xref: /openbmc/qemu/hw/i386/acpi-build.c (revision 9ff7f5bddbe5814bafe5e798d2cf1087b58dc7b6)
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 "acpi-build.h"
24 #include <stddef.h>
25 #include <glib.h>
26 #include "qemu-common.h"
27 #include "qemu/bitmap.h"
28 #include "qemu/osdep.h"
29 #include "qemu/error-report.h"
30 #include "hw/pci/pci.h"
31 #include "qom/cpu.h"
32 #include "hw/i386/pc.h"
33 #include "target-i386/cpu.h"
34 #include "hw/timer/hpet.h"
35 #include "hw/acpi/acpi-defs.h"
36 #include "hw/acpi/acpi.h"
37 #include "hw/nvram/fw_cfg.h"
38 #include "hw/acpi/bios-linker-loader.h"
39 #include "hw/loader.h"
40 #include "hw/isa/isa.h"
41 #include "hw/acpi/memory_hotplug.h"
42 #include "sysemu/tpm.h"
43 #include "hw/acpi/tpm.h"
44 #include "sysemu/tpm_backend.h"
45 
46 /* Supported chipsets: */
47 #include "hw/acpi/piix4.h"
48 #include "hw/acpi/pcihp.h"
49 #include "hw/i386/ich9.h"
50 #include "hw/pci/pci_bus.h"
51 #include "hw/pci-host/q35.h"
52 #include "hw/i386/intel_iommu.h"
53 
54 #include "hw/i386/q35-acpi-dsdt.hex"
55 #include "hw/i386/acpi-dsdt.hex"
56 
57 #include "hw/acpi/aml-build.h"
58 
59 #include "qapi/qmp/qint.h"
60 #include "qom/qom-qobject.h"
61 
62 /* These are used to size the ACPI tables for -M pc-i440fx-1.7 and
63  * -M pc-i440fx-2.0.  Even if the actual amount of AML generated grows
64  * a little bit, there should be plenty of free space since the DSDT
65  * shrunk by ~1.5k between QEMU 2.0 and QEMU 2.1.
66  */
67 #define ACPI_BUILD_LEGACY_CPU_AML_SIZE    97
68 #define ACPI_BUILD_ALIGN_SIZE             0x1000
69 
70 #define ACPI_BUILD_TABLE_SIZE             0x20000
71 
72 /* #define DEBUG_ACPI_BUILD */
73 #ifdef DEBUG_ACPI_BUILD
74 #define ACPI_BUILD_DPRINTF(fmt, ...)        \
75     do {printf("ACPI_BUILD: " fmt, ## __VA_ARGS__); } while (0)
76 #else
77 #define ACPI_BUILD_DPRINTF(fmt, ...)
78 #endif
79 
80 typedef struct AcpiCpuInfo {
81     DECLARE_BITMAP(found_cpus, ACPI_CPU_HOTPLUG_ID_LIMIT);
82 } AcpiCpuInfo;
83 
84 typedef struct AcpiMcfgInfo {
85     uint64_t mcfg_base;
86     uint32_t mcfg_size;
87 } AcpiMcfgInfo;
88 
89 typedef struct AcpiPmInfo {
90     bool s3_disabled;
91     bool s4_disabled;
92     bool pcihp_bridge_en;
93     uint8_t s4_val;
94     uint16_t sci_int;
95     uint8_t acpi_enable_cmd;
96     uint8_t acpi_disable_cmd;
97     uint32_t gpe0_blk;
98     uint32_t gpe0_blk_len;
99     uint32_t io_base;
100     uint16_t cpu_hp_io_base;
101     uint16_t cpu_hp_io_len;
102     uint16_t mem_hp_io_base;
103     uint16_t mem_hp_io_len;
104     uint16_t pcihp_io_base;
105     uint16_t pcihp_io_len;
106 } AcpiPmInfo;
107 
108 typedef struct AcpiMiscInfo {
109     bool has_hpet;
110     TPMVersion tpm_version;
111     const unsigned char *dsdt_code;
112     unsigned dsdt_size;
113     uint16_t pvpanic_port;
114     uint16_t applesmc_io_base;
115 } AcpiMiscInfo;
116 
117 typedef struct AcpiBuildPciBusHotplugState {
118     GArray *device_table;
119     GArray *notify_table;
120     struct AcpiBuildPciBusHotplugState *parent;
121     bool pcihp_bridge_en;
122 } AcpiBuildPciBusHotplugState;
123 
124 static void acpi_get_dsdt(AcpiMiscInfo *info)
125 {
126     Object *piix = piix4_pm_find();
127     Object *lpc = ich9_lpc_find();
128     assert(!!piix != !!lpc);
129 
130     if (piix) {
131         info->dsdt_code = AcpiDsdtAmlCode;
132         info->dsdt_size = sizeof AcpiDsdtAmlCode;
133     }
134     if (lpc) {
135         info->dsdt_code = Q35AcpiDsdtAmlCode;
136         info->dsdt_size = sizeof Q35AcpiDsdtAmlCode;
137     }
138 }
139 
140 static
141 int acpi_add_cpu_info(Object *o, void *opaque)
142 {
143     AcpiCpuInfo *cpu = opaque;
144     uint64_t apic_id;
145 
146     if (object_dynamic_cast(o, TYPE_CPU)) {
147         apic_id = object_property_get_int(o, "apic-id", NULL);
148         assert(apic_id < ACPI_CPU_HOTPLUG_ID_LIMIT);
149 
150         set_bit(apic_id, cpu->found_cpus);
151     }
152 
153     object_child_foreach(o, acpi_add_cpu_info, opaque);
154     return 0;
155 }
156 
157 static void acpi_get_cpu_info(AcpiCpuInfo *cpu)
158 {
159     Object *root = object_get_root();
160 
161     memset(cpu->found_cpus, 0, sizeof cpu->found_cpus);
162     object_child_foreach(root, acpi_add_cpu_info, cpu);
163 }
164 
165 static void acpi_get_pm_info(AcpiPmInfo *pm)
166 {
167     Object *piix = piix4_pm_find();
168     Object *lpc = ich9_lpc_find();
169     Object *obj = NULL;
170     QObject *o;
171 
172     pm->pcihp_io_base = 0;
173     pm->pcihp_io_len = 0;
174     if (piix) {
175         obj = piix;
176         pm->cpu_hp_io_base = PIIX4_CPU_HOTPLUG_IO_BASE;
177         pm->pcihp_io_base =
178             object_property_get_int(obj, ACPI_PCIHP_IO_BASE_PROP, NULL);
179         pm->pcihp_io_len =
180             object_property_get_int(obj, ACPI_PCIHP_IO_LEN_PROP, NULL);
181     }
182     if (lpc) {
183         obj = lpc;
184         pm->cpu_hp_io_base = ICH9_CPU_HOTPLUG_IO_BASE;
185     }
186     assert(obj);
187 
188     pm->cpu_hp_io_len = ACPI_GPE_PROC_LEN;
189     pm->mem_hp_io_base = ACPI_MEMORY_HOTPLUG_BASE;
190     pm->mem_hp_io_len = ACPI_MEMORY_HOTPLUG_IO_LEN;
191 
192     /* Fill in optional s3/s4 related properties */
193     o = object_property_get_qobject(obj, ACPI_PM_PROP_S3_DISABLED, NULL);
194     if (o) {
195         pm->s3_disabled = qint_get_int(qobject_to_qint(o));
196     } else {
197         pm->s3_disabled = false;
198     }
199     qobject_decref(o);
200     o = object_property_get_qobject(obj, ACPI_PM_PROP_S4_DISABLED, NULL);
201     if (o) {
202         pm->s4_disabled = qint_get_int(qobject_to_qint(o));
203     } else {
204         pm->s4_disabled = false;
205     }
206     qobject_decref(o);
207     o = object_property_get_qobject(obj, ACPI_PM_PROP_S4_VAL, NULL);
208     if (o) {
209         pm->s4_val = qint_get_int(qobject_to_qint(o));
210     } else {
211         pm->s4_val = false;
212     }
213     qobject_decref(o);
214 
215     /* Fill in mandatory properties */
216     pm->sci_int = object_property_get_int(obj, ACPI_PM_PROP_SCI_INT, NULL);
217 
218     pm->acpi_enable_cmd = object_property_get_int(obj,
219                                                   ACPI_PM_PROP_ACPI_ENABLE_CMD,
220                                                   NULL);
221     pm->acpi_disable_cmd = object_property_get_int(obj,
222                                                   ACPI_PM_PROP_ACPI_DISABLE_CMD,
223                                                   NULL);
224     pm->io_base = object_property_get_int(obj, ACPI_PM_PROP_PM_IO_BASE,
225                                           NULL);
226     pm->gpe0_blk = object_property_get_int(obj, ACPI_PM_PROP_GPE0_BLK,
227                                            NULL);
228     pm->gpe0_blk_len = object_property_get_int(obj, ACPI_PM_PROP_GPE0_BLK_LEN,
229                                                NULL);
230     pm->pcihp_bridge_en =
231         object_property_get_bool(obj, "acpi-pci-hotplug-with-bridge-support",
232                                  NULL);
233 }
234 
235 static void acpi_get_misc_info(AcpiMiscInfo *info)
236 {
237     info->has_hpet = hpet_find();
238     info->tpm_version = tpm_get_version();
239     info->pvpanic_port = pvpanic_port();
240     info->applesmc_io_base = applesmc_port();
241 }
242 
243 static void acpi_get_pci_info(PcPciInfo *info)
244 {
245     Object *pci_host;
246     bool ambiguous;
247 
248     pci_host = object_resolve_path_type("", TYPE_PCI_HOST_BRIDGE, &ambiguous);
249     g_assert(!ambiguous);
250     g_assert(pci_host);
251 
252     info->w32.begin = object_property_get_int(pci_host,
253                                               PCI_HOST_PROP_PCI_HOLE_START,
254                                               NULL);
255     info->w32.end = object_property_get_int(pci_host,
256                                             PCI_HOST_PROP_PCI_HOLE_END,
257                                             NULL);
258     info->w64.begin = object_property_get_int(pci_host,
259                                               PCI_HOST_PROP_PCI_HOLE64_START,
260                                               NULL);
261     info->w64.end = object_property_get_int(pci_host,
262                                             PCI_HOST_PROP_PCI_HOLE64_END,
263                                             NULL);
264 }
265 
266 #define ACPI_PORT_SMI_CMD           0x00b2 /* TODO: this is APM_CNT_IOPORT */
267 
268 static void acpi_align_size(GArray *blob, unsigned align)
269 {
270     /* Align size to multiple of given size. This reduces the chance
271      * we need to change size in the future (breaking cross version migration).
272      */
273     g_array_set_size(blob, ROUND_UP(acpi_data_len(blob), align));
274 }
275 
276 /* FACS */
277 static void
278 build_facs(GArray *table_data, GArray *linker, PcGuestInfo *guest_info)
279 {
280     AcpiFacsDescriptorRev1 *facs = acpi_data_push(table_data, sizeof *facs);
281     memcpy(&facs->signature, "FACS", 4);
282     facs->length = cpu_to_le32(sizeof(*facs));
283 }
284 
285 /* Load chipset information in FADT */
286 static void fadt_setup(AcpiFadtDescriptorRev1 *fadt, AcpiPmInfo *pm)
287 {
288     fadt->model = 1;
289     fadt->reserved1 = 0;
290     fadt->sci_int = cpu_to_le16(pm->sci_int);
291     fadt->smi_cmd = cpu_to_le32(ACPI_PORT_SMI_CMD);
292     fadt->acpi_enable = pm->acpi_enable_cmd;
293     fadt->acpi_disable = pm->acpi_disable_cmd;
294     /* EVT, CNT, TMR offset matches hw/acpi/core.c */
295     fadt->pm1a_evt_blk = cpu_to_le32(pm->io_base);
296     fadt->pm1a_cnt_blk = cpu_to_le32(pm->io_base + 0x04);
297     fadt->pm_tmr_blk = cpu_to_le32(pm->io_base + 0x08);
298     fadt->gpe0_blk = cpu_to_le32(pm->gpe0_blk);
299     /* EVT, CNT, TMR length matches hw/acpi/core.c */
300     fadt->pm1_evt_len = 4;
301     fadt->pm1_cnt_len = 2;
302     fadt->pm_tmr_len = 4;
303     fadt->gpe0_blk_len = pm->gpe0_blk_len;
304     fadt->plvl2_lat = cpu_to_le16(0xfff); /* C2 state not supported */
305     fadt->plvl3_lat = cpu_to_le16(0xfff); /* C3 state not supported */
306     fadt->flags = cpu_to_le32((1 << ACPI_FADT_F_WBINVD) |
307                               (1 << ACPI_FADT_F_PROC_C1) |
308                               (1 << ACPI_FADT_F_SLP_BUTTON) |
309                               (1 << ACPI_FADT_F_RTC_S4));
310     fadt->flags |= cpu_to_le32(1 << ACPI_FADT_F_USE_PLATFORM_CLOCK);
311     /* APIC destination mode ("Flat Logical") has an upper limit of 8 CPUs
312      * For more than 8 CPUs, "Clustered Logical" mode has to be used
313      */
314     if (max_cpus > 8) {
315         fadt->flags |= cpu_to_le32(1 << ACPI_FADT_F_FORCE_APIC_CLUSTER_MODEL);
316     }
317 }
318 
319 
320 /* FADT */
321 static void
322 build_fadt(GArray *table_data, GArray *linker, AcpiPmInfo *pm,
323            unsigned facs, unsigned dsdt)
324 {
325     AcpiFadtDescriptorRev1 *fadt = acpi_data_push(table_data, sizeof(*fadt));
326 
327     fadt->firmware_ctrl = cpu_to_le32(facs);
328     /* FACS address to be filled by Guest linker */
329     bios_linker_loader_add_pointer(linker, ACPI_BUILD_TABLE_FILE,
330                                    ACPI_BUILD_TABLE_FILE,
331                                    table_data, &fadt->firmware_ctrl,
332                                    sizeof fadt->firmware_ctrl);
333 
334     fadt->dsdt = cpu_to_le32(dsdt);
335     /* DSDT address to be filled by Guest linker */
336     bios_linker_loader_add_pointer(linker, ACPI_BUILD_TABLE_FILE,
337                                    ACPI_BUILD_TABLE_FILE,
338                                    table_data, &fadt->dsdt,
339                                    sizeof fadt->dsdt);
340 
341     fadt_setup(fadt, pm);
342 
343     build_header(linker, table_data,
344                  (void *)fadt, "FACP", sizeof(*fadt), 1);
345 }
346 
347 static void
348 build_madt(GArray *table_data, GArray *linker, AcpiCpuInfo *cpu,
349            PcGuestInfo *guest_info)
350 {
351     int madt_start = table_data->len;
352 
353     AcpiMultipleApicTable *madt;
354     AcpiMadtIoApic *io_apic;
355     AcpiMadtIntsrcovr *intsrcovr;
356     AcpiMadtLocalNmi *local_nmi;
357     int i;
358 
359     madt = acpi_data_push(table_data, sizeof *madt);
360     madt->local_apic_address = cpu_to_le32(APIC_DEFAULT_ADDRESS);
361     madt->flags = cpu_to_le32(1);
362 
363     for (i = 0; i < guest_info->apic_id_limit; i++) {
364         AcpiMadtProcessorApic *apic = acpi_data_push(table_data, sizeof *apic);
365         apic->type = ACPI_APIC_PROCESSOR;
366         apic->length = sizeof(*apic);
367         apic->processor_id = i;
368         apic->local_apic_id = i;
369         if (test_bit(i, cpu->found_cpus)) {
370             apic->flags = cpu_to_le32(1);
371         } else {
372             apic->flags = cpu_to_le32(0);
373         }
374     }
375     io_apic = acpi_data_push(table_data, sizeof *io_apic);
376     io_apic->type = ACPI_APIC_IO;
377     io_apic->length = sizeof(*io_apic);
378 #define ACPI_BUILD_IOAPIC_ID 0x0
379     io_apic->io_apic_id = ACPI_BUILD_IOAPIC_ID;
380     io_apic->address = cpu_to_le32(IO_APIC_DEFAULT_ADDRESS);
381     io_apic->interrupt = cpu_to_le32(0);
382 
383     if (guest_info->apic_xrupt_override) {
384         intsrcovr = acpi_data_push(table_data, sizeof *intsrcovr);
385         intsrcovr->type   = ACPI_APIC_XRUPT_OVERRIDE;
386         intsrcovr->length = sizeof(*intsrcovr);
387         intsrcovr->source = 0;
388         intsrcovr->gsi    = cpu_to_le32(2);
389         intsrcovr->flags  = cpu_to_le16(0); /* conforms to bus specifications */
390     }
391     for (i = 1; i < 16; i++) {
392 #define ACPI_BUILD_PCI_IRQS ((1<<5) | (1<<9) | (1<<10) | (1<<11))
393         if (!(ACPI_BUILD_PCI_IRQS & (1 << i))) {
394             /* No need for a INT source override structure. */
395             continue;
396         }
397         intsrcovr = acpi_data_push(table_data, sizeof *intsrcovr);
398         intsrcovr->type   = ACPI_APIC_XRUPT_OVERRIDE;
399         intsrcovr->length = sizeof(*intsrcovr);
400         intsrcovr->source = i;
401         intsrcovr->gsi    = cpu_to_le32(i);
402         intsrcovr->flags  = cpu_to_le16(0xd); /* active high, level triggered */
403     }
404 
405     local_nmi = acpi_data_push(table_data, sizeof *local_nmi);
406     local_nmi->type         = ACPI_APIC_LOCAL_NMI;
407     local_nmi->length       = sizeof(*local_nmi);
408     local_nmi->processor_id = 0xff; /* all processors */
409     local_nmi->flags        = cpu_to_le16(0);
410     local_nmi->lint         = 1; /* ACPI_LINT1 */
411 
412     build_header(linker, table_data,
413                  (void *)(table_data->data + madt_start), "APIC",
414                  table_data->len - madt_start, 1);
415 }
416 
417 #include "hw/i386/ssdt-tpm.hex"
418 #include "hw/i386/ssdt-tpm2.hex"
419 
420 /* Assign BSEL property to all buses.  In the future, this can be changed
421  * to only assign to buses that support hotplug.
422  */
423 static void *acpi_set_bsel(PCIBus *bus, void *opaque)
424 {
425     unsigned *bsel_alloc = opaque;
426     unsigned *bus_bsel;
427 
428     if (qbus_is_hotpluggable(BUS(bus))) {
429         bus_bsel = g_malloc(sizeof *bus_bsel);
430 
431         *bus_bsel = (*bsel_alloc)++;
432         object_property_add_uint32_ptr(OBJECT(bus), ACPI_PCIHP_PROP_BSEL,
433                                        bus_bsel, NULL);
434     }
435 
436     return bsel_alloc;
437 }
438 
439 static void acpi_set_pci_info(void)
440 {
441     PCIBus *bus = find_i440fx(); /* TODO: Q35 support */
442     unsigned bsel_alloc = 0;
443 
444     if (bus) {
445         /* Scan all PCI buses. Set property to enable acpi based hotplug. */
446         pci_for_each_bus_depth_first(bus, acpi_set_bsel, NULL, &bsel_alloc);
447     }
448 }
449 
450 static void build_append_pcihp_notify_entry(Aml *method, int slot)
451 {
452     Aml *if_ctx;
453     int32_t devfn = PCI_DEVFN(slot, 0);
454 
455     if_ctx = aml_if(aml_and(aml_arg(0), aml_int(0x1U << slot)));
456     aml_append(if_ctx, aml_notify(aml_name("S%.02X", devfn), aml_arg(1)));
457     aml_append(method, if_ctx);
458 }
459 
460 static void build_append_pci_bus_devices(Aml *parent_scope, PCIBus *bus,
461                                          bool pcihp_bridge_en)
462 {
463     Aml *dev, *notify_method, *method;
464     QObject *bsel;
465     PCIBus *sec;
466     int i;
467 
468     bsel = object_property_get_qobject(OBJECT(bus), ACPI_PCIHP_PROP_BSEL, NULL);
469     if (bsel) {
470         int64_t bsel_val = qint_get_int(qobject_to_qint(bsel));
471 
472         aml_append(parent_scope, aml_name_decl("BSEL", aml_int(bsel_val)));
473         notify_method = aml_method("DVNT", 2);
474     }
475 
476     for (i = 0; i < ARRAY_SIZE(bus->devices); i += PCI_FUNC_MAX) {
477         DeviceClass *dc;
478         PCIDeviceClass *pc;
479         PCIDevice *pdev = bus->devices[i];
480         int slot = PCI_SLOT(i);
481         bool hotplug_enabled_dev;
482         bool bridge_in_acpi;
483 
484         if (!pdev) {
485             if (bsel) { /* add hotplug slots for non present devices */
486                 dev = aml_device("S%.02X", PCI_DEVFN(slot, 0));
487                 aml_append(dev, aml_name_decl("_SUN", aml_int(slot)));
488                 aml_append(dev, aml_name_decl("_ADR", aml_int(slot << 16)));
489                 method = aml_method("_EJ0", 1);
490                 aml_append(method,
491                     aml_call2("PCEJ", aml_name("BSEL"), aml_name("_SUN"))
492                 );
493                 aml_append(dev, method);
494                 aml_append(parent_scope, dev);
495 
496                 build_append_pcihp_notify_entry(notify_method, slot);
497             }
498             continue;
499         }
500 
501         pc = PCI_DEVICE_GET_CLASS(pdev);
502         dc = DEVICE_GET_CLASS(pdev);
503 
504         /* When hotplug for bridges is enabled, bridges are
505          * described in ACPI separately (see build_pci_bus_end).
506          * In this case they aren't themselves hot-pluggable.
507          * Hotplugged bridges *are* hot-pluggable.
508          */
509         bridge_in_acpi = pc->is_bridge && pcihp_bridge_en &&
510             !DEVICE(pdev)->hotplugged;
511 
512         hotplug_enabled_dev = bsel && dc->hotpluggable && !bridge_in_acpi;
513 
514         if (pc->class_id == PCI_CLASS_BRIDGE_ISA) {
515             continue;
516         }
517 
518         /* start to compose PCI slot descriptor */
519         dev = aml_device("S%.02X", PCI_DEVFN(slot, 0));
520         aml_append(dev, aml_name_decl("_ADR", aml_int(slot << 16)));
521 
522         if (pc->class_id == PCI_CLASS_DISPLAY_VGA) {
523             /* add VGA specific AML methods */
524             int s3d;
525 
526             if (object_dynamic_cast(OBJECT(pdev), "qxl-vga")) {
527                 s3d = 3;
528             } else {
529                 s3d = 0;
530             }
531 
532             method = aml_method("_S1D", 0);
533             aml_append(method, aml_return(aml_int(0)));
534             aml_append(dev, method);
535 
536             method = aml_method("_S2D", 0);
537             aml_append(method, aml_return(aml_int(0)));
538             aml_append(dev, method);
539 
540             method = aml_method("_S3D", 0);
541             aml_append(method, aml_return(aml_int(s3d)));
542             aml_append(dev, method);
543         } else if (hotplug_enabled_dev) {
544             /* add _SUN/_EJ0 to make slot hotpluggable  */
545             aml_append(dev, aml_name_decl("_SUN", aml_int(slot)));
546 
547             method = aml_method("_EJ0", 1);
548             aml_append(method,
549                 aml_call2("PCEJ", aml_name("BSEL"), aml_name("_SUN"))
550             );
551             aml_append(dev, method);
552 
553             if (bsel) {
554                 build_append_pcihp_notify_entry(notify_method, slot);
555             }
556         } else if (bridge_in_acpi) {
557             /*
558              * device is coldplugged bridge,
559              * add child device descriptions into its scope
560              */
561             PCIBus *sec_bus = pci_bridge_get_sec_bus(PCI_BRIDGE(pdev));
562 
563             build_append_pci_bus_devices(dev, sec_bus, pcihp_bridge_en);
564         }
565         /* slot descriptor has been composed, add it into parent context */
566         aml_append(parent_scope, dev);
567     }
568 
569     if (bsel) {
570         aml_append(parent_scope, notify_method);
571     }
572 
573     /* Append PCNT method to notify about events on local and child buses.
574      * Add unconditionally for root since DSDT expects it.
575      */
576     method = aml_method("PCNT", 0);
577 
578     /* If bus supports hotplug select it and notify about local events */
579     if (bsel) {
580         int64_t bsel_val = qint_get_int(qobject_to_qint(bsel));
581         aml_append(method, aml_store(aml_int(bsel_val), aml_name("BNUM")));
582         aml_append(method,
583             aml_call2("DVNT", aml_name("PCIU"), aml_int(1) /* Device Check */)
584         );
585         aml_append(method,
586             aml_call2("DVNT", aml_name("PCID"), aml_int(3)/* Eject Request */)
587         );
588     }
589 
590     /* Notify about child bus events in any case */
591     if (pcihp_bridge_en) {
592         QLIST_FOREACH(sec, &bus->child, sibling) {
593             int32_t devfn = sec->parent_dev->devfn;
594 
595             aml_append(method, aml_name("^S%.02X.PCNT", devfn));
596         }
597     }
598     aml_append(parent_scope, method);
599     qobject_decref(bsel);
600 }
601 
602 static void
603 build_ssdt(GArray *table_data, GArray *linker,
604            AcpiCpuInfo *cpu, AcpiPmInfo *pm, AcpiMiscInfo *misc,
605            PcPciInfo *pci, PcGuestInfo *guest_info)
606 {
607     MachineState *machine = MACHINE(qdev_get_machine());
608     uint32_t nr_mem = machine->ram_slots;
609     unsigned acpi_cpus = guest_info->apic_id_limit;
610     Aml *ssdt, *sb_scope, *scope, *pkg, *dev, *method, *crs, *field, *ifctx;
611     int i;
612 
613     ssdt = init_aml_allocator();
614     /* The current AML generator can cover the APIC ID range [0..255],
615      * inclusive, for VCPU hotplug. */
616     QEMU_BUILD_BUG_ON(ACPI_CPU_HOTPLUG_ID_LIMIT > 256);
617     g_assert(acpi_cpus <= ACPI_CPU_HOTPLUG_ID_LIMIT);
618 
619     /* Reserve space for header */
620     acpi_data_push(ssdt->buf, sizeof(AcpiTableHeader));
621 
622     scope = aml_scope("\\_SB.PCI0");
623     /* build PCI0._CRS */
624     crs = aml_resource_template();
625     aml_append(crs,
626         aml_word_bus_number(AML_MIN_FIXED, AML_MAX_FIXED, AML_POS_DECODE,
627                             0x0000, 0x0000, 0x00FF, 0x0000, 0x0100));
628     aml_append(crs, aml_io(AML_DECODE16, 0x0CF8, 0x0CF8, 0x01, 0x08));
629 
630     aml_append(crs,
631         aml_word_io(AML_MIN_FIXED, AML_MAX_FIXED,
632                     AML_POS_DECODE, AML_ENTIRE_RANGE,
633                     0x0000, 0x0000, 0x0CF7, 0x0000, 0x0CF8));
634     aml_append(crs,
635         aml_word_io(AML_MIN_FIXED, AML_MAX_FIXED,
636                     AML_POS_DECODE, AML_ENTIRE_RANGE,
637                     0x0000, 0x0D00, 0xFFFF, 0x0000, 0xF300));
638     aml_append(crs,
639         aml_dword_memory(AML_POS_DECODE, AML_MIN_FIXED, AML_MAX_FIXED,
640                          AML_CACHEABLE, AML_READ_WRITE,
641                          0, 0x000A0000, 0x000BFFFF, 0, 0x00020000));
642     aml_append(crs,
643         aml_dword_memory(AML_POS_DECODE, AML_MIN_FIXED, AML_MAX_FIXED,
644                          AML_NON_CACHEABLE, AML_READ_WRITE,
645                          0, pci->w32.begin, pci->w32.end - 1, 0,
646                          pci->w32.end - pci->w32.begin));
647     if (pci->w64.begin) {
648         aml_append(crs,
649             aml_qword_memory(AML_POS_DECODE, AML_MIN_FIXED, AML_MAX_FIXED,
650                              AML_CACHEABLE, AML_READ_WRITE,
651                              0, pci->w64.begin, pci->w64.end - 1, 0,
652                              pci->w64.end - pci->w64.begin));
653     }
654     aml_append(scope, aml_name_decl("_CRS", crs));
655 
656     /* reserve GPE0 block resources */
657     dev = aml_device("GPE0");
658     aml_append(dev, aml_name_decl("_HID", aml_string("PNP0A06")));
659     aml_append(dev, aml_name_decl("_UID", aml_string("GPE0 resources")));
660     /* device present, functioning, decoding, not shown in UI */
661     aml_append(dev, aml_name_decl("_STA", aml_int(0xB)));
662     crs = aml_resource_template();
663     aml_append(crs,
664         aml_io(AML_DECODE16, pm->gpe0_blk, pm->gpe0_blk, 1, pm->gpe0_blk_len)
665     );
666     aml_append(dev, aml_name_decl("_CRS", crs));
667     aml_append(scope, dev);
668 
669     /* reserve PCIHP resources */
670     if (pm->pcihp_io_len) {
671         dev = aml_device("PHPR");
672         aml_append(dev, aml_name_decl("_HID", aml_string("PNP0A06")));
673         aml_append(dev,
674             aml_name_decl("_UID", aml_string("PCI Hotplug resources")));
675         /* device present, functioning, decoding, not shown in UI */
676         aml_append(dev, aml_name_decl("_STA", aml_int(0xB)));
677         crs = aml_resource_template();
678         aml_append(crs,
679             aml_io(AML_DECODE16, pm->pcihp_io_base, pm->pcihp_io_base, 1,
680                    pm->pcihp_io_len)
681         );
682         aml_append(dev, aml_name_decl("_CRS", crs));
683         aml_append(scope, dev);
684     }
685     aml_append(ssdt, scope);
686 
687     /*  create S3_ / S4_ / S5_ packages if necessary */
688     scope = aml_scope("\\");
689     if (!pm->s3_disabled) {
690         pkg = aml_package(4);
691         aml_append(pkg, aml_int(1)); /* PM1a_CNT.SLP_TYP */
692         aml_append(pkg, aml_int(1)); /* PM1b_CNT.SLP_TYP, FIXME: not impl. */
693         aml_append(pkg, aml_int(0)); /* reserved */
694         aml_append(pkg, aml_int(0)); /* reserved */
695         aml_append(scope, aml_name_decl("_S3", pkg));
696     }
697 
698     if (!pm->s4_disabled) {
699         pkg = aml_package(4);
700         aml_append(pkg, aml_int(pm->s4_val)); /* PM1a_CNT.SLP_TYP */
701         /* PM1b_CNT.SLP_TYP, FIXME: not impl. */
702         aml_append(pkg, aml_int(pm->s4_val));
703         aml_append(pkg, aml_int(0)); /* reserved */
704         aml_append(pkg, aml_int(0)); /* reserved */
705         aml_append(scope, aml_name_decl("_S4", pkg));
706     }
707 
708     pkg = aml_package(4);
709     aml_append(pkg, aml_int(0)); /* PM1a_CNT.SLP_TYP */
710     aml_append(pkg, aml_int(0)); /* PM1b_CNT.SLP_TYP not impl. */
711     aml_append(pkg, aml_int(0)); /* reserved */
712     aml_append(pkg, aml_int(0)); /* reserved */
713     aml_append(scope, aml_name_decl("_S5", pkg));
714     aml_append(ssdt, scope);
715 
716     if (misc->applesmc_io_base) {
717         scope = aml_scope("\\_SB.PCI0.ISA");
718         dev = aml_device("SMC");
719 
720         aml_append(dev, aml_name_decl("_HID", aml_eisaid("APP0001")));
721         /* device present, functioning, decoding, not shown in UI */
722         aml_append(dev, aml_name_decl("_STA", aml_int(0xB)));
723 
724         crs = aml_resource_template();
725         aml_append(crs,
726             aml_io(AML_DECODE16, misc->applesmc_io_base, misc->applesmc_io_base,
727                    0x01, APPLESMC_MAX_DATA_LENGTH)
728         );
729         aml_append(crs, aml_irq_no_flags(6));
730         aml_append(dev, aml_name_decl("_CRS", crs));
731 
732         aml_append(scope, dev);
733         aml_append(ssdt, scope);
734     }
735 
736     if (misc->pvpanic_port) {
737         scope = aml_scope("\\_SB.PCI0.ISA");
738 
739         dev = aml_device("PEVT");
740         aml_append(dev, aml_name_decl("_HID", aml_string("QEMU0001")));
741 
742         crs = aml_resource_template();
743         aml_append(crs,
744             aml_io(AML_DECODE16, misc->pvpanic_port, misc->pvpanic_port, 1, 1)
745         );
746         aml_append(dev, aml_name_decl("_CRS", crs));
747 
748         aml_append(dev, aml_operation_region("PEOR", AML_SYSTEM_IO,
749                                               misc->pvpanic_port, 1));
750         field = aml_field("PEOR", AML_BYTE_ACC, AML_PRESERVE);
751         aml_append(field, aml_named_field("PEPT", 8));
752         aml_append(dev, field);
753 
754         /* device present, functioning, decoding, not shown in UI */
755         aml_append(dev, aml_name_decl("_STA", aml_int(0xB)));
756 
757         method = aml_method("RDPT", 0);
758         aml_append(method, aml_store(aml_name("PEPT"), aml_local(0)));
759         aml_append(method, aml_return(aml_local(0)));
760         aml_append(dev, method);
761 
762         method = aml_method("WRPT", 1);
763         aml_append(method, aml_store(aml_arg(0), aml_name("PEPT")));
764         aml_append(dev, method);
765 
766         aml_append(scope, dev);
767         aml_append(ssdt, scope);
768     }
769 
770     sb_scope = aml_scope("\\_SB");
771     {
772         /* create PCI0.PRES device and its _CRS to reserve CPU hotplug MMIO */
773         dev = aml_device("PCI0." stringify(CPU_HOTPLUG_RESOURCE_DEVICE));
774         aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0A06")));
775         aml_append(dev,
776             aml_name_decl("_UID", aml_string("CPU Hotplug resources"))
777         );
778         /* device present, functioning, decoding, not shown in UI */
779         aml_append(dev, aml_name_decl("_STA", aml_int(0xB)));
780         crs = aml_resource_template();
781         aml_append(crs,
782             aml_io(AML_DECODE16, pm->cpu_hp_io_base, pm->cpu_hp_io_base, 1,
783                    pm->cpu_hp_io_len)
784         );
785         aml_append(dev, aml_name_decl("_CRS", crs));
786         aml_append(sb_scope, dev);
787         /* declare CPU hotplug MMIO region and PRS field to access it */
788         aml_append(sb_scope, aml_operation_region(
789             "PRST", AML_SYSTEM_IO, pm->cpu_hp_io_base, pm->cpu_hp_io_len));
790         field = aml_field("PRST", AML_BYTE_ACC, AML_PRESERVE);
791         aml_append(field, aml_named_field("PRS", 256));
792         aml_append(sb_scope, field);
793 
794         /* build Processor object for each processor */
795         for (i = 0; i < acpi_cpus; i++) {
796             dev = aml_processor(i, 0, 0, "CP%.02X", i);
797 
798             method = aml_method("_MAT", 0);
799             aml_append(method, aml_return(aml_call1("CPMA", aml_int(i))));
800             aml_append(dev, method);
801 
802             method = aml_method("_STA", 0);
803             aml_append(method, aml_return(aml_call1("CPST", aml_int(i))));
804             aml_append(dev, method);
805 
806             method = aml_method("_EJ0", 1);
807             aml_append(method,
808                 aml_return(aml_call2("CPEJ", aml_int(i), aml_arg(0)))
809             );
810             aml_append(dev, method);
811 
812             aml_append(sb_scope, dev);
813         }
814 
815         /* build this code:
816          *   Method(NTFY, 2) {If (LEqual(Arg0, 0x00)) {Notify(CP00, Arg1)} ...}
817          */
818         /* Arg0 = Processor ID = APIC ID */
819         method = aml_method("NTFY", 2);
820         for (i = 0; i < acpi_cpus; i++) {
821             ifctx = aml_if(aml_equal(aml_arg(0), aml_int(i)));
822             aml_append(ifctx,
823                 aml_notify(aml_name("CP%.02X", i), aml_arg(1))
824             );
825             aml_append(method, ifctx);
826         }
827         aml_append(sb_scope, method);
828 
829         /* build "Name(CPON, Package() { One, One, ..., Zero, Zero, ... })"
830          *
831          * Note: The ability to create variable-sized packages was first
832          * introduced in ACPI 2.0. ACPI 1.0 only allowed fixed-size packages
833          * ith up to 255 elements. Windows guests up to win2k8 fail when
834          * VarPackageOp is used.
835          */
836         pkg = acpi_cpus <= 255 ? aml_package(acpi_cpus) :
837                                  aml_varpackage(acpi_cpus);
838 
839         for (i = 0; i < acpi_cpus; i++) {
840             uint8_t b = test_bit(i, cpu->found_cpus) ? 0x01 : 0x00;
841             aml_append(pkg, aml_int(b));
842         }
843         aml_append(sb_scope, aml_name_decl("CPON", pkg));
844 
845         /* build memory devices */
846         assert(nr_mem <= ACPI_MAX_RAM_SLOTS);
847         scope = aml_scope("\\_SB.PCI0." stringify(MEMORY_HOTPLUG_DEVICE));
848         aml_append(scope,
849             aml_name_decl(stringify(MEMORY_SLOTS_NUMBER), aml_int(nr_mem))
850         );
851 
852         crs = aml_resource_template();
853         aml_append(crs,
854             aml_io(AML_DECODE16, pm->mem_hp_io_base, pm->mem_hp_io_base, 0,
855                    pm->mem_hp_io_len)
856         );
857         aml_append(scope, aml_name_decl("_CRS", crs));
858 
859         aml_append(scope, aml_operation_region(
860             stringify(MEMORY_HOTPLUG_IO_REGION), AML_SYSTEM_IO,
861             pm->mem_hp_io_base, pm->mem_hp_io_len)
862         );
863 
864         field = aml_field(stringify(MEMORY_HOTPLUG_IO_REGION), AML_DWORD_ACC,
865                           AML_PRESERVE);
866         aml_append(field, /* read only */
867             aml_named_field(stringify(MEMORY_SLOT_ADDR_LOW), 32));
868         aml_append(field, /* read only */
869             aml_named_field(stringify(MEMORY_SLOT_ADDR_HIGH), 32));
870         aml_append(field, /* read only */
871             aml_named_field(stringify(MEMORY_SLOT_SIZE_LOW), 32));
872         aml_append(field, /* read only */
873             aml_named_field(stringify(MEMORY_SLOT_SIZE_HIGH), 32));
874         aml_append(field, /* read only */
875             aml_named_field(stringify(MEMORY_SLOT_PROXIMITY), 32));
876         aml_append(scope, field);
877 
878         field = aml_field(stringify(MEMORY_HOTPLUG_IO_REGION), AML_BYTE_ACC,
879                           AML_WRITE_AS_ZEROS);
880         aml_append(field, aml_reserved_field(160 /* bits, Offset(20) */));
881         aml_append(field, /* 1 if enabled, read only */
882             aml_named_field(stringify(MEMORY_SLOT_ENABLED), 1));
883         aml_append(field,
884             /*(read) 1 if has a insert event. (write) 1 to clear event */
885             aml_named_field(stringify(MEMORY_SLOT_INSERT_EVENT), 1));
886         aml_append(field,
887             /* (read) 1 if has a remove event. (write) 1 to clear event */
888             aml_named_field(stringify(MEMORY_SLOT_REMOVE_EVENT), 1));
889         aml_append(field,
890             /* initiates device eject, write only */
891             aml_named_field(stringify(MEMORY_SLOT_EJECT), 1));
892         aml_append(scope, field);
893 
894         field = aml_field(stringify(MEMORY_HOTPLUG_IO_REGION), AML_DWORD_ACC,
895                           AML_PRESERVE);
896         aml_append(field, /* DIMM selector, write only */
897             aml_named_field(stringify(MEMORY_SLOT_SLECTOR), 32));
898         aml_append(field, /* _OST event code, write only */
899             aml_named_field(stringify(MEMORY_SLOT_OST_EVENT), 32));
900         aml_append(field, /* _OST status code, write only */
901             aml_named_field(stringify(MEMORY_SLOT_OST_STATUS), 32));
902         aml_append(scope, field);
903 
904         aml_append(sb_scope, scope);
905 
906         for (i = 0; i < nr_mem; i++) {
907             #define BASEPATH "\\_SB.PCI0." stringify(MEMORY_HOTPLUG_DEVICE) "."
908             const char *s;
909 
910             dev = aml_device("MP%02X", i);
911             aml_append(dev, aml_name_decl("_UID", aml_string("0x%02X", i)));
912             aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0C80")));
913 
914             method = aml_method("_CRS", 0);
915             s = BASEPATH stringify(MEMORY_SLOT_CRS_METHOD);
916             aml_append(method, aml_return(aml_call1(s, aml_name("_UID"))));
917             aml_append(dev, method);
918 
919             method = aml_method("_STA", 0);
920             s = BASEPATH stringify(MEMORY_SLOT_STATUS_METHOD);
921             aml_append(method, aml_return(aml_call1(s, aml_name("_UID"))));
922             aml_append(dev, method);
923 
924             method = aml_method("_PXM", 0);
925             s = BASEPATH stringify(MEMORY_SLOT_PROXIMITY_METHOD);
926             aml_append(method, aml_return(aml_call1(s, aml_name("_UID"))));
927             aml_append(dev, method);
928 
929             method = aml_method("_OST", 3);
930             s = BASEPATH stringify(MEMORY_SLOT_OST_METHOD);
931             aml_append(method, aml_return(aml_call4(
932                 s, aml_name("_UID"), aml_arg(0), aml_arg(1), aml_arg(2)
933             )));
934             aml_append(dev, method);
935 
936             method = aml_method("_EJ0", 1);
937             s = BASEPATH stringify(MEMORY_SLOT_EJECT_METHOD);
938             aml_append(method, aml_return(aml_call2(
939                        s, aml_name("_UID"), aml_arg(0))));
940             aml_append(dev, method);
941 
942             aml_append(sb_scope, dev);
943         }
944 
945         /* build Method(MEMORY_SLOT_NOTIFY_METHOD, 2) {
946          *     If (LEqual(Arg0, 0x00)) {Notify(MP00, Arg1)} ... }
947          */
948         method = aml_method(stringify(MEMORY_SLOT_NOTIFY_METHOD), 2);
949         for (i = 0; i < nr_mem; i++) {
950             ifctx = aml_if(aml_equal(aml_arg(0), aml_int(i)));
951             aml_append(ifctx,
952                 aml_notify(aml_name("MP%.02X", i), aml_arg(1))
953             );
954             aml_append(method, ifctx);
955         }
956         aml_append(sb_scope, method);
957 
958         {
959             Object *pci_host;
960             PCIBus *bus = NULL;
961             bool ambiguous;
962 
963             pci_host = object_resolve_path_type("", TYPE_PCI_HOST_BRIDGE, &ambiguous);
964             if (!ambiguous && pci_host) {
965                 bus = PCI_HOST_BRIDGE(pci_host)->bus;
966             }
967 
968             if (bus) {
969                 Aml *scope = aml_scope("PCI0");
970                 /* Scan all PCI buses. Generate tables to support hotplug. */
971                 build_append_pci_bus_devices(scope, bus, pm->pcihp_bridge_en);
972                 aml_append(sb_scope, scope);
973             }
974         }
975         aml_append(ssdt, sb_scope);
976     }
977 
978     /* copy AML table into ACPI tables blob and patch header there */
979     g_array_append_vals(table_data, ssdt->buf->data, ssdt->buf->len);
980     build_header(linker, table_data,
981         (void *)(table_data->data + table_data->len - ssdt->buf->len),
982         "SSDT", ssdt->buf->len, 1);
983     free_aml_allocator();
984 }
985 
986 static void
987 build_hpet(GArray *table_data, GArray *linker)
988 {
989     Acpi20Hpet *hpet;
990 
991     hpet = acpi_data_push(table_data, sizeof(*hpet));
992     /* Note timer_block_id value must be kept in sync with value advertised by
993      * emulated hpet
994      */
995     hpet->timer_block_id = cpu_to_le32(0x8086a201);
996     hpet->addr.address = cpu_to_le64(HPET_BASE);
997     build_header(linker, table_data,
998                  (void *)hpet, "HPET", sizeof(*hpet), 1);
999 }
1000 
1001 static void
1002 build_tpm_tcpa(GArray *table_data, GArray *linker, GArray *tcpalog)
1003 {
1004     Acpi20Tcpa *tcpa = acpi_data_push(table_data, sizeof *tcpa);
1005     uint64_t log_area_start_address = acpi_data_len(tcpalog);
1006 
1007     tcpa->platform_class = cpu_to_le16(TPM_TCPA_ACPI_CLASS_CLIENT);
1008     tcpa->log_area_minimum_length = cpu_to_le32(TPM_LOG_AREA_MINIMUM_SIZE);
1009     tcpa->log_area_start_address = cpu_to_le64(log_area_start_address);
1010 
1011     bios_linker_loader_alloc(linker, ACPI_BUILD_TPMLOG_FILE, 1,
1012                              false /* high memory */);
1013 
1014     /* log area start address to be filled by Guest linker */
1015     bios_linker_loader_add_pointer(linker, ACPI_BUILD_TABLE_FILE,
1016                                    ACPI_BUILD_TPMLOG_FILE,
1017                                    table_data, &tcpa->log_area_start_address,
1018                                    sizeof(tcpa->log_area_start_address));
1019 
1020     build_header(linker, table_data,
1021                  (void *)tcpa, "TCPA", sizeof(*tcpa), 2);
1022 
1023     acpi_data_push(tcpalog, TPM_LOG_AREA_MINIMUM_SIZE);
1024 }
1025 
1026 static void
1027 build_tpm_ssdt(GArray *table_data, GArray *linker)
1028 {
1029     void *tpm_ptr;
1030 
1031     tpm_ptr = acpi_data_push(table_data, sizeof(ssdt_tpm_aml));
1032     memcpy(tpm_ptr, ssdt_tpm_aml, sizeof(ssdt_tpm_aml));
1033 }
1034 
1035 static void
1036 build_tpm2(GArray *table_data, GArray *linker)
1037 {
1038     Acpi20TPM2 *tpm2_ptr;
1039     void *tpm_ptr;
1040 
1041     tpm_ptr = acpi_data_push(table_data, sizeof(ssdt_tpm2_aml));
1042     memcpy(tpm_ptr, ssdt_tpm2_aml, sizeof(ssdt_tpm2_aml));
1043 
1044     tpm2_ptr = acpi_data_push(table_data, sizeof *tpm2_ptr);
1045 
1046     tpm2_ptr->platform_class = cpu_to_le16(TPM2_ACPI_CLASS_CLIENT);
1047     tpm2_ptr->control_area_address = cpu_to_le64(0);
1048     tpm2_ptr->start_method = cpu_to_le32(TPM2_START_METHOD_MMIO);
1049 
1050     build_header(linker, table_data,
1051                  (void *)tpm2_ptr, "TPM2", sizeof(*tpm2_ptr), 4);
1052 }
1053 
1054 typedef enum {
1055     MEM_AFFINITY_NOFLAGS      = 0,
1056     MEM_AFFINITY_ENABLED      = (1 << 0),
1057     MEM_AFFINITY_HOTPLUGGABLE = (1 << 1),
1058     MEM_AFFINITY_NON_VOLATILE = (1 << 2),
1059 } MemoryAffinityFlags;
1060 
1061 static void
1062 acpi_build_srat_memory(AcpiSratMemoryAffinity *numamem, uint64_t base,
1063                        uint64_t len, int node, MemoryAffinityFlags flags)
1064 {
1065     numamem->type = ACPI_SRAT_MEMORY;
1066     numamem->length = sizeof(*numamem);
1067     memset(numamem->proximity, 0, 4);
1068     numamem->proximity[0] = node;
1069     numamem->flags = cpu_to_le32(flags);
1070     numamem->base_addr = cpu_to_le64(base);
1071     numamem->range_length = cpu_to_le64(len);
1072 }
1073 
1074 static void
1075 build_srat(GArray *table_data, GArray *linker, PcGuestInfo *guest_info)
1076 {
1077     AcpiSystemResourceAffinityTable *srat;
1078     AcpiSratProcessorAffinity *core;
1079     AcpiSratMemoryAffinity *numamem;
1080 
1081     int i;
1082     uint64_t curnode;
1083     int srat_start, numa_start, slots;
1084     uint64_t mem_len, mem_base, next_base;
1085     PCMachineState *pcms = PC_MACHINE(qdev_get_machine());
1086     ram_addr_t hotplugabble_address_space_size =
1087         object_property_get_int(OBJECT(pcms), PC_MACHINE_MEMHP_REGION_SIZE,
1088                                 NULL);
1089 
1090     srat_start = table_data->len;
1091 
1092     srat = acpi_data_push(table_data, sizeof *srat);
1093     srat->reserved1 = cpu_to_le32(1);
1094     core = (void *)(srat + 1);
1095 
1096     for (i = 0; i < guest_info->apic_id_limit; ++i) {
1097         core = acpi_data_push(table_data, sizeof *core);
1098         core->type = ACPI_SRAT_PROCESSOR;
1099         core->length = sizeof(*core);
1100         core->local_apic_id = i;
1101         curnode = guest_info->node_cpu[i];
1102         core->proximity_lo = curnode;
1103         memset(core->proximity_hi, 0, 3);
1104         core->local_sapic_eid = 0;
1105         core->flags = cpu_to_le32(1);
1106     }
1107 
1108 
1109     /* the memory map is a bit tricky, it contains at least one hole
1110      * from 640k-1M and possibly another one from 3.5G-4G.
1111      */
1112     next_base = 0;
1113     numa_start = table_data->len;
1114 
1115     numamem = acpi_data_push(table_data, sizeof *numamem);
1116     acpi_build_srat_memory(numamem, 0, 640*1024, 0, MEM_AFFINITY_ENABLED);
1117     next_base = 1024 * 1024;
1118     for (i = 1; i < guest_info->numa_nodes + 1; ++i) {
1119         mem_base = next_base;
1120         mem_len = guest_info->node_mem[i - 1];
1121         if (i == 1) {
1122             mem_len -= 1024 * 1024;
1123         }
1124         next_base = mem_base + mem_len;
1125 
1126         /* Cut out the ACPI_PCI hole */
1127         if (mem_base <= guest_info->ram_size_below_4g &&
1128             next_base > guest_info->ram_size_below_4g) {
1129             mem_len -= next_base - guest_info->ram_size_below_4g;
1130             if (mem_len > 0) {
1131                 numamem = acpi_data_push(table_data, sizeof *numamem);
1132                 acpi_build_srat_memory(numamem, mem_base, mem_len, i - 1,
1133                                        MEM_AFFINITY_ENABLED);
1134             }
1135             mem_base = 1ULL << 32;
1136             mem_len = next_base - guest_info->ram_size_below_4g;
1137             next_base += (1ULL << 32) - guest_info->ram_size_below_4g;
1138         }
1139         numamem = acpi_data_push(table_data, sizeof *numamem);
1140         acpi_build_srat_memory(numamem, mem_base, mem_len, i - 1,
1141                                MEM_AFFINITY_ENABLED);
1142     }
1143     slots = (table_data->len - numa_start) / sizeof *numamem;
1144     for (; slots < guest_info->numa_nodes + 2; slots++) {
1145         numamem = acpi_data_push(table_data, sizeof *numamem);
1146         acpi_build_srat_memory(numamem, 0, 0, 0, MEM_AFFINITY_NOFLAGS);
1147     }
1148 
1149     /*
1150      * Entry is required for Windows to enable memory hotplug in OS.
1151      * Memory devices may override proximity set by this entry,
1152      * providing _PXM method if necessary.
1153      */
1154     if (hotplugabble_address_space_size) {
1155         numamem = acpi_data_push(table_data, sizeof *numamem);
1156         acpi_build_srat_memory(numamem, pcms->hotplug_memory_base,
1157                                hotplugabble_address_space_size, 0,
1158                                MEM_AFFINITY_HOTPLUGGABLE |
1159                                MEM_AFFINITY_ENABLED);
1160     }
1161 
1162     build_header(linker, table_data,
1163                  (void *)(table_data->data + srat_start),
1164                  "SRAT",
1165                  table_data->len - srat_start, 1);
1166 }
1167 
1168 static void
1169 build_mcfg_q35(GArray *table_data, GArray *linker, AcpiMcfgInfo *info)
1170 {
1171     AcpiTableMcfg *mcfg;
1172     const char *sig;
1173     int len = sizeof(*mcfg) + 1 * sizeof(mcfg->allocation[0]);
1174 
1175     mcfg = acpi_data_push(table_data, len);
1176     mcfg->allocation[0].address = cpu_to_le64(info->mcfg_base);
1177     /* Only a single allocation so no need to play with segments */
1178     mcfg->allocation[0].pci_segment = cpu_to_le16(0);
1179     mcfg->allocation[0].start_bus_number = 0;
1180     mcfg->allocation[0].end_bus_number = PCIE_MMCFG_BUS(info->mcfg_size - 1);
1181 
1182     /* MCFG is used for ECAM which can be enabled or disabled by guest.
1183      * To avoid table size changes (which create migration issues),
1184      * always create the table even if there are no allocations,
1185      * but set the signature to a reserved value in this case.
1186      * ACPI spec requires OSPMs to ignore such tables.
1187      */
1188     if (info->mcfg_base == PCIE_BASE_ADDR_UNMAPPED) {
1189         /* Reserved signature: ignored by OSPM */
1190         sig = "QEMU";
1191     } else {
1192         sig = "MCFG";
1193     }
1194     build_header(linker, table_data, (void *)mcfg, sig, len, 1);
1195 }
1196 
1197 static void
1198 build_dmar_q35(GArray *table_data, GArray *linker)
1199 {
1200     int dmar_start = table_data->len;
1201 
1202     AcpiTableDmar *dmar;
1203     AcpiDmarHardwareUnit *drhd;
1204 
1205     dmar = acpi_data_push(table_data, sizeof(*dmar));
1206     dmar->host_address_width = VTD_HOST_ADDRESS_WIDTH - 1;
1207     dmar->flags = 0;    /* No intr_remap for now */
1208 
1209     /* DMAR Remapping Hardware Unit Definition structure */
1210     drhd = acpi_data_push(table_data, sizeof(*drhd));
1211     drhd->type = cpu_to_le16(ACPI_DMAR_TYPE_HARDWARE_UNIT);
1212     drhd->length = cpu_to_le16(sizeof(*drhd));   /* No device scope now */
1213     drhd->flags = ACPI_DMAR_INCLUDE_PCI_ALL;
1214     drhd->pci_segment = cpu_to_le16(0);
1215     drhd->address = cpu_to_le64(Q35_HOST_BRIDGE_IOMMU_ADDR);
1216 
1217     build_header(linker, table_data, (void *)(table_data->data + dmar_start),
1218                  "DMAR", table_data->len - dmar_start, 1);
1219 }
1220 
1221 static void
1222 build_dsdt(GArray *table_data, GArray *linker, AcpiMiscInfo *misc)
1223 {
1224     AcpiTableHeader *dsdt;
1225 
1226     assert(misc->dsdt_code && misc->dsdt_size);
1227 
1228     dsdt = acpi_data_push(table_data, misc->dsdt_size);
1229     memcpy(dsdt, misc->dsdt_code, misc->dsdt_size);
1230 
1231     memset(dsdt, 0, sizeof *dsdt);
1232     build_header(linker, table_data, dsdt, "DSDT",
1233                  misc->dsdt_size, 1);
1234 }
1235 
1236 static GArray *
1237 build_rsdp(GArray *rsdp_table, GArray *linker, unsigned rsdt)
1238 {
1239     AcpiRsdpDescriptor *rsdp = acpi_data_push(rsdp_table, sizeof *rsdp);
1240 
1241     bios_linker_loader_alloc(linker, ACPI_BUILD_RSDP_FILE, 16,
1242                              true /* fseg memory */);
1243 
1244     memcpy(&rsdp->signature, "RSD PTR ", 8);
1245     memcpy(rsdp->oem_id, ACPI_BUILD_APPNAME6, 6);
1246     rsdp->rsdt_physical_address = cpu_to_le32(rsdt);
1247     /* Address to be filled by Guest linker */
1248     bios_linker_loader_add_pointer(linker, ACPI_BUILD_RSDP_FILE,
1249                                    ACPI_BUILD_TABLE_FILE,
1250                                    rsdp_table, &rsdp->rsdt_physical_address,
1251                                    sizeof rsdp->rsdt_physical_address);
1252     rsdp->checksum = 0;
1253     /* Checksum to be filled by Guest linker */
1254     bios_linker_loader_add_checksum(linker, ACPI_BUILD_RSDP_FILE,
1255                                     rsdp, rsdp, sizeof *rsdp, &rsdp->checksum);
1256 
1257     return rsdp_table;
1258 }
1259 
1260 typedef
1261 struct AcpiBuildState {
1262     /* Copy of table in RAM (for patching). */
1263     MemoryRegion *table_mr;
1264     /* Is table patched? */
1265     uint8_t patched;
1266     PcGuestInfo *guest_info;
1267     void *rsdp;
1268     MemoryRegion *rsdp_mr;
1269     MemoryRegion *linker_mr;
1270 } AcpiBuildState;
1271 
1272 static bool acpi_get_mcfg(AcpiMcfgInfo *mcfg)
1273 {
1274     Object *pci_host;
1275     QObject *o;
1276     bool ambiguous;
1277 
1278     pci_host = object_resolve_path_type("", TYPE_PCI_HOST_BRIDGE, &ambiguous);
1279     g_assert(!ambiguous);
1280     g_assert(pci_host);
1281 
1282     o = object_property_get_qobject(pci_host, PCIE_HOST_MCFG_BASE, NULL);
1283     if (!o) {
1284         return false;
1285     }
1286     mcfg->mcfg_base = qint_get_int(qobject_to_qint(o));
1287     qobject_decref(o);
1288 
1289     o = object_property_get_qobject(pci_host, PCIE_HOST_MCFG_SIZE, NULL);
1290     assert(o);
1291     mcfg->mcfg_size = qint_get_int(qobject_to_qint(o));
1292     qobject_decref(o);
1293     return true;
1294 }
1295 
1296 static bool acpi_has_iommu(void)
1297 {
1298     bool ambiguous;
1299     Object *intel_iommu;
1300 
1301     intel_iommu = object_resolve_path_type("", TYPE_INTEL_IOMMU_DEVICE,
1302                                            &ambiguous);
1303     return intel_iommu && !ambiguous;
1304 }
1305 
1306 static
1307 void acpi_build(PcGuestInfo *guest_info, AcpiBuildTables *tables)
1308 {
1309     GArray *table_offsets;
1310     unsigned facs, ssdt, dsdt, rsdt;
1311     AcpiCpuInfo cpu;
1312     AcpiPmInfo pm;
1313     AcpiMiscInfo misc;
1314     AcpiMcfgInfo mcfg;
1315     PcPciInfo pci;
1316     uint8_t *u;
1317     size_t aml_len = 0;
1318     GArray *tables_blob = tables->table_data;
1319 
1320     acpi_get_cpu_info(&cpu);
1321     acpi_get_pm_info(&pm);
1322     acpi_get_dsdt(&misc);
1323     acpi_get_misc_info(&misc);
1324     acpi_get_pci_info(&pci);
1325 
1326     table_offsets = g_array_new(false, true /* clear */,
1327                                         sizeof(uint32_t));
1328     ACPI_BUILD_DPRINTF("init ACPI tables\n");
1329 
1330     bios_linker_loader_alloc(tables->linker, ACPI_BUILD_TABLE_FILE,
1331                              64 /* Ensure FACS is aligned */,
1332                              false /* high memory */);
1333 
1334     /*
1335      * FACS is pointed to by FADT.
1336      * We place it first since it's the only table that has alignment
1337      * requirements.
1338      */
1339     facs = tables_blob->len;
1340     build_facs(tables_blob, tables->linker, guest_info);
1341 
1342     /* DSDT is pointed to by FADT */
1343     dsdt = tables_blob->len;
1344     build_dsdt(tables_blob, tables->linker, &misc);
1345 
1346     /* Count the size of the DSDT and SSDT, we will need it for legacy
1347      * sizing of ACPI tables.
1348      */
1349     aml_len += tables_blob->len - dsdt;
1350 
1351     /* ACPI tables pointed to by RSDT */
1352     acpi_add_table(table_offsets, tables_blob);
1353     build_fadt(tables_blob, tables->linker, &pm, facs, dsdt);
1354 
1355     ssdt = tables_blob->len;
1356     acpi_add_table(table_offsets, tables_blob);
1357     build_ssdt(tables_blob, tables->linker, &cpu, &pm, &misc, &pci,
1358                guest_info);
1359     aml_len += tables_blob->len - ssdt;
1360 
1361     acpi_add_table(table_offsets, tables_blob);
1362     build_madt(tables_blob, tables->linker, &cpu, guest_info);
1363 
1364     if (misc.has_hpet) {
1365         acpi_add_table(table_offsets, tables_blob);
1366         build_hpet(tables_blob, tables->linker);
1367     }
1368     if (misc.tpm_version != TPM_VERSION_UNSPEC) {
1369         acpi_add_table(table_offsets, tables_blob);
1370         build_tpm_tcpa(tables_blob, tables->linker, tables->tcpalog);
1371 
1372         acpi_add_table(table_offsets, tables_blob);
1373         switch (misc.tpm_version) {
1374         case TPM_VERSION_1_2:
1375             build_tpm_ssdt(tables_blob, tables->linker);
1376             break;
1377         case TPM_VERSION_2_0:
1378             build_tpm2(tables_blob, tables->linker);
1379             break;
1380         default:
1381             assert(false);
1382         }
1383     }
1384     if (guest_info->numa_nodes) {
1385         acpi_add_table(table_offsets, tables_blob);
1386         build_srat(tables_blob, tables->linker, guest_info);
1387     }
1388     if (acpi_get_mcfg(&mcfg)) {
1389         acpi_add_table(table_offsets, tables_blob);
1390         build_mcfg_q35(tables_blob, tables->linker, &mcfg);
1391     }
1392     if (acpi_has_iommu()) {
1393         acpi_add_table(table_offsets, tables_blob);
1394         build_dmar_q35(tables_blob, tables->linker);
1395     }
1396 
1397     /* Add tables supplied by user (if any) */
1398     for (u = acpi_table_first(); u; u = acpi_table_next(u)) {
1399         unsigned len = acpi_table_len(u);
1400 
1401         acpi_add_table(table_offsets, tables_blob);
1402         g_array_append_vals(tables_blob, u, len);
1403     }
1404 
1405     /* RSDT is pointed to by RSDP */
1406     rsdt = tables_blob->len;
1407     build_rsdt(tables_blob, tables->linker, table_offsets);
1408 
1409     /* RSDP is in FSEG memory, so allocate it separately */
1410     build_rsdp(tables->rsdp, tables->linker, rsdt);
1411 
1412     /* We'll expose it all to Guest so we want to reduce
1413      * chance of size changes.
1414      *
1415      * We used to align the tables to 4k, but of course this would
1416      * too simple to be enough.  4k turned out to be too small an
1417      * alignment very soon, and in fact it is almost impossible to
1418      * keep the table size stable for all (max_cpus, max_memory_slots)
1419      * combinations.  So the table size is always 64k for pc-i440fx-2.1
1420      * and we give an error if the table grows beyond that limit.
1421      *
1422      * We still have the problem of migrating from "-M pc-i440fx-2.0".  For
1423      * that, we exploit the fact that QEMU 2.1 generates _smaller_ tables
1424      * than 2.0 and we can always pad the smaller tables with zeros.  We can
1425      * then use the exact size of the 2.0 tables.
1426      *
1427      * All this is for PIIX4, since QEMU 2.0 didn't support Q35 migration.
1428      */
1429     if (guest_info->legacy_acpi_table_size) {
1430         /* Subtracting aml_len gives the size of fixed tables.  Then add the
1431          * size of the PIIX4 DSDT/SSDT in QEMU 2.0.
1432          */
1433         int legacy_aml_len =
1434             guest_info->legacy_acpi_table_size +
1435             ACPI_BUILD_LEGACY_CPU_AML_SIZE * max_cpus;
1436         int legacy_table_size =
1437             ROUND_UP(tables_blob->len - aml_len + legacy_aml_len,
1438                      ACPI_BUILD_ALIGN_SIZE);
1439         if (tables_blob->len > legacy_table_size) {
1440             /* Should happen only with PCI bridges and -M pc-i440fx-2.0.  */
1441             error_report("Warning: migration may not work.");
1442         }
1443         g_array_set_size(tables_blob, legacy_table_size);
1444     } else {
1445         /* Make sure we have a buffer in case we need to resize the tables. */
1446         if (tables_blob->len > ACPI_BUILD_TABLE_SIZE / 2) {
1447             /* As of QEMU 2.1, this fires with 160 VCPUs and 255 memory slots.  */
1448             error_report("Warning: ACPI tables are larger than 64k.");
1449             error_report("Warning: migration may not work.");
1450             error_report("Warning: please remove CPUs, NUMA nodes, "
1451                          "memory slots or PCI bridges.");
1452         }
1453         acpi_align_size(tables_blob, ACPI_BUILD_TABLE_SIZE);
1454     }
1455 
1456     acpi_align_size(tables->linker, ACPI_BUILD_ALIGN_SIZE);
1457 
1458     /* Cleanup memory that's no longer used. */
1459     g_array_free(table_offsets, true);
1460 }
1461 
1462 static void acpi_ram_update(MemoryRegion *mr, GArray *data)
1463 {
1464     uint32_t size = acpi_data_len(data);
1465 
1466     /* Make sure RAM size is correct - in case it got changed e.g. by migration */
1467     memory_region_ram_resize(mr, size, &error_abort);
1468 
1469     memcpy(memory_region_get_ram_ptr(mr), data->data, size);
1470     memory_region_set_dirty(mr, 0, size);
1471 }
1472 
1473 static void acpi_build_update(void *build_opaque, uint32_t offset)
1474 {
1475     AcpiBuildState *build_state = build_opaque;
1476     AcpiBuildTables tables;
1477 
1478     /* No state to update or already patched? Nothing to do. */
1479     if (!build_state || build_state->patched) {
1480         return;
1481     }
1482     build_state->patched = 1;
1483 
1484     acpi_build_tables_init(&tables);
1485 
1486     acpi_build(build_state->guest_info, &tables);
1487 
1488     acpi_ram_update(build_state->table_mr, tables.table_data);
1489 
1490     if (build_state->rsdp) {
1491         memcpy(build_state->rsdp, tables.rsdp->data, acpi_data_len(tables.rsdp));
1492     } else {
1493         acpi_ram_update(build_state->rsdp_mr, tables.rsdp);
1494     }
1495 
1496     acpi_ram_update(build_state->linker_mr, tables.linker);
1497     acpi_build_tables_cleanup(&tables, true);
1498 }
1499 
1500 static void acpi_build_reset(void *build_opaque)
1501 {
1502     AcpiBuildState *build_state = build_opaque;
1503     build_state->patched = 0;
1504 }
1505 
1506 static MemoryRegion *acpi_add_rom_blob(AcpiBuildState *build_state,
1507                                        GArray *blob, const char *name,
1508                                        uint64_t max_size)
1509 {
1510     return rom_add_blob(name, blob->data, acpi_data_len(blob), max_size, -1,
1511                         name, acpi_build_update, build_state);
1512 }
1513 
1514 static const VMStateDescription vmstate_acpi_build = {
1515     .name = "acpi_build",
1516     .version_id = 1,
1517     .minimum_version_id = 1,
1518     .fields = (VMStateField[]) {
1519         VMSTATE_UINT8(patched, AcpiBuildState),
1520         VMSTATE_END_OF_LIST()
1521     },
1522 };
1523 
1524 void acpi_setup(PcGuestInfo *guest_info)
1525 {
1526     AcpiBuildTables tables;
1527     AcpiBuildState *build_state;
1528 
1529     if (!guest_info->fw_cfg) {
1530         ACPI_BUILD_DPRINTF("No fw cfg. Bailing out.\n");
1531         return;
1532     }
1533 
1534     if (!guest_info->has_acpi_build) {
1535         ACPI_BUILD_DPRINTF("ACPI build disabled. Bailing out.\n");
1536         return;
1537     }
1538 
1539     if (!acpi_enabled) {
1540         ACPI_BUILD_DPRINTF("ACPI disabled. Bailing out.\n");
1541         return;
1542     }
1543 
1544     build_state = g_malloc0(sizeof *build_state);
1545 
1546     build_state->guest_info = guest_info;
1547 
1548     acpi_set_pci_info();
1549 
1550     acpi_build_tables_init(&tables);
1551     acpi_build(build_state->guest_info, &tables);
1552 
1553     /* Now expose it all to Guest */
1554     build_state->table_mr = acpi_add_rom_blob(build_state, tables.table_data,
1555                                                ACPI_BUILD_TABLE_FILE,
1556                                                ACPI_BUILD_TABLE_MAX_SIZE);
1557     assert(build_state->table_mr != NULL);
1558 
1559     build_state->linker_mr =
1560         acpi_add_rom_blob(build_state, tables.linker, "etc/table-loader", 0);
1561 
1562     fw_cfg_add_file(guest_info->fw_cfg, ACPI_BUILD_TPMLOG_FILE,
1563                     tables.tcpalog->data, acpi_data_len(tables.tcpalog));
1564 
1565     if (!guest_info->rsdp_in_ram) {
1566         /*
1567          * Keep for compatibility with old machine types.
1568          * Though RSDP is small, its contents isn't immutable, so
1569          * we'll update it along with the rest of tables on guest access.
1570          */
1571         uint32_t rsdp_size = acpi_data_len(tables.rsdp);
1572 
1573         build_state->rsdp = g_memdup(tables.rsdp->data, rsdp_size);
1574         fw_cfg_add_file_callback(guest_info->fw_cfg, ACPI_BUILD_RSDP_FILE,
1575                                  acpi_build_update, build_state,
1576                                  build_state->rsdp, rsdp_size);
1577         build_state->rsdp_mr = NULL;
1578     } else {
1579         build_state->rsdp = NULL;
1580         build_state->rsdp_mr = acpi_add_rom_blob(build_state, tables.rsdp,
1581                                                   ACPI_BUILD_RSDP_FILE, 0);
1582     }
1583 
1584     qemu_register_reset(acpi_build_reset, build_state);
1585     acpi_build_reset(build_state);
1586     vmstate_register(NULL, 0, &vmstate_acpi_build, build_state);
1587 
1588     /* Cleanup tables but don't free the memory: we track it
1589      * in build_state.
1590      */
1591     acpi_build_tables_cleanup(&tables, false);
1592 }
1593