xref: /openbmc/qemu/hw/i386/acpi-build.c (revision 93dd625f)
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->irq0));
971     /* FIXME: newer HyperV gets by with only one IRQ */
972     aml_append(crs, aml_irq_no_flags(vmbus_bridge->irq1));
973     aml_append(dev, aml_name_decl("_CRS", crs));
974 
975     return dev;
976 }
977 
978 static void build_isa_devices_aml(Aml *table)
979 {
980     VMBusBridge *vmbus_bridge = vmbus_bridge_find();
981     bool ambiguous;
982     Object *obj = object_resolve_path_type("", TYPE_ISA_BUS, &ambiguous);
983     Aml *scope;
984 
985     assert(obj && !ambiguous);
986 
987     scope = aml_scope("_SB.PCI0.ISA");
988     build_acpi_ipmi_devices(scope, BUS(obj), "\\_SB.PCI0.ISA");
989     isa_build_aml(ISA_BUS(obj), scope);
990 
991     if (vmbus_bridge) {
992         aml_append(scope, build_vmbus_device_aml(vmbus_bridge));
993     }
994 
995     aml_append(table, scope);
996 }
997 
998 static void build_dbg_aml(Aml *table)
999 {
1000     Aml *field;
1001     Aml *method;
1002     Aml *while_ctx;
1003     Aml *scope = aml_scope("\\");
1004     Aml *buf = aml_local(0);
1005     Aml *len = aml_local(1);
1006     Aml *idx = aml_local(2);
1007 
1008     aml_append(scope,
1009        aml_operation_region("DBG", AML_SYSTEM_IO, aml_int(0x0402), 0x01));
1010     field = aml_field("DBG", AML_BYTE_ACC, AML_NOLOCK, AML_PRESERVE);
1011     aml_append(field, aml_named_field("DBGB", 8));
1012     aml_append(scope, field);
1013 
1014     method = aml_method("DBUG", 1, AML_NOTSERIALIZED);
1015 
1016     aml_append(method, aml_to_hexstring(aml_arg(0), buf));
1017     aml_append(method, aml_to_buffer(buf, buf));
1018     aml_append(method, aml_subtract(aml_sizeof(buf), aml_int(1), len));
1019     aml_append(method, aml_store(aml_int(0), idx));
1020 
1021     while_ctx = aml_while(aml_lless(idx, len));
1022     aml_append(while_ctx,
1023         aml_store(aml_derefof(aml_index(buf, idx)), aml_name("DBGB")));
1024     aml_append(while_ctx, aml_increment(idx));
1025     aml_append(method, while_ctx);
1026 
1027     aml_append(method, aml_store(aml_int(0x0A), aml_name("DBGB")));
1028     aml_append(scope, method);
1029 
1030     aml_append(table, scope);
1031 }
1032 
1033 static Aml *build_link_dev(const char *name, uint8_t uid, Aml *reg)
1034 {
1035     Aml *dev;
1036     Aml *crs;
1037     Aml *method;
1038     uint32_t irqs[] = {5, 10, 11};
1039 
1040     dev = aml_device("%s", name);
1041     aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0C0F")));
1042     aml_append(dev, aml_name_decl("_UID", aml_int(uid)));
1043 
1044     crs = aml_resource_template();
1045     aml_append(crs, aml_interrupt(AML_CONSUMER, AML_LEVEL, AML_ACTIVE_HIGH,
1046                                   AML_SHARED, irqs, ARRAY_SIZE(irqs)));
1047     aml_append(dev, aml_name_decl("_PRS", crs));
1048 
1049     method = aml_method("_STA", 0, AML_NOTSERIALIZED);
1050     aml_append(method, aml_return(aml_call1("IQST", reg)));
1051     aml_append(dev, method);
1052 
1053     method = aml_method("_DIS", 0, AML_NOTSERIALIZED);
1054     aml_append(method, aml_or(reg, aml_int(0x80), reg));
1055     aml_append(dev, method);
1056 
1057     method = aml_method("_CRS", 0, AML_NOTSERIALIZED);
1058     aml_append(method, aml_return(aml_call1("IQCR", reg)));
1059     aml_append(dev, method);
1060 
1061     method = aml_method("_SRS", 1, AML_NOTSERIALIZED);
1062     aml_append(method, aml_create_dword_field(aml_arg(0), aml_int(5), "PRRI"));
1063     aml_append(method, aml_store(aml_name("PRRI"), reg));
1064     aml_append(dev, method);
1065 
1066     return dev;
1067  }
1068 
1069 static Aml *build_gsi_link_dev(const char *name, uint8_t uid, uint8_t gsi)
1070 {
1071     Aml *dev;
1072     Aml *crs;
1073     Aml *method;
1074     uint32_t irqs;
1075 
1076     dev = aml_device("%s", name);
1077     aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0C0F")));
1078     aml_append(dev, aml_name_decl("_UID", aml_int(uid)));
1079 
1080     crs = aml_resource_template();
1081     irqs = gsi;
1082     aml_append(crs, aml_interrupt(AML_CONSUMER, AML_LEVEL, AML_ACTIVE_HIGH,
1083                                   AML_SHARED, &irqs, 1));
1084     aml_append(dev, aml_name_decl("_PRS", crs));
1085 
1086     aml_append(dev, aml_name_decl("_CRS", crs));
1087 
1088     /*
1089      * _DIS can be no-op because the interrupt cannot be disabled.
1090      */
1091     method = aml_method("_DIS", 0, AML_NOTSERIALIZED);
1092     aml_append(dev, method);
1093 
1094     method = aml_method("_SRS", 1, AML_NOTSERIALIZED);
1095     aml_append(dev, method);
1096 
1097     return dev;
1098 }
1099 
1100 /* _CRS method - get current settings */
1101 static Aml *build_iqcr_method(bool is_piix4)
1102 {
1103     Aml *if_ctx;
1104     uint32_t irqs;
1105     Aml *method = aml_method("IQCR", 1, AML_SERIALIZED);
1106     Aml *crs = aml_resource_template();
1107 
1108     irqs = 0;
1109     aml_append(crs, aml_interrupt(AML_CONSUMER, AML_LEVEL,
1110                                   AML_ACTIVE_HIGH, AML_SHARED, &irqs, 1));
1111     aml_append(method, aml_name_decl("PRR0", crs));
1112 
1113     aml_append(method,
1114         aml_create_dword_field(aml_name("PRR0"), aml_int(5), "PRRI"));
1115 
1116     if (is_piix4) {
1117         if_ctx = aml_if(aml_lless(aml_arg(0), aml_int(0x80)));
1118         aml_append(if_ctx, aml_store(aml_arg(0), aml_name("PRRI")));
1119         aml_append(method, if_ctx);
1120     } else {
1121         aml_append(method,
1122             aml_store(aml_and(aml_arg(0), aml_int(0xF), NULL),
1123                       aml_name("PRRI")));
1124     }
1125 
1126     aml_append(method, aml_return(aml_name("PRR0")));
1127     return method;
1128 }
1129 
1130 /* _STA method - get status */
1131 static Aml *build_irq_status_method(void)
1132 {
1133     Aml *if_ctx;
1134     Aml *method = aml_method("IQST", 1, AML_NOTSERIALIZED);
1135 
1136     if_ctx = aml_if(aml_and(aml_int(0x80), aml_arg(0), NULL));
1137     aml_append(if_ctx, aml_return(aml_int(0x09)));
1138     aml_append(method, if_ctx);
1139     aml_append(method, aml_return(aml_int(0x0B)));
1140     return method;
1141 }
1142 
1143 static void build_piix4_pci0_int(Aml *table)
1144 {
1145     Aml *dev;
1146     Aml *crs;
1147     Aml *field;
1148     Aml *method;
1149     uint32_t irqs;
1150     Aml *sb_scope = aml_scope("_SB");
1151     Aml *pci0_scope = aml_scope("PCI0");
1152 
1153     aml_append(pci0_scope, build_prt(true));
1154     aml_append(sb_scope, pci0_scope);
1155 
1156     field = aml_field("PCI0.ISA.P40C", AML_BYTE_ACC, AML_NOLOCK, AML_PRESERVE);
1157     aml_append(field, aml_named_field("PRQ0", 8));
1158     aml_append(field, aml_named_field("PRQ1", 8));
1159     aml_append(field, aml_named_field("PRQ2", 8));
1160     aml_append(field, aml_named_field("PRQ3", 8));
1161     aml_append(sb_scope, field);
1162 
1163     aml_append(sb_scope, build_irq_status_method());
1164     aml_append(sb_scope, build_iqcr_method(true));
1165 
1166     aml_append(sb_scope, build_link_dev("LNKA", 0, aml_name("PRQ0")));
1167     aml_append(sb_scope, build_link_dev("LNKB", 1, aml_name("PRQ1")));
1168     aml_append(sb_scope, build_link_dev("LNKC", 2, aml_name("PRQ2")));
1169     aml_append(sb_scope, build_link_dev("LNKD", 3, aml_name("PRQ3")));
1170 
1171     dev = aml_device("LNKS");
1172     {
1173         aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0C0F")));
1174         aml_append(dev, aml_name_decl("_UID", aml_int(4)));
1175 
1176         crs = aml_resource_template();
1177         irqs = 9;
1178         aml_append(crs, aml_interrupt(AML_CONSUMER, AML_LEVEL,
1179                                       AML_ACTIVE_HIGH, AML_SHARED,
1180                                       &irqs, 1));
1181         aml_append(dev, aml_name_decl("_PRS", crs));
1182 
1183         /* The SCI cannot be disabled and is always attached to GSI 9,
1184          * so these are no-ops.  We only need this link to override the
1185          * polarity to active high and match the content of the MADT.
1186          */
1187         method = aml_method("_STA", 0, AML_NOTSERIALIZED);
1188         aml_append(method, aml_return(aml_int(0x0b)));
1189         aml_append(dev, method);
1190 
1191         method = aml_method("_DIS", 0, AML_NOTSERIALIZED);
1192         aml_append(dev, method);
1193 
1194         method = aml_method("_CRS", 0, AML_NOTSERIALIZED);
1195         aml_append(method, aml_return(aml_name("_PRS")));
1196         aml_append(dev, method);
1197 
1198         method = aml_method("_SRS", 1, AML_NOTSERIALIZED);
1199         aml_append(dev, method);
1200     }
1201     aml_append(sb_scope, dev);
1202 
1203     aml_append(table, sb_scope);
1204 }
1205 
1206 static void append_q35_prt_entry(Aml *ctx, uint32_t nr, const char *name)
1207 {
1208     int i;
1209     int head;
1210     Aml *pkg;
1211     char base = name[3] < 'E' ? 'A' : 'E';
1212     char *s = g_strdup(name);
1213     Aml *a_nr = aml_int((nr << 16) | 0xffff);
1214 
1215     assert(strlen(s) == 4);
1216 
1217     head = name[3] - base;
1218     for (i = 0; i < 4; i++) {
1219         if (head + i > 3) {
1220             head = i * -1;
1221         }
1222         s[3] = base + head + i;
1223         pkg = aml_package(4);
1224         aml_append(pkg, a_nr);
1225         aml_append(pkg, aml_int(i));
1226         aml_append(pkg, aml_name("%s", s));
1227         aml_append(pkg, aml_int(0));
1228         aml_append(ctx, pkg);
1229     }
1230     g_free(s);
1231 }
1232 
1233 static Aml *build_q35_routing_table(const char *str)
1234 {
1235     int i;
1236     Aml *pkg;
1237     char *name = g_strdup_printf("%s ", str);
1238 
1239     pkg = aml_package(128);
1240     for (i = 0; i < 0x18; i++) {
1241             name[3] = 'E' + (i & 0x3);
1242             append_q35_prt_entry(pkg, i, name);
1243     }
1244 
1245     name[3] = 'E';
1246     append_q35_prt_entry(pkg, 0x18, name);
1247 
1248     /* INTA -> PIRQA for slot 25 - 31, see the default value of D<N>IR */
1249     for (i = 0x0019; i < 0x1e; i++) {
1250         name[3] = 'A';
1251         append_q35_prt_entry(pkg, i, name);
1252     }
1253 
1254     /* PCIe->PCI bridge. use PIRQ[E-H] */
1255     name[3] = 'E';
1256     append_q35_prt_entry(pkg, 0x1e, name);
1257     name[3] = 'A';
1258     append_q35_prt_entry(pkg, 0x1f, name);
1259 
1260     g_free(name);
1261     return pkg;
1262 }
1263 
1264 static void build_q35_pci0_int(Aml *table)
1265 {
1266     Aml *field;
1267     Aml *method;
1268     Aml *sb_scope = aml_scope("_SB");
1269     Aml *pci0_scope = aml_scope("PCI0");
1270 
1271     /* Zero => PIC mode, One => APIC Mode */
1272     aml_append(table, aml_name_decl("PICF", aml_int(0)));
1273     method = aml_method("_PIC", 1, AML_NOTSERIALIZED);
1274     {
1275         aml_append(method, aml_store(aml_arg(0), aml_name("PICF")));
1276     }
1277     aml_append(table, method);
1278 
1279     aml_append(pci0_scope,
1280         aml_name_decl("PRTP", build_q35_routing_table("LNK")));
1281     aml_append(pci0_scope,
1282         aml_name_decl("PRTA", build_q35_routing_table("GSI")));
1283 
1284     method = aml_method("_PRT", 0, AML_NOTSERIALIZED);
1285     {
1286         Aml *if_ctx;
1287         Aml *else_ctx;
1288 
1289         /* PCI IRQ routing table, example from ACPI 2.0a specification,
1290            section 6.2.8.1 */
1291         /* Note: we provide the same info as the PCI routing
1292            table of the Bochs BIOS */
1293         if_ctx = aml_if(aml_equal(aml_name("PICF"), aml_int(0)));
1294         aml_append(if_ctx, aml_return(aml_name("PRTP")));
1295         aml_append(method, if_ctx);
1296         else_ctx = aml_else();
1297         aml_append(else_ctx, aml_return(aml_name("PRTA")));
1298         aml_append(method, else_ctx);
1299     }
1300     aml_append(pci0_scope, method);
1301     aml_append(sb_scope, pci0_scope);
1302 
1303     field = aml_field("PCI0.ISA.PIRQ", AML_BYTE_ACC, AML_NOLOCK, AML_PRESERVE);
1304     aml_append(field, aml_named_field("PRQA", 8));
1305     aml_append(field, aml_named_field("PRQB", 8));
1306     aml_append(field, aml_named_field("PRQC", 8));
1307     aml_append(field, aml_named_field("PRQD", 8));
1308     aml_append(field, aml_reserved_field(0x20));
1309     aml_append(field, aml_named_field("PRQE", 8));
1310     aml_append(field, aml_named_field("PRQF", 8));
1311     aml_append(field, aml_named_field("PRQG", 8));
1312     aml_append(field, aml_named_field("PRQH", 8));
1313     aml_append(sb_scope, field);
1314 
1315     aml_append(sb_scope, build_irq_status_method());
1316     aml_append(sb_scope, build_iqcr_method(false));
1317 
1318     aml_append(sb_scope, build_link_dev("LNKA", 0, aml_name("PRQA")));
1319     aml_append(sb_scope, build_link_dev("LNKB", 1, aml_name("PRQB")));
1320     aml_append(sb_scope, build_link_dev("LNKC", 2, aml_name("PRQC")));
1321     aml_append(sb_scope, build_link_dev("LNKD", 3, aml_name("PRQD")));
1322     aml_append(sb_scope, build_link_dev("LNKE", 4, aml_name("PRQE")));
1323     aml_append(sb_scope, build_link_dev("LNKF", 5, aml_name("PRQF")));
1324     aml_append(sb_scope, build_link_dev("LNKG", 6, aml_name("PRQG")));
1325     aml_append(sb_scope, build_link_dev("LNKH", 7, aml_name("PRQH")));
1326 
1327     aml_append(sb_scope, build_gsi_link_dev("GSIA", 0x10, 0x10));
1328     aml_append(sb_scope, build_gsi_link_dev("GSIB", 0x11, 0x11));
1329     aml_append(sb_scope, build_gsi_link_dev("GSIC", 0x12, 0x12));
1330     aml_append(sb_scope, build_gsi_link_dev("GSID", 0x13, 0x13));
1331     aml_append(sb_scope, build_gsi_link_dev("GSIE", 0x14, 0x14));
1332     aml_append(sb_scope, build_gsi_link_dev("GSIF", 0x15, 0x15));
1333     aml_append(sb_scope, build_gsi_link_dev("GSIG", 0x16, 0x16));
1334     aml_append(sb_scope, build_gsi_link_dev("GSIH", 0x17, 0x17));
1335 
1336     aml_append(table, sb_scope);
1337 }
1338 
1339 static void build_q35_isa_bridge(Aml *table)
1340 {
1341     Aml *dev;
1342     Aml *scope;
1343 
1344     scope =  aml_scope("_SB.PCI0");
1345     dev = aml_device("ISA");
1346     aml_append(dev, aml_name_decl("_ADR", aml_int(0x001F0000)));
1347 
1348     /* ICH9 PCI to ISA irq remapping */
1349     aml_append(dev, aml_operation_region("PIRQ", AML_PCI_CONFIG,
1350                                          aml_int(0x60), 0x0C));
1351 
1352     aml_append(scope, dev);
1353     aml_append(table, scope);
1354 }
1355 
1356 static void build_piix4_isa_bridge(Aml *table)
1357 {
1358     Aml *dev;
1359     Aml *scope;
1360 
1361     scope =  aml_scope("_SB.PCI0");
1362     dev = aml_device("ISA");
1363     aml_append(dev, aml_name_decl("_ADR", aml_int(0x00010000)));
1364 
1365     /* PIIX PCI to ISA irq remapping */
1366     aml_append(dev, aml_operation_region("P40C", AML_PCI_CONFIG,
1367                                          aml_int(0x60), 0x04));
1368 
1369     aml_append(scope, dev);
1370     aml_append(table, scope);
1371 }
1372 
1373 static void build_piix4_pci_hotplug(Aml *table)
1374 {
1375     Aml *scope;
1376     Aml *field;
1377     Aml *method;
1378 
1379     scope =  aml_scope("_SB.PCI0");
1380 
1381     aml_append(scope,
1382         aml_operation_region("PCST", AML_SYSTEM_IO, aml_int(0xae00), 0x08));
1383     field = aml_field("PCST", AML_DWORD_ACC, AML_NOLOCK, AML_WRITE_AS_ZEROS);
1384     aml_append(field, aml_named_field("PCIU", 32));
1385     aml_append(field, aml_named_field("PCID", 32));
1386     aml_append(scope, field);
1387 
1388     aml_append(scope,
1389         aml_operation_region("SEJ", AML_SYSTEM_IO, aml_int(0xae08), 0x04));
1390     field = aml_field("SEJ", AML_DWORD_ACC, AML_NOLOCK, AML_WRITE_AS_ZEROS);
1391     aml_append(field, aml_named_field("B0EJ", 32));
1392     aml_append(scope, field);
1393 
1394     aml_append(scope,
1395         aml_operation_region("BNMR", AML_SYSTEM_IO, aml_int(0xae10), 0x04));
1396     field = aml_field("BNMR", AML_DWORD_ACC, AML_NOLOCK, AML_WRITE_AS_ZEROS);
1397     aml_append(field, aml_named_field("BNUM", 32));
1398     aml_append(scope, field);
1399 
1400     aml_append(scope, aml_mutex("BLCK", 0));
1401 
1402     method = aml_method("PCEJ", 2, AML_NOTSERIALIZED);
1403     aml_append(method, aml_acquire(aml_name("BLCK"), 0xFFFF));
1404     aml_append(method, aml_store(aml_arg(0), aml_name("BNUM")));
1405     aml_append(method,
1406         aml_store(aml_shiftleft(aml_int(1), aml_arg(1)), aml_name("B0EJ")));
1407     aml_append(method, aml_release(aml_name("BLCK")));
1408     aml_append(method, aml_return(aml_int(0)));
1409     aml_append(scope, method);
1410 
1411     aml_append(table, scope);
1412 }
1413 
1414 static Aml *build_q35_osc_method(void)
1415 {
1416     Aml *if_ctx;
1417     Aml *if_ctx2;
1418     Aml *else_ctx;
1419     Aml *method;
1420     Aml *a_cwd1 = aml_name("CDW1");
1421     Aml *a_ctrl = aml_local(0);
1422 
1423     method = aml_method("_OSC", 4, AML_NOTSERIALIZED);
1424     aml_append(method, aml_create_dword_field(aml_arg(3), aml_int(0), "CDW1"));
1425 
1426     if_ctx = aml_if(aml_equal(
1427         aml_arg(0), aml_touuid("33DB4D5B-1FF7-401C-9657-7441C03DD766")));
1428     aml_append(if_ctx, aml_create_dword_field(aml_arg(3), aml_int(4), "CDW2"));
1429     aml_append(if_ctx, aml_create_dword_field(aml_arg(3), aml_int(8), "CDW3"));
1430 
1431     aml_append(if_ctx, aml_store(aml_name("CDW3"), a_ctrl));
1432 
1433     /*
1434      * Always allow native PME, AER (no dependencies)
1435      * Allow SHPC (PCI bridges can have SHPC controller)
1436      */
1437     aml_append(if_ctx, aml_and(a_ctrl, aml_int(0x1F), a_ctrl));
1438 
1439     if_ctx2 = aml_if(aml_lnot(aml_equal(aml_arg(1), aml_int(1))));
1440     /* Unknown revision */
1441     aml_append(if_ctx2, aml_or(a_cwd1, aml_int(0x08), a_cwd1));
1442     aml_append(if_ctx, if_ctx2);
1443 
1444     if_ctx2 = aml_if(aml_lnot(aml_equal(aml_name("CDW3"), a_ctrl)));
1445     /* Capabilities bits were masked */
1446     aml_append(if_ctx2, aml_or(a_cwd1, aml_int(0x10), a_cwd1));
1447     aml_append(if_ctx, if_ctx2);
1448 
1449     /* Update DWORD3 in the buffer */
1450     aml_append(if_ctx, aml_store(a_ctrl, aml_name("CDW3")));
1451     aml_append(method, if_ctx);
1452 
1453     else_ctx = aml_else();
1454     /* Unrecognized UUID */
1455     aml_append(else_ctx, aml_or(a_cwd1, aml_int(4), a_cwd1));
1456     aml_append(method, else_ctx);
1457 
1458     aml_append(method, aml_return(aml_arg(3)));
1459     return method;
1460 }
1461 
1462 static void build_smb0(Aml *table, I2CBus *smbus, int devnr, int func)
1463 {
1464     Aml *scope = aml_scope("_SB.PCI0");
1465     Aml *dev = aml_device("SMB0");
1466 
1467     aml_append(dev, aml_name_decl("_ADR", aml_int(devnr << 16 | func)));
1468     build_acpi_ipmi_devices(dev, BUS(smbus), "\\_SB.PCI0.SMB0");
1469     aml_append(scope, dev);
1470     aml_append(table, scope);
1471 }
1472 
1473 static void
1474 build_dsdt(GArray *table_data, BIOSLinker *linker,
1475            AcpiPmInfo *pm, AcpiMiscInfo *misc,
1476            Range *pci_hole, Range *pci_hole64, MachineState *machine)
1477 {
1478     CrsRangeEntry *entry;
1479     Aml *dsdt, *sb_scope, *scope, *dev, *method, *field, *pkg, *crs;
1480     CrsRangeSet crs_range_set;
1481     PCMachineState *pcms = PC_MACHINE(machine);
1482     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(machine);
1483     X86MachineState *x86ms = X86_MACHINE(machine);
1484     AcpiMcfgInfo mcfg;
1485     uint32_t nr_mem = machine->ram_slots;
1486     int root_bus_limit = 0xFF;
1487     PCIBus *bus = NULL;
1488     TPMIf *tpm = tpm_find();
1489     int i;
1490 
1491     dsdt = init_aml_allocator();
1492 
1493     /* Reserve space for header */
1494     acpi_data_push(dsdt->buf, sizeof(AcpiTableHeader));
1495 
1496     build_dbg_aml(dsdt);
1497     if (misc->is_piix4) {
1498         sb_scope = aml_scope("_SB");
1499         dev = aml_device("PCI0");
1500         aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0A03")));
1501         aml_append(dev, aml_name_decl("_ADR", aml_int(0)));
1502         aml_append(dev, aml_name_decl("_UID", aml_int(1)));
1503         aml_append(sb_scope, dev);
1504         aml_append(dsdt, sb_scope);
1505 
1506         build_hpet_aml(dsdt);
1507         build_piix4_isa_bridge(dsdt);
1508         build_isa_devices_aml(dsdt);
1509         build_piix4_pci_hotplug(dsdt);
1510         build_piix4_pci0_int(dsdt);
1511     } else {
1512         sb_scope = aml_scope("_SB");
1513         dev = aml_device("PCI0");
1514         aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0A08")));
1515         aml_append(dev, aml_name_decl("_CID", aml_eisaid("PNP0A03")));
1516         aml_append(dev, aml_name_decl("_ADR", aml_int(0)));
1517         aml_append(dev, aml_name_decl("_UID", aml_int(1)));
1518         aml_append(dev, build_q35_osc_method());
1519         aml_append(sb_scope, dev);
1520         aml_append(dsdt, sb_scope);
1521 
1522         build_hpet_aml(dsdt);
1523         build_q35_isa_bridge(dsdt);
1524         build_isa_devices_aml(dsdt);
1525         build_q35_pci0_int(dsdt);
1526         if (pcms->smbus && !pcmc->do_not_add_smb_acpi) {
1527             build_smb0(dsdt, pcms->smbus, ICH9_SMB_DEV, ICH9_SMB_FUNC);
1528         }
1529     }
1530 
1531     if (pcmc->legacy_cpu_hotplug) {
1532         build_legacy_cpu_hotplug_aml(dsdt, machine, pm->cpu_hp_io_base);
1533     } else {
1534         CPUHotplugFeatures opts = {
1535             .acpi_1_compatible = true, .has_legacy_cphp = true
1536         };
1537         build_cpus_aml(dsdt, machine, opts, pm->cpu_hp_io_base,
1538                        "\\_SB.PCI0", "\\_GPE._E02");
1539     }
1540 
1541     if (pcms->memhp_io_base && nr_mem) {
1542         build_memory_hotplug_aml(dsdt, nr_mem, "\\_SB.PCI0",
1543                                  "\\_GPE._E03", AML_SYSTEM_IO,
1544                                  pcms->memhp_io_base);
1545     }
1546 
1547     scope =  aml_scope("_GPE");
1548     {
1549         aml_append(scope, aml_name_decl("_HID", aml_string("ACPI0006")));
1550 
1551         if (misc->is_piix4) {
1552             method = aml_method("_E01", 0, AML_NOTSERIALIZED);
1553             aml_append(method,
1554                 aml_acquire(aml_name("\\_SB.PCI0.BLCK"), 0xFFFF));
1555             aml_append(method, aml_call0("\\_SB.PCI0.PCNT"));
1556             aml_append(method, aml_release(aml_name("\\_SB.PCI0.BLCK")));
1557             aml_append(scope, method);
1558         }
1559 
1560         if (machine->nvdimms_state->is_enabled) {
1561             method = aml_method("_E04", 0, AML_NOTSERIALIZED);
1562             aml_append(method, aml_notify(aml_name("\\_SB.NVDR"),
1563                                           aml_int(0x80)));
1564             aml_append(scope, method);
1565         }
1566     }
1567     aml_append(dsdt, scope);
1568 
1569     crs_range_set_init(&crs_range_set);
1570     bus = PC_MACHINE(machine)->bus;
1571     if (bus) {
1572         QLIST_FOREACH(bus, &bus->child, sibling) {
1573             uint8_t bus_num = pci_bus_num(bus);
1574             uint8_t numa_node = pci_bus_numa_node(bus);
1575 
1576             /* look only for expander root buses */
1577             if (!pci_bus_is_root(bus)) {
1578                 continue;
1579             }
1580 
1581             if (bus_num < root_bus_limit) {
1582                 root_bus_limit = bus_num - 1;
1583             }
1584 
1585             scope = aml_scope("\\_SB");
1586             dev = aml_device("PC%.02X", bus_num);
1587             aml_append(dev, aml_name_decl("_UID", aml_int(bus_num)));
1588             aml_append(dev, aml_name_decl("_BBN", aml_int(bus_num)));
1589             if (pci_bus_is_express(bus)) {
1590                 aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0A08")));
1591                 aml_append(dev, aml_name_decl("_CID", aml_eisaid("PNP0A03")));
1592                 aml_append(dev, build_q35_osc_method());
1593             } else {
1594                 aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0A03")));
1595             }
1596 
1597             if (numa_node != NUMA_NODE_UNASSIGNED) {
1598                 aml_append(dev, aml_name_decl("_PXM", aml_int(numa_node)));
1599             }
1600 
1601             aml_append(dev, build_prt(false));
1602             crs = build_crs(PCI_HOST_BRIDGE(BUS(bus)->parent), &crs_range_set);
1603             aml_append(dev, aml_name_decl("_CRS", crs));
1604             aml_append(scope, dev);
1605             aml_append(dsdt, scope);
1606         }
1607     }
1608 
1609     /*
1610      * At this point crs_range_set has all the ranges used by pci
1611      * busses *other* than PCI0.  These ranges will be excluded from
1612      * the PCI0._CRS.  Add mmconfig to the set so it will be excluded
1613      * too.
1614      */
1615     if (acpi_get_mcfg(&mcfg)) {
1616         crs_range_insert(crs_range_set.mem_ranges,
1617                          mcfg.base, mcfg.base + mcfg.size - 1);
1618     }
1619 
1620     scope = aml_scope("\\_SB.PCI0");
1621     /* build PCI0._CRS */
1622     crs = aml_resource_template();
1623     aml_append(crs,
1624         aml_word_bus_number(AML_MIN_FIXED, AML_MAX_FIXED, AML_POS_DECODE,
1625                             0x0000, 0x0, root_bus_limit,
1626                             0x0000, root_bus_limit + 1));
1627     aml_append(crs, aml_io(AML_DECODE16, 0x0CF8, 0x0CF8, 0x01, 0x08));
1628 
1629     aml_append(crs,
1630         aml_word_io(AML_MIN_FIXED, AML_MAX_FIXED,
1631                     AML_POS_DECODE, AML_ENTIRE_RANGE,
1632                     0x0000, 0x0000, 0x0CF7, 0x0000, 0x0CF8));
1633 
1634     crs_replace_with_free_ranges(crs_range_set.io_ranges, 0x0D00, 0xFFFF);
1635     for (i = 0; i < crs_range_set.io_ranges->len; i++) {
1636         entry = g_ptr_array_index(crs_range_set.io_ranges, i);
1637         aml_append(crs,
1638             aml_word_io(AML_MIN_FIXED, AML_MAX_FIXED,
1639                         AML_POS_DECODE, AML_ENTIRE_RANGE,
1640                         0x0000, entry->base, entry->limit,
1641                         0x0000, entry->limit - entry->base + 1));
1642     }
1643 
1644     aml_append(crs,
1645         aml_dword_memory(AML_POS_DECODE, AML_MIN_FIXED, AML_MAX_FIXED,
1646                          AML_CACHEABLE, AML_READ_WRITE,
1647                          0, 0x000A0000, 0x000BFFFF, 0, 0x00020000));
1648 
1649     crs_replace_with_free_ranges(crs_range_set.mem_ranges,
1650                                  range_lob(pci_hole),
1651                                  range_upb(pci_hole));
1652     for (i = 0; i < crs_range_set.mem_ranges->len; i++) {
1653         entry = g_ptr_array_index(crs_range_set.mem_ranges, i);
1654         aml_append(crs,
1655             aml_dword_memory(AML_POS_DECODE, AML_MIN_FIXED, AML_MAX_FIXED,
1656                              AML_NON_CACHEABLE, AML_READ_WRITE,
1657                              0, entry->base, entry->limit,
1658                              0, entry->limit - entry->base + 1));
1659     }
1660 
1661     if (!range_is_empty(pci_hole64)) {
1662         crs_replace_with_free_ranges(crs_range_set.mem_64bit_ranges,
1663                                      range_lob(pci_hole64),
1664                                      range_upb(pci_hole64));
1665         for (i = 0; i < crs_range_set.mem_64bit_ranges->len; i++) {
1666             entry = g_ptr_array_index(crs_range_set.mem_64bit_ranges, i);
1667             aml_append(crs,
1668                        aml_qword_memory(AML_POS_DECODE, AML_MIN_FIXED,
1669                                         AML_MAX_FIXED,
1670                                         AML_CACHEABLE, AML_READ_WRITE,
1671                                         0, entry->base, entry->limit,
1672                                         0, entry->limit - entry->base + 1));
1673         }
1674     }
1675 
1676     if (TPM_IS_TIS_ISA(tpm_find())) {
1677         aml_append(crs, aml_memory32_fixed(TPM_TIS_ADDR_BASE,
1678                    TPM_TIS_ADDR_SIZE, AML_READ_WRITE));
1679     }
1680     aml_append(scope, aml_name_decl("_CRS", crs));
1681 
1682     /* reserve GPE0 block resources */
1683     dev = aml_device("GPE0");
1684     aml_append(dev, aml_name_decl("_HID", aml_string("PNP0A06")));
1685     aml_append(dev, aml_name_decl("_UID", aml_string("GPE0 resources")));
1686     /* device present, functioning, decoding, not shown in UI */
1687     aml_append(dev, aml_name_decl("_STA", aml_int(0xB)));
1688     crs = aml_resource_template();
1689     aml_append(crs,
1690         aml_io(
1691                AML_DECODE16,
1692                pm->fadt.gpe0_blk.address,
1693                pm->fadt.gpe0_blk.address,
1694                1,
1695                pm->fadt.gpe0_blk.bit_width / 8)
1696     );
1697     aml_append(dev, aml_name_decl("_CRS", crs));
1698     aml_append(scope, dev);
1699 
1700     crs_range_set_free(&crs_range_set);
1701 
1702     /* reserve PCIHP resources */
1703     if (pm->pcihp_io_len) {
1704         dev = aml_device("PHPR");
1705         aml_append(dev, aml_name_decl("_HID", aml_string("PNP0A06")));
1706         aml_append(dev,
1707             aml_name_decl("_UID", aml_string("PCI Hotplug resources")));
1708         /* device present, functioning, decoding, not shown in UI */
1709         aml_append(dev, aml_name_decl("_STA", aml_int(0xB)));
1710         crs = aml_resource_template();
1711         aml_append(crs,
1712             aml_io(AML_DECODE16, pm->pcihp_io_base, pm->pcihp_io_base, 1,
1713                    pm->pcihp_io_len)
1714         );
1715         aml_append(dev, aml_name_decl("_CRS", crs));
1716         aml_append(scope, dev);
1717     }
1718     aml_append(dsdt, scope);
1719 
1720     /*  create S3_ / S4_ / S5_ packages if necessary */
1721     scope = aml_scope("\\");
1722     if (!pm->s3_disabled) {
1723         pkg = aml_package(4);
1724         aml_append(pkg, aml_int(1)); /* PM1a_CNT.SLP_TYP */
1725         aml_append(pkg, aml_int(1)); /* PM1b_CNT.SLP_TYP, FIXME: not impl. */
1726         aml_append(pkg, aml_int(0)); /* reserved */
1727         aml_append(pkg, aml_int(0)); /* reserved */
1728         aml_append(scope, aml_name_decl("_S3", pkg));
1729     }
1730 
1731     if (!pm->s4_disabled) {
1732         pkg = aml_package(4);
1733         aml_append(pkg, aml_int(pm->s4_val)); /* PM1a_CNT.SLP_TYP */
1734         /* PM1b_CNT.SLP_TYP, FIXME: not impl. */
1735         aml_append(pkg, aml_int(pm->s4_val));
1736         aml_append(pkg, aml_int(0)); /* reserved */
1737         aml_append(pkg, aml_int(0)); /* reserved */
1738         aml_append(scope, aml_name_decl("_S4", pkg));
1739     }
1740 
1741     pkg = aml_package(4);
1742     aml_append(pkg, aml_int(0)); /* PM1a_CNT.SLP_TYP */
1743     aml_append(pkg, aml_int(0)); /* PM1b_CNT.SLP_TYP not impl. */
1744     aml_append(pkg, aml_int(0)); /* reserved */
1745     aml_append(pkg, aml_int(0)); /* reserved */
1746     aml_append(scope, aml_name_decl("_S5", pkg));
1747     aml_append(dsdt, scope);
1748 
1749     /* create fw_cfg node, unconditionally */
1750     {
1751         scope = aml_scope("\\_SB.PCI0");
1752         fw_cfg_add_acpi_dsdt(scope, x86ms->fw_cfg);
1753         aml_append(dsdt, scope);
1754     }
1755 
1756     if (misc->applesmc_io_base) {
1757         scope = aml_scope("\\_SB.PCI0.ISA");
1758         dev = aml_device("SMC");
1759 
1760         aml_append(dev, aml_name_decl("_HID", aml_eisaid("APP0001")));
1761         /* device present, functioning, decoding, not shown in UI */
1762         aml_append(dev, aml_name_decl("_STA", aml_int(0xB)));
1763 
1764         crs = aml_resource_template();
1765         aml_append(crs,
1766             aml_io(AML_DECODE16, misc->applesmc_io_base, misc->applesmc_io_base,
1767                    0x01, APPLESMC_MAX_DATA_LENGTH)
1768         );
1769         aml_append(crs, aml_irq_no_flags(6));
1770         aml_append(dev, aml_name_decl("_CRS", crs));
1771 
1772         aml_append(scope, dev);
1773         aml_append(dsdt, scope);
1774     }
1775 
1776     if (misc->pvpanic_port) {
1777         scope = aml_scope("\\_SB.PCI0.ISA");
1778 
1779         dev = aml_device("PEVT");
1780         aml_append(dev, aml_name_decl("_HID", aml_string("QEMU0001")));
1781 
1782         crs = aml_resource_template();
1783         aml_append(crs,
1784             aml_io(AML_DECODE16, misc->pvpanic_port, misc->pvpanic_port, 1, 1)
1785         );
1786         aml_append(dev, aml_name_decl("_CRS", crs));
1787 
1788         aml_append(dev, aml_operation_region("PEOR", AML_SYSTEM_IO,
1789                                               aml_int(misc->pvpanic_port), 1));
1790         field = aml_field("PEOR", AML_BYTE_ACC, AML_NOLOCK, AML_PRESERVE);
1791         aml_append(field, aml_named_field("PEPT", 8));
1792         aml_append(dev, field);
1793 
1794         /* device present, functioning, decoding, shown in UI */
1795         aml_append(dev, aml_name_decl("_STA", aml_int(0xF)));
1796 
1797         method = aml_method("RDPT", 0, AML_NOTSERIALIZED);
1798         aml_append(method, aml_store(aml_name("PEPT"), aml_local(0)));
1799         aml_append(method, aml_return(aml_local(0)));
1800         aml_append(dev, method);
1801 
1802         method = aml_method("WRPT", 1, AML_NOTSERIALIZED);
1803         aml_append(method, aml_store(aml_arg(0), aml_name("PEPT")));
1804         aml_append(dev, method);
1805 
1806         aml_append(scope, dev);
1807         aml_append(dsdt, scope);
1808     }
1809 
1810     sb_scope = aml_scope("\\_SB");
1811     {
1812         Object *pci_host;
1813         PCIBus *bus = NULL;
1814 
1815         pci_host = acpi_get_i386_pci_host();
1816         if (pci_host) {
1817             bus = PCI_HOST_BRIDGE(pci_host)->bus;
1818         }
1819 
1820         if (bus) {
1821             Aml *scope = aml_scope("PCI0");
1822             /* Scan all PCI buses. Generate tables to support hotplug. */
1823             build_append_pci_bus_devices(scope, bus, pm->pcihp_bridge_en);
1824 
1825             if (TPM_IS_TIS_ISA(tpm)) {
1826                 if (misc->tpm_version == TPM_VERSION_2_0) {
1827                     dev = aml_device("TPM");
1828                     aml_append(dev, aml_name_decl("_HID",
1829                                                   aml_string("MSFT0101")));
1830                 } else {
1831                     dev = aml_device("ISA.TPM");
1832                     aml_append(dev, aml_name_decl("_HID",
1833                                                   aml_eisaid("PNP0C31")));
1834                 }
1835 
1836                 aml_append(dev, aml_name_decl("_STA", aml_int(0xF)));
1837                 crs = aml_resource_template();
1838                 aml_append(crs, aml_memory32_fixed(TPM_TIS_ADDR_BASE,
1839                            TPM_TIS_ADDR_SIZE, AML_READ_WRITE));
1840                 /*
1841                     FIXME: TPM_TIS_IRQ=5 conflicts with PNP0C0F irqs,
1842                     Rewrite to take IRQ from TPM device model and
1843                     fix default IRQ value there to use some unused IRQ
1844                  */
1845                 /* aml_append(crs, aml_irq_no_flags(TPM_TIS_IRQ)); */
1846                 aml_append(dev, aml_name_decl("_CRS", crs));
1847 
1848                 tpm_build_ppi_acpi(tpm, dev);
1849 
1850                 aml_append(scope, dev);
1851             }
1852 
1853             aml_append(sb_scope, scope);
1854         }
1855     }
1856 
1857     if (TPM_IS_CRB(tpm)) {
1858         dev = aml_device("TPM");
1859         aml_append(dev, aml_name_decl("_HID", aml_string("MSFT0101")));
1860         crs = aml_resource_template();
1861         aml_append(crs, aml_memory32_fixed(TPM_CRB_ADDR_BASE,
1862                                            TPM_CRB_ADDR_SIZE, AML_READ_WRITE));
1863         aml_append(dev, aml_name_decl("_CRS", crs));
1864 
1865         aml_append(dev, aml_name_decl("_STA", aml_int(0xf)));
1866 
1867         tpm_build_ppi_acpi(tpm, dev);
1868 
1869         aml_append(sb_scope, dev);
1870     }
1871 
1872     aml_append(dsdt, sb_scope);
1873 
1874     /* copy AML table into ACPI tables blob and patch header there */
1875     g_array_append_vals(table_data, dsdt->buf->data, dsdt->buf->len);
1876     build_header(linker, table_data,
1877         (void *)(table_data->data + table_data->len - dsdt->buf->len),
1878         "DSDT", dsdt->buf->len, 1, NULL, NULL);
1879     free_aml_allocator();
1880 }
1881 
1882 static void
1883 build_hpet(GArray *table_data, BIOSLinker *linker)
1884 {
1885     Acpi20Hpet *hpet;
1886 
1887     hpet = acpi_data_push(table_data, sizeof(*hpet));
1888     /* Note timer_block_id value must be kept in sync with value advertised by
1889      * emulated hpet
1890      */
1891     hpet->timer_block_id = cpu_to_le32(0x8086a201);
1892     hpet->addr.address = cpu_to_le64(HPET_BASE);
1893     build_header(linker, table_data,
1894                  (void *)hpet, "HPET", sizeof(*hpet), 1, NULL, NULL);
1895 }
1896 
1897 static void
1898 build_tpm_tcpa(GArray *table_data, BIOSLinker *linker, GArray *tcpalog)
1899 {
1900     Acpi20Tcpa *tcpa = acpi_data_push(table_data, sizeof *tcpa);
1901     unsigned log_addr_size = sizeof(tcpa->log_area_start_address);
1902     unsigned log_addr_offset =
1903         (char *)&tcpa->log_area_start_address - table_data->data;
1904 
1905     tcpa->platform_class = cpu_to_le16(TPM_TCPA_ACPI_CLASS_CLIENT);
1906     tcpa->log_area_minimum_length = cpu_to_le32(TPM_LOG_AREA_MINIMUM_SIZE);
1907     acpi_data_push(tcpalog, le32_to_cpu(tcpa->log_area_minimum_length));
1908 
1909     bios_linker_loader_alloc(linker, ACPI_BUILD_TPMLOG_FILE, tcpalog, 1,
1910                              false /* high memory */);
1911 
1912     /* log area start address to be filled by Guest linker */
1913     bios_linker_loader_add_pointer(linker,
1914         ACPI_BUILD_TABLE_FILE, log_addr_offset, log_addr_size,
1915         ACPI_BUILD_TPMLOG_FILE, 0);
1916 
1917     build_header(linker, table_data,
1918                  (void *)tcpa, "TCPA", sizeof(*tcpa), 2, NULL, NULL);
1919 }
1920 
1921 #define HOLE_640K_START  (640 * KiB)
1922 #define HOLE_640K_END   (1 * MiB)
1923 
1924 static void
1925 build_srat(GArray *table_data, BIOSLinker *linker, MachineState *machine)
1926 {
1927     AcpiSystemResourceAffinityTable *srat;
1928     AcpiSratMemoryAffinity *numamem;
1929 
1930     int i;
1931     int srat_start, numa_start, slots;
1932     uint64_t mem_len, mem_base, next_base;
1933     MachineClass *mc = MACHINE_GET_CLASS(machine);
1934     X86MachineState *x86ms = X86_MACHINE(machine);
1935     const CPUArchIdList *apic_ids = mc->possible_cpu_arch_ids(machine);
1936     PCMachineState *pcms = PC_MACHINE(machine);
1937     ram_addr_t hotplugabble_address_space_size =
1938         object_property_get_int(OBJECT(pcms), PC_MACHINE_DEVMEM_REGION_SIZE,
1939                                 NULL);
1940 
1941     srat_start = table_data->len;
1942 
1943     srat = acpi_data_push(table_data, sizeof *srat);
1944     srat->reserved1 = cpu_to_le32(1);
1945 
1946     for (i = 0; i < apic_ids->len; i++) {
1947         int node_id = apic_ids->cpus[i].props.node_id;
1948         uint32_t apic_id = apic_ids->cpus[i].arch_id;
1949 
1950         if (apic_id < 255) {
1951             AcpiSratProcessorAffinity *core;
1952 
1953             core = acpi_data_push(table_data, sizeof *core);
1954             core->type = ACPI_SRAT_PROCESSOR_APIC;
1955             core->length = sizeof(*core);
1956             core->local_apic_id = apic_id;
1957             core->proximity_lo = node_id;
1958             memset(core->proximity_hi, 0, 3);
1959             core->local_sapic_eid = 0;
1960             core->flags = cpu_to_le32(1);
1961         } else {
1962             AcpiSratProcessorX2ApicAffinity *core;
1963 
1964             core = acpi_data_push(table_data, sizeof *core);
1965             core->type = ACPI_SRAT_PROCESSOR_x2APIC;
1966             core->length = sizeof(*core);
1967             core->x2apic_id = cpu_to_le32(apic_id);
1968             core->proximity_domain = cpu_to_le32(node_id);
1969             core->flags = cpu_to_le32(1);
1970         }
1971     }
1972 
1973 
1974     /* the memory map is a bit tricky, it contains at least one hole
1975      * from 640k-1M and possibly another one from 3.5G-4G.
1976      */
1977     next_base = 0;
1978     numa_start = table_data->len;
1979 
1980     for (i = 1; i < pcms->numa_nodes + 1; ++i) {
1981         mem_base = next_base;
1982         mem_len = pcms->node_mem[i - 1];
1983         next_base = mem_base + mem_len;
1984 
1985         /* Cut out the 640K hole */
1986         if (mem_base <= HOLE_640K_START &&
1987             next_base > HOLE_640K_START) {
1988             mem_len -= next_base - HOLE_640K_START;
1989             if (mem_len > 0) {
1990                 numamem = acpi_data_push(table_data, sizeof *numamem);
1991                 build_srat_memory(numamem, mem_base, mem_len, i - 1,
1992                                   MEM_AFFINITY_ENABLED);
1993             }
1994 
1995             /* Check for the rare case: 640K < RAM < 1M */
1996             if (next_base <= HOLE_640K_END) {
1997                 next_base = HOLE_640K_END;
1998                 continue;
1999             }
2000             mem_base = HOLE_640K_END;
2001             mem_len = next_base - HOLE_640K_END;
2002         }
2003 
2004         /* Cut out the ACPI_PCI hole */
2005         if (mem_base <= x86ms->below_4g_mem_size &&
2006             next_base > x86ms->below_4g_mem_size) {
2007             mem_len -= next_base - x86ms->below_4g_mem_size;
2008             if (mem_len > 0) {
2009                 numamem = acpi_data_push(table_data, sizeof *numamem);
2010                 build_srat_memory(numamem, mem_base, mem_len, i - 1,
2011                                   MEM_AFFINITY_ENABLED);
2012             }
2013             mem_base = 1ULL << 32;
2014             mem_len = next_base - x86ms->below_4g_mem_size;
2015             next_base = mem_base + mem_len;
2016         }
2017 
2018         if (mem_len > 0) {
2019             numamem = acpi_data_push(table_data, sizeof *numamem);
2020             build_srat_memory(numamem, mem_base, mem_len, i - 1,
2021                               MEM_AFFINITY_ENABLED);
2022         }
2023     }
2024 
2025     if (machine->nvdimms_state->is_enabled) {
2026         nvdimm_build_srat(table_data);
2027     }
2028 
2029     slots = (table_data->len - numa_start) / sizeof *numamem;
2030     for (; slots < pcms->numa_nodes + 2; slots++) {
2031         numamem = acpi_data_push(table_data, sizeof *numamem);
2032         build_srat_memory(numamem, 0, 0, 0, MEM_AFFINITY_NOFLAGS);
2033     }
2034 
2035     /*
2036      * Entry is required for Windows to enable memory hotplug in OS
2037      * and for Linux to enable SWIOTLB when booted with less than
2038      * 4G of RAM. Windows works better if the entry sets proximity
2039      * to the highest NUMA node in the machine.
2040      * Memory devices may override proximity set by this entry,
2041      * providing _PXM method if necessary.
2042      */
2043     if (hotplugabble_address_space_size) {
2044         numamem = acpi_data_push(table_data, sizeof *numamem);
2045         build_srat_memory(numamem, machine->device_memory->base,
2046                           hotplugabble_address_space_size, pcms->numa_nodes - 1,
2047                           MEM_AFFINITY_HOTPLUGGABLE | MEM_AFFINITY_ENABLED);
2048     }
2049 
2050     build_header(linker, table_data,
2051                  (void *)(table_data->data + srat_start),
2052                  "SRAT",
2053                  table_data->len - srat_start, 1, NULL, NULL);
2054 }
2055 
2056 /*
2057  * VT-d spec 8.1 DMA Remapping Reporting Structure
2058  * (version Oct. 2014 or later)
2059  */
2060 static void
2061 build_dmar_q35(GArray *table_data, BIOSLinker *linker)
2062 {
2063     int dmar_start = table_data->len;
2064 
2065     AcpiTableDmar *dmar;
2066     AcpiDmarHardwareUnit *drhd;
2067     AcpiDmarRootPortATS *atsr;
2068     uint8_t dmar_flags = 0;
2069     X86IOMMUState *iommu = x86_iommu_get_default();
2070     AcpiDmarDeviceScope *scope = NULL;
2071     /* Root complex IOAPIC use one path[0] only */
2072     size_t ioapic_scope_size = sizeof(*scope) + sizeof(scope->path[0]);
2073     IntelIOMMUState *intel_iommu = INTEL_IOMMU_DEVICE(iommu);
2074 
2075     assert(iommu);
2076     if (x86_iommu_ir_supported(iommu)) {
2077         dmar_flags |= 0x1;      /* Flags: 0x1: INT_REMAP */
2078     }
2079 
2080     dmar = acpi_data_push(table_data, sizeof(*dmar));
2081     dmar->host_address_width = intel_iommu->aw_bits - 1;
2082     dmar->flags = dmar_flags;
2083 
2084     /* DMAR Remapping Hardware Unit Definition structure */
2085     drhd = acpi_data_push(table_data, sizeof(*drhd) + ioapic_scope_size);
2086     drhd->type = cpu_to_le16(ACPI_DMAR_TYPE_HARDWARE_UNIT);
2087     drhd->length = cpu_to_le16(sizeof(*drhd) + ioapic_scope_size);
2088     drhd->flags = ACPI_DMAR_INCLUDE_PCI_ALL;
2089     drhd->pci_segment = cpu_to_le16(0);
2090     drhd->address = cpu_to_le64(Q35_HOST_BRIDGE_IOMMU_ADDR);
2091 
2092     /* Scope definition for the root-complex IOAPIC. See VT-d spec
2093      * 8.3.1 (version Oct. 2014 or later). */
2094     scope = &drhd->scope[0];
2095     scope->entry_type = 0x03;   /* Type: 0x03 for IOAPIC */
2096     scope->length = ioapic_scope_size;
2097     scope->enumeration_id = ACPI_BUILD_IOAPIC_ID;
2098     scope->bus = Q35_PSEUDO_BUS_PLATFORM;
2099     scope->path[0].device = PCI_SLOT(Q35_PSEUDO_DEVFN_IOAPIC);
2100     scope->path[0].function = PCI_FUNC(Q35_PSEUDO_DEVFN_IOAPIC);
2101 
2102     if (iommu->dt_supported) {
2103         atsr = acpi_data_push(table_data, sizeof(*atsr));
2104         atsr->type = cpu_to_le16(ACPI_DMAR_TYPE_ATSR);
2105         atsr->length = cpu_to_le16(sizeof(*atsr));
2106         atsr->flags = ACPI_DMAR_ATSR_ALL_PORTS;
2107         atsr->pci_segment = cpu_to_le16(0);
2108     }
2109 
2110     build_header(linker, table_data, (void *)(table_data->data + dmar_start),
2111                  "DMAR", table_data->len - dmar_start, 1, NULL, NULL);
2112 }
2113 
2114 /*
2115  * Windows ACPI Emulated Devices Table
2116  * (Version 1.0 - April 6, 2009)
2117  * Spec: http://download.microsoft.com/download/7/E/7/7E7662CF-CBEA-470B-A97E-CE7CE0D98DC2/WAET.docx
2118  *
2119  * Helpful to speedup Windows guests and ignored by others.
2120  */
2121 static void
2122 build_waet(GArray *table_data, BIOSLinker *linker)
2123 {
2124     int waet_start = table_data->len;
2125 
2126     /* WAET header */
2127     acpi_data_push(table_data, sizeof(AcpiTableHeader));
2128     /*
2129      * Set "ACPI PM timer good" flag.
2130      *
2131      * Tells Windows guests that our ACPI PM timer is reliable in the
2132      * sense that guest can read it only once to obtain a reliable value.
2133      * Which avoids costly VMExits caused by guest re-reading it unnecessarily.
2134      */
2135     build_append_int_noprefix(table_data, 1 << 1 /* ACPI PM timer good */, 4);
2136 
2137     build_header(linker, table_data, (void *)(table_data->data + waet_start),
2138                  "WAET", table_data->len - waet_start, 1, NULL, NULL);
2139 }
2140 
2141 /*
2142  *   IVRS table as specified in AMD IOMMU Specification v2.62, Section 5.2
2143  *   accessible here http://support.amd.com/TechDocs/48882_IOMMU.pdf
2144  */
2145 #define IOAPIC_SB_DEVID   (uint64_t)PCI_BUILD_BDF(0, PCI_DEVFN(0x14, 0))
2146 
2147 /*
2148  * Insert IVHD entry for device and recurse, insert alias, or insert range as
2149  * necessary for the PCI topology.
2150  */
2151 static void
2152 insert_ivhd(PCIBus *bus, PCIDevice *dev, void *opaque)
2153 {
2154     GArray *table_data = opaque;
2155     uint32_t entry;
2156 
2157     /* "Select" IVHD entry, type 0x2 */
2158     entry = PCI_BUILD_BDF(pci_bus_num(bus), dev->devfn) << 8 | 0x2;
2159     build_append_int_noprefix(table_data, entry, 4);
2160 
2161     if (object_dynamic_cast(OBJECT(dev), TYPE_PCI_BRIDGE)) {
2162         PCIBus *sec_bus = pci_bridge_get_sec_bus(PCI_BRIDGE(dev));
2163         uint8_t sec = pci_bus_num(sec_bus);
2164         uint8_t sub = dev->config[PCI_SUBORDINATE_BUS];
2165 
2166         if (pci_bus_is_express(sec_bus)) {
2167             /*
2168              * Walk the bus if there are subordinates, otherwise use a range
2169              * to cover an entire leaf bus.  We could potentially also use a
2170              * range for traversed buses, but we'd need to take care not to
2171              * create both Select and Range entries covering the same device.
2172              * This is easier and potentially more compact.
2173              *
2174              * An example bare metal system seems to use Select entries for
2175              * root ports without a slot (ie. built-ins) and Range entries
2176              * when there is a slot.  The same system also only hard-codes
2177              * the alias range for an onboard PCIe-to-PCI bridge, apparently
2178              * making no effort to support nested bridges.  We attempt to
2179              * be more thorough here.
2180              */
2181             if (sec == sub) { /* leaf bus */
2182                 /* "Start of Range" IVHD entry, type 0x3 */
2183                 entry = PCI_BUILD_BDF(sec, PCI_DEVFN(0, 0)) << 8 | 0x3;
2184                 build_append_int_noprefix(table_data, entry, 4);
2185                 /* "End of Range" IVHD entry, type 0x4 */
2186                 entry = PCI_BUILD_BDF(sub, PCI_DEVFN(31, 7)) << 8 | 0x4;
2187                 build_append_int_noprefix(table_data, entry, 4);
2188             } else {
2189                 pci_for_each_device(sec_bus, sec, insert_ivhd, table_data);
2190             }
2191         } else {
2192             /*
2193              * If the secondary bus is conventional, then we need to create an
2194              * Alias range for everything downstream.  The range covers the
2195              * first devfn on the secondary bus to the last devfn on the
2196              * subordinate bus.  The alias target depends on legacy versus
2197              * express bridges, just as in pci_device_iommu_address_space().
2198              * DeviceIDa vs DeviceIDb as per the AMD IOMMU spec.
2199              */
2200             uint16_t dev_id_a, dev_id_b;
2201 
2202             dev_id_a = PCI_BUILD_BDF(sec, PCI_DEVFN(0, 0));
2203 
2204             if (pci_is_express(dev) &&
2205                 pcie_cap_get_type(dev) == PCI_EXP_TYPE_PCI_BRIDGE) {
2206                 dev_id_b = dev_id_a;
2207             } else {
2208                 dev_id_b = PCI_BUILD_BDF(pci_bus_num(bus), dev->devfn);
2209             }
2210 
2211             /* "Alias Start of Range" IVHD entry, type 0x43, 8 bytes */
2212             build_append_int_noprefix(table_data, dev_id_a << 8 | 0x43, 4);
2213             build_append_int_noprefix(table_data, dev_id_b << 8 | 0x0, 4);
2214 
2215             /* "End of Range" IVHD entry, type 0x4 */
2216             entry = PCI_BUILD_BDF(sub, PCI_DEVFN(31, 7)) << 8 | 0x4;
2217             build_append_int_noprefix(table_data, entry, 4);
2218         }
2219     }
2220 }
2221 
2222 /* For all PCI host bridges, walk and insert IVHD entries */
2223 static int
2224 ivrs_host_bridges(Object *obj, void *opaque)
2225 {
2226     GArray *ivhd_blob = opaque;
2227 
2228     if (object_dynamic_cast(obj, TYPE_PCI_HOST_BRIDGE)) {
2229         PCIBus *bus = PCI_HOST_BRIDGE(obj)->bus;
2230 
2231         if (bus) {
2232             pci_for_each_device(bus, pci_bus_num(bus), insert_ivhd, ivhd_blob);
2233         }
2234     }
2235 
2236     return 0;
2237 }
2238 
2239 static void
2240 build_amd_iommu(GArray *table_data, BIOSLinker *linker)
2241 {
2242     int ivhd_table_len = 24;
2243     int iommu_start = table_data->len;
2244     AMDVIState *s = AMD_IOMMU_DEVICE(x86_iommu_get_default());
2245     GArray *ivhd_blob = g_array_new(false, true, 1);
2246 
2247     /* IVRS header */
2248     acpi_data_push(table_data, sizeof(AcpiTableHeader));
2249     /* IVinfo - IO virtualization information common to all
2250      * IOMMU units in a system
2251      */
2252     build_append_int_noprefix(table_data, 40UL << 8/* PASize */, 4);
2253     /* reserved */
2254     build_append_int_noprefix(table_data, 0, 8);
2255 
2256     /* IVHD definition - type 10h */
2257     build_append_int_noprefix(table_data, 0x10, 1);
2258     /* virtualization flags */
2259     build_append_int_noprefix(table_data,
2260                              (1UL << 0) | /* HtTunEn      */
2261                              (1UL << 4) | /* iotblSup     */
2262                              (1UL << 6) | /* PrefSup      */
2263                              (1UL << 7),  /* PPRSup       */
2264                              1);
2265 
2266     /*
2267      * A PCI bus walk, for each PCI host bridge, is necessary to create a
2268      * complete set of IVHD entries.  Do this into a separate blob so that we
2269      * can calculate the total IVRS table length here and then append the new
2270      * blob further below.  Fall back to an entry covering all devices, which
2271      * is sufficient when no aliases are present.
2272      */
2273     object_child_foreach_recursive(object_get_root(),
2274                                    ivrs_host_bridges, ivhd_blob);
2275 
2276     if (!ivhd_blob->len) {
2277         /*
2278          *   Type 1 device entry reporting all devices
2279          *   These are 4-byte device entries currently reporting the range of
2280          *   Refer to Spec - Table 95:IVHD Device Entry Type Codes(4-byte)
2281          */
2282         build_append_int_noprefix(ivhd_blob, 0x0000001, 4);
2283     }
2284 
2285     ivhd_table_len += ivhd_blob->len;
2286 
2287     /*
2288      * When interrupt remapping is supported, we add a special IVHD device
2289      * for type IO-APIC.
2290      */
2291     if (x86_iommu_ir_supported(x86_iommu_get_default())) {
2292         ivhd_table_len += 8;
2293     }
2294 
2295     /* IVHD length */
2296     build_append_int_noprefix(table_data, ivhd_table_len, 2);
2297     /* DeviceID */
2298     build_append_int_noprefix(table_data, s->devid, 2);
2299     /* Capability offset */
2300     build_append_int_noprefix(table_data, s->capab_offset, 2);
2301     /* IOMMU base address */
2302     build_append_int_noprefix(table_data, s->mmio.addr, 8);
2303     /* PCI Segment Group */
2304     build_append_int_noprefix(table_data, 0, 2);
2305     /* IOMMU info */
2306     build_append_int_noprefix(table_data, 0, 2);
2307     /* IOMMU Feature Reporting */
2308     build_append_int_noprefix(table_data,
2309                              (48UL << 30) | /* HATS   */
2310                              (48UL << 28) | /* GATS   */
2311                              (1UL << 2)   | /* GTSup  */
2312                              (1UL << 6),    /* GASup  */
2313                              4);
2314 
2315     /* IVHD entries as found above */
2316     g_array_append_vals(table_data, ivhd_blob->data, ivhd_blob->len);
2317     g_array_free(ivhd_blob, TRUE);
2318 
2319     /*
2320      * Add a special IVHD device type.
2321      * Refer to spec - Table 95: IVHD device entry type codes
2322      *
2323      * Linux IOMMU driver checks for the special IVHD device (type IO-APIC).
2324      * See Linux kernel commit 'c2ff5cf5294bcbd7fa50f7d860e90a66db7e5059'
2325      */
2326     if (x86_iommu_ir_supported(x86_iommu_get_default())) {
2327         build_append_int_noprefix(table_data,
2328                                  (0x1ull << 56) |           /* type IOAPIC */
2329                                  (IOAPIC_SB_DEVID << 40) |  /* IOAPIC devid */
2330                                  0x48,                      /* special device */
2331                                  8);
2332     }
2333 
2334     build_header(linker, table_data, (void *)(table_data->data + iommu_start),
2335                  "IVRS", table_data->len - iommu_start, 1, NULL, NULL);
2336 }
2337 
2338 typedef
2339 struct AcpiBuildState {
2340     /* Copy of table in RAM (for patching). */
2341     MemoryRegion *table_mr;
2342     /* Is table patched? */
2343     uint8_t patched;
2344     void *rsdp;
2345     MemoryRegion *rsdp_mr;
2346     MemoryRegion *linker_mr;
2347 } AcpiBuildState;
2348 
2349 static bool acpi_get_mcfg(AcpiMcfgInfo *mcfg)
2350 {
2351     Object *pci_host;
2352     QObject *o;
2353 
2354     pci_host = acpi_get_i386_pci_host();
2355     g_assert(pci_host);
2356 
2357     o = object_property_get_qobject(pci_host, PCIE_HOST_MCFG_BASE, NULL);
2358     if (!o) {
2359         return false;
2360     }
2361     mcfg->base = qnum_get_uint(qobject_to(QNum, o));
2362     qobject_unref(o);
2363     if (mcfg->base == PCIE_BASE_ADDR_UNMAPPED) {
2364         return false;
2365     }
2366 
2367     o = object_property_get_qobject(pci_host, PCIE_HOST_MCFG_SIZE, NULL);
2368     assert(o);
2369     mcfg->size = qnum_get_uint(qobject_to(QNum, o));
2370     qobject_unref(o);
2371     return true;
2372 }
2373 
2374 static
2375 void acpi_build(AcpiBuildTables *tables, MachineState *machine)
2376 {
2377     PCMachineState *pcms = PC_MACHINE(machine);
2378     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
2379     X86MachineState *x86ms = X86_MACHINE(machine);
2380     GArray *table_offsets;
2381     unsigned facs, dsdt, rsdt, fadt;
2382     AcpiPmInfo pm;
2383     AcpiMiscInfo misc;
2384     AcpiMcfgInfo mcfg;
2385     Range pci_hole, pci_hole64;
2386     uint8_t *u;
2387     size_t aml_len = 0;
2388     GArray *tables_blob = tables->table_data;
2389     AcpiSlicOem slic_oem = { .id = NULL, .table_id = NULL };
2390     Object *vmgenid_dev;
2391 
2392     acpi_get_pm_info(machine, &pm);
2393     acpi_get_misc_info(&misc);
2394     acpi_get_pci_holes(&pci_hole, &pci_hole64);
2395     acpi_get_slic_oem(&slic_oem);
2396 
2397     table_offsets = g_array_new(false, true /* clear */,
2398                                         sizeof(uint32_t));
2399     ACPI_BUILD_DPRINTF("init ACPI tables\n");
2400 
2401     bios_linker_loader_alloc(tables->linker,
2402                              ACPI_BUILD_TABLE_FILE, tables_blob,
2403                              64 /* Ensure FACS is aligned */,
2404                              false /* high memory */);
2405 
2406     /*
2407      * FACS is pointed to by FADT.
2408      * We place it first since it's the only table that has alignment
2409      * requirements.
2410      */
2411     facs = tables_blob->len;
2412     build_facs(tables_blob);
2413 
2414     /* DSDT is pointed to by FADT */
2415     dsdt = tables_blob->len;
2416     build_dsdt(tables_blob, tables->linker, &pm, &misc,
2417                &pci_hole, &pci_hole64, machine);
2418 
2419     /* Count the size of the DSDT and SSDT, we will need it for legacy
2420      * sizing of ACPI tables.
2421      */
2422     aml_len += tables_blob->len - dsdt;
2423 
2424     /* ACPI tables pointed to by RSDT */
2425     fadt = tables_blob->len;
2426     acpi_add_table(table_offsets, tables_blob);
2427     pm.fadt.facs_tbl_offset = &facs;
2428     pm.fadt.dsdt_tbl_offset = &dsdt;
2429     pm.fadt.xdsdt_tbl_offset = &dsdt;
2430     build_fadt(tables_blob, tables->linker, &pm.fadt,
2431                slic_oem.id, slic_oem.table_id);
2432     aml_len += tables_blob->len - fadt;
2433 
2434     acpi_add_table(table_offsets, tables_blob);
2435     acpi_build_madt(tables_blob, tables->linker, x86ms,
2436                     ACPI_DEVICE_IF(pcms->acpi_dev), true);
2437 
2438     vmgenid_dev = find_vmgenid_dev();
2439     if (vmgenid_dev) {
2440         acpi_add_table(table_offsets, tables_blob);
2441         vmgenid_build_acpi(VMGENID(vmgenid_dev), tables_blob,
2442                            tables->vmgenid, tables->linker);
2443     }
2444 
2445     if (misc.has_hpet) {
2446         acpi_add_table(table_offsets, tables_blob);
2447         build_hpet(tables_blob, tables->linker);
2448     }
2449     if (misc.tpm_version != TPM_VERSION_UNSPEC) {
2450         if (misc.tpm_version == TPM_VERSION_1_2) {
2451             acpi_add_table(table_offsets, tables_blob);
2452             build_tpm_tcpa(tables_blob, tables->linker, tables->tcpalog);
2453         } else { /* TPM_VERSION_2_0 */
2454             acpi_add_table(table_offsets, tables_blob);
2455             build_tpm2(tables_blob, tables->linker, tables->tcpalog);
2456         }
2457     }
2458     if (pcms->numa_nodes) {
2459         acpi_add_table(table_offsets, tables_blob);
2460         build_srat(tables_blob, tables->linker, machine);
2461         if (machine->numa_state->have_numa_distance) {
2462             acpi_add_table(table_offsets, tables_blob);
2463             build_slit(tables_blob, tables->linker, machine);
2464         }
2465         if (machine->numa_state->hmat_enabled) {
2466             acpi_add_table(table_offsets, tables_blob);
2467             build_hmat(tables_blob, tables->linker, machine->numa_state);
2468         }
2469     }
2470     if (acpi_get_mcfg(&mcfg)) {
2471         acpi_add_table(table_offsets, tables_blob);
2472         build_mcfg(tables_blob, tables->linker, &mcfg);
2473     }
2474     if (x86_iommu_get_default()) {
2475         IommuType IOMMUType = x86_iommu_get_type();
2476         if (IOMMUType == TYPE_AMD) {
2477             acpi_add_table(table_offsets, tables_blob);
2478             build_amd_iommu(tables_blob, tables->linker);
2479         } else if (IOMMUType == TYPE_INTEL) {
2480             acpi_add_table(table_offsets, tables_blob);
2481             build_dmar_q35(tables_blob, tables->linker);
2482         }
2483     }
2484     if (machine->nvdimms_state->is_enabled) {
2485         nvdimm_build_acpi(table_offsets, tables_blob, tables->linker,
2486                           machine->nvdimms_state, machine->ram_slots);
2487     }
2488 
2489     acpi_add_table(table_offsets, tables_blob);
2490     build_waet(tables_blob, tables->linker);
2491 
2492     /* Add tables supplied by user (if any) */
2493     for (u = acpi_table_first(); u; u = acpi_table_next(u)) {
2494         unsigned len = acpi_table_len(u);
2495 
2496         acpi_add_table(table_offsets, tables_blob);
2497         g_array_append_vals(tables_blob, u, len);
2498     }
2499 
2500     /* RSDT is pointed to by RSDP */
2501     rsdt = tables_blob->len;
2502     build_rsdt(tables_blob, tables->linker, table_offsets,
2503                slic_oem.id, slic_oem.table_id);
2504 
2505     /* RSDP is in FSEG memory, so allocate it separately */
2506     {
2507         AcpiRsdpData rsdp_data = {
2508             .revision = 0,
2509             .oem_id = ACPI_BUILD_APPNAME6,
2510             .xsdt_tbl_offset = NULL,
2511             .rsdt_tbl_offset = &rsdt,
2512         };
2513         build_rsdp(tables->rsdp, tables->linker, &rsdp_data);
2514         if (!pcmc->rsdp_in_ram) {
2515             /* We used to allocate some extra space for RSDP revision 2 but
2516              * only used the RSDP revision 0 space. The extra bytes were
2517              * zeroed out and not used.
2518              * Here we continue wasting those extra 16 bytes to make sure we
2519              * don't break migration for machine types 2.2 and older due to
2520              * RSDP blob size mismatch.
2521              */
2522             build_append_int_noprefix(tables->rsdp, 0, 16);
2523         }
2524     }
2525 
2526     /* We'll expose it all to Guest so we want to reduce
2527      * chance of size changes.
2528      *
2529      * We used to align the tables to 4k, but of course this would
2530      * too simple to be enough.  4k turned out to be too small an
2531      * alignment very soon, and in fact it is almost impossible to
2532      * keep the table size stable for all (max_cpus, max_memory_slots)
2533      * combinations.  So the table size is always 64k for pc-i440fx-2.1
2534      * and we give an error if the table grows beyond that limit.
2535      *
2536      * We still have the problem of migrating from "-M pc-i440fx-2.0".  For
2537      * that, we exploit the fact that QEMU 2.1 generates _smaller_ tables
2538      * than 2.0 and we can always pad the smaller tables with zeros.  We can
2539      * then use the exact size of the 2.0 tables.
2540      *
2541      * All this is for PIIX4, since QEMU 2.0 didn't support Q35 migration.
2542      */
2543     if (pcmc->legacy_acpi_table_size) {
2544         /* Subtracting aml_len gives the size of fixed tables.  Then add the
2545          * size of the PIIX4 DSDT/SSDT in QEMU 2.0.
2546          */
2547         int legacy_aml_len =
2548             pcmc->legacy_acpi_table_size +
2549             ACPI_BUILD_LEGACY_CPU_AML_SIZE * x86ms->apic_id_limit;
2550         int legacy_table_size =
2551             ROUND_UP(tables_blob->len - aml_len + legacy_aml_len,
2552                      ACPI_BUILD_ALIGN_SIZE);
2553         if (tables_blob->len > legacy_table_size) {
2554             /* Should happen only with PCI bridges and -M pc-i440fx-2.0.  */
2555             warn_report("ACPI table size %u exceeds %d bytes,"
2556                         " migration may not work",
2557                         tables_blob->len, legacy_table_size);
2558             error_printf("Try removing CPUs, NUMA nodes, memory slots"
2559                          " or PCI bridges.");
2560         }
2561         g_array_set_size(tables_blob, legacy_table_size);
2562     } else {
2563         /* Make sure we have a buffer in case we need to resize the tables. */
2564         if (tables_blob->len > ACPI_BUILD_TABLE_SIZE / 2) {
2565             /* As of QEMU 2.1, this fires with 160 VCPUs and 255 memory slots.  */
2566             warn_report("ACPI table size %u exceeds %d bytes,"
2567                         " migration may not work",
2568                         tables_blob->len, ACPI_BUILD_TABLE_SIZE / 2);
2569             error_printf("Try removing CPUs, NUMA nodes, memory slots"
2570                          " or PCI bridges.");
2571         }
2572         acpi_align_size(tables_blob, ACPI_BUILD_TABLE_SIZE);
2573     }
2574 
2575     acpi_align_size(tables->linker->cmd_blob, ACPI_BUILD_ALIGN_SIZE);
2576 
2577     /* Cleanup memory that's no longer used. */
2578     g_array_free(table_offsets, true);
2579 }
2580 
2581 static void acpi_ram_update(MemoryRegion *mr, GArray *data)
2582 {
2583     uint32_t size = acpi_data_len(data);
2584 
2585     /* Make sure RAM size is correct - in case it got changed e.g. by migration */
2586     memory_region_ram_resize(mr, size, &error_abort);
2587 
2588     memcpy(memory_region_get_ram_ptr(mr), data->data, size);
2589     memory_region_set_dirty(mr, 0, size);
2590 }
2591 
2592 static void acpi_build_update(void *build_opaque)
2593 {
2594     AcpiBuildState *build_state = build_opaque;
2595     AcpiBuildTables tables;
2596 
2597     /* No state to update or already patched? Nothing to do. */
2598     if (!build_state || build_state->patched) {
2599         return;
2600     }
2601     build_state->patched = 1;
2602 
2603     acpi_build_tables_init(&tables);
2604 
2605     acpi_build(&tables, MACHINE(qdev_get_machine()));
2606 
2607     acpi_ram_update(build_state->table_mr, tables.table_data);
2608 
2609     if (build_state->rsdp) {
2610         memcpy(build_state->rsdp, tables.rsdp->data, acpi_data_len(tables.rsdp));
2611     } else {
2612         acpi_ram_update(build_state->rsdp_mr, tables.rsdp);
2613     }
2614 
2615     acpi_ram_update(build_state->linker_mr, tables.linker->cmd_blob);
2616     acpi_build_tables_cleanup(&tables, true);
2617 }
2618 
2619 static void acpi_build_reset(void *build_opaque)
2620 {
2621     AcpiBuildState *build_state = build_opaque;
2622     build_state->patched = 0;
2623 }
2624 
2625 static const VMStateDescription vmstate_acpi_build = {
2626     .name = "acpi_build",
2627     .version_id = 1,
2628     .minimum_version_id = 1,
2629     .fields = (VMStateField[]) {
2630         VMSTATE_UINT8(patched, AcpiBuildState),
2631         VMSTATE_END_OF_LIST()
2632     },
2633 };
2634 
2635 void acpi_setup(void)
2636 {
2637     PCMachineState *pcms = PC_MACHINE(qdev_get_machine());
2638     PCMachineClass *pcmc = PC_MACHINE_GET_CLASS(pcms);
2639     X86MachineState *x86ms = X86_MACHINE(pcms);
2640     AcpiBuildTables tables;
2641     AcpiBuildState *build_state;
2642     Object *vmgenid_dev;
2643     TPMIf *tpm;
2644     static FwCfgTPMConfig tpm_config;
2645 
2646     if (!x86ms->fw_cfg) {
2647         ACPI_BUILD_DPRINTF("No fw cfg. Bailing out.\n");
2648         return;
2649     }
2650 
2651     if (!pcms->acpi_build_enabled) {
2652         ACPI_BUILD_DPRINTF("ACPI build disabled. Bailing out.\n");
2653         return;
2654     }
2655 
2656     if (!x86_machine_is_acpi_enabled(X86_MACHINE(pcms))) {
2657         ACPI_BUILD_DPRINTF("ACPI disabled. Bailing out.\n");
2658         return;
2659     }
2660 
2661     build_state = g_malloc0(sizeof *build_state);
2662 
2663     acpi_build_tables_init(&tables);
2664     acpi_build(&tables, MACHINE(pcms));
2665 
2666     /* Now expose it all to Guest */
2667     build_state->table_mr = acpi_add_rom_blob(acpi_build_update,
2668                                               build_state, tables.table_data,
2669                                               ACPI_BUILD_TABLE_FILE,
2670                                               ACPI_BUILD_TABLE_MAX_SIZE);
2671     assert(build_state->table_mr != NULL);
2672 
2673     build_state->linker_mr =
2674         acpi_add_rom_blob(acpi_build_update, build_state,
2675                           tables.linker->cmd_blob, ACPI_BUILD_LOADER_FILE, 0);
2676 
2677     fw_cfg_add_file(x86ms->fw_cfg, ACPI_BUILD_TPMLOG_FILE,
2678                     tables.tcpalog->data, acpi_data_len(tables.tcpalog));
2679 
2680     tpm = tpm_find();
2681     if (tpm && object_property_get_bool(OBJECT(tpm), "ppi", &error_abort)) {
2682         tpm_config = (FwCfgTPMConfig) {
2683             .tpmppi_address = cpu_to_le32(TPM_PPI_ADDR_BASE),
2684             .tpm_version = tpm_get_version(tpm),
2685             .tpmppi_version = TPM_PPI_VERSION_1_30
2686         };
2687         fw_cfg_add_file(x86ms->fw_cfg, "etc/tpm/config",
2688                         &tpm_config, sizeof tpm_config);
2689     }
2690 
2691     vmgenid_dev = find_vmgenid_dev();
2692     if (vmgenid_dev) {
2693         vmgenid_add_fw_cfg(VMGENID(vmgenid_dev), x86ms->fw_cfg,
2694                            tables.vmgenid);
2695     }
2696 
2697     if (!pcmc->rsdp_in_ram) {
2698         /*
2699          * Keep for compatibility with old machine types.
2700          * Though RSDP is small, its contents isn't immutable, so
2701          * we'll update it along with the rest of tables on guest access.
2702          */
2703         uint32_t rsdp_size = acpi_data_len(tables.rsdp);
2704 
2705         build_state->rsdp = g_memdup(tables.rsdp->data, rsdp_size);
2706         fw_cfg_add_file_callback(x86ms->fw_cfg, ACPI_BUILD_RSDP_FILE,
2707                                  acpi_build_update, NULL, build_state,
2708                                  build_state->rsdp, rsdp_size, true);
2709         build_state->rsdp_mr = NULL;
2710     } else {
2711         build_state->rsdp = NULL;
2712         build_state->rsdp_mr = acpi_add_rom_blob(acpi_build_update,
2713                                                  build_state, tables.rsdp,
2714                                                  ACPI_BUILD_RSDP_FILE, 0);
2715     }
2716 
2717     qemu_register_reset(acpi_build_reset, build_state);
2718     acpi_build_reset(build_state);
2719     vmstate_register(NULL, 0, &vmstate_acpi_build, build_state);
2720 
2721     /* Cleanup tables but don't free the memory: we track it
2722      * in build_state.
2723      */
2724     acpi_build_tables_cleanup(&tables, false);
2725 }
2726