope, PCIBus *bus,
340                                          bool pcihp_bridge_en)
341 {
342     Aml *dev, *notify_method = NULL, *method;
343     QObject *bsel;
344     PCIBus *sec;
345     int i;
346 
347     bsel = object_property_get_qobject(OBJECT(bus), ACPI_PCIHP_PROP_BSEL, NULL);
348     if (bsel) {
349         uint64_t bsel_val = qnum_get_uint(qobject_to(QNum, bsel));
350 
351         aml_append(parent_scope, aml_name_decl("BSEL", aml_int(bsel_val)));
352         notify_method = aml_method("DVNT", 2, AML_NOTSERIALIZED);
353     }
354 
355     for (i = 0; i < ARRAY_SIZE(bus->devices); i += PCI_FUNC_MAX) {
356         DeviceClass *dc;
357         PCIDeviceClass *pc;
358         PCIDevice *pdev = bus->devices[i];
359         int slot = PCI_SLOT(i);
360         bool hotplug_enabled_dev;
361         bool bridge_in_acpi;
362 
363         if (!pdev) {
364             if (bsel) { /* add hotplug slots for non present devices */
365                 dev = aml_device("S%.02X", PCI_DEVFN(slot, 0));
366                 aml_append(dev, aml_name_decl("_SUN", aml_int(slot)));
367                 aml_append(dev, aml_name_decl("_ADR", aml_int(slot << 16)));
368                 method = aml_method("_EJ0", 1, AML_NOTSERIALIZED);
369                 aml_append(method,
370                     aml_call2("PCEJ", aml_name("BSEL"), aml_name("_SUN"))
371                 );
372                 aml_append(dev, method);
373                 aml_append(parent_scope, dev);
374 
375                 build_append_pcihp_notify_entry(notify_method, slot);
376             }
377             continue;
378         }
379 
380         pc = PCI_DEVICE_GET_CLASS(pdev);
381         dc = DEVICE_GET_CLASS(pdev);
382 
383         /* When hotplug for bridges is enabled, bridges are
384          * described in ACPI separately (see build_pci_bus_end).
385          * In this case they aren't themselves hot-pluggable.
386          * Hotplugged bridges *are* hot-pluggable.
387          */
388         bridge_in_acpi = pc->is_bridge && pcihp_bridge_en &&
389             !DEVICE(pdev)->hotplugged;
390 
391         hotplug_enabled_dev = bsel && dc->hotpluggable && !bridge_in_acpi;
392 
393         if (pc->class_id == PCI_CLASS_BRIDGE_ISA) {
394             continue;
395         }
396 
397         /* start to compose PCI slot descriptor */
398         dev = aml_device("S%.02X", PCI_DEVFN(slot, 0));
399         aml_append(dev, aml_name_decl("_ADR", aml_int(slot << 16)));
400 
401         if (pc->class_id == PCI_CLASS_DISPLAY_VGA) {
402             /* add VGA specific AML methods */
403             int s3d;
404 
405             if (object_dynamic_cast(OBJECT(pdev), "qxl-vga")) {
406                 s3d = 3;
407             } else {
408                 s3d = 0;
409             }
410 
411             method = aml_method("_S1D", 0, AML_NOTSERIALIZED);
412             aml_append(method, aml_return(aml_int(0)));
413             aml_append(dev, method);
414 
415             method = aml_method("_S2D", 0, AML_NOTSERIALIZED);
416             aml_append(method, aml_return(aml_int(0)));
417             aml_append(dev, method);
418 
419             method = aml_method("_S3D", 0, AML_NOTSERIALIZED);
420             aml_append(method, aml_return(aml_int(s3d)));
421             aml_append(dev, method);
422         } else if (hotplug_enabled_dev) {
423             /* add _SUN/_EJ0 to make slot hotpluggable  */
424             aml_append(dev, aml_name_decl("_SUN", aml_int(slot)));
425 
426             method = aml_method("_EJ0", 1, AML_NOTSERIALIZED);
427             aml_append(method,
428                 aml_call2("PCEJ", aml_name("BSEL"), aml_name("_SUN"))
429             );
430             aml_append(dev, method);
431 
432             if (bsel) {
433                 build_append_pcihp_notify_entry(notify_method, slot);
434             }
435         } else if (bridge_in_acpi) {
436             /*
437              * device is coldplugged bridge,
438              * add child device descriptions into its scope
439              */
440             PCIBus *sec_bus = pci_bridge_get_sec_bus(PCI_BRIDGE(pdev));
441 
442             build_append_pci_bus_devices(dev, sec_bus, pcihp_bridge_en);
443         }
444         /* slot descriptor has been composed, add it into parent context */
445         aml_append(parent_scope, dev);
446     }
447 
448     if (bsel) {
449         aml_append(parent_scope, notify_method);
450     }
451 
452     /* Append PCNT method to notify about events on local and child buses.
453      * Add unconditionally for root since DSDT expects it.
454      */
455     method = aml_method("PCNT", 0, AML_NOTSERIALIZED);
456 
457     /* If bus supports hotplug select it and notify about local events */
458     if (bsel) {
459         uint64_t bsel_val = qnum_get_uint(qobject_to(QNum, bsel));
460 
461         aml_append(method, aml_store(aml_int(bsel_val), aml_name("BNUM")));
462         aml_append(method,
463             aml_call2("DVNT", aml_name("PCIU"), aml_int(1) /* Device Check */)
464         );
465         aml_append(method,
466             aml_call2("DVNT", aml_name("PCID"), aml_int(3)/* Eject Request */)
467         );
468     }
469 
470     /* Notify about child bus events in any case */
471     if (pcihp_bridge_en) {
472         QLIST_FOREACH(sec, &bus->child, sibling) {
473             int32_t devfn = sec->parent_dev->devfn;
474 
475             if (pci_bus_is_root(sec) || pci_bus_is_express(sec)) {
476                 continue;
477             }
478 
479             aml_append(method, aml_name("^S%.02X.PCNT", devfn));
480         }
481     }
482     aml_append(parent_scope, method);
483     qobject_unref(bsel);
484 }
485 
486 /**
487  * build_prt_entry:
488  * @link_name: link name for PCI route entry
489  *
490  * build AML package containing a PCI route entry for @link_name
491  */
492 static Aml *build_prt_entry(const char *link_name)
493 {
494     Aml *a_zero = aml_int(0);
495     Aml *pkg = aml_package(4);
496     aml_append(pkg, a_zero);
497     aml_append(pkg, a_zero);
498     aml_append(pkg, aml_name("%s", link_name));
499     aml_append(pkg, a_zero);
500     return pkg;
501 }
502 
503 /*
504  * initialize_route - Initialize the interrupt routing rule
505  * through a specific LINK:
506  *  if (lnk_idx == idx)
507  *      route using link 'link_name'
508  */
509 static Aml *initialize_route(Aml *route, const char *link_name,
510                              Aml *lnk_idx, int idx)
511 {
512     Aml *if_ctx = aml_if(aml_equal(lnk_idx, aml_int(idx)));
513     Aml *pkg = build_prt_entry(link_name);
514 
515     aml_append(if_ctx, aml_store(pkg, route));
516 
517     return if_ctx;
518 }
519 
520 /*
521  * build_prt - Define interrupt rounting rules
522  *
523  * Returns an array of 128 routes, one for each device,
524  * based on device location.
525  * The main goal is to equaly distribute the interrupts
526  * over the 4 existing ACPI links (works only for i440fx).
527  * The hash function is  (slot + pin) & 3 -> "LNK[D|A|B|C]".
528  *
529  */
530 static Aml *build_prt(bool is_pci0_prt)
531 {
532     Aml *method, *while_ctx, *pin, *res;
533 
534     method = aml_method("_PRT", 0, AML_NOTSERIALIZED);
535     res = aml_local(0);
536     pin = aml_local(1);
537     aml_append(method, aml_store(aml_package(128), res));
538     aml_append(method, aml_store(aml_int(0), pin));
539 
540     /* while (pin < 128) */
541     while_ctx = aml_while(aml_lless(pin, aml_int(128)));
542     {
543         Aml *slot = aml_local(2);
544         Aml *lnk_idx = aml_local(3);
545         Aml *route = aml_local(4);
546 
547         /* slot = pin >> 2 */
548         aml_append(while_ctx,
549                    aml_store(aml_shiftright(pin, aml_int(2), NULL), slot));
550         /* lnk_idx = (slot + pin) & 3 */
551         aml_append(while_ctx,
552             aml_store(aml_and(aml_add(pin, slot, NULL), aml_int(3), NULL),
553                       lnk_idx));
554 
555         /* route[2] = "LNK[D|A|B|C]", selection based on pin % 3  */
556         aml_append(while_ctx, initialize_route(route, "LNKD", lnk_idx, 0));
557         if (is_pci0_prt) {
558             Aml *if_device_1, *if_pin_4, *else_pin_4;
559 
560             /* device 1 is the power-management device, needs SCI */
561             if_device_1 = aml_if(aml_equal(lnk_idx, aml_int(1)));
562             {
563                 if_pin_4 = aml_if(aml_equal(pin, aml_int(4)));
564                 {
565                     aml_append(if_pin_4,
566                         aml_store(build_prt_entry("LNKS"), route));
567                 }
568                 aml_append(if_device_1, if_pin_4);
569                 else_pin_4 = aml_else();
570                 {
571                     aml_append(else_pin_4,
572                         aml_store(build_prt_entry("LNKA"), route));
573                 }
574                 aml_append(if_device_1, else_pin_4);
575             }
576             aml_append(while_ctx, if_device_1);
577         } else {
578             aml_append(while_ctx, initialize_route(route, "LNKA", lnk_idx, 1));
579         }
580         aml_append(while_ctx, initialize_route(route, "LNKB", lnk_idx, 2));
581         aml_append(while_ctx, initialize_route(route, "LNKC", lnk_idx, 3));
582 
583         /* route[0] = 0x[slot]FFFF */
584         aml_append(while_ctx,
585             aml_store(aml_or(aml_shiftleft(slot, aml_int(16)), aml_int(0xFFFF),
586                              NULL),
587                       aml_index(route, aml_int(0))));
588         /* route[1] = pin & 3 */
589         aml_append(while_ctx,
590             aml_store(aml_and(pin, aml_int(3), NULL),
591                       aml_index(route, aml_int(1))));
592         /* res[pin] = route */
593         aml_append(while_ctx, aml_store(route, aml_index(res, pin)));
594         /* pin++ */
595         aml_append(while_ctx, aml_increment(pin));
596     }
597     aml_append(method, while_ctx);
598     /* return res*/
599     aml_append(method, aml_return(res));
600 
601     return method;
602 }
603 
604 typedef struct CrsRangeEntry {
605     uint64_t base;
606     uint64_t limit;
607 } CrsRangeEntry;
608 
609 static void crs_range_insert(GPtrArray *ranges, uint64_t base, uint64_t limit)
610 {
611     CrsRangeEntry *entry;
612 
613     entry = g_malloc(sizeof(*entry));
614     entry->base = base;
615     entry->limit = limit;
616 
617     g_ptr_array_add(ranges, entry);
618 }
619 
620 static void crs_range_free(gpointer data)
621 {
622     CrsRangeEntry *entry = (CrsRangeEntry *)data;
623     g_free(entry);
624 }
625 
626 typedef struct CrsRangeSet {
627     GPtrArray *io_ranges;
628     GPtrArray *mem_ranges;
629     GPtrArray *mem_64bit_ranges;
630  } CrsRangeSet;
631 
632 static void crs_range_set_init(CrsRangeSet *range_set)
633 {
634     range_set->io_ranges = g_ptr_array_new_with_free_func(crs_range_free);
635     range_set->mem_ranges = g_ptr_array_new_with_free_func(crs_range_free);
636     range_set->mem_64bit_ranges =
637             g_ptr_array_new_with_free_func(crs_range_free);
638 }
639 
640 static void crs_range_set_free(CrsRangeSet *range_set)
641 {
642     g_ptr_array_free(range_set->io_ranges, true);
643     g_ptr_array_free(range_set->mem_ranges, true);
644     g_ptr_array_free(range_set->mem_64bit_ranges, true);
645 }
646 
647 static gint crs_range_compare(gconstpointer a, gconstpointer b)
648 {
649     CrsRangeEntry *entry_a = *(CrsRangeEntry **)a;
650     CrsRangeEntry *entry_b = *(CrsRangeEntry **)b;
651 
652     if (entry_a->base < entry_b->base) {
653         return -1;
654     } else if (entry_a->base > entry_b->base) {
655         return 1;
656     } else {
657         return 0;
658     }
659 }
660 
661 /*
662  * crs_replace_with_free_ranges - given the 'used' ranges within [start - end]
663  * interval, computes the 'free' ranges from the same interval.
664  * Example: If the input array is { [a1 - a2],[b1 - b2] }, the function
665  * will return { [base - a1], [a2 - b1], [b2 - limit] }.
666  */
667 static void crs_replace_with_free_ranges(GPtrArray *ranges,
668                                          uint64_t start, uint64_t end)
669 {
670     GPtrArray *free_ranges = g_ptr_array_new();
671     uint64_t free_base = start;
672     int i;
673 
674     g_ptr_array_sort(ranges, crs_range_compare);
675     for (i = 0; i < ranges->len; i++) {
676         CrsRangeEntry *used = g_ptr_array_index(ranges, i);
677 
678         if (free_base < used->base) {
679             crs_range_insert(free_ranges, free_base, used->base - 1);
680         }
681 
682         free_base = used->limit + 1;
683     }
684 
685     if (free_base < end) {
686         crs_range_insert(free_ranges, free_base, end);
687     }
688 
689     g_ptr_array_set_size(ranges, 0);
690     for (i = 0; i < free_ranges->len; i++) {
691         g_ptr_array_add(ranges, g_ptr_array_index(free_ranges, i));
692     }
693 
694     g_ptr_array_free(free_ranges, true);
695 }
696 
697 /*
698  * crs_range_merge - merges adjacent ranges in the given array.
699  * Array elements are deleted and replaced with the merged ranges.
700  */
701 static void crs_range_merge(GPtrArray *range)
702 {
703     GPtrArray *tmp =  g_ptr_array_new_with_free_func(crs_range_free);
704     CrsRangeEntry *entry;
705     uint64_t range_base, range_limit;
706     int i;
707 
708     if (!range->len) {
709         return;
710     }
711 
712     g_ptr_array_sort(range, crs_range_compare);
713 
714     entry = g_ptr_array_index(range, 0);
715     range_base = entry->base;
716     range_limit = entry->limit;
717     for (i = 1; i < range->len; i++) {
718         entry = g_ptr_array_index(range, i);
719         if (entry->base - 1 == range_limit) {
720             range_limit = entry->limit;
721         } else {
722             crs_range_insert(tmp, range_base, range_limit);
723             range_base = entry->base;
724             range_limit = entry->limit;
725         }
726     }
727     crs_range_insert(tmp, range_base, range_limit);
728 
729     g_ptr_array_set_size(range, 0);
730     for (i = 0; i < tmp->len; i++) {
731         entry = g_ptr_array_index(tmp, i);
732         crs_range_insert(range, entry->base, entry->limit);
733     }
734     g_ptr_array_free(tmp, true);
735 }
736 
737 static Aml *build_crs(PCIHostState *host, CrsRangeSet *range_set)
738 {
739     Aml *crs = aml_resource_template();
740     CrsRangeSet temp_range_set;
741     CrsRangeEntry *entry;
742     uint8_t max_bus = pci_bus_num(host->bus);
743     uint8_t type;
744     int devfn;
745     int i;
746 
747     crs_range_set_init(&temp_range_set);
748     for (devfn = 0; devfn < ARRAY_SIZE(host->bus->devices); devfn++) {
749         uint64_t range_base, range_limit;
750         PCIDevice *dev = host->bus->devices[devfn];
751 
752         if (!dev) {
753             continue;
754         }
755 
756         for (i = 0; i < PCI_NUM_REGIONS; i++) {
757             PCIIORegion *r = &dev->io_regions[i];
758 
759             range_base = r->addr;
760             range_limit = r->addr + r->size - 1;
761 
762             /*
763              * Work-around for old bioses
764              * that do not support multiple root buses
765              */
766             if (!range_base || range_base > range_limit) {
767                 continue;
768             }
769 
770             if (r->type & PCI_BASE_ADDRESS_SPACE_IO) {
771                 crs_range_insert(temp_range_set.io_ranges,
772                                  range_base, range_limit);
773             } else { /* "memory" */
774                 crs_range_insert(temp_range_set.mem_ranges,
775                                  range_base, range_limit);
776             }
777         }
778 
779         type = dev->config[PCI_HEADER_TYPE] & ~PCI_HEADER_TYPE_MULTI_FUNCTION;
780         if (type == PCI_HEADER_TYPE_BRIDGE) {
781             uint8_t subordinate = dev->config[PCI_SUBORDINATE_BUS];
782             if (subordinate > max_bus) {
783                 max_bus = subordinate;
784             }
785 
786             range_base = pci_bridge_get_base(dev, PCI_BASE_ADDRESS_SPACE_IO);
787             range_limit = pci_bridge_get_limit(dev, PCI_BASE_ADDRESS_SPACE_IO);
788 
789             /*
790              * Work-around for old bioses
791              * that do not support multiple root buses
792              */
793             if (range_base && range_base <= range_limit) {
794                 crs_range_insert(temp_range_set.io_ranges,
795                                  range_base, range_limit);
796             }
797 
798             range_base =
799                 pci_bridge_get_base(dev, PCI_BASE_ADDRESS_SPACE_MEMORY);
800             range_limit =
801                 pci_bridge_get_limit(dev, PCI_BASE_ADDRESS_SPACE_MEMORY);
802 
803             /*
804              * Work-around for old bioses
805              * that do not support multiple root buses
806              */
807             if (range_base && range_base <= range_limit) {
808                 uint64_t length = range_limit - range_base + 1;
809                 if (range_limit <= UINT32_MAX && length <= UINT32_MAX) {
810                     crs_range_insert(temp_range_set.mem_ranges,
811                                      range_base, range_limit);
812                 } else {
813                     crs_range_insert(temp_range_set.mem_64bit_ranges,
814                                      range_base, range_limit);
815                 }
816             }
817 
818             range_base =
819                 pci_bridge_get_base(dev, PCI_BASE_ADDRESS_MEM_PREFETCH);
820             range_limit =
821                 pci_bridge_get_limit(dev, PCI_BASE_ADDRESS_MEM_PREFETCH);
822 
823             /*
824              * Work-around for old bioses
825              * that do not support multiple root buses
826              */
827             if (range_base && range_base <= range_limit) {
828                 uint64_t length = range_limit - range_base + 1;
829                 if (range_limit <= UINT32_MAX && length <= UINT32_MAX) {
830                     crs_range_insert(temp_range_set.mem_ranges,
831                                      range_base, range_limit);
832                 } else {
833                     crs_range_insert(temp_range_set.mem_64bit_ranges,
834                                      range_base, range_limit);
835                 }
836             }
837         }
838     }
839 
840     crs_range_merge(temp_range_set.io_ranges);
841     for (i = 0; i < temp_range_set.io_ranges->len; i++) {
842         entry = g_ptr_array_index(temp_range_set.io_ranges, i);
843         aml_append(crs,
844                    aml_word_io(AML_MIN_FIXED, AML_MAX_FIXED,
845                                AML_POS_DECODE, AML_ENTIRE_RANGE,
846                                0, entry->base, entry->limit, 0,
847                                entry->limit - entry->base + 1));
848         crs_range_insert(range_set->io_ranges, entry->base, entry->limit);
849     }
850 
851     crs_range_merge(temp_range_set.mem_ranges);
852     for (i = 0; i < temp_range_set.mem_ranges->len; i++) {
853         entry = g_ptr_array_index(temp_range_set.mem_ranges, i);
854         aml_append(crs,
855                    aml_dword_memory(AML_POS_DECODE, AML_MIN_FIXED,
856                                     AML_MAX_FIXED, AML_NON_CACHEABLE,
857                                     AML_READ_WRITE,
858                                     0, entry->base, entry->limit, 0,
859                                     entry->limit - entry->base + 1));
860         crs_range_insert(range_set->mem_ranges, entry->base, entry->limit);
861     }
862 
863     crs_range_merge(temp_range_set.mem_64bit_ranges);
864     for (i = 0; i < temp_range_set.mem_64bit_ranges->len; i++) {
865         entry = g_ptr_array_index(temp_range_set.mem_64bit_ranges, i);
866         aml_append(crs,
867                    aml_qword_memory(AML_POS_DECODE, AML_MIN_FIXED,
868                                     AML_MAX_FIXED, AML_NON_CACHEABLE,
869                                     AML_READ_WRITE,
870                                     0, entry->base, entry->limit, 0,
871                                     entry->limit - entry->base + 1));
872         crs_range_insert(range_set->mem_64bit_ranges,
873                          entry->base, entry->limit);
874     }
875 
876     crs_range_set_free(&temp_range_set);
877 
878     aml_append(crs,
879         aml_word_bus_number(AML_MIN_FIXED, AML_MAX_FIXED, AML_POS_DECODE,
880                             0,
881                             pci_bus_num(host->bus),
882                             max_bus,
883                             0,
884                             max_bus - pci_bus_num(host->bus) + 1));
885 
886     return crs;
887 }
888 
889 static void build_hpet_aml(Aml *table)
890 {
891     Aml *crs;
892     Aml *field;
893     Aml *method;
894     Aml *if_ctx;
895     Aml *scope = aml_scope("_SB");
896     Aml *dev = aml_device("HPET");
897     Aml *zero = aml_int(0);
898     Aml *id = aml_local(0);
899     Aml *period = aml_local(1);
900 
901     aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0103")));
902     aml_append(dev, aml_name_decl("_UID", zero));
903 
904     aml_append(dev,
905         aml_operation_region("HPTM", AML_SYSTEM_MEMORY, aml_int(HPET_BASE),
906                              HPET_LEN));
907     field = aml_field("HPTM", AML_DWORD_ACC, AML_LOCK, AML_PRESERVE);
908     aml_append(field, aml_named_field("VEND", 32));
909     aml_append(field, aml_named_field("PRD", 32));
910     aml_append(dev, field);
911 
912     method = aml_method("_STA", 0, AML_NOTSERIALIZED);
913     aml_append(method, aml_store(aml_name("VEND"), id));
914     aml_append(method, aml_store(aml_name("PRD"), period));
915     aml_append(method, aml_shiftright(id, aml_int(16), id));
916     if_ctx = aml_if(aml_lor(aml_equal(id, zero),
917                             aml_equal(id, aml_int(0xffff))));
918     {
919         aml_append(if_ctx, aml_return(zero));
920     }
921     aml_append(method, if_ctx);
922 
923     if_ctx = aml_if(aml_lor(aml_equal(period, zero),
924                             aml_lgreater(period, aml_int(100000000))));
925     {
926         aml_append(if_ctx, aml_return(zero));
927     }
928     aml_append(method, if_ctx);
929 
930     aml_append(method, aml_return(aml_int(0x0F)));
931     aml_append(dev, method);
932 
933     crs = aml_resource_template();
934     aml_append(crs, aml_memory32_fixed(HPET_BASE, HPET_LEN, AML_READ_ONLY));
935     aml_append(dev, aml_name_decl("_CRS", crs));
936 
937     aml_append(scope, dev);
938     aml_append(table, scope);
939 }
940 
941 static Aml *build_vmbus_device_aml(VMBusBridge *vmbus_bridge)
942 {
943     Aml *dev;
944     Aml *method;
945     Aml *crs;
946 
947     dev = aml_device("VMBS");
948     aml_append(dev, aml_name_decl("STA", aml_int(0xF)));
949     aml_append(dev, aml_name_decl("_HID", aml_string("VMBus")));
950     aml_append(dev, aml_name_decl("_UID", aml_int(0x0)));
951     aml_append(dev, aml_name_decl("_DDN", aml_string("VMBUS")));
952 
953     method = aml_method("_DIS", 0, AML_NOTSERIALIZED);
954     aml_append(method, aml_store(aml_and(aml_name("STA"), aml_int(0xD), NULL),
955                                      aml_name("STA")));
956     aml_append(dev, method);
957 
958     method = aml_method("_PS0", 0, AML_NOTSERIALIZED);
959     aml_append(method, aml_store(aml_or(aml_name("STA"), aml_int(0xF), NULL),
960                                      aml_name("STA")));
961     aml_append(dev, method);
962 
963     method = aml_method("_STA", 0, AML_NOTSERIALIZED);
964     aml_append(method, aml_return(aml_name("STA")));
965     aml_append(dev, method);
966 
967     aml_append(dev, aml_name_decl("_PS3", aml_int(0x0)));
968 
969     crs = aml_resource_template();
970     aml_append(crs, aml_irq_no_flags(vmbus_bridge->irq));
971     aml_append(dev, aml_name_decl("_CRS", crs));
972 
973     return dev;
974 }
975 
976 static void build_isa_devices_aml(Aml *table)
977 {
978     VMBusBridge *vmbus_bridge = vmbus_bridge_find();
979     bool ambiguous;
980     Object *obj = object_resolve_path_type("", TYPE_ISA_BUS, &ambiguous);
981     Aml *scope;
982 
983     assert(obj && !ambiguous);
984 
985     scope = aml_scope("_SB.PCI0.ISA");
986     build_acpi_ipmi_devices(scope, BUS(obj), "\\_SB.PCI0.ISA");
987     isa_build_aml(ISA_BUS(obj), scope);
988 
989     if (vmbus_bridge) {
990         aml_append(scope, build_vmbus_device_aml(vmbus_bridge));
991     }
992 
993     aml_append(table, scope);
994 }
995 
996 static void build_dbg_aml(Aml *table)
997 {
998     Aml *field;
999     Aml *method;
1000     Aml *while_ctx;
1001     Aml *scope = aml_scope("\\");
1002     Aml *buf = aml_local(0);
1003     Aml *len = aml_local(1);
1004     Aml *idx = aml_local(2);
1005 
1006     aml_append(scope,
1007        aml_operation_region("DBG", AML_SYSTEM_IO, aml_int(0x0402), 0x01));
1008     field = aml_field("DBG", AML_BYTE_ACC, AML_NOLOCK, AML_PRESERVE);
1009     aml_append(field, aml_named_field("DBGB", 8));
1010     aml_append(scope, field);
1011 
1012     method = aml_method("DBUG", 1, AML_NOTSERIALIZED);
1013 
1014     aml_append(method, aml_to_hexstring(aml_arg(0), buf));
1015     aml_append(method, aml_to_buffer(buf, buf));
1016     aml_append(method, aml_subtract(aml_sizeof(buf), aml_int(1), len));
1017     aml_append(method, aml_store(aml_int(0), idx));
1018 
1019     while_ctx = aml_while(aml_lless(idx, len));
1020     aml_append(while_ctx,
1021         aml_store(aml_derefof(aml_index(buf, idx)), aml_name("DBGB")));
1022     aml_append(while_ctx, aml_increment(idx));
1023     aml_append(method, while_ctx);
1024 
1025     aml_append(method, aml_store(aml_int(0x0A), aml_name("DBGB")));
1026     aml_append(scope, method);
1027 
1028     aml_append(table, scope);
1029 }
1030 
1031 static Aml *build_link_dev(const char *name, uint8_t uid, Aml *reg)
1032 {
1033     Aml *dev;
1034     Aml *crs;
1035     Aml *method;
1036     uint32_t irqs[] = {5, 10, 11};
1037 
1038     dev = aml_device("%s", name);
1039     aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0C0F")));
1040     aml_append(dev, aml_name_decl("_UID", aml_int(uid)));
1041 
1042     crs = aml_resource_template();
1043     aml_append(crs, aml_interrupt(AML_CONSUMER, AML_LEVEL, AML_ACTIVE_HIGH,
1044                                   AML_SHARED, irqs, ARRAY_SIZE(irqs)));
1045     aml_append(dev, aml_name_decl("_PRS", crs));
1046 
1047     method = aml_method("_STA", 0, AML_NOTSERIALIZED);
1048     aml_append(method, aml_return(aml_call1("IQST", reg)));
1049     aml_append(dev, method);
1050 
1051     method = aml_method("_DIS", 0, AML_NOTSERIALIZED);
1052     aml_append(method, aml_or(reg, aml_int(0x80), reg));
1053     aml_append(dev, method);
1054 
1055     method = aml_method("_CRS", 0, AML_NOTSERIALIZED);
1056     aml_append(method, aml_return(aml_call1("IQCR", reg)));
1057     aml_append(dev, method);
1058 
1059     method = aml_method("_SRS", 1, AML_NOTSERIALIZED);
1060     aml_append(method, aml_create_dword_field(aml_arg(0), aml_int(5), "PRRI"));
1061     aml_append(method, aml_store(aml_name("PRRI"), reg));
1062     aml_append(dev, method);
1063 
1064     return dev;
1065  }
1066 
1067 static Aml *build_gsi_link_dev(const char *name, uint8_t uid, uint8_t gsi)
1068 {
1069     Aml *dev;
1070     Aml *crs;
1071     Aml *method;
1072     uint32_t irqs;
1073 
1074     dev = aml_device("%s", name);
1075     aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0C0F")));
1076     aml_append(dev, aml_name_decl("_UID", aml_int(uid)));
1077 
1078     crs = aml_resource_template();
1079     irqs = gsi;
1080     aml_append(crs, aml_interrupt(AML_CONSUMER, AML_LEVEL, AML_ACTIVE_HIGH,
1081                                   AML_SHARED, &irqs, 1));
1082     aml_append(dev, aml_name_decl("_PRS", crs));
1083 
1084     aml_append(dev, aml_name_decl("_CRS", crs));
1085 
1086     /*
1087      * _DIS can be no-op because the interrupt cannot be disabled.
1088      */
1089     method = aml_method("_DIS", 0, AML_NOTSERIALIZED);
1090     aml_append(dev, method);
1091 
1092     method = aml_method("_SRS", 1, AML_NOTSERIALIZED);
1093     aml_append(dev, method);
1094 
1095     return dev;
1096 }
1097 
1098 /* _CRS method - get current settings */
1099 static Aml *build_iqcr_method(bool is_piix4)
1100 {
1101     Aml *if_ctx;
1102     uint32_t irqs;
1103     Aml *method = aml_method("IQCR", 1, AML_SERIALIZED);
1104     Aml *crs = aml_resource_template();
1105 
1106     irqs = 0;
1107     aml_append(crs, aml_interrupt(AML_CONSUMER, AML_LEVEL,
1108                                   AML_ACTIVE_HIGH, AML_SHARED, &irqs, 1));
1109     aml_append(method, aml_name_decl("PRR0", crs));
1110 
1111     aml_append(method,
1112         aml_create_dword_field(aml_name("PRR0"), aml_int(5), "PRRI"));
1113 
1114     if (is_piix4) {
1115         if_ctx = aml_if(aml_lless(aml_arg(0), aml_int(0x80)));
1116         aml_append(if_ctx, aml_store(aml_arg(0), aml_name("PRRI")));
1117         aml_append(method, if_ctx);
1118     } else {
1119         aml_append(method,
1120             aml_store(aml_and(aml_arg(0), aml_int(0xF), NULL),
1121                       aml_name("PRRI")));
1122     }
1123 
1124     aml_append(method, aml_return(aml_name("PRR0")));
1125     return method;
1126 }
1127 
1128 /* _STA method - get status */
1129 static Aml *build_irq_status_method(void)
1130 {
1131     Aml *if_ctx;
1132     Aml *method = aml_method("IQST", 1, AML_NOTSERIALIZED);
1133 
1134     if_ctx = aml_if(aml_and(aml_int(0x80), aml_arg(0), NULL));
1135     aml_append(if_ctx, aml_return(aml_int(0x09)));
1136     aml_append(method, if_ctx);
1137     aml_append(method, aml_return(aml_int(0x0B)));
1138     return method;
1139 }
1140 
1141 static void build_piix4_pci0_int(Aml *table)
1142 {
1143     Aml *dev;
1144     Aml *crs;
1145     Aml *field;
1146     Aml *method;
1147     uint32_t irqs;
1148     Aml *sb_scope = aml_scope("_SB");
1149     Aml *pci0_scope = aml_scope("PCI0");
1150 
1151     aml_append(pci0_scope, build_prt(true));
1152     aml_append(sb_scope, pci0_scope);
1153 
1154     field = aml_field("PCI0.ISA.P40C", AML_BYTE_ACC, AML_NOLOCK, AML_PRESERVE);
1155     aml_append(field, aml_named_field("PRQ0", 8));
1156     aml_append(field, aml_named_field("PRQ1", 8));
1157     aml_append(field, aml_named_field("PRQ2", 8));
1158     aml_append(field, aml_named_field("PRQ3", 8));
1159     aml_append(sb_scope, field);
1160 
1161     aml_append(sb_scope, build_irq_status_method());
1162     aml_append(sb_scope, build_iqcr_method(true));
1163 
1164     aml_append(sb_scope, build_link_dev("LNKA", 0, aml_name("PRQ0")));
1165     aml_append(sb_scope, build_link_dev("LNKB", 1, aml_name("PRQ1")));
1166     aml_append(sb_scope, build_link_dev("LNKC", 2, aml_name("PRQ2")));
1167     aml_append(sb_scope, build_link_dev("LNKD", 3, aml_name("PRQ3")));
1168 
1169     dev = aml_device("LNKS");
1170     {
1171         aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0C0F")));
1172         aml_append(dev, aml_name_decl("_UID", aml_int(4)));
1173 
1174         crs = aml_resource_template();
1175         irqs = 9;
1176         aml_append(crs, aml_interrupt(AML_CONSUMER, AML_LEVEL,
1177                                       AML_ACTIVE_HIGH, AML_SHARED,
1178                                       &irqs, 1));
1179         aml_append(dev, aml_name_decl("_PRS", crs));
1180 
1181         /* The SCI cannot be disabled and is always attached to GSI 9,
1182          * so these are no-ops.  We only need this link to override the
1183          * polarity to active high and match the content of the MADT.
1184          */
1185         method = aml_method("_STA", 0, AML_NOTSERIALIZED);
1186         aml_append(method, aml_return(aml_int(0x0b)));
1187         aml_append(dev, method);
1188 
1189         method = aml_method("_DIS", 0, AML_NOTSERIALIZED);
1190         aml_append(dev, method);
1191 
1192         method = aml_method("_CRS", 0, AML_NOTSERIALIZED);
1193         aml_append(method, aml_return(aml_name("_PRS")));
1194         aml_append(dev, method);
1195 
1196         method = aml_method("_SRS", 1, AML_NOTSERIALIZED);
1197         aml_append(dev, method);
1198     }
1199     aml_append(sb_scope, dev);
1200 
1201     aml_append(table, sb_scope);
1202 }
1203 
1204 static void append_q35_prt_entry(Aml *ctx, uint32_t nr, const char *name)
1205 {
1206     int i;
1207     int head;
1208     Aml *pkg;
1209     char base = name[3] < 'E' ? 'A' : 'E';
1210     char *s = g_strdup(name);
1211     Aml *a_nr = aml_int((nr << 16) | 0xffff);
1212 
1213     assert(strlen(s) == 4);
1214 
1215     head = name[3] - base;
1216     for (i = 0; i < 4; i++) {
1217         if (head + i > 3) {
1218             head = i * -1;
1219         }
1220         s[3] = base + head + i;
1221         pkg = aml_package(4);
1222         aml_append(pkg, a_nr);
1223         aml_append(pkg, aml_int(i));
1224         aml_append(pkg, aml_name("%s", s));
1225         aml_append(pkg, aml_int(0));
1226         aml_append(ctx, pkg);
1227     }
1228     g_free(s);
1229 }
1230 
1231 static Aml *build_q35_routing_table(const char *str)
1232 {
1233     int i;
1234     Aml *pkg;
1235     char *name = g_strdup_printf("%s ", str);
1236 
1237     pkg = aml_package(128);
1238     for (i = 0; i < 0x18; i++) {
1239             name[3] = 'E' + (i & 0x3);
1240             append_q35_prt_entry(pkg, i, name);
1241     }
1242 
1243     name[3] = 'E';
1244     append_q35_prt_entry(pkg, 0x18, name);
1245 
1246     /* INTA -> PIRQA for slot 25 - 31, see the default value of D<N>IR */
1247     for (i = 0x0019; i < 0x1e; i++) {
1248         name[3] = 'A';
1249         append_q35_prt_entry(pkg, i, name);
1250     }
1251 
1252     /* PCIe->PCI bridge. use PIRQ[E-H] */
1253     name[3] = 'E';
1254     append_q35_prt_entry(pkg, 0x1e, name);
1255     name[3] = 'A';
1256     append_q35_prt_entry(pkg, 0x1f, name);
1257 
1258     g_free(name);
1259     return pkg;
1260 }
1261 
1262 static void build_q35_pci0_int(Aml *table)
1263 {
1264     Aml *field;
1265     Aml *method;
1266     Aml *sb_scope = aml_scope("_SB");
1267     Aml *pci0_scope = aml_scope("PCI0");
1268 
1269     /* Zero => PIC mode, One => APIC Mode */
1270     aml_append(table, aml_name_decl("PICF", aml_int(0)));
1271     method = aml_method("_PIC", 1, AML_NOTSERIALIZED);
1272     {
1273         aml_append(method, aml_store(aml_arg(0), aml_name("PICF")));
1274     }
1275     aml_append(table, method);
1276 
1277     aml_append(pci0_scope,
1278         aml_name_decl("PRTP", build_q35_routing_table("LNK")));
1279     aml_append(pci0_scope,
1280         aml_name_decl("PRTA", build_q35_routing_table("GSI")));
1281 
1282     method = aml_method("_PRT", 0, AML_NOTSERIALIZED);
1283     {
1284         Aml *if_ctx;
1285         Aml *else_ctx;
1286 
1287         /* PCI IRQ routing table, example from ACPI 2.0a specification,
1288            section 6.2.8.1 */
1289         /* Note: we provide the same info as the PCI routing
1290            table of the Bochs BIOS */
1291         if_ctx = aml_if(aml_equal(aml_name("PICF"), aml_int(0)));
1292         aml_append(if_ctx, aml_return(aml_name("PRTP")));
1293         aml_append(method, if_ctx);
1294         else_ctx = aml_else();
1295         aml_append(else_ctx, aml_return(aml_name("PRTA")));
1296         aml_append(method, else_ctx);
1297     }
1298     aml_append(pci0_scope, method);
1299     aml_append(sb_scope, pci0_scope);
1300 
1301     field = aml_field("PCI0.ISA.PIRQ", AML_BYTE_ACC, AML_NOLOCK, AML_PRESERVE);
1302     aml_append(field, aml_named_field("PRQA", 8));
1303     aml_append(field, aml_named_field("PRQB", 8));
1304     aml_append(field, aml_named_field("PRQC", 8));
1305     aml_append(field, aml_named_field("PRQD", 8));
1306     aml_append(field, aml_reserved_field(0x20));
1307     aml_append(field, aml_named_field("PRQE", 8));
1308     aml_append(field, aml_named_field("PRQF", 8));
1309     aml_append(field, aml_named_field("PRQG", 8));
1310     aml_append(field, aml_named_field("PRQH", 8));
1311     aml_append(sb_scope, field);
1312 
1313     aml_append(sb_scope, build_irq_status_method());
1314     aml_append(sb_scope, build_iqcr_method(false));
1315 
1316     aml_append(sb_scope, build_link_dev("LNKA", 0, aml_name("PRQA")));
1317     aml_append(sb_scope, build_link_dev("LNKB", 1, aml_name("PRQB")));
1318     aml_append(sb_scope, build_link_dev("LNKC", 2, aml_name("PRQC")));
1319     aml_append(sb_scope, build_link_dev("LNKD", 3, aml_name("PRQD")));
1320     aml_append(sb_scope, build_link_dev("LNKE", 4, aml_name("PRQE")));
1321     aml_append(sb_scope, build_link_dev("LNKF", 5, aml_name("PRQF")));
1322     aml_append(sb_scope, build_link_dev("LNKG", 6, aml_name("PRQG")));
1323     aml_append(sb_scope, build_link_dev("LNKH", 7, aml_name("PRQH")));
1324 
1325     aml_append(sb_scope, build_gsi_link_dev("GSIA", 0x10, 0x10));
1326     aml_append(sb_scope, build_gsi_link_dev("GSIB", 0x11, 0x11));
1327     aml_append(sb_scope, build_gsi_link_dev("GSIC", 0x12, 0x12));
1328     aml_append(sb_scope, build_gsi_link_dev("GSID", 0x13, 0x13));
1329     aml_append(sb_scope, build_gsi_link_dev("GSIE", 0x14, 0x14));
1330     aml_append(sb_scope, build_gsi_link_dev("GSIF", 0x15, 0x15));
1331     aml_append(sb_scope, build_gsi_link_dev("GSIG", 0x16, 0x16));
1332     aml_append(sb_scope, build_gsi_link_dev("GSIH", 0x17, 0x17));
1333 
1334     aml_append(table, sb_scope);
1335 }
1336 
1337 static void build_q35_isa_bridge(Aml *table)
1338 {
1339     Aml *dev;
1340     Aml *scope;
1341 
1342     scope =  aml_scope("_SB.PCI0");
1343     dev = aml_device("ISA");
1344     aml_append(dev, aml_name_decl("_ADR", aml_int(0x001F0000)));
1345 
1346     /* ICH9 PCI to ISA irq remapping */
1347     aml_append(dev, aml_operation_region("PIRQ", AML_PCI_CONFIG,
1348                                          aml_int(0x60), 0x0C));
1349 
1350     aml_append(scope, dev);
1351     aml_append(table, scope);
1352 }
1353 
1354 static void build_piix4_isa_bridge(Aml *table)
1355 {
1356     Aml *dev;
1357     Aml *scope;
1358 
1359     scope =  aml_scope("_SB.PCI0");
1360     dev = aml_device("ISA");
1361     aml_append(dev, aml_name_decl("_ADR", aml_int(0x00010000)));
1362 
1363     /* PIIX PCI to ISA irq remapping */
1364     aml_append(dev, aml_operation_region("P40C", AML_PCI_CONFIG,
1365                                          aml_int(0x60), 0x04));
1366 
1367     aml_append(scope, dev);
1368     aml_append(table, scope);
1369 }
1370 
1371 static void build_piix4_pci_hotplug(Aml *table)
1372 {
1373     Aml *scope;
1374     Aml *field;
1375     Aml *method;
1376 
1377     scope =  aml_scope("_SB.PCI0");
1378 
1379     aml_append(scope,
1380         aml_operation_region("PCST", AML_SYSTEM_IO, aml_int(0xae00), 0x08));
1381     field = aml_field("PCST", AML_DWORD_ACC, AML_NOLOCK, AML_WRITE_AS_ZEROS);
1382     aml_append(field, aml_named_field("PCIU", 32));
1383     aml_append(field, aml_named_field("PCID", 32));
1384     aml_append(scope, field);
1385 
1386     aml_append(scope,
1387         aml_operation_region("SEJ", AML_SYSTEM_IO, aml_int(0xae08), 0x04));
1388     field = aml_field("SEJ", AML_DWORD_ACC, AML_NOLOCK, AML_WRITE_AS_ZEROS);
1389     aml_append(field, aml_named_field("B0EJ", 32));
1390     aml_append(scope, field);
1391 
1392     aml_append(scope,
1393         aml_operation_region("BNMR", AML_SYSTEM_IO, aml_int(0xae10), 0x04));
1394     field = aml_field("BNMR", AML_DWORD_ACC, AML_NOLOCK, AML_WRITE_AS_ZEROS);
1395     aml_append(field, aml_named_field("BNUM", 32));
1396     aml_append(scope, field);
1397 
1398     aml_append(scope, aml_mutex("BLCK", 0));
1399 
1400     method = aml_method("PCEJ", 2, AML_NOTSERIALIZED);
1401     aml_append(method, aml_acquire(aml_name("BLCK"), 0xFFFF));
1402     aml_append(method, aml_store(aml_arg(0), aml_name("BNUM")));
1403     aml_append(method,
1404         aml_store(aml_shiftleft(aml_int(1), aml_arg(1)), aml_name("B0EJ")));
1405     aml_append(method, aml_release(aml_name("BLCK")));
1406     aml_append(method, aml_return(aml_int(0)));
1407     aml_append(scope, method);
1408 
1409     aml_append(table, scope);
1410 }
1411 
1412 static Aml *build_q35_osc_method(void)
1413 {
1414     Aml *if_ctx;
1415     Aml *if_ctx2;
1416     Aml *else_ctx;
1417     Aml *method;
1418     Aml *a_cwd1 = aml_name("CDW1");
1419     Aml *a_ctrl = aml_local(0);
1420 
1421     method = aml_method("_OSC", 4, AML_NOTSERIALIZED);
1422     aml_append(method, aml_create_dword_field(aml_arg(3), aml_int(0), "CDW1"));
1423 
1424     if_ctx = aml_if(aml_equal(
1425         aml_arg(0), aml_touuid("33DB4D5B-1FF7-401C-9657-7441C03DD766")));
1426     aml_append(if_ctx, aml_create_dword_field(aml_arg(3), aml_int(4), "CDW2"));
1427     aml_append(if_ctx, aml_create_dword_field(aml_arg(3), aml_int(8), "CDW3"));
1428 
1429     aml_append(if_ctx, aml_store(aml_name("CDW3"), a_ctrl));
1430 
1431     /*
1432      * Always allow native PME, AER (no dependencies)
1433      * Allow SHPC (PCI bridges can have SHPC controller)
1434      */
1435     aml_append(if_ctx, aml_and(a_ctrl, aml_int(0x1F), a_ctrl));
1436 
1437     if_ctx2 = aml_if(aml_lnot(aml_equal(aml_arg(1), aml_int(1))));
1438     /* Unknown revision */
1439     aml_append(if_ctx2, aml_or(a_cwd1, aml_int(0x08), a_cwd1));
1440     aml_append(if_ctx, if_ctx2);
1441 
1442     if_ctx2 = aml_if(aml_lnot(aml_equal(aml_name("CDW3"), a_ctrl)));
1443     /* Capabilities bits were masked */
1444     aml_append(if_ctx2, aml_or(a_cwd1, aml_int(0x10), a_cwd1));
1445     aml_append(if_ctx, if_ctx2);
1446 
1447     /* Update DWORD3 in the buffer */
1448     aml_append(if_ctx, aml_store(a_ctrl, aml_name("CDW3")));
1449     aml_append(method, if_ctx);
1450 
1451     else_ctx = aml_else();
1452     /* Unrecognized UUID */
1453     aml_append(else_ctx, aml_or(a_cwd1, aml_int(4), a_cwd1));
1454     aml_append(method, else_ctx);
1455 
1456     aml_append(method, aml_return(aml_arg(3)));
1457     return method;
1458 }
1459 
1460 static void build_smb0(Aml *table, I2CBus *smbus, int devnr, int func)
1461 {
1462     Aml *scope = aml_scope("_SB.PCI0");
1463     Aml *dev = aml_device("SMB0");
1464 
1465     aml_append(dev, aml_name_decl("_ADR", aml_int(devnr << 16 | func)));
1466     build_acpi_ipmi_devices(dev, BUS(smbus), "\\_SB.PCI0.SMB0");
1467     aml_append(scope, dev);
1468     aml_append(table, scope);
1469 }
1470 
1471 static void
1472 build_dsdt(GArray *table_data, BIOSLinker *linker,
1473            AcpiPmInfo *pm, AcpiMiscInfo *misc,
1474            Range *pci_hole, Range *pci_hole64, MachineState *machine)
1475 {
1476     CrsRangeEntry *entry;
1477     Aml *dsdt, *sb_scope, *scope, *dev, *method, *field, *pkg, *crs;
1478     CrsRangeSet crs_range_set;
1479     PCMachineState *pcms = PC_MACHINE(machine);
1480     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(machine);
1481     X86MachineState *x86ms = X86_MACHINE(machine);
1482     AcpiMcfgInfo mcfg;
1483     uint32_t nr_mem = machine->ram_slots;
1484     int root_bus_limit = 0xFF;
1485     PCIBus *bus = NULL;
1486     TPMIf *tpm = tpm_find();
1487     int i;
1488 
1489     dsdt = init_aml_allocator();
1490 
1491     /* Reserve space for header */
1492     acpi_data_push(dsdt->buf, sizeof(AcpiTableHeader));
1493 
1494     build_dbg_aml(dsdt);
1495     if (misc->is_piix4) {
1496         sb_scope = aml_scope("_SB");
1497         dev = aml_device("PCI0");
1498         aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0A03")));
1499         aml_append(dev, aml_name_decl("_ADR", aml_int(0)));
1500         aml_append(dev, aml_name_decl("_UID", aml_int(0)));
1501         aml_append(sb_scope, dev);
1502         aml_append(dsdt, sb_scope);
1503 
1504         build_hpet_aml(dsdt);
1505         build_piix4_isa_bridge(dsdt);
1506         build_isa_devices_aml(dsdt);
1507         build_piix4_pci_hotplug(dsdt);
1508         build_piix4_pci0_int(dsdt);
1509     } else {
1510         sb_scope = aml_scope("_SB");
1511         dev = aml_device("PCI0");
1512         aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0A08")));
1513         aml_append(dev, aml_name_decl("_CID", aml_eisaid("PNP0A03")));
1514         aml_append(dev, aml_name_decl("_ADR", aml_int(0)));
1515         aml_append(dev, aml_name_decl("_UID", aml_int(0)));
1516         aml_append(dev, build_q35_osc_method());
1517         aml_append(sb_scope, dev);
1518         aml_append(dsdt, sb_scope);
1519 
1520         build_hpet_aml(dsdt);
1521         build_q35_isa_bridge(dsdt);
1522         build_isa_devices_aml(dsdt);
1523         build_q35_pci0_int(dsdt);
1524         if (pcms->smbus && !pcmc->do_not_add_smb_acpi) {
1525             build_smb0(dsdt, pcms->smbus, ICH9_SMB_DEV, ICH9_SMB_FUNC);
1526         }
1527     }
1528 
1529     if (pcmc->legacy_cpu_hotplug) {
1530         build_legacy_cpu_hotplug_aml(dsdt, machine, pm->cpu_hp_io_base);
1531     } else {
1532         CPUHotplugFeatures opts = {
1533             .acpi_1_compatible = true, .has_legacy_cphp = true
1534         };
1535         build_cpus_aml(dsdt, machine, opts, pm->cpu_hp_io_base,
1536                        "\\_SB.PCI0", "\\_GPE._E02");
1537     }
1538 
1539     if (pcms->memhp_io_base && nr_mem) {
1540         build_memory_hotplug_aml(dsdt, nr_mem, "\\_SB.PCI0",
1541                                  "\\_GPE._E03", AML_SYSTEM_IO,
1542                                  pcms->memhp_io_base);
1543     }
1544 
1545     scope =  aml_scope("_GPE");
1546     {
1547         aml_append(scope, aml_name_decl("_HID", aml_string("ACPI0006")));
1548 
1549         if (misc->is_piix4) {
1550             method = aml_method("_E01", 0, AML_NOTSERIALIZED);
1551             aml_append(method,
1552                 aml_acquire(aml_name("\\_SB.PCI0.BLCK"), 0xFFFF));
1553             aml_append(method, aml_call0("\\_SB.PCI0.PCNT"));
1554             aml_append(method, aml_release(aml_name("\\_SB.PCI0.BLCK")));
1555             aml_append(scope, method);
1556         }
1557 
1558         if (machine->nvdimms_state->is_enabled) {
1559             method = aml_method("_E04", 0, AML_NOTSERIALIZED);
1560             aml_append(method, aml_notify(aml_name("\\_SB.NVDR"),
1561                                           aml_int(0x80)));
1562             aml_append(scope, method);
1563         }
1564     }
1565     aml_append(dsdt, scope);
1566 
1567     crs_range_set_init(&crs_range_set);
1568     bus = PC_MACHINE(machine)->bus;
1569     if (bus) {
1570         QLIST_FOREACH(bus, &bus->child, sibling) {
1571             uint8_t bus_num = pci_bus_num(bus);
1572             uint8_t numa_node = pci_bus_numa_node(bus);
1573 
1574             /* look only for expander root buses */
1575             if (!pci_bus_is_root(bus)) {
1576                 continue;
1577             }
1578 
1579             if (bus_num < root_bus_limit) {
1580                 root_bus_limit = bus_num - 1;
1581             }
1582 
1583             scope = aml_scope("\\_SB");
1584             dev = aml_device("PC%.02X", bus_num);
1585             aml_append(dev, aml_name_decl("_UID", aml_int(bus_num)));
1586             aml_append(dev, aml_name_decl("_BBN", aml_int(bus_num)));
1587             if (pci_bus_is_express(bus)) {
1588                 aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0A08")));
1589                 aml_append(dev, aml_name_decl("_CID", aml_eisaid("PNP0A03")));
1590                 aml_append(dev, build_q35_osc_method());
1591             } else {
1592                 aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0A03")));
1593             }
1594 
1595             if (numa_node != NUMA_NODE_UNASSIGNED) {
1596                 aml_append(dev, aml_name_decl("_PXM", aml_int(numa_node)));
1597             }
1598 
1599             aml_append(dev, build_prt(false));
1600             crs = build_crs(PCI_HOST_BRIDGE(BUS(bus)->parent), &crs_range_set);
1601             aml_append(dev, aml_name_decl("_CRS", crs));
1602             aml_append(scope, dev);
1603             aml_append(dsdt, scope);
1604         }
1605     }
1606 
1607     /*
1608      * At this point crs_range_set has all the ranges used by pci
1609      * busses *other* than PCI0.  These ranges will be excluded from
1610      * the PCI0._CRS.  Add mmconfig to the set so it will be excluded
1611      * too.
1612      */
1613     if (acpi_get_mcfg(&mcfg)) {
1614         crs_range_insert(crs_range_set.mem_ranges,
1615                          mcfg.base, mcfg.base + mcfg.size - 1);
1616     }
1617 
1618     scope = aml_scope("\\_SB.PCI0");
1619     /* build PCI0._CRS */
1620     crs = aml_resource_template();
1621     aml_append(crs,
1622         aml_word_bus_number(AML_MIN_FIXED, AML_MAX_FIXED, AML_POS_DECODE,
1623                             0x0000, 0x0, root_bus_limit,
1624                             0x0000, root_bus_limit + 1));
1625     aml_append(crs, aml_io(AML_DECODE16, 0x0CF8, 0x0CF8, 0x01, 0x08));
1626 
1627     aml_append(crs,
1628         aml_word_io(AML_MIN_FIXED, AML_MAX_FIXED,
1629                     AML_POS_DECODE, AML_ENTIRE_RANGE,
1630                     0x0000, 0x0000, 0x0CF7, 0x0000, 0x0CF8));
1631 
1632     crs_replace_with_free_ranges(crs_range_set.io_ranges, 0x0D00, 0xFFFF);
1633     for (i = 0; i < crs_range_set.io_ranges->len; i++) {
1634         entry = g_ptr_array_index(crs_range_set.io_ranges, i);
1635         aml_append(crs,
1636             aml_word_io(AML_MIN_FIXED, AML_MAX_FIXED,
1637                         AML_POS_DECODE, AML_ENTIRE_RANGE,
1638                         0x0000, entry->base, entry->limit,
1639                         0x0000, entry->limit - entry->base + 1));
1640     }
1641 
1642     aml_append(crs,
1643         aml_dword_memory(AML_POS_DECODE, AML_MIN_FIXED, AML_MAX_FIXED,
1644                          AML_CACHEABLE, AML_READ_WRITE,
1645                          0, 0x000A0000, 0x000BFFFF, 0, 0x00020000));
1646 
1647     crs_replace_with_free_ranges(crs_range_set.mem_ranges,
1648                                  range_lob(pci_hole),
1649                                  range_upb(pci_hole));
1650     for (i = 0; i < crs_range_set.mem_ranges->len; i++) {
1651         entry = g_ptr_array_index(crs_range_set.mem_ranges, i);
1652         aml_append(crs,
1653             aml_dword_memory(AML_POS_DECODE, AML_MIN_FIXED, AML_MAX_FIXED,
1654                              AML_NON_CACHEABLE, AML_READ_WRITE,
1655                              0, entry->base, entry->limit,
1656                              0, entry->limit - entry->base + 1));
1657     }
1658 
1659     if (!range_is_empty(pci_hole64)) {
1660         crs_replace_with_free_ranges(crs_range_set.mem_64bit_ranges,
1661                                      range_lob(pci_hole64),
1662                                      range_upb(pci_hole64));
1663         for (i = 0; i < crs_range_set.mem_64bit_ranges->len; i++) {
1664             entry = g_ptr_array_index(crs_range_set.mem_64bit_ranges, i);
1665             aml_append(crs,
1666                        aml_qword_memory(AML_POS_DECODE, AML_MIN_FIXED,
1667                                         AML_MAX_FIXED,
1668                                         AML_CACHEABLE, AML_READ_WRITE,
1669                                         0, entry->base, entry->limit,
1670                                         0, entry->limit - entry->base + 1));
1671         }
1672     }
1673 
1674     if (TPM_IS_TIS_ISA(tpm_find())) {
1675         aml_append(crs, aml_memory32_fixed(TPM_TIS_ADDR_BASE,
1676                    TPM_TIS_ADDR_SIZE, AML_READ_WRITE));
1677     }
1678     aml_append(scope, aml_name_decl("_CRS", crs));
1679 
1680     /* reserve GPE0 block resources */
1681     dev = aml_device("GPE0");
1682     aml_append(dev, aml_name_decl("_HID", aml_string("PNP0A06")));
1683     aml_append(dev, aml_name_decl("_UID", aml_string("GPE0 resources")));
1684     /* device present, functioning, decoding, not shown in UI */
1685     aml_append(dev, aml_name_decl("_STA", aml_int(0xB)));
1686     crs = aml_resource_template();
1687     aml_append(crs,
1688         aml_io(
1689                AML_DECODE16,
1690                pm->fadt.gpe0_blk.address,
1691                pm->fadt.gpe0_blk.address,
1692                1,
1693                pm->fadt.gpe0_blk.bit_width / 8)
1694     );
1695     aml_append(dev, aml_name_decl("_CRS", crs));
1696     aml_append(scope, dev);
1697 
1698     crs_range_set_free(&crs_range_set);
1699 
1700     /* reserve PCIHP resources */
1701     if (pm->pcihp_io_len) {
1702         dev = aml_device("PHPR");
1703         aml_append(dev, aml_name_decl("_HID", aml_string("PNP0A06")));
1704         aml_append(dev,
1705             aml_name_decl("_UID", aml_string("PCI Hotplug resources")));
1706         /* device present, functioning, decoding, not shown in UI */
1707         aml_append(dev, aml_name_decl("_STA", aml_int(0xB)));
1708         crs = aml_resource_template();
1709         aml_append(crs,
1710             aml_io(AML_DECODE16, pm->pcihp_io_base, pm->pcihp_io_base, 1,
1711                    pm->pcihp_io_len)
1712         );
1713         aml_append(dev, aml_name_decl("_CRS", crs));
1714         aml_append(scope, dev);
1715     }
1716     aml_append(dsdt, scope);
1717 
1718     /*  create S3_ / S4_ / S5_ packages if necessary */
1719     scope = aml_scope("\\");
1720     if (!pm->s3_disabled) {
1721         pkg = aml_package(4);
1722         aml_append(pkg, aml_int(1)); /* PM1a_CNT.SLP_TYP */
1723         aml_append(pkg, aml_int(1)); /* PM1b_CNT.SLP_TYP, FIXME: not impl. */
1724         aml_append(pkg, aml_int(0)); /* reserved */
1725         aml_append(pkg, aml_int(0)); /* reserved */
1726         aml_append(scope, aml_name_decl("_S3", pkg));
1727     }
1728 
1729     if (!pm->s4_disabled) {
1730         pkg = aml_package(4);
1731         aml_append(pkg, aml_int(pm->s4_val)); /* PM1a_CNT.SLP_TYP */
1732         /* PM1b_CNT.SLP_TYP, FIXME: not impl. */
1733         aml_append(pkg, aml_int(pm->s4_val));
1734         aml_append(pkg, aml_int(0)); /* reserved */
1735         aml_append(pkg, aml_int(0)); /* reserved */
1736         aml_append(scope, aml_name_decl("_S4", pkg));
1737     }
1738 
1739     pkg = aml_package(4);
1740     aml_append(pkg, aml_int(0)); /* PM1a_CNT.SLP_TYP */
1741     aml_append(pkg, aml_int(0)); /* PM1b_CNT.SLP_TYP not impl. */
1742     aml_append(pkg, aml_int(0)); /* reserved */
1743     aml_append(pkg, aml_int(0)); /* reserved */
1744     aml_append(scope, aml_name_decl("_S5", pkg));
1745     aml_append(dsdt, scope);
1746 
1747     /* create fw_cfg node, unconditionally */
1748     {
1749         scope = aml_scope("\\_SB.PCI0");
1750         fw_cfg_add_acpi_dsdt(scope, x86ms->fw_cfg);
1751         aml_append(dsdt, scope);
1752     }
1753 
1754     if (misc->applesmc_io_base) {
1755         scope = aml_scope("\\_SB.PCI0.ISA");
1756         dev = aml_device("SMC");
1757 
1758         aml_append(dev, aml_name_decl("_HID", aml_eisaid("APP0001")));
1759         /* device present, functioning, decoding, not shown in UI */
1760         aml_append(dev, aml_name_decl("_STA", aml_int(0xB)));
1761 
1762         crs = aml_resource_template();
1763         aml_append(crs,
1764             aml_io(AML_DECODE16, misc->applesmc_io_base, misc->applesmc_io_base,
1765                    0x01, APPLESMC_MAX_DATA_LENGTH)
1766         );
1767         aml_append(crs, aml_irq_no_flags(6));
1768         aml_append(dev, aml_name_decl("_CRS", crs));
1769 
1770         aml_append(scope, dev);
1771         aml_append(dsdt, scope);
1772     }
1773 
1774     if (misc->pvpanic_port) {
1775         scope = aml_scope("\\_SB.PCI0.ISA");
1776 
1777         dev = aml_device("PEVT");
1778         aml_append(dev, aml_name_decl("_HID", aml_string("QEMU0001")));
1779 
1780         crs = aml_resource_template();
1781         aml_append(crs,
1782             aml_io(AML_DECODE16, misc->pvpanic_port, misc->pvpanic_port, 1, 1)
1783         );
1784         aml_append(dev, aml_name_decl("_CRS", crs));
1785 
1786         aml_append(dev, aml_operation_region("PEOR", AML_SYSTEM_IO,
1787                                               aml_int(misc->pvpanic_port), 1));
1788         field = aml_field("PEOR", AML_BYTE_ACC, AML_NOLOCK, AML_PRESERVE);
1789         aml_append(field, aml_named_field("PEPT", 8));
1790         aml_append(dev, field);
1791 
1792         /* device present, functioning, decoding, shown in UI */
1793         aml_append(dev, aml_name_decl("_STA", aml_int(0xF)));
1794 
1795         method = aml_method("RDPT", 0, AML_NOTSERIALIZED);
1796         aml_append(method, aml_store(aml_name("PEPT"), aml_local(0)));
1797         aml_append(method, aml_return(aml_local(0)));
1798         aml_append(dev, method);
1799 
1800         method = aml_method("WRPT", 1, AML_NOTSERIALIZED);
1801         aml_append(method, aml_store(aml_arg(0), aml_name("PEPT")));
1802         aml_append(dev, method);
1803 
1804         aml_append(scope, dev);
1805         aml_append(dsdt, scope);
1806     }
1807 
1808     sb_scope = aml_scope("\\_SB");
1809     {
1810         Object *pci_host;
1811         PCIBus *bus = NULL;
1812 
1813         pci_host = acpi_get_i386_pci_host();
1814         if (pci_host) {
1815             bus = PCI_HOST_BRIDGE(pci_host)->bus;
1816         }
1817 
1818         if (bus) {
1819             Aml *scope = aml_scope("PCI0");
1820             /* Scan all PCI buses. Generate tables to support hotplug. */
1821             build_append_pci_bus_devices(scope, bus, pm->pcihp_bridge_en);
1822 
1823             if (TPM_IS_TIS_ISA(tpm)) {
1824                 if (misc->tpm_version == TPM_VERSION_2_0) {
1825                     dev = aml_device("TPM");
1826                     aml_append(dev, aml_name_decl("_HID",
1827                                                   aml_string("MSFT0101")));
1828                 } else {
1829                     dev = aml_device("ISA.TPM");
1830                     aml_append(dev, aml_name_decl("_HID",
1831                                                   aml_eisaid("PNP0C31")));
1832                 }
1833 
1834                 aml_append(dev, aml_name_decl("_STA", aml_int(0xF)));
1835                 crs = aml_resource_template();
1836                 aml_append(crs, aml_memory32_fixed(TPM_TIS_ADDR_BASE,
1837                            TPM_TIS_ADDR_SIZE, AML_READ_WRITE));
1838                 /*
1839                     FIXME: TPM_TIS_IRQ=5 conflicts with PNP0C0F irqs,
1840                     Rewrite to take IRQ from TPM device model and
1841                     fix default IRQ value there to use some unused IRQ
1842                  */
1843                 /* aml_append(crs, aml_irq_no_flags(TPM_TIS_IRQ)); */
1844                 aml_append(dev, aml_name_decl("_CRS", crs));
1845 
1846                 tpm_build_ppi_acpi(tpm, dev);
1847 
1848                 aml_append(scope, dev);
1849             }
1850 
1851             aml_append(sb_scope, scope);
1852         }
1853     }
1854 
1855     if (TPM_IS_CRB(tpm)) {
1856         dev = aml_device("TPM");
1857         aml_append(dev, aml_name_decl("_HID", aml_string("MSFT0101")));
1858         crs = aml_resource_template();
1859         aml_append(crs, aml_memory32_fixed(TPM_CRB_ADDR_BASE,
1860                                            TPM_CRB_ADDR_SIZE, AML_READ_WRITE));
1861         aml_append(dev, aml_name_decl("_CRS", crs));
1862 
1863         aml_append(dev, aml_name_decl("_STA", aml_int(0xf)));
1864 
1865         tpm_build_ppi_acpi(tpm, dev);
1866 
1867         aml_append(sb_scope, dev);
1868     }
1869 
1870     aml_append(dsdt, sb_scope);
1871 
1872     /* copy AML table into ACPI tables blob and patch header there */
1873     g_array_append_vals(table_data, dsdt->buf->data, dsdt->buf->len);
1874     build_header(linker, table_data,
1875         (void *)(table_data->data + table_data->len - dsdt->buf->len),
1876         "DSDT", dsdt->buf->len, 1, NULL, NULL);
1877     free_aml_allocator();
1878 }
1879 
1880 static void
1881 build_hpet(GArray *table_data, BIOSLinker *linker)
1882 {
1883     Acpi20Hpet *hpet;
1884 
1885     hpet = acpi_data_push(table_data, sizeof(*hpet));
1886     /* Note timer_block_id value must be kept in sync with value advertised by
1887      * emulated hpet
1888      */
1889     hpet->timer_block_id = cpu_to_le32(0x8086a201);
1890     hpet->addr.address = cpu_to_le64(HPET_BASE);
1891     build_header(linker, table_data,
1892                  (void *)hpet, "HPET", sizeof(*hpet), 1, NULL, NULL);
1893 }
1894 
1895 static void
1896 build_tpm_tcpa(GArray *table_data, BIOSLinker *linker, GArray *tcpalog)
1897 {
1898     Acpi20Tcpa *tcpa = acpi_data_push(table_data, sizeof *tcpa);
1899     unsigned log_addr_size = sizeof(tcpa->log_area_start_address);
1900     unsigned log_addr_offset =
1901         (char *)&tcpa->log_area_start_address - table_data->data;
1902 
1903     tcpa->platform_class = cpu_to_le16(TPM_TCPA_ACPI_CLASS_CLIENT);
1904     tcpa->log_area_minimum_length = cpu_to_le32(TPM_LOG_AREA_MINIMUM_SIZE);
1905     acpi_data_push(tcpalog, le32_to_cpu(tcpa->log_area_minimum_length));
1906 
1907     bios_linker_loader_alloc(linker, ACPI_BUILD_TPMLOG_FILE, tcpalog, 1,
1908                              false /* high memory */);
1909 
1910     /* log area start address to be filled by Guest linker */
1911     bios_linker_loader_add_pointer(linker,
1912         ACPI_BUILD_TABLE_FILE, log_addr_offset, log_addr_size,
1913         ACPI_BUILD_TPMLOG_FILE, 0);
1914 
1915     build_header(linker, table_data,
1916                  (void *)tcpa, "TCPA", sizeof(*tcpa), 2, NULL, NULL);
1917 }
1918 
1919 #define HOLE_640K_START  (640 * KiB)
1920 #define HOLE_640K_END   (1 * MiB)
1921 
1922 static void
1923 build_srat(GArray *table_data, BIOSLinker *linker, MachineState *machine)
1924 {
1925     AcpiSystemResourceAffinityTable *srat;
1926     AcpiSratMemoryAffinity *numamem;
1927 
1928     int i;
1929     int srat_start, numa_start, slots;
1930     uint64_t mem_len, mem_base, next_base;
1931     MachineClass *mc = MACHINE_GET_CLASS(machine);
1932     X86MachineState *x86ms = X86_MACHINE(machine);
1933     const CPUArchIdList *apic_ids = mc->possible_cpu_arch_ids(machine);
1934     PCMachineState *pcms = PC_MACHINE(machine);
1935     ram_addr_t hotplugabble_address_space_size =
1936         object_property_get_int(OBJECT(pcms), PC_MACHINE_DEVMEM_REGION_SIZE,
1937                                 NULL);
1938 
1939     srat_start = table_data->len;
1940 
1941     srat = acpi_data_push(table_data, sizeof *srat);
1942     srat->reserved1 = cpu_to_le32(1);
1943 
1944     for (i = 0; i < apic_ids->len; i++) {
1945         int node_id = apic_ids->cpus[i].props.node_id;
1946         uint32_t apic_id = apic_ids->cpus[i].arch_id;
1947 
1948         if (apic_id < 255) {
1949             AcpiSratProcessorAffinity *core;
1950 
1951             core = acpi_data_push(table_data, sizeof *core);
1952             core->type = ACPI_SRAT_PROCESSOR_APIC;
1953             core->length = sizeof(*core);
1954             core->local_apic_id = apic_id;
1955             core->proximity_lo = node_id;
1956             memset(core->proximity_hi, 0, 3);
1957             core->local_sapic_eid = 0;
1958             core->flags = cpu_to_le32(1);
1959         } else {
1960             AcpiSratProcessorX2ApicAffinity *core;
1961 
1962             core = acpi_data_push(table_data, sizeof *core);
1963             core->type = ACPI_SRAT_PROCESSOR_x2APIC;
1964             core->length = sizeof(*core);
1965             core->x2apic_id = cpu_to_le32(apic_id);
1966             core->proximity_domain = cpu_to_le32(node_id);
1967             core->flags = cpu_to_le32(1);
1968         }
1969     }
1970 
1971 
1972     /* the memory map is a bit tricky, it contains at least one hole
1973      * from 640k-1M and possibly another one from 3.5G-4G.
1974      */
1975     next_base = 0;
1976     numa_start = table_data->len;
1977 
1978     for (i = 1; i < pcms->numa_nodes + 1; ++i) {
1979         mem_base = next_base;
1980         mem_len = pcms->node_mem[i - 1];
1981         next_base = mem_base + mem_len;
1982 
1983         /* Cut out the 640K hole */
1984         if (mem_base <= HOLE_640K_START &&
1985             next_base > HOLE_640K_START) {
1986             mem_len -= next_base - HOLE_640K_START;
1987             if (mem_len > 0) {
1988                 numamem = acpi_data_push(table_data, sizeof *numamem);
1989                 build_srat_memory(numamem, mem_base, mem_len, i - 1,
1990                                   MEM_AFFINITY_ENABLED);
1991             }
1992 
1993             /* Check for the rare case: 640K < RAM < 1M */
1994             if (next_base <= HOLE_640K_END) {
1995                 next_base = HOLE_640K_END;
1996                 continue;
1997             }
1998             mem_base = HOLE_640K_END;
1999             mem_len = next_base - HOLE_640K_END;
2000         }
2001 
2002         /* Cut out the ACPI_PCI hole */
2003         if (mem_base <= x86ms->below_4g_mem_size &&
2004             next_base > x86ms->below_4g_mem_size) {
2005             mem_len -= next_base - x86ms->below_4g_mem_size;
2006             if (mem_len > 0) {
2007                 numamem = acpi_data_push(table_data, sizeof *numamem);
2008                 build_srat_memory(numamem, mem_base, mem_len, i - 1,
2009                                   MEM_AFFINITY_ENABLED);
2010             }
2011             mem_base = 1ULL << 32;
2012             mem_len = next_base - x86ms->below_4g_mem_size;
2013             next_base = mem_base + mem_len;
2014         }
2015 
2016         if (mem_len > 0) {
2017             numamem = acpi_data_push(table_data, sizeof *numamem);
2018             build_srat_memory(numamem, mem_base, mem_len, i - 1,
2019                               MEM_AFFINITY_ENABLED);
2020         }
2021     }
2022 
2023     if (machine->nvdimms_state->is_enabled) {
2024         nvdimm_build_srat(table_data);
2025     }
2026 
2027     slots = (table_data->len - numa_start) / sizeof *numamem;
2028     for (; slots < pcms->numa_nodes + 2; slots++) {
2029         numamem = acpi_data_push(table_data, sizeof *numamem);
2030         build_srat_memory(numamem, 0, 0, 0, MEM_AFFINITY_NOFLAGS);
2031     }
2032 
2033     /*
2034      * Entry is required for Windows to enable memory hotplug in OS
2035      * and for Linux to enable SWIOTLB when booted with less than
2036      * 4G of RAM. Windows works better if the entry sets proximity
2037      * to the highest NUMA node in the machine.
2038      * Memory devices may override proximity set by this entry,
2039      * providing _PXM method if necessary.
2040      */
2041     if (hotplugabble_address_space_size) {
2042         numamem = acpi_data_push(table_data, sizeof *numamem);
2043         build_srat_memory(numamem, machine->device_memory->base,
2044                           hotplugabble_address_space_size, pcms->numa_nodes - 1,
2045                           MEM_AFFINITY_HOTPLUGGABLE | MEM_AFFINITY_ENABLED);
2046     }
2047 
2048     build_header(linker, table_data,
2049                  (void *)(table_data->data + srat_start),
2050                  "SRAT",
2051                  table_data->len - srat_start, 1, NULL, NULL);
2052 }
2053 
2054 /*
2055  * VT-d spec 8.1 DMA Remapping Reporting Structure
2056  * (version Oct. 2014 or later)
2057  */
2058 static void
2059 build_dmar_q35(GArray *table_data, BIOSLinker *linker)
2060 {
2061     int dmar_start = table_data->len;
2062 
2063     AcpiTableDmar *dmar;
2064     AcpiDmarHardwareUnit *drhd;
2065     AcpiDmarRootPortATS *atsr;
2066     uint8_t dmar_flags = 0;
2067     X86IOMMUState *iommu = x86_iommu_get_default();
2068     AcpiDmarDeviceScope *scope = NULL;
2069     /* Root complex IOAPIC use one path[0] only */
2070     size_t ioapic_scope_size = sizeof(*scope) + sizeof(scope->path[0]);
2071     IntelIOMMUState *intel_iommu = INTEL_IOMMU_DEVICE(iommu);
2072 
2073     assert(iommu);
2074     if (x86_iommu_ir_supported(iommu)) {
2075         dmar_flags |= 0x1;      /* Flags: 0x1: INT_REMAP */
2076     }
2077 
2078     dmar = acpi_data_push(table_data, sizeof(*dmar));
2079     dmar->host_address_width = intel_iommu->aw_bits - 1;
2080     dmar->flags = dmar_flags;
2081 
2082     /* DMAR Remapping Hardware Unit Definition structure */
2083     drhd = acpi_data_push(table_data, sizeof(*drhd) + ioapic_scope_size);
2084     drhd->type = cpu_to_le16(ACPI_DMAR_TYPE_HARDWARE_UNIT);
2085     drhd->length = cpu_to_le16(sizeof(*drhd) + ioapic_scope_size);
2086     drhd->flags = ACPI_DMAR_INCLUDE_PCI_ALL;
2087     drhd->pci_segment = cpu_to_le16(0);
2088     drhd->address = cpu_to_le64(Q35_HOST_BRIDGE_IOMMU_ADDR);
2089 
2090     /* Scope definition for the root-complex IOAPIC. See VT-d spec
2091      * 8.3.1 (version Oct. 2014 or later). */
2092     scope = &drhd->scope[0];
2093     scope->entry_type = 0x03;   /* Type: 0x03 for IOAPIC */
2094     scope->length = ioapic_scope_size;
2095     scope->enumeration_id = ACPI_BUILD_IOAPIC_ID;
2096     scope->bus = Q35_PSEUDO_BUS_PLATFORM;
2097     scope->path[0].device = PCI_SLOT(Q35_PSEUDO_DEVFN_IOAPIC);
2098     scope->path[0].function = PCI_FUNC(Q35_PSEUDO_DEVFN_IOAPIC);
2099 
2100     if (iommu->dt_supported) {
2101         atsr = acpi_data_push(table_data, sizeof(*atsr));
2102         atsr->type = cpu_to_le16(ACPI_DMAR_TYPE_ATSR);
2103         atsr->length = cpu_to_le16(sizeof(*atsr));
2104         atsr->flags = ACPI_DMAR_ATSR_ALL_PORTS;
2105         atsr->pci_segment = cpu_to_le16(0);
2106     }
2107 
2108     build_header(linker, table_data, (void *)(table_data->data + dmar_start),
2109                  "DMAR", table_data->len - dmar_start, 1, NULL, NULL);
2110 }
2111 
2112 /*
2113  * Windows ACPI Emulated Devices Table
2114  * (Version 1.0 - April 6, 2009)
2115  * Spec: http://download.microsoft.com/download/7/E/7/7E7662CF-CBEA-470B-A97E-CE7CE0D98DC2/WAET.docx
2116  *
2117  * Helpful to speedup Windows guests and ignored by others.
2118  */
2119 static void
2120 build_waet(GArray *table_data, BIOSLinker *linker)
2121 {
2122     int waet_start = table_data->len;
2123 
2124     /* WAET header */
2125     acpi_data_push(table_data, sizeof(AcpiTableHeader));
2126     /*
2127      * Set "ACPI PM timer good" flag.
2128      *
2129      * Tells Windows guests that our ACPI PM timer is reliable in the
2130      * sense that guest can read it only once to obtain a reliable value.
2131      * Which avoids costly VMExits caused by guest re-reading it unnecessarily.
2132      */
2133     build_append_int_noprefix(table_data, 1 << 1 /* ACPI PM timer good */, 4);
2134 
2135     build_header(linker, table_data, (void *)(table_data->data + waet_start),
2136                  "WAET", table_data->len - waet_start, 1, NULL, NULL);
2137 }
2138 
2139 /*
2140  *   IVRS table as specified in AMD IOMMU Specification v2.62, Section 5.2
2141  *   accessible here http://support.amd.com/TechDocs/48882_IOMMU.pdf
2142  */
2143 #define IOAPIC_SB_DEVID   (uint64_t)PCI_BUILD_BDF(0, PCI_DEVFN(0x14, 0))
2144 
2145 /*
2146  * Insert IVHD entry for device and recurse, insert alias, or insert range as
2147  * necessary for the PCI topology.
2148  */
2149 static void
2150 insert_ivhd(PCIBus *bus, PCIDevice *dev, void *opaque)
2151 {
2152     GArray *table_data = opaque;
2153     uint32_t entry;
2154 
2155     /* "Select" IVHD entry, type 0x2 */
2156     entry = PCI_BUILD_BDF(pci_bus_num(bus), dev->devfn) << 8 | 0x2;
2157     build_append_int_noprefix(table_data, entry, 4);
2158 
2159     if (object_dynamic_cast(OBJECT(dev), TYPE_PCI_BRIDGE)) {
2160         PCIBus *sec_bus = pci_bridge_get_sec_bus(PCI_BRIDGE(dev));
2161         uint8_t sec = pci_bus_num(sec_bus);
2162         uint8_t sub = dev->config[PCI_SUBORDINATE_BUS];
2163 
2164         if (pci_bus_is_express(sec_bus)) {
2165             /*
2166              * Walk the bus if there are subordinates, otherwise use a range
2167              * to cover an entire leaf bus.  We could potentially also use a
2168              * range for traversed buses, but we'd need to take care not to
2169              * create both Select and Range entries covering the same device.
2170              * This is easier and potentially more compact.
2171              *
2172              * An example bare metal system seems to use Select entries for
2173              * root ports without a slot (ie. built-ins) and Range entries
2174              * when there is a slot.  The same system also only hard-codes
2175              * the alias range for an onboard PCIe-to-PCI bridge, apparently
2176              * making no effort to support nested bridges.  We attempt to
2177              * be more thorough here.
2178              */
2179             if (sec == sub) { /* leaf bus */
2180                 /* "Start of Range" IVHD entry, type 0x3 */
2181                 entry = PCI_BUILD_BDF(sec, PCI_DEVFN(0, 0)) << 8 | 0x3;
2182                 build_append_int_noprefix(table_data, entry, 4);
2183                 /* "End of Range" IVHD entry, type 0x4 */
2184                 entry = PCI_BUILD_BDF(sub, PCI_DEVFN(31, 7)) << 8 | 0x4;
2185                 build_append_int_noprefix(table_data, entry, 4);
2186             } else {
2187                 pci_for_each_device(sec_bus, sec, insert_ivhd, table_data);
2188             }
2189         } else {
2190             /*
2191              * If the secondary bus is conventional, then we need to create an
2192              * Alias range for everything downstream.  The range covers the
2193              * first devfn on the secondary bus to the last devfn on the
2194              * subordinate bus.  The alias target depends on legacy versus
2195              * express bridges, just as in pci_device_iommu_address_space().
2196              * DeviceIDa vs DeviceIDb as per the AMD IOMMU spec.
2197              */
2198             uint16_t dev_id_a, dev_id_b;
2199 
2200             dev_id_a = PCI_BUILD_BDF(sec, PCI_DEVFN(0, 0));
2201 
2202             if (pci_is_express(dev) &&
2203                 pcie_cap_get_type(dev) == PCI_EXP_TYPE_PCI_BRIDGE) {
2204                 dev_id_b = dev_id_a;
2205             } else {
2206                 dev_id_b = PCI_BUILD_BDF(pci_bus_num(bus), dev->devfn);
2207             }
2208 
2209             /* "Alias Start of Range" IVHD entry, type 0x43, 8 bytes */
2210             build_append_int_noprefix(table_data, dev_id_a << 8 | 0x43, 4);
2211             build_append_int_noprefix(table_data, dev_id_b << 8 | 0x0, 4);
2212 
2213             /* "End of Range" IVHD entry, type 0x4 */
2214             entry = PCI_BUILD_BDF(sub, PCI_DEVFN(31, 7)) << 8 | 0x4;
2215             build_append_int_noprefix(table_data, entry, 4);
2216         }
2217     }
2218 }
2219 
2220 /* For all PCI host bridges, walk and insert IVHD entries */
2221 static int
2222 ivrs_host_bridges(Object *obj, void *opaque)
2223 {
2224     GArray *ivhd_blob = opaque;
2225 
2226     if (object_dynamic_cast(obj, TYPE_PCI_HOST_BRIDGE)) {
2227         PCIBus *bus = PCI_HOST_BRIDGE(obj)->bus;
2228 
2229         if (bus) {
2230             pci_for_each_device(bus, pci_bus_num(bus), insert_ivhd, ivhd_blob);
2231         }
2232     }
2233 
2234     return 0;
2235 }
2236 
2237 static void
2238 build_amd_iommu(GArray *table_data, BIOSLinker *linker)
2239 {
2240     int ivhd_table_len = 24;
2241     int iommu_start = table_data->len;
2242     AMDVIState *s = AMD_IOMMU_DEVICE(x86_iommu_get_default());
2243     GArray *ivhd_blob = g_array_new(false, true, 1);
2244 
2245     /* IVRS header */
2246     acpi_data_push(table_data, sizeof(AcpiTableHeader));
2247     /* IVinfo - IO virtualization information common to all
2248      * IOMMU units in a system
2249      */
2250     build_append_int_noprefix(table_data, 40UL << 8/* PASize */, 4);
2251     /* reserved */
2252     build_append_int_noprefix(table_data, 0, 8);
2253 
2254     /* IVHD definition - type 10h */
2255     build_append_int_noprefix(table_data, 0x10, 1);
2256     /* virtualization flags */
2257     build_append_int_noprefix(table_data,
2258                              (1UL << 0) | /* HtTunEn      */
2259                              (1UL << 4) | /* iotblSup     */
2260                              (1UL << 6) | /* PrefSup      */
2261                              (1UL << 7),  /* PPRSup       */
2262                              1);
2263 
2264     /*
2265      * A PCI bus walk, for each PCI host bridge, is necessary to create a
2266      * complete set of IVHD entries.  Do this into a separate blob so that we
2267      * can calculate the total IVRS table length here and then append the new
2268      * blob further below.  Fall back to an entry covering all devices, which
2269      * is sufficient when no aliases are present.
2270      */
2271     object_child_foreach_recursive(object_get_root(),
2272                                    ivrs_host_bridges, ivhd_blob);
2273 
2274     if (!ivhd_blob->len) {
2275         /*
2276          *   Type 1 device entry reporting all devices
2277          *   These are 4-byte device entries currently reporting the range of
2278          *   Refer to Spec - Table 95:IVHD Device Entry Type Codes(4-byte)
2279          */
2280         build_append_int_noprefix(ivhd_blob, 0x0000001, 4);
2281     }
2282 
2283     ivhd_table_len += ivhd_blob->len;
2284 
2285     /*
2286      * When interrupt remapping is supported, we add a special IVHD device
2287      * for type IO-APIC.
2288      */
2289     if (x86_iommu_ir_supported(x86_iommu_get_default())) {
2290         ivhd_table_len += 8;
2291     }
2292 
2293     /* IVHD length */
2294     build_append_int_noprefix(table_data, ivhd_table_len, 2);
2295     /* DeviceID */
2296     build_append_int_noprefix(table_data, s->devid, 2);
2297     /* Capability offset */
2298     build_append_int_noprefix(table_data, s->capab_offset, 2);
2299     /* IOMMU base address */
2300     build_append_int_noprefix(table_data, s->mmio.addr, 8);
2301     /* PCI Segment Group */
2302     build_append_int_noprefix(table_data, 0, 2);
2303     /* IOMMU info */
2304     build_append_int_noprefix(table_data, 0, 2);
2305     /* IOMMU Feature Reporting */
2306     build_append_int_noprefix(table_data,
2307                              (48UL << 30) | /* HATS   */
2308                              (48UL << 28) | /* GATS   */
2309                              (1UL << 2)   | /* GTSup  */
2310                              (1UL << 6),    /* GASup  */
2311                              4);
2312 
2313     /* IVHD entries as found above */
2314     g_array_append_vals(table_data, ivhd_blob->data, ivhd_blob->len);
2315     g_array_free(ivhd_blob, TRUE);
2316 
2317     /*
2318      * Add a special IVHD device type.
2319      * Refer to spec - Table 95: IVHD device entry type codes
2320      *
2321      * Linux IOMMU driver checks for the special IVHD device (type IO-APIC).
2322      * See Linux kernel commit 'c2ff5cf5294bcbd7fa50f7d860e90a66db7e5059'
2323      */
2324     if (x86_iommu_ir_supported(x86_iommu_get_default())) {
2325         build_append_int_noprefix(table_data,
2326                                  (0x1ull << 56) |           /* type IOAPIC */
2327                                  (IOAPIC_SB_DEVID << 40) |  /* IOAPIC devid */
2328                                  0x48,                      /* special device */
2329                                  8);
2330     }
2331 
2332     build_header(linker, table_data, (void *)(table_data->data + iommu_start),
2333                  "IVRS", table_data->len - iommu_start, 1, NULL, NULL);
2334 }
2335 
2336 typedef
2337 struct AcpiBuildState {
2338     /* Copy of table in RAM (for patching). */
2339     MemoryRegion *table_mr;
2340     /* Is table patched? */
2341     uint8_t patched;
2342     void *rsdp;
2343     MemoryRegion *rsdp_mr;
2344     MemoryRegion *linker_mr;
2345 } AcpiBuildState;
2346 
2347 static bool acpi_get_mcfg(AcpiMcfgInfo *mcfg)
2348 {
2349     Object *pci_host;
2350     QObject *o;
2351 
2352     pci_host = acpi_get_i386_pci_host();
2353     g_assert(pci_host);
2354 
2355     o = object_property_get_qobject(pci_host, PCIE_HOST_MCFG_BASE, NULL);
2356     if (!o) {
2357         return false;
2358     }
2359     mcfg->base = qnum_get_uint(qobject_to(QNum, o));
2360     qobject_unref(o);
2361     if (mcfg->base == PCIE_BASE_ADDR_UNMAPPED) {
2362         return false;
2363     }
2364 
2365     o = object_property_get_qobject(pci_host, PCIE_HOST_MCFG_SIZE, NULL);
2366     assert(o);
2367     mcfg->size = qnum_get_uint(qobject_to(QNum, o));
2368     qobject_unref(o);
2369     return true;
2370 }
2371 
2372 static
2373 void acpi_build(AcpiBuildTables *tables, MachineState *machine)
2374 {
2375     PCMachineState *pcms = PC_MACHINE(machine);
2376     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
2377     X86MachineState *x86ms = X86_MACHINE(machine);
2378     GArray *table_offsets;
2379     unsigned facs, dsdt, rsdt, fadt;
2380     AcpiPmInfo pm;
2381     AcpiMiscInfo misc;
2382     AcpiMcfgInfo mcfg;
2383     Range pci_hole, pci_hole64;
2384     uint8_t *u;
2385     size_t aml_len = 0;
2386     GArray *tables_blob = tables->table_data;
2387     AcpiSlicOem slic_oem = { .id = NULL, .table_id = NULL };
2388     Object *vmgenid_dev;
2389 
2390     acpi_get_pm_info(machine, &pm);
2391     acpi_get_misc_info(&misc);
2392     acpi_get_pci_holes(&pci_hole, &pci_hole64);
2393     acpi_get_slic_oem(&slic_oem);
2394 
2395     table_offsets = g_array_new(false, true /* clear */,
2396                                         sizeof(uint32_t));
2397     ACPI_BUILD_DPRINTF("init ACPI tables\n");
2398 
2399     bios_linker_loader_alloc(tables->linker,
2400                              ACPI_BUILD_TABLE_FILE, tables_blob,
2401                              64 /* Ensure FACS is aligned */,
2402                              false /* high memory */);
2403 
2404     /*
2405      * FACS is pointed to by FADT.
2406      * We place it first since it's the only table that has alignment
2407      * requirements.
2408      */
2409     facs = tables_blob->len;
2410     build_facs(tables_blob);
2411 
2412     /* DSDT is pointed to by FADT */
2413     dsdt = tables_blob->len;
2414     build_dsdt(tables_blob, tables->linker, &pm, &misc,
2415                &pci_hole, &pci_hole64, machine);
2416 
2417     /* Count the size of the DSDT and SSDT, we will need it for legacy
2418      * sizing of ACPI tables.
2419      */
2420     aml_len += tables_blob->len - dsdt;
2421 
2422     /* ACPI tables pointed to by RSDT */
2423     fadt = tables_blob->len;
2424     acpi_add_table(table_offsets, tables_blob);
2425     pm.fadt.facs_tbl_offset = &facs;
2426     pm.fadt.dsdt_tbl_offset = &dsdt;
2427     pm.fadt.xdsdt_tbl_offset = &dsdt;
2428     build_fadt(tables_blob, tables->linker, &pm.fadt,
2429                slic_oem.id, slic_oem.table_id);
2430     aml_len += tables_blob->len - fadt;
2431 
2432     acpi_add_table(table_offsets, tables_blob);
2433     acpi_build_madt(tables_blob, tables->linker, x86ms,
2434                     ACPI_DEVICE_IF(x86ms->acpi_dev), true);
2435 
2436     vmgenid_dev = find_vmgenid_dev();
2437     if (vmgenid_dev) {
2438         acpi_add_table(table_offsets, tables_blob);
2439         vmgenid_build_acpi(VMGENID(vmgenid_dev), tables_blob,
2440                            tables->vmgenid, tables->linker);
2441     }
2442 
2443     if (misc.has_hpet) {
2444         acpi_add_table(table_offsets, tables_blob);
2445         build_hpet(tables_blob, tables->linker);
2446     }
2447     if (misc.tpm_version != TPM_VERSION_UNSPEC) {
2448         if (misc.tpm_version == TPM_VERSION_1_2) {
2449             acpi_add_table(table_offsets, tables_blob);
2450             build_tpm_tcpa(tables_blob, tables->linker, tables->tcpalog);
2451         } else { /* TPM_VERSION_2_0 */
2452             acpi_add_table(table_offsets, tables_blob);
2453             build_tpm2(tables_blob, tables->linker, tables->tcpalog);
2454         }
2455     }
2456     if (pcms->numa_nodes) {
2457         acpi_add_table(table_offsets, tables_blob);
2458         build_srat(tables_blob, tables->linker, machine);
2459         if (machine->numa_state->have_numa_distance) {
2460             acpi_add_table(table_offsets, tables_blob);
2461             build_slit(tables_blob, tables->linker, machine);
2462         }
2463         if (machine->numa_state->hmat_enabled) {
2464             acpi_add_table(table_offsets, tables_blob);
2465             build_hmat(tables_blob, tables->linker, machine->numa_state);
2466         }
2467     }
2468     if (acpi_get_mcfg(&mcfg)) {
2469         acpi_add_table(table_offsets, tables_blob);
2470         build_mcfg(tables_blob, tables->linker, &mcfg);
2471     }
2472     if (x86_iommu_get_default()) {
2473         IommuType IOMMUType = x86_iommu_get_type();
2474         if (IOMMUType == TYPE_AMD) {
2475             acpi_add_table(table_offsets, tables_blob);
2476             build_amd_iommu(tables_blob, tables->linker);
2477         } else if (IOMMUType == TYPE_INTEL) {
2478             acpi_add_table(table_offsets, tables_blob);
2479             build_dmar_q35(tables_blob, tables->linker);
2480         }
2481     }
2482     if (machine->nvdimms_state->is_enabled) {
2483         nvdimm_build_acpi(table_offsets, tables_blob, tables->linker,
2484                           machine->nvdimms_state, machine->ram_slots);
2485     }
2486 
2487     acpi_add_table(table_offsets, tables_blob);
2488     build_waet(tables_blob, tables->linker);
2489 
2490     /* Add tables supplied by user (if any) */
2491     for (u = acpi_table_first(); u; u = acpi_table_next(u)) {
2492         unsigned len = acpi_table_len(u);
2493 
2494         acpi_add_table(table_offsets, tables_blob);
2495         g_array_append_vals(tables_blob, u, len);
2496     }
2497 
2498     /* RSDT is pointed to by RSDP */
2499     rsdt = tables_blob->len;
2500     build_rsdt(tables_blob, tables->linker, table_offsets,
2501                slic_oem.id, slic_oem.table_id);
2502 
2503     /* RSDP is in FSEG memory, so allocate it separately */
2504     {
2505         AcpiRsdpData rsdp_data = {
2506             .revision = 0,
2507             .oem_id = ACPI_BUILD_APPNAME6,
2508             .xsdt_tbl_offset = NULL,
2509             .rsdt_tbl_offset = &rsdt,
2510         };
2511         build_rsdp(tables->rsdp, tables->linker, &rsdp_data);
2512         if (!pcmc->rsdp_in_ram) {
2513             /* We used to allocate some extra space for RSDP revision 2 but
2514              * only used the RSDP revision 0 space. The extra bytes were
2515              * zeroed out and not used.
2516              * Here we continue wasting those extra 16 bytes to make sure we
2517              * don't break migration for machine types 2.2 and older due to
2518              * RSDP blob size mismatch.
2519              */
2520             build_append_int_noprefix(tables->rsdp, 0, 16);
2521         }
2522     }
2523 
2524     /* We'll expose it all to Guest so we want to reduce
2525      * chance of size changes.
2526      *
2527      * We used to align the tables to 4k, but of course this would
2528      * too simple to be enough.  4k turned out to be too small an
2529      * alignment very soon, and in fact it is almost impossible to
2530      * keep the table size stable for all (max_cpus, max_memory_slots)
2531      * combinations.  So the table size is always 64k for pc-i440fx-2.1
2532      * and we give an error if the table grows beyond that limit.
2533      *
2534      * We still have the problem of migrating from "-M pc-i440fx-2.0".  For
2535      * that, we exploit the fact that QEMU 2.1 generates _smaller_ tables
2536      * than 2.0 and we can always pad the smaller tables with zeros.  We can
2537      * then use the exact size of the 2.0 tables.
2538      *
2539      * All this is for PIIX4, since QEMU 2.0 didn't support Q35 migration.
2540      */
2541     if (pcmc->legacy_acpi_table_size) {
2542         /* Subtracting aml_len gives the size of fixed tables.  Then add the
2543          * size of the PIIX4 DSDT/SSDT in QEMU 2.0.
2544          */
2545         int legacy_aml_len =
2546             pcmc->legacy_acpi_table_size +
2547             ACPI_BUILD_LEGACY_CPU_AML_SIZE * x86ms->apic_id_limit;
2548         int legacy_table_size =
2549             ROUND_UP(tables_blob->len - aml_len + legacy_aml_len,
2550                      ACPI_BUILD_ALIGN_SIZE);
2551         if (tables_blob->len > legacy_table_size) {
2552             /* Should happen only with PCI bridges and -M pc-i440fx-2.0.  */
2553             warn_report("ACPI table size %u exceeds %d bytes,"
2554                         " migration may not work",
2555                         tables_blob->len, legacy_table_size);
2556             error_printf("Try removing CPUs, NUMA nodes, memory slots"
2557                          " or PCI bridges.");
2558         }
2559         g_array_set_size(tables_blob, legacy_table_size);
2560     } else {
2561         /* Make sure we have a buffer in case we need to resize the tables. */
2562         if (tables_blob->len > ACPI_BUILD_TABLE_SIZE / 2) {
2563             /* As of QEMU 2.1, this fires with 160 VCPUs and 255 memory slots.  */
2564             warn_report("ACPI table size %u exceeds %d bytes,"
2565                         " migration may not work",
2566                         tables_blob->len, ACPI_BUILD_TABLE_SIZE / 2);
2567             error_printf("Try removing CPUs, NUMA nodes, memory slots"
2568                          " or PCI bridges.");
2569         }
2570         acpi_align_size(tables_blob, ACPI_BUILD_TABLE_SIZE);
2571     }
2572 
2573     acpi_align_size(tables->linker->cmd_blob, ACPI_BUILD_ALIGN_SIZE);
2574 
2575     /* Cleanup memory that's no longer used. */
2576     g_array_free(table_offsets, true);
2577 }
2578 
2579 static void acpi_ram_update(MemoryRegion *mr, GArray *data)
2580 {
2581     uint32_t size = acpi_data_len(data);
2582 
2583     /* Make sure RAM size is correct - in case it got changed e.g. by migration */
2584     memory_region_ram_resize(mr, size, &error_abort);
2585 
2586     memcpy(memory_region_get_ram_ptr(mr), data->data, size);
2587     memory_region_set_dirty(mr, 0, size);
2588 }
2589 
2590 static void acpi_build_update(void *build_opaque)
2591 {
2592     AcpiBuildState *build_state = build_opaque;
2593     AcpiBuildTables tables;
2594 
2595     /* No state to update or already patched? Nothing to do. */
2596     if (!build_state || build_state->patched) {
2597         return;
2598     }
2599     build_state->patched = 1;
2600 
2601     acpi_build_tables_init(&tables);
2602 
2603     acpi_build(&tables, MACHINE(qdev_get_machine()));
2604 
2605     acpi_ram_update(build_state->table_mr, tables.table_data);
2606 
2607     if (build_state->rsdp) {
2608         memcpy(build_state->rsdp, tables.rsdp->data, acpi_data_len(tables.rsdp));
2609     } else {
2610         acpi_ram_update(build_state->rsdp_mr, tables.rsdp);
2611     }
2612 
2613     acpi_ram_update(build_state->linker_mr, tables.linker->cmd_blob);
2614     acpi_build_tables_cleanup(&tables, true);
2615 }
2616 
2617 static void acpi_build_reset(void *build_opaque)
2618 {
2619     AcpiBuildState *build_state = build_opaque;
2620     build_state->patched = 0;
2621 }
2622 
2623 static const VMStateDescription vmstate_acpi_build = {
2624     .name = "acpi_build",
2625     .version_id = 1,
2626     .minimum_version_id = 1,
2627     .fields = (VMStateField[]) {
2628         VMSTATE_UINT8(patched, AcpiBuildState),
2629         VMSTATE_END_OF_LIST()
2630     },
2631 };
2632 
2633 void acpi_setup(void)
2634 {
2635     PCMachineState *pcms = PC_MACHINE(qdev_get_machine());
2636     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
2637     X86MachineState *x86ms = X86_MACHINE(pcms);
2638     AcpiBuildTables tables;
2639     AcpiBuildState *build_state;
2640     Object *vmgenid_dev;
2641     TPMIf *tpm;
2642     static FwCfgTPMConfig tpm_config;
2643 
2644     if (!x86ms->fw_cfg) {
2645         ACPI_BUILD_DPRINTF("No fw cfg. Bailing out.\n");
2646         return;
2647     }
2648 
2649     if (!pcms->acpi_build_enabled) {
2650         ACPI_BUILD_DPRINTF("ACPI build disabled. Bailing out.\n");
2651         return;
2652     }
2653 
2654     if (!x86_machine_is_acpi_enabled(X86_MACHINE(pcms))) {
2655         ACPI_BUILD_DPRINTF("ACPI disabled. Bailing out.\n");
2656         return;
2657     }
2658 
2659     build_state = g_malloc0(sizeof *build_state);
2660 
2661     acpi_build_tables_init(&tables);
2662     acpi_build(&tables, MACHINE(pcms));
2663 
2664     /* Now expose it all to Guest */
2665     build_state->table_mr = acpi_add_rom_blob(acpi_build_update,
2666                                               build_state, tables.table_data,
2667                                               ACPI_BUILD_TABLE_FILE,
2668                                               ACPI_BUILD_TABLE_MAX_SIZE);
2669     assert(build_state->table_mr != NULL);
2670 
2671     build_state->linker_mr =
2672         acpi_add_rom_blob(acpi_build_update, build_state,
2673                           tables.linker->cmd_blob, ACPI_BUILD_LOADER_FILE, 0);
2674 
2675     fw_cfg_add_file(x86ms->fw_cfg, ACPI_BUILD_TPMLOG_FILE,
2676                     tables.tcpalog->data, acpi_data_len(tables.tcpalog));
2677 
2678     tpm = tpm_find();
2679     if (tpm && object_property_get_bool(OBJECT(tpm), "ppi", &error_abort)) {
2680         tpm_config = (FwCfgTPMConfig) {
2681             .tpmppi_address = cpu_to_le32(TPM_PPI_ADDR_BASE),
2682             .tpm_version = tpm_get_version(tpm),
2683             .tpmppi_version = TPM_PPI_VERSION_1_30
2684         };
2685         fw_cfg_add_file(x86ms->fw_cfg, "etc/tpm/config",
2686                         &tpm_config, sizeof tpm_config);
2687     }
2688 
2689     vmgenid_dev = find_vmgenid_dev();
2690     if (vmgenid_dev) {
2691         vmgenid_add_fw_cfg(VMGENID(vmgenid_dev), x86ms->fw_cfg,
2692                            tables.vmgenid);
2693     }
2694 
2695     if (!pcmc->rsdp_in_ram) {
2696         /*
2697          * Keep for compatibility with old machine types.
2698          * Though RSDP is small, its contents isn't immutable, so
2699          * we'll update it along with the rest of tables on guest access.
2700          */
2701         uint32_t rsdp_size = acpi_data_len(tables.rsdp);
2702 
2703         build_state->rsdp = g_memdup(tables.rsdp->data, rsdp_size);
2704         fw_cfg_add_file_callback(x86ms->fw_cfg, ACPI_BUILD_RSDP_FILE,
2705                                  acpi_build_update, NULL, build_state,
2706                                  build_state->rsdp, rsdp_size, true);
2707         build_state->rsdp_mr = NULL;
2708     } else {
2709         build_state->rsdp = NULL;
2710         build_state->rsdp_mr = acpi_add_rom_blob(acpi_build_update,
2711                                                  build_state, tables.rsdp,
2712                                                  ACPI_BUILD_RSDP_FILE, 0);
2713     }
2714 
2715     qemu_register_reset(acpi_build_reset, build_state);
2716     acpi_build_reset(build_state);
2717     vmstate_register(NULL, 0, &vmstate_acpi_build, build_state);
2718 
2719     /* Cleanup tables but don't free the memory: we track it
2720      * in build_state.
2721      */
2722     acpi_build_tables_cleanup(&tables, false);
2723 }
2724