xref: /openbmc/qemu/hw/i386/acpi-build.c (revision 77a8257e)
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/range.h"
30 #include "qemu/error-report.h"
31 #include "hw/pci/pci.h"
32 #include "qom/cpu.h"
33 #include "hw/i386/pc.h"
34 #include "target-i386/cpu.h"
35 #include "hw/timer/hpet.h"
36 #include "hw/i386/acpi-defs.h"
37 #include "hw/acpi/acpi.h"
38 #include "hw/nvram/fw_cfg.h"
39 #include "hw/acpi/bios-linker-loader.h"
40 #include "hw/loader.h"
41 #include "hw/isa/isa.h"
42 #include "hw/acpi/memory_hotplug.h"
43 #include "sysemu/tpm.h"
44 #include "hw/acpi/tpm.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 #include "exec/ram_addr.h"
62 
63 /* These are used to size the ACPI tables for -M pc-i440fx-1.7 and
64  * -M pc-i440fx-2.0.  Even if the actual amount of AML generated grows
65  * a little bit, there should be plenty of free space since the DSDT
66  * shrunk by ~1.5k between QEMU 2.0 and QEMU 2.1.
67  */
68 #define ACPI_BUILD_LEGACY_CPU_AML_SIZE    97
69 #define ACPI_BUILD_ALIGN_SIZE             0x1000
70 
71 #define ACPI_BUILD_TABLE_SIZE             0x20000
72 
73 /* Reserve RAM space for tables: add another order of magnitude. */
74 #define ACPI_BUILD_TABLE_MAX_SIZE         0x200000
75 
76 /* #define DEBUG_ACPI_BUILD */
77 #ifdef DEBUG_ACPI_BUILD
78 #define ACPI_BUILD_DPRINTF(fmt, ...)        \
79     do {printf("ACPI_BUILD: " fmt, ## __VA_ARGS__); } while (0)
80 #else
81 #define ACPI_BUILD_DPRINTF(fmt, ...)
82 #endif
83 
84 typedef struct AcpiCpuInfo {
85     DECLARE_BITMAP(found_cpus, ACPI_CPU_HOTPLUG_ID_LIMIT);
86 } AcpiCpuInfo;
87 
88 typedef struct AcpiMcfgInfo {
89     uint64_t mcfg_base;
90     uint32_t mcfg_size;
91 } AcpiMcfgInfo;
92 
93 typedef struct AcpiPmInfo {
94     bool s3_disabled;
95     bool s4_disabled;
96     bool pcihp_bridge_en;
97     uint8_t s4_val;
98     uint16_t sci_int;
99     uint8_t acpi_enable_cmd;
100     uint8_t acpi_disable_cmd;
101     uint32_t gpe0_blk;
102     uint32_t gpe0_blk_len;
103     uint32_t io_base;
104     uint16_t cpu_hp_io_base;
105     uint16_t cpu_hp_io_len;
106     uint16_t mem_hp_io_base;
107     uint16_t mem_hp_io_len;
108     uint16_t pcihp_io_base;
109     uint16_t pcihp_io_len;
110 } AcpiPmInfo;
111 
112 typedef struct AcpiMiscInfo {
113     bool has_hpet;
114     bool has_tpm;
115     const unsigned char *dsdt_code;
116     unsigned dsdt_size;
117     uint16_t pvpanic_port;
118     uint16_t applesmc_io_base;
119 } AcpiMiscInfo;
120 
121 typedef struct AcpiBuildPciBusHotplugState {
122     GArray *device_table;
123     GArray *notify_table;
124     struct AcpiBuildPciBusHotplugState *parent;
125     bool pcihp_bridge_en;
126 } AcpiBuildPciBusHotplugState;
127 
128 static void acpi_get_dsdt(AcpiMiscInfo *info)
129 {
130     Object *piix = piix4_pm_find();
131     Object *lpc = ich9_lpc_find();
132     assert(!!piix != !!lpc);
133 
134     if (piix) {
135         info->dsdt_code = AcpiDsdtAmlCode;
136         info->dsdt_size = sizeof AcpiDsdtAmlCode;
137     }
138     if (lpc) {
139         info->dsdt_code = Q35AcpiDsdtAmlCode;
140         info->dsdt_size = sizeof Q35AcpiDsdtAmlCode;
141     }
142 }
143 
144 static
145 int acpi_add_cpu_info(Object *o, void *opaque)
146 {
147     AcpiCpuInfo *cpu = opaque;
148     uint64_t apic_id;
149 
150     if (object_dynamic_cast(o, TYPE_CPU)) {
151         apic_id = object_property_get_int(o, "apic-id", NULL);
152         assert(apic_id < ACPI_CPU_HOTPLUG_ID_LIMIT);
153 
154         set_bit(apic_id, cpu->found_cpus);
155     }
156 
157     object_child_foreach(o, acpi_add_cpu_info, opaque);
158     return 0;
159 }
160 
161 static void acpi_get_cpu_info(AcpiCpuInfo *cpu)
162 {
163     Object *root = object_get_root();
164 
165     memset(cpu->found_cpus, 0, sizeof cpu->found_cpus);
166     object_child_foreach(root, acpi_add_cpu_info, cpu);
167 }
168 
169 static void acpi_get_pm_info(AcpiPmInfo *pm)
170 {
171     Object *piix = piix4_pm_find();
172     Object *lpc = ich9_lpc_find();
173     Object *obj = NULL;
174     QObject *o;
175 
176     pm->pcihp_io_base = 0;
177     pm->pcihp_io_len = 0;
178     if (piix) {
179         obj = piix;
180         pm->cpu_hp_io_base = PIIX4_CPU_HOTPLUG_IO_BASE;
181         pm->pcihp_io_base =
182             object_property_get_int(obj, ACPI_PCIHP_IO_BASE_PROP, NULL);
183         pm->pcihp_io_len =
184             object_property_get_int(obj, ACPI_PCIHP_IO_LEN_PROP, NULL);
185     }
186     if (lpc) {
187         obj = lpc;
188         pm->cpu_hp_io_base = ICH9_CPU_HOTPLUG_IO_BASE;
189     }
190     assert(obj);
191 
192     pm->cpu_hp_io_len = ACPI_GPE_PROC_LEN;
193     pm->mem_hp_io_base = ACPI_MEMORY_HOTPLUG_BASE;
194     pm->mem_hp_io_len = ACPI_MEMORY_HOTPLUG_IO_LEN;
195 
196     /* Fill in optional s3/s4 related properties */
197     o = object_property_get_qobject(obj, ACPI_PM_PROP_S3_DISABLED, NULL);
198     if (o) {
199         pm->s3_disabled = qint_get_int(qobject_to_qint(o));
200     } else {
201         pm->s3_disabled = false;
202     }
203     qobject_decref(o);
204     o = object_property_get_qobject(obj, ACPI_PM_PROP_S4_DISABLED, NULL);
205     if (o) {
206         pm->s4_disabled = qint_get_int(qobject_to_qint(o));
207     } else {
208         pm->s4_disabled = false;
209     }
210     qobject_decref(o);
211     o = object_property_get_qobject(obj, ACPI_PM_PROP_S4_VAL, NULL);
212     if (o) {
213         pm->s4_val = qint_get_int(qobject_to_qint(o));
214     } else {
215         pm->s4_val = false;
216     }
217     qobject_decref(o);
218 
219     /* Fill in mandatory properties */
220     pm->sci_int = object_property_get_int(obj, ACPI_PM_PROP_SCI_INT, NULL);
221 
222     pm->acpi_enable_cmd = object_property_get_int(obj,
223                                                   ACPI_PM_PROP_ACPI_ENABLE_CMD,
224                                                   NULL);
225     pm->acpi_disable_cmd = object_property_get_int(obj,
226                                                   ACPI_PM_PROP_ACPI_DISABLE_CMD,
227                                                   NULL);
228     pm->io_base = object_property_get_int(obj, ACPI_PM_PROP_PM_IO_BASE,
229                                           NULL);
230     pm->gpe0_blk = object_property_get_int(obj, ACPI_PM_PROP_GPE0_BLK,
231                                            NULL);
232     pm->gpe0_blk_len = object_property_get_int(obj, ACPI_PM_PROP_GPE0_BLK_LEN,
233                                                NULL);
234     pm->pcihp_bridge_en =
235         object_property_get_bool(obj, "acpi-pci-hotplug-with-bridge-support",
236                                  NULL);
237 }
238 
239 static void acpi_get_misc_info(AcpiMiscInfo *info)
240 {
241     info->has_hpet = hpet_find();
242     info->has_tpm = tpm_find();
243     info->pvpanic_port = pvpanic_port();
244     info->applesmc_io_base = applesmc_port();
245 }
246 
247 static void acpi_get_pci_info(PcPciInfo *info)
248 {
249     Object *pci_host;
250     bool ambiguous;
251 
252     pci_host = object_resolve_path_type("", TYPE_PCI_HOST_BRIDGE, &ambiguous);
253     g_assert(!ambiguous);
254     g_assert(pci_host);
255 
256     info->w32.begin = object_property_get_int(pci_host,
257                                               PCI_HOST_PROP_PCI_HOLE_START,
258                                               NULL);
259     info->w32.end = object_property_get_int(pci_host,
260                                             PCI_HOST_PROP_PCI_HOLE_END,
261                                             NULL);
262     info->w64.begin = object_property_get_int(pci_host,
263                                               PCI_HOST_PROP_PCI_HOLE64_START,
264                                               NULL);
265     info->w64.end = object_property_get_int(pci_host,
266                                             PCI_HOST_PROP_PCI_HOLE64_END,
267                                             NULL);
268 }
269 
270 #define ACPI_BUILD_APPNAME  "Bochs"
271 #define ACPI_BUILD_APPNAME6 "BOCHS "
272 #define ACPI_BUILD_APPNAME4 "BXPC"
273 
274 #define ACPI_BUILD_TABLE_FILE "etc/acpi/tables"
275 #define ACPI_BUILD_RSDP_FILE "etc/acpi/rsdp"
276 #define ACPI_BUILD_TPMLOG_FILE "etc/tpm/log"
277 
278 static void
279 build_header(GArray *linker, GArray *table_data,
280              AcpiTableHeader *h, const char *sig, int len, uint8_t rev)
281 {
282     memcpy(&h->signature, sig, 4);
283     h->length = cpu_to_le32(len);
284     h->revision = rev;
285     memcpy(h->oem_id, ACPI_BUILD_APPNAME6, 6);
286     memcpy(h->oem_table_id, ACPI_BUILD_APPNAME4, 4);
287     memcpy(h->oem_table_id + 4, sig, 4);
288     h->oem_revision = cpu_to_le32(1);
289     memcpy(h->asl_compiler_id, ACPI_BUILD_APPNAME4, 4);
290     h->asl_compiler_revision = cpu_to_le32(1);
291     h->checksum = 0;
292     /* Checksum to be filled in by Guest linker */
293     bios_linker_loader_add_checksum(linker, ACPI_BUILD_TABLE_FILE,
294                                     table_data->data, h, len, &h->checksum);
295 }
296 
297 /* End here */
298 #define ACPI_PORT_SMI_CMD           0x00b2 /* TODO: this is APM_CNT_IOPORT */
299 
300 static inline void *acpi_data_push(GArray *table_data, unsigned size)
301 {
302     unsigned off = table_data->len;
303     g_array_set_size(table_data, off + size);
304     return table_data->data + off;
305 }
306 
307 static unsigned acpi_data_len(GArray *table)
308 {
309 #if GLIB_CHECK_VERSION(2, 22, 0)
310     assert(g_array_get_element_size(table) == 1);
311 #endif
312     return table->len;
313 }
314 
315 static void acpi_align_size(GArray *blob, unsigned align)
316 {
317     /* Align size to multiple of given size. This reduces the chance
318      * we need to change size in the future (breaking cross version migration).
319      */
320     g_array_set_size(blob, ROUND_UP(acpi_data_len(blob), align));
321 }
322 
323 static inline void acpi_add_table(GArray *table_offsets, GArray *table_data)
324 {
325     uint32_t offset = cpu_to_le32(table_data->len);
326     g_array_append_val(table_offsets, offset);
327 }
328 
329 /* FACS */
330 static void
331 build_facs(GArray *table_data, GArray *linker, PcGuestInfo *guest_info)
332 {
333     AcpiFacsDescriptorRev1 *facs = acpi_data_push(table_data, sizeof *facs);
334     memcpy(&facs->signature, "FACS", 4);
335     facs->length = cpu_to_le32(sizeof(*facs));
336 }
337 
338 /* Load chipset information in FADT */
339 static void fadt_setup(AcpiFadtDescriptorRev1 *fadt, AcpiPmInfo *pm)
340 {
341     fadt->model = 1;
342     fadt->reserved1 = 0;
343     fadt->sci_int = cpu_to_le16(pm->sci_int);
344     fadt->smi_cmd = cpu_to_le32(ACPI_PORT_SMI_CMD);
345     fadt->acpi_enable = pm->acpi_enable_cmd;
346     fadt->acpi_disable = pm->acpi_disable_cmd;
347     /* EVT, CNT, TMR offset matches hw/acpi/core.c */
348     fadt->pm1a_evt_blk = cpu_to_le32(pm->io_base);
349     fadt->pm1a_cnt_blk = cpu_to_le32(pm->io_base + 0x04);
350     fadt->pm_tmr_blk = cpu_to_le32(pm->io_base + 0x08);
351     fadt->gpe0_blk = cpu_to_le32(pm->gpe0_blk);
352     /* EVT, CNT, TMR length matches hw/acpi/core.c */
353     fadt->pm1_evt_len = 4;
354     fadt->pm1_cnt_len = 2;
355     fadt->pm_tmr_len = 4;
356     fadt->gpe0_blk_len = pm->gpe0_blk_len;
357     fadt->plvl2_lat = cpu_to_le16(0xfff); /* C2 state not supported */
358     fadt->plvl3_lat = cpu_to_le16(0xfff); /* C3 state not supported */
359     fadt->flags = cpu_to_le32((1 << ACPI_FADT_F_WBINVD) |
360                               (1 << ACPI_FADT_F_PROC_C1) |
361                               (1 << ACPI_FADT_F_SLP_BUTTON) |
362                               (1 << ACPI_FADT_F_RTC_S4));
363     fadt->flags |= cpu_to_le32(1 << ACPI_FADT_F_USE_PLATFORM_CLOCK);
364     /* APIC destination mode ("Flat Logical") has an upper limit of 8 CPUs
365      * For more than 8 CPUs, "Clustered Logical" mode has to be used
366      */
367     if (max_cpus > 8) {
368         fadt->flags |= cpu_to_le32(1 << ACPI_FADT_F_FORCE_APIC_CLUSTER_MODEL);
369     }
370 }
371 
372 
373 /* FADT */
374 static void
375 build_fadt(GArray *table_data, GArray *linker, AcpiPmInfo *pm,
376            unsigned facs, unsigned dsdt)
377 {
378     AcpiFadtDescriptorRev1 *fadt = acpi_data_push(table_data, sizeof(*fadt));
379 
380     fadt->firmware_ctrl = cpu_to_le32(facs);
381     /* FACS address to be filled by Guest linker */
382     bios_linker_loader_add_pointer(linker, ACPI_BUILD_TABLE_FILE,
383                                    ACPI_BUILD_TABLE_FILE,
384                                    table_data, &fadt->firmware_ctrl,
385                                    sizeof fadt->firmware_ctrl);
386 
387     fadt->dsdt = cpu_to_le32(dsdt);
388     /* DSDT address to be filled by Guest linker */
389     bios_linker_loader_add_pointer(linker, ACPI_BUILD_TABLE_FILE,
390                                    ACPI_BUILD_TABLE_FILE,
391                                    table_data, &fadt->dsdt,
392                                    sizeof fadt->dsdt);
393 
394     fadt_setup(fadt, pm);
395 
396     build_header(linker, table_data,
397                  (void *)fadt, "FACP", sizeof(*fadt), 1);
398 }
399 
400 static void
401 build_madt(GArray *table_data, GArray *linker, AcpiCpuInfo *cpu,
402            PcGuestInfo *guest_info)
403 {
404     int madt_start = table_data->len;
405 
406     AcpiMultipleApicTable *madt;
407     AcpiMadtIoApic *io_apic;
408     AcpiMadtIntsrcovr *intsrcovr;
409     AcpiMadtLocalNmi *local_nmi;
410     int i;
411 
412     madt = acpi_data_push(table_data, sizeof *madt);
413     madt->local_apic_address = cpu_to_le32(APIC_DEFAULT_ADDRESS);
414     madt->flags = cpu_to_le32(1);
415 
416     for (i = 0; i < guest_info->apic_id_limit; i++) {
417         AcpiMadtProcessorApic *apic = acpi_data_push(table_data, sizeof *apic);
418         apic->type = ACPI_APIC_PROCESSOR;
419         apic->length = sizeof(*apic);
420         apic->processor_id = i;
421         apic->local_apic_id = i;
422         if (test_bit(i, cpu->found_cpus)) {
423             apic->flags = cpu_to_le32(1);
424         } else {
425             apic->flags = cpu_to_le32(0);
426         }
427     }
428     io_apic = acpi_data_push(table_data, sizeof *io_apic);
429     io_apic->type = ACPI_APIC_IO;
430     io_apic->length = sizeof(*io_apic);
431 #define ACPI_BUILD_IOAPIC_ID 0x0
432     io_apic->io_apic_id = ACPI_BUILD_IOAPIC_ID;
433     io_apic->address = cpu_to_le32(IO_APIC_DEFAULT_ADDRESS);
434     io_apic->interrupt = cpu_to_le32(0);
435 
436     if (guest_info->apic_xrupt_override) {
437         intsrcovr = acpi_data_push(table_data, sizeof *intsrcovr);
438         intsrcovr->type   = ACPI_APIC_XRUPT_OVERRIDE;
439         intsrcovr->length = sizeof(*intsrcovr);
440         intsrcovr->source = 0;
441         intsrcovr->gsi    = cpu_to_le32(2);
442         intsrcovr->flags  = cpu_to_le16(0); /* conforms to bus specifications */
443     }
444     for (i = 1; i < 16; i++) {
445 #define ACPI_BUILD_PCI_IRQS ((1<<5) | (1<<9) | (1<<10) | (1<<11))
446         if (!(ACPI_BUILD_PCI_IRQS & (1 << i))) {
447             /* No need for a INT source override structure. */
448             continue;
449         }
450         intsrcovr = acpi_data_push(table_data, sizeof *intsrcovr);
451         intsrcovr->type   = ACPI_APIC_XRUPT_OVERRIDE;
452         intsrcovr->length = sizeof(*intsrcovr);
453         intsrcovr->source = i;
454         intsrcovr->gsi    = cpu_to_le32(i);
455         intsrcovr->flags  = cpu_to_le16(0xd); /* active high, level triggered */
456     }
457 
458     local_nmi = acpi_data_push(table_data, sizeof *local_nmi);
459     local_nmi->type         = ACPI_APIC_LOCAL_NMI;
460     local_nmi->length       = sizeof(*local_nmi);
461     local_nmi->processor_id = 0xff; /* all processors */
462     local_nmi->flags        = cpu_to_le16(0);
463     local_nmi->lint         = 1; /* ACPI_LINT1 */
464 
465     build_header(linker, table_data,
466                  (void *)(table_data->data + madt_start), "APIC",
467                  table_data->len - madt_start, 1);
468 }
469 
470 #include "hw/i386/ssdt-tpm.hex"
471 
472 /* Assign BSEL property to all buses.  In the future, this can be changed
473  * to only assign to buses that support hotplug.
474  */
475 static void *acpi_set_bsel(PCIBus *bus, void *opaque)
476 {
477     unsigned *bsel_alloc = opaque;
478     unsigned *bus_bsel;
479 
480     if (qbus_is_hotpluggable(BUS(bus))) {
481         bus_bsel = g_malloc(sizeof *bus_bsel);
482 
483         *bus_bsel = (*bsel_alloc)++;
484         object_property_add_uint32_ptr(OBJECT(bus), ACPI_PCIHP_PROP_BSEL,
485                                        bus_bsel, NULL);
486     }
487 
488     return bsel_alloc;
489 }
490 
491 static void acpi_set_pci_info(void)
492 {
493     PCIBus *bus = find_i440fx(); /* TODO: Q35 support */
494     unsigned bsel_alloc = 0;
495 
496     if (bus) {
497         /* Scan all PCI buses. Set property to enable acpi based hotplug. */
498         pci_for_each_bus_depth_first(bus, acpi_set_bsel, NULL, &bsel_alloc);
499     }
500 }
501 
502 static void build_append_pcihp_notify_entry(Aml *method, int slot)
503 {
504     Aml *if_ctx;
505     int32_t devfn = PCI_DEVFN(slot, 0);
506 
507     if_ctx = aml_if(aml_and(aml_arg(0), aml_int(0x1U << slot)));
508     aml_append(if_ctx, aml_notify(aml_name("S%.02X", devfn), aml_arg(1)));
509     aml_append(method, if_ctx);
510 }
511 
512 static void build_append_pci_bus_devices(Aml *parent_scope, PCIBus *bus,
513                                          bool pcihp_bridge_en)
514 {
515     Aml *dev, *notify_method, *method;
516     QObject *bsel;
517     PCIBus *sec;
518     int i;
519 
520     bsel = object_property_get_qobject(OBJECT(bus), ACPI_PCIHP_PROP_BSEL, NULL);
521     if (bsel) {
522         int64_t bsel_val = qint_get_int(qobject_to_qint(bsel));
523 
524         aml_append(parent_scope, aml_name_decl("BSEL", aml_int(bsel_val)));
525         notify_method = aml_method("DVNT", 2);
526     }
527 
528     for (i = 0; i < ARRAY_SIZE(bus->devices); i += PCI_FUNC_MAX) {
529         DeviceClass *dc;
530         PCIDeviceClass *pc;
531         PCIDevice *pdev = bus->devices[i];
532         int slot = PCI_SLOT(i);
533         bool hotplug_enabled_dev;
534         bool bridge_in_acpi;
535 
536         if (!pdev) {
537             if (bsel) { /* add hotplug slots for non present devices */
538                 dev = aml_device("S%.02X", PCI_DEVFN(slot, 0));
539                 aml_append(dev, aml_name_decl("_SUN", aml_int(slot)));
540                 aml_append(dev, aml_name_decl("_ADR", aml_int(slot << 16)));
541                 method = aml_method("_EJ0", 1);
542                 aml_append(method,
543                     aml_call2("PCEJ", aml_name("BSEL"), aml_name("_SUN"))
544                 );
545                 aml_append(dev, method);
546                 aml_append(parent_scope, dev);
547 
548                 build_append_pcihp_notify_entry(notify_method, slot);
549             }
550             continue;
551         }
552 
553         pc = PCI_DEVICE_GET_CLASS(pdev);
554         dc = DEVICE_GET_CLASS(pdev);
555 
556         /* When hotplug for bridges is enabled, bridges are
557          * described in ACPI separately (see build_pci_bus_end).
558          * In this case they aren't themselves hot-pluggable.
559          * Hotplugged bridges *are* hot-pluggable.
560          */
561         bridge_in_acpi = pc->is_bridge && pcihp_bridge_en &&
562             !DEVICE(pdev)->hotplugged;
563 
564         hotplug_enabled_dev = bsel && dc->hotpluggable && !bridge_in_acpi;
565 
566         if (pc->class_id == PCI_CLASS_BRIDGE_ISA) {
567             continue;
568         }
569 
570         /* start to compose PCI slot descriptor */
571         dev = aml_device("S%.02X", PCI_DEVFN(slot, 0));
572         aml_append(dev, aml_name_decl("_ADR", aml_int(slot << 16)));
573 
574         if (pc->class_id == PCI_CLASS_DISPLAY_VGA) {
575             /* add VGA specific AML methods */
576             int s3d;
577 
578             if (object_dynamic_cast(OBJECT(pdev), "qxl-vga")) {
579                 s3d = 3;
580             } else {
581                 s3d = 0;
582             }
583 
584             method = aml_method("_S1D", 0);
585             aml_append(method, aml_return(aml_int(0)));
586             aml_append(dev, method);
587 
588             method = aml_method("_S2D", 0);
589             aml_append(method, aml_return(aml_int(0)));
590             aml_append(dev, method);
591 
592             method = aml_method("_S3D", 0);
593             aml_append(method, aml_return(aml_int(s3d)));
594             aml_append(dev, method);
595         } else if (hotplug_enabled_dev) {
596             /* add _SUN/_EJ0 to make slot hotpluggable  */
597             aml_append(dev, aml_name_decl("_SUN", aml_int(slot)));
598 
599             method = aml_method("_EJ0", 1);
600             aml_append(method,
601                 aml_call2("PCEJ", aml_name("BSEL"), aml_name("_SUN"))
602             );
603             aml_append(dev, method);
604 
605             if (bsel) {
606                 build_append_pcihp_notify_entry(notify_method, slot);
607             }
608         } else if (bridge_in_acpi) {
609             /*
610              * device is coldplugged bridge,
611              * add child device descriptions into its scope
612              */
613             PCIBus *sec_bus = pci_bridge_get_sec_bus(PCI_BRIDGE(pdev));
614 
615             build_append_pci_bus_devices(dev, sec_bus, pcihp_bridge_en);
616         }
617         /* slot descriptor has been composed, add it into parent context */
618         aml_append(parent_scope, dev);
619     }
620 
621     if (bsel) {
622         aml_append(parent_scope, notify_method);
623     }
624 
625     /* Append PCNT method to notify about events on local and child buses.
626      * Add unconditionally for root since DSDT expects it.
627      */
628     method = aml_method("PCNT", 0);
629 
630     /* If bus supports hotplug select it and notify about local events */
631     if (bsel) {
632         int64_t bsel_val = qint_get_int(qobject_to_qint(bsel));
633         aml_append(method, aml_store(aml_int(bsel_val), aml_name("BNUM")));
634         aml_append(method,
635             aml_call2("DVNT", aml_name("PCIU"), aml_int(1) /* Device Check */)
636         );
637         aml_append(method,
638             aml_call2("DVNT", aml_name("PCID"), aml_int(3)/* Eject Request */)
639         );
640     }
641 
642     /* Notify about child bus events in any case */
643     if (pcihp_bridge_en) {
644         QLIST_FOREACH(sec, &bus->child, sibling) {
645             int32_t devfn = sec->parent_dev->devfn;
646 
647             aml_append(method, aml_name("^S%.02X.PCNT", devfn));
648         }
649     }
650     aml_append(parent_scope, method);
651 }
652 
653 static void
654 build_ssdt(GArray *table_data, GArray *linker,
655            AcpiCpuInfo *cpu, AcpiPmInfo *pm, AcpiMiscInfo *misc,
656            PcPciInfo *pci, PcGuestInfo *guest_info)
657 {
658     MachineState *machine = MACHINE(qdev_get_machine());
659     uint32_t nr_mem = machine->ram_slots;
660     unsigned acpi_cpus = guest_info->apic_id_limit;
661     Aml *ssdt, *sb_scope, *scope, *pkg, *dev, *method, *crs, *field, *ifctx;
662     int i;
663 
664     ssdt = init_aml_allocator();
665     /* The current AML generator can cover the APIC ID range [0..255],
666      * inclusive, for VCPU hotplug. */
667     QEMU_BUILD_BUG_ON(ACPI_CPU_HOTPLUG_ID_LIMIT > 256);
668     g_assert(acpi_cpus <= ACPI_CPU_HOTPLUG_ID_LIMIT);
669 
670     /* Reserve space for header */
671     acpi_data_push(ssdt->buf, sizeof(AcpiTableHeader));
672 
673     scope = aml_scope("\\_SB.PCI0");
674     /* build PCI0._CRS */
675     crs = aml_resource_template();
676     aml_append(crs,
677         aml_word_bus_number(aml_min_fixed, aml_max_fixed, aml_pos_decode,
678                             0x0000, 0x0000, 0x00FF, 0x0000, 0x0100));
679     aml_append(crs, aml_io(aml_decode16, 0x0CF8, 0x0CF8, 0x01, 0x08));
680 
681     aml_append(crs,
682         aml_word_io(aml_min_fixed, aml_max_fixed,
683                     aml_pos_decode, aml_entire_range,
684                     0x0000, 0x0000, 0x0CF7, 0x0000, 0x0CF8));
685     aml_append(crs,
686         aml_word_io(aml_min_fixed, aml_max_fixed,
687                     aml_pos_decode, aml_entire_range,
688                     0x0000, 0x0D00, 0xFFFF, 0x0000, 0xF300));
689     aml_append(crs,
690         aml_dword_memory(aml_pos_decode, aml_min_fixed, aml_max_fixed,
691                          aml_cacheable, aml_ReadWrite,
692                          0, 0x000A0000, 0x000BFFFF, 0, 0x00020000));
693     aml_append(crs,
694         aml_dword_memory(aml_pos_decode, aml_min_fixed, aml_max_fixed,
695                          aml_non_cacheable, aml_ReadWrite,
696                          0, pci->w32.begin, pci->w32.end - 1, 0,
697                          pci->w32.end - pci->w32.begin));
698     if (pci->w64.begin) {
699         aml_append(crs,
700             aml_qword_memory(aml_pos_decode, aml_min_fixed, aml_max_fixed,
701                              aml_cacheable, aml_ReadWrite,
702                              0, pci->w64.begin, pci->w64.end - 1, 0,
703                              pci->w64.end - pci->w64.begin));
704     }
705     aml_append(scope, aml_name_decl("_CRS", crs));
706 
707     /* reserve GPE0 block resources */
708     dev = aml_device("GPE0");
709     aml_append(dev, aml_name_decl("_HID", aml_string("PNP0A06")));
710     aml_append(dev, aml_name_decl("_UID", aml_string("GPE0 resources")));
711     /* device present, functioning, decoding, not shown in UI */
712     aml_append(dev, aml_name_decl("_STA", aml_int(0xB)));
713     crs = aml_resource_template();
714     aml_append(crs,
715         aml_io(aml_decode16, pm->gpe0_blk, pm->gpe0_blk, 1, pm->gpe0_blk_len)
716     );
717     aml_append(dev, aml_name_decl("_CRS", crs));
718     aml_append(scope, dev);
719 
720     /* reserve PCIHP resources */
721     if (pm->pcihp_io_len) {
722         dev = aml_device("PHPR");
723         aml_append(dev, aml_name_decl("_HID", aml_string("PNP0A06")));
724         aml_append(dev,
725             aml_name_decl("_UID", aml_string("PCI Hotplug resources")));
726         /* device present, functioning, decoding, not shown in UI */
727         aml_append(dev, aml_name_decl("_STA", aml_int(0xB)));
728         crs = aml_resource_template();
729         aml_append(crs,
730             aml_io(aml_decode16, pm->pcihp_io_base, pm->pcihp_io_base, 1,
731                    pm->pcihp_io_len)
732         );
733         aml_append(dev, aml_name_decl("_CRS", crs));
734         aml_append(scope, dev);
735     }
736     aml_append(ssdt, scope);
737 
738     /*  create S3_ / S4_ / S5_ packages if necessary */
739     scope = aml_scope("\\");
740     if (!pm->s3_disabled) {
741         pkg = aml_package(4);
742         aml_append(pkg, aml_int(1)); /* PM1a_CNT.SLP_TYP */
743         aml_append(pkg, aml_int(1)); /* PM1b_CNT.SLP_TYP, FIXME: not impl. */
744         aml_append(pkg, aml_int(0)); /* reserved */
745         aml_append(pkg, aml_int(0)); /* reserved */
746         aml_append(scope, aml_name_decl("_S3", pkg));
747     }
748 
749     if (!pm->s4_disabled) {
750         pkg = aml_package(4);
751         aml_append(pkg, aml_int(pm->s4_val)); /* PM1a_CNT.SLP_TYP */
752         /* PM1b_CNT.SLP_TYP, FIXME: not impl. */
753         aml_append(pkg, aml_int(pm->s4_val));
754         aml_append(pkg, aml_int(0)); /* reserved */
755         aml_append(pkg, aml_int(0)); /* reserved */
756         aml_append(scope, aml_name_decl("_S4", pkg));
757     }
758 
759     pkg = aml_package(4);
760     aml_append(pkg, aml_int(0)); /* PM1a_CNT.SLP_TYP */
761     aml_append(pkg, aml_int(0)); /* PM1b_CNT.SLP_TYP not impl. */
762     aml_append(pkg, aml_int(0)); /* reserved */
763     aml_append(pkg, aml_int(0)); /* reserved */
764     aml_append(scope, aml_name_decl("_S5", pkg));
765     aml_append(ssdt, scope);
766 
767     if (misc->applesmc_io_base) {
768         scope = aml_scope("\\_SB.PCI0.ISA");
769         dev = aml_device("SMC");
770 
771         aml_append(dev, aml_name_decl("_HID", aml_eisaid("APP0001")));
772         /* device present, functioning, decoding, not shown in UI */
773         aml_append(dev, aml_name_decl("_STA", aml_int(0xB)));
774 
775         crs = aml_resource_template();
776         aml_append(crs,
777             aml_io(aml_decode16, misc->applesmc_io_base, misc->applesmc_io_base,
778                    0x01, APPLESMC_MAX_DATA_LENGTH)
779         );
780         aml_append(crs, aml_irq_no_flags(6));
781         aml_append(dev, aml_name_decl("_CRS", crs));
782 
783         aml_append(scope, dev);
784         aml_append(ssdt, scope);
785     }
786 
787     if (misc->pvpanic_port) {
788         scope = aml_scope("\\_SB.PCI0.ISA");
789 
790         dev = aml_device("PEVR");
791         aml_append(dev, aml_name_decl("_HID", aml_string("QEMU0002")));
792 
793         crs = aml_resource_template();
794         aml_append(crs,
795             aml_io(aml_decode16, misc->pvpanic_port, misc->pvpanic_port, 1, 1)
796         );
797         aml_append(dev, aml_name_decl("_CRS", crs));
798 
799         aml_append(dev, aml_operation_region("PEOR", aml_system_io,
800                                               misc->pvpanic_port, 1));
801         field = aml_field("PEOR", aml_byte_acc);
802         aml_append(field, aml_named_field("PEPT", 8));
803         aml_append(dev, field);
804 
805         method = aml_method("RDPT", 0);
806         aml_append(method, aml_store(aml_name("PEPT"), aml_local(0)));
807         aml_append(method, aml_return(aml_local(0)));
808         aml_append(dev, method);
809 
810         method = aml_method("WRPT", 1);
811         aml_append(method, aml_store(aml_arg(0), aml_name("PEPT")));
812         aml_append(dev, method);
813 
814         aml_append(scope, dev);
815         aml_append(ssdt, scope);
816     }
817 
818     sb_scope = aml_scope("_SB");
819     {
820         /* create PCI0.PRES device and its _CRS to reserve CPU hotplug MMIO */
821         dev = aml_device("PCI0." stringify(CPU_HOTPLUG_RESOURCE_DEVICE));
822         aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0A06")));
823         aml_append(dev,
824             aml_name_decl("_UID", aml_string("CPU Hotplug resources"))
825         );
826         /* device present, functioning, decoding, not shown in UI */
827         aml_append(dev, aml_name_decl("_STA", aml_int(0xB)));
828         crs = aml_resource_template();
829         aml_append(crs,
830             aml_io(aml_decode16, pm->cpu_hp_io_base, pm->cpu_hp_io_base, 1,
831                    pm->cpu_hp_io_len)
832         );
833         aml_append(dev, aml_name_decl("_CRS", crs));
834         aml_append(sb_scope, dev);
835         /* declare CPU hotplug MMIO region and PRS field to access it */
836         aml_append(sb_scope, aml_operation_region(
837             "PRST", aml_system_io, pm->cpu_hp_io_base, pm->cpu_hp_io_len));
838         field = aml_field("PRST", aml_byte_acc);
839         aml_append(field, aml_named_field("PRS", 256));
840         aml_append(sb_scope, field);
841 
842         /* build Processor object for each processor */
843         for (i = 0; i < acpi_cpus; i++) {
844             dev = aml_processor(i, 0, 0, "CP%.02X", i);
845 
846             method = aml_method("_MAT", 0);
847             aml_append(method, aml_return(aml_call1("CPMA", aml_int(i))));
848             aml_append(dev, method);
849 
850             method = aml_method("_STA", 0);
851             aml_append(method, aml_return(aml_call1("CPST", aml_int(i))));
852             aml_append(dev, method);
853 
854             method = aml_method("_EJ0", 1);
855             aml_append(method,
856                 aml_return(aml_call2("CPEJ", aml_int(i), aml_arg(0)))
857             );
858             aml_append(dev, method);
859 
860             aml_append(sb_scope, dev);
861         }
862 
863         /* build this code:
864          *   Method(NTFY, 2) {If (LEqual(Arg0, 0x00)) {Notify(CP00, Arg1)} ...}
865          */
866         /* Arg0 = Processor ID = APIC ID */
867         method = aml_method("NTFY", 2);
868         for (i = 0; i < acpi_cpus; i++) {
869             ifctx = aml_if(aml_equal(aml_arg(0), aml_int(i)));
870             aml_append(ifctx,
871                 aml_notify(aml_name("CP%.02X", i), aml_arg(1))
872             );
873             aml_append(method, ifctx);
874         }
875         aml_append(sb_scope, method);
876 
877         /* build "Name(CPON, Package() { One, One, ..., Zero, Zero, ... })"
878          *
879          * Note: The ability to create variable-sized packages was first
880          * introduced in ACPI 2.0. ACPI 1.0 only allowed fixed-size packages
881          * ith up to 255 elements. Windows guests up to win2k8 fail when
882          * VarPackageOp is used.
883          */
884         pkg = acpi_cpus <= 255 ? aml_package(acpi_cpus) :
885                                  aml_varpackage(acpi_cpus);
886 
887         for (i = 0; i < acpi_cpus; i++) {
888             uint8_t b = test_bit(i, cpu->found_cpus) ? 0x01 : 0x00;
889             aml_append(pkg, aml_int(b));
890         }
891         aml_append(sb_scope, aml_name_decl("CPON", pkg));
892 
893         /* build memory devices */
894         assert(nr_mem <= ACPI_MAX_RAM_SLOTS);
895         scope = aml_scope("\\_SB.PCI0." stringify(MEMORY_HOTPLUG_DEVICE));
896         aml_append(scope,
897             aml_name_decl(stringify(MEMORY_SLOTS_NUMBER), aml_int(nr_mem))
898         );
899 
900         crs = aml_resource_template();
901         aml_append(crs,
902             aml_io(aml_decode16, pm->mem_hp_io_base, pm->mem_hp_io_base, 0,
903                    pm->mem_hp_io_len)
904         );
905         aml_append(scope, aml_name_decl("_CRS", crs));
906 
907         aml_append(scope, aml_operation_region(
908             stringify(MEMORY_HOTPLUG_IO_REGION), aml_system_io,
909             pm->mem_hp_io_base, pm->mem_hp_io_len)
910         );
911 
912         field = aml_field(stringify(MEMORY_HOTPLUG_IO_REGION), aml_dword_acc);
913         aml_append(field, /* read only */
914             aml_named_field(stringify(MEMORY_SLOT_ADDR_LOW), 32));
915         aml_append(field, /* read only */
916             aml_named_field(stringify(MEMORY_SLOT_ADDR_HIGH), 32));
917         aml_append(field, /* read only */
918             aml_named_field(stringify(MEMORY_SLOT_SIZE_LOW), 32));
919         aml_append(field, /* read only */
920             aml_named_field(stringify(MEMORY_SLOT_SIZE_HIGH), 32));
921         aml_append(field, /* read only */
922             aml_named_field(stringify(MEMORY_SLOT_PROXIMITY), 32));
923         aml_append(scope, field);
924 
925         field = aml_field(stringify(MEMORY_HOTPLUG_IO_REGION), aml_byte_acc);
926         aml_append(field, aml_reserved_field(160 /* bits, Offset(20) */));
927         aml_append(field, /* 1 if enabled, read only */
928             aml_named_field(stringify(MEMORY_SLOT_ENABLED), 1));
929         aml_append(field,
930             /*(read) 1 if has a insert event. (write) 1 to clear event */
931             aml_named_field(stringify(MEMORY_SLOT_INSERT_EVENT), 1));
932         aml_append(scope, field);
933 
934         field = aml_field(stringify(MEMORY_HOTPLUG_IO_REGION), aml_dword_acc);
935         aml_append(field, /* DIMM selector, write only */
936             aml_named_field(stringify(MEMORY_SLOT_SLECTOR), 32));
937         aml_append(field, /* _OST event code, write only */
938             aml_named_field(stringify(MEMORY_SLOT_OST_EVENT), 32));
939         aml_append(field, /* _OST status code, write only */
940             aml_named_field(stringify(MEMORY_SLOT_OST_STATUS), 32));
941         aml_append(scope, field);
942 
943         aml_append(sb_scope, scope);
944 
945         for (i = 0; i < nr_mem; i++) {
946             #define BASEPATH "\\_SB.PCI0." stringify(MEMORY_HOTPLUG_DEVICE) "."
947             const char *s;
948 
949             dev = aml_device("MP%02X", i);
950             aml_append(dev, aml_name_decl("_UID", aml_string("0x%02X", i)));
951             aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0C80")));
952 
953             method = aml_method("_CRS", 0);
954             s = BASEPATH stringify(MEMORY_SLOT_CRS_METHOD);
955             aml_append(method, aml_return(aml_call1(s, aml_name("_UID"))));
956             aml_append(dev, method);
957 
958             method = aml_method("_STA", 0);
959             s = BASEPATH stringify(MEMORY_SLOT_STATUS_METHOD);
960             aml_append(method, aml_return(aml_call1(s, aml_name("_UID"))));
961             aml_append(dev, method);
962 
963             method = aml_method("_PXM", 0);
964             s = BASEPATH stringify(MEMORY_SLOT_PROXIMITY_METHOD);
965             aml_append(method, aml_return(aml_call1(s, aml_name("_UID"))));
966             aml_append(dev, method);
967 
968             method = aml_method("_OST", 3);
969             s = BASEPATH stringify(MEMORY_SLOT_OST_METHOD);
970             aml_append(method, aml_return(aml_call4(
971                 s, aml_name("_UID"), aml_arg(0), aml_arg(1), aml_arg(2)
972             )));
973             aml_append(dev, method);
974 
975             aml_append(sb_scope, dev);
976         }
977 
978         /* build Method(MEMORY_SLOT_NOTIFY_METHOD, 2) {
979          *     If (LEqual(Arg0, 0x00)) {Notify(MP00, Arg1)} ...
980          */
981         method = aml_method(stringify(MEMORY_SLOT_NOTIFY_METHOD), 2);
982         for (i = 0; i < nr_mem; i++) {
983             ifctx = aml_if(aml_equal(aml_arg(0), aml_int(i)));
984             aml_append(ifctx,
985                 aml_notify(aml_name("MP%.02X", i), aml_arg(1))
986             );
987             aml_append(method, ifctx);
988         }
989         aml_append(sb_scope, method);
990 
991         {
992             Object *pci_host;
993             PCIBus *bus = NULL;
994             bool ambiguous;
995 
996             pci_host = object_resolve_path_type("", TYPE_PCI_HOST_BRIDGE, &ambiguous);
997             if (!ambiguous && pci_host) {
998                 bus = PCI_HOST_BRIDGE(pci_host)->bus;
999             }
1000 
1001             if (bus) {
1002                 Aml *scope = aml_scope("PCI0");
1003                 /* Scan all PCI buses. Generate tables to support hotplug. */
1004                 build_append_pci_bus_devices(scope, bus, pm->pcihp_bridge_en);
1005                 aml_append(sb_scope, scope);
1006             }
1007         }
1008         aml_append(ssdt, sb_scope);
1009     }
1010 
1011     /* copy AML table into ACPI tables blob and patch header there */
1012     g_array_append_vals(table_data, ssdt->buf->data, ssdt->buf->len);
1013     build_header(linker, table_data,
1014         (void *)(table_data->data + table_data->len - ssdt->buf->len),
1015         "SSDT", ssdt->buf->len, 1);
1016     free_aml_allocator();
1017 }
1018 
1019 static void
1020 build_hpet(GArray *table_data, GArray *linker)
1021 {
1022     Acpi20Hpet *hpet;
1023 
1024     hpet = acpi_data_push(table_data, sizeof(*hpet));
1025     /* Note timer_block_id value must be kept in sync with value advertised by
1026      * emulated hpet
1027      */
1028     hpet->timer_block_id = cpu_to_le32(0x8086a201);
1029     hpet->addr.address = cpu_to_le64(HPET_BASE);
1030     build_header(linker, table_data,
1031                  (void *)hpet, "HPET", sizeof(*hpet), 1);
1032 }
1033 
1034 static void
1035 build_tpm_tcpa(GArray *table_data, GArray *linker, GArray *tcpalog)
1036 {
1037     Acpi20Tcpa *tcpa = acpi_data_push(table_data, sizeof *tcpa);
1038     uint64_t log_area_start_address = acpi_data_len(tcpalog);
1039 
1040     tcpa->platform_class = cpu_to_le16(TPM_TCPA_ACPI_CLASS_CLIENT);
1041     tcpa->log_area_minimum_length = cpu_to_le32(TPM_LOG_AREA_MINIMUM_SIZE);
1042     tcpa->log_area_start_address = cpu_to_le64(log_area_start_address);
1043 
1044     bios_linker_loader_alloc(linker, ACPI_BUILD_TPMLOG_FILE, 1,
1045                              false /* high memory */);
1046 
1047     /* log area start address to be filled by Guest linker */
1048     bios_linker_loader_add_pointer(linker, ACPI_BUILD_TABLE_FILE,
1049                                    ACPI_BUILD_TPMLOG_FILE,
1050                                    table_data, &tcpa->log_area_start_address,
1051                                    sizeof(tcpa->log_area_start_address));
1052 
1053     build_header(linker, table_data,
1054                  (void *)tcpa, "TCPA", sizeof(*tcpa), 2);
1055 
1056     acpi_data_push(tcpalog, TPM_LOG_AREA_MINIMUM_SIZE);
1057 }
1058 
1059 static void
1060 build_tpm_ssdt(GArray *table_data, GArray *linker)
1061 {
1062     void *tpm_ptr;
1063 
1064     tpm_ptr = acpi_data_push(table_data, sizeof(ssdt_tpm_aml));
1065     memcpy(tpm_ptr, ssdt_tpm_aml, sizeof(ssdt_tpm_aml));
1066 }
1067 
1068 typedef enum {
1069     MEM_AFFINITY_NOFLAGS      = 0,
1070     MEM_AFFINITY_ENABLED      = (1 << 0),
1071     MEM_AFFINITY_HOTPLUGGABLE = (1 << 1),
1072     MEM_AFFINITY_NON_VOLATILE = (1 << 2),
1073 } MemoryAffinityFlags;
1074 
1075 static void
1076 acpi_build_srat_memory(AcpiSratMemoryAffinity *numamem, uint64_t base,
1077                        uint64_t len, int node, MemoryAffinityFlags flags)
1078 {
1079     numamem->type = ACPI_SRAT_MEMORY;
1080     numamem->length = sizeof(*numamem);
1081     memset(numamem->proximity, 0, 4);
1082     numamem->proximity[0] = node;
1083     numamem->flags = cpu_to_le32(flags);
1084     numamem->base_addr = cpu_to_le64(base);
1085     numamem->range_length = cpu_to_le64(len);
1086 }
1087 
1088 static void
1089 build_srat(GArray *table_data, GArray *linker, PcGuestInfo *guest_info)
1090 {
1091     AcpiSystemResourceAffinityTable *srat;
1092     AcpiSratProcessorAffinity *core;
1093     AcpiSratMemoryAffinity *numamem;
1094 
1095     int i;
1096     uint64_t curnode;
1097     int srat_start, numa_start, slots;
1098     uint64_t mem_len, mem_base, next_base;
1099     PCMachineState *pcms = PC_MACHINE(qdev_get_machine());
1100     ram_addr_t hotplugabble_address_space_size =
1101         object_property_get_int(OBJECT(pcms), PC_MACHINE_MEMHP_REGION_SIZE,
1102                                 NULL);
1103 
1104     srat_start = table_data->len;
1105 
1106     srat = acpi_data_push(table_data, sizeof *srat);
1107     srat->reserved1 = cpu_to_le32(1);
1108     core = (void *)(srat + 1);
1109 
1110     for (i = 0; i < guest_info->apic_id_limit; ++i) {
1111         core = acpi_data_push(table_data, sizeof *core);
1112         core->type = ACPI_SRAT_PROCESSOR;
1113         core->length = sizeof(*core);
1114         core->local_apic_id = i;
1115         curnode = guest_info->node_cpu[i];
1116         core->proximity_lo = curnode;
1117         memset(core->proximity_hi, 0, 3);
1118         core->local_sapic_eid = 0;
1119         core->flags = cpu_to_le32(1);
1120     }
1121 
1122 
1123     /* the memory map is a bit tricky, it contains at least one hole
1124      * from 640k-1M and possibly another one from 3.5G-4G.
1125      */
1126     next_base = 0;
1127     numa_start = table_data->len;
1128 
1129     numamem = acpi_data_push(table_data, sizeof *numamem);
1130     acpi_build_srat_memory(numamem, 0, 640*1024, 0, MEM_AFFINITY_ENABLED);
1131     next_base = 1024 * 1024;
1132     for (i = 1; i < guest_info->numa_nodes + 1; ++i) {
1133         mem_base = next_base;
1134         mem_len = guest_info->node_mem[i - 1];
1135         if (i == 1) {
1136             mem_len -= 1024 * 1024;
1137         }
1138         next_base = mem_base + mem_len;
1139 
1140         /* Cut out the ACPI_PCI hole */
1141         if (mem_base <= guest_info->ram_size_below_4g &&
1142             next_base > guest_info->ram_size_below_4g) {
1143             mem_len -= next_base - guest_info->ram_size_below_4g;
1144             if (mem_len > 0) {
1145                 numamem = acpi_data_push(table_data, sizeof *numamem);
1146                 acpi_build_srat_memory(numamem, mem_base, mem_len, i - 1,
1147                                        MEM_AFFINITY_ENABLED);
1148             }
1149             mem_base = 1ULL << 32;
1150             mem_len = next_base - guest_info->ram_size_below_4g;
1151             next_base += (1ULL << 32) - guest_info->ram_size_below_4g;
1152         }
1153         numamem = acpi_data_push(table_data, sizeof *numamem);
1154         acpi_build_srat_memory(numamem, mem_base, mem_len, i - 1,
1155                                MEM_AFFINITY_ENABLED);
1156     }
1157     slots = (table_data->len - numa_start) / sizeof *numamem;
1158     for (; slots < guest_info->numa_nodes + 2; slots++) {
1159         numamem = acpi_data_push(table_data, sizeof *numamem);
1160         acpi_build_srat_memory(numamem, 0, 0, 0, MEM_AFFINITY_NOFLAGS);
1161     }
1162 
1163     /*
1164      * Entry is required for Windows to enable memory hotplug in OS.
1165      * Memory devices may override proximity set by this entry,
1166      * providing _PXM method if necessary.
1167      */
1168     if (hotplugabble_address_space_size) {
1169         numamem = acpi_data_push(table_data, sizeof *numamem);
1170         acpi_build_srat_memory(numamem, pcms->hotplug_memory_base,
1171                                hotplugabble_address_space_size, 0,
1172                                MEM_AFFINITY_HOTPLUGGABLE |
1173                                MEM_AFFINITY_ENABLED);
1174     }
1175 
1176     build_header(linker, table_data,
1177                  (void *)(table_data->data + srat_start),
1178                  "SRAT",
1179                  table_data->len - srat_start, 1);
1180 }
1181 
1182 static void
1183 build_mcfg_q35(GArray *table_data, GArray *linker, AcpiMcfgInfo *info)
1184 {
1185     AcpiTableMcfg *mcfg;
1186     const char *sig;
1187     int len = sizeof(*mcfg) + 1 * sizeof(mcfg->allocation[0]);
1188 
1189     mcfg = acpi_data_push(table_data, len);
1190     mcfg->allocation[0].address = cpu_to_le64(info->mcfg_base);
1191     /* Only a single allocation so no need to play with segments */
1192     mcfg->allocation[0].pci_segment = cpu_to_le16(0);
1193     mcfg->allocation[0].start_bus_number = 0;
1194     mcfg->allocation[0].end_bus_number = PCIE_MMCFG_BUS(info->mcfg_size - 1);
1195 
1196     /* MCFG is used for ECAM which can be enabled or disabled by guest.
1197      * To avoid table size changes (which create migration issues),
1198      * always create the table even if there are no allocations,
1199      * but set the signature to a reserved value in this case.
1200      * ACPI spec requires OSPMs to ignore such tables.
1201      */
1202     if (info->mcfg_base == PCIE_BASE_ADDR_UNMAPPED) {
1203         /* Reserved signature: ignored by OSPM */
1204         sig = "QEMU";
1205     } else {
1206         sig = "MCFG";
1207     }
1208     build_header(linker, table_data, (void *)mcfg, sig, len, 1);
1209 }
1210 
1211 static void
1212 build_dmar_q35(GArray *table_data, GArray *linker)
1213 {
1214     int dmar_start = table_data->len;
1215 
1216     AcpiTableDmar *dmar;
1217     AcpiDmarHardwareUnit *drhd;
1218 
1219     dmar = acpi_data_push(table_data, sizeof(*dmar));
1220     dmar->host_address_width = VTD_HOST_ADDRESS_WIDTH - 1;
1221     dmar->flags = 0;    /* No intr_remap for now */
1222 
1223     /* DMAR Remapping Hardware Unit Definition structure */
1224     drhd = acpi_data_push(table_data, sizeof(*drhd));
1225     drhd->type = cpu_to_le16(ACPI_DMAR_TYPE_HARDWARE_UNIT);
1226     drhd->length = cpu_to_le16(sizeof(*drhd));   /* No device scope now */
1227     drhd->flags = ACPI_DMAR_INCLUDE_PCI_ALL;
1228     drhd->pci_segment = cpu_to_le16(0);
1229     drhd->address = cpu_to_le64(Q35_HOST_BRIDGE_IOMMU_ADDR);
1230 
1231     build_header(linker, table_data, (void *)(table_data->data + dmar_start),
1232                  "DMAR", table_data->len - dmar_start, 1);
1233 }
1234 
1235 static void
1236 build_dsdt(GArray *table_data, GArray *linker, AcpiMiscInfo *misc)
1237 {
1238     AcpiTableHeader *dsdt;
1239 
1240     assert(misc->dsdt_code && misc->dsdt_size);
1241 
1242     dsdt = acpi_data_push(table_data, misc->dsdt_size);
1243     memcpy(dsdt, misc->dsdt_code, misc->dsdt_size);
1244 
1245     memset(dsdt, 0, sizeof *dsdt);
1246     build_header(linker, table_data, dsdt, "DSDT",
1247                  misc->dsdt_size, 1);
1248 }
1249 
1250 /* Build final rsdt table */
1251 static void
1252 build_rsdt(GArray *table_data, GArray *linker, GArray *table_offsets)
1253 {
1254     AcpiRsdtDescriptorRev1 *rsdt;
1255     size_t rsdt_len;
1256     int i;
1257 
1258     rsdt_len = sizeof(*rsdt) + sizeof(uint32_t) * table_offsets->len;
1259     rsdt = acpi_data_push(table_data, rsdt_len);
1260     memcpy(rsdt->table_offset_entry, table_offsets->data,
1261            sizeof(uint32_t) * table_offsets->len);
1262     for (i = 0; i < table_offsets->len; ++i) {
1263         /* rsdt->table_offset_entry to be filled by Guest linker */
1264         bios_linker_loader_add_pointer(linker,
1265                                        ACPI_BUILD_TABLE_FILE,
1266                                        ACPI_BUILD_TABLE_FILE,
1267                                        table_data, &rsdt->table_offset_entry[i],
1268                                        sizeof(uint32_t));
1269     }
1270     build_header(linker, table_data,
1271                  (void *)rsdt, "RSDT", rsdt_len, 1);
1272 }
1273 
1274 static GArray *
1275 build_rsdp(GArray *rsdp_table, GArray *linker, unsigned rsdt)
1276 {
1277     AcpiRsdpDescriptor *rsdp = acpi_data_push(rsdp_table, sizeof *rsdp);
1278 
1279     bios_linker_loader_alloc(linker, ACPI_BUILD_RSDP_FILE, 16,
1280                              true /* fseg memory */);
1281 
1282     memcpy(&rsdp->signature, "RSD PTR ", 8);
1283     memcpy(rsdp->oem_id, ACPI_BUILD_APPNAME6, 6);
1284     rsdp->rsdt_physical_address = cpu_to_le32(rsdt);
1285     /* Address to be filled by Guest linker */
1286     bios_linker_loader_add_pointer(linker, ACPI_BUILD_RSDP_FILE,
1287                                    ACPI_BUILD_TABLE_FILE,
1288                                    rsdp_table, &rsdp->rsdt_physical_address,
1289                                    sizeof rsdp->rsdt_physical_address);
1290     rsdp->checksum = 0;
1291     /* Checksum to be filled by Guest linker */
1292     bios_linker_loader_add_checksum(linker, ACPI_BUILD_RSDP_FILE,
1293                                     rsdp, rsdp, sizeof *rsdp, &rsdp->checksum);
1294 
1295     return rsdp_table;
1296 }
1297 
1298 typedef
1299 struct AcpiBuildTables {
1300     GArray *table_data;
1301     GArray *rsdp;
1302     GArray *tcpalog;
1303     GArray *linker;
1304 } AcpiBuildTables;
1305 
1306 static inline void acpi_build_tables_init(AcpiBuildTables *tables)
1307 {
1308     tables->rsdp = g_array_new(false, true /* clear */, 1);
1309     tables->table_data = g_array_new(false, true /* clear */, 1);
1310     tables->tcpalog = g_array_new(false, true /* clear */, 1);
1311     tables->linker = bios_linker_loader_init();
1312 }
1313 
1314 static inline void acpi_build_tables_cleanup(AcpiBuildTables *tables, bool mfre)
1315 {
1316     void *linker_data = bios_linker_loader_cleanup(tables->linker);
1317     g_free(linker_data);
1318     g_array_free(tables->rsdp, true);
1319     g_array_free(tables->table_data, true);
1320     g_array_free(tables->tcpalog, mfre);
1321 }
1322 
1323 typedef
1324 struct AcpiBuildState {
1325     /* Copy of table in RAM (for patching). */
1326     ram_addr_t table_ram;
1327     /* Is table patched? */
1328     uint8_t patched;
1329     PcGuestInfo *guest_info;
1330     void *rsdp;
1331     ram_addr_t rsdp_ram;
1332     ram_addr_t linker_ram;
1333 } AcpiBuildState;
1334 
1335 static bool acpi_get_mcfg(AcpiMcfgInfo *mcfg)
1336 {
1337     Object *pci_host;
1338     QObject *o;
1339     bool ambiguous;
1340 
1341     pci_host = object_resolve_path_type("", TYPE_PCI_HOST_BRIDGE, &ambiguous);
1342     g_assert(!ambiguous);
1343     g_assert(pci_host);
1344 
1345     o = object_property_get_qobject(pci_host, PCIE_HOST_MCFG_BASE, NULL);
1346     if (!o) {
1347         return false;
1348     }
1349     mcfg->mcfg_base = qint_get_int(qobject_to_qint(o));
1350     qobject_decref(o);
1351 
1352     o = object_property_get_qobject(pci_host, PCIE_HOST_MCFG_SIZE, NULL);
1353     assert(o);
1354     mcfg->mcfg_size = qint_get_int(qobject_to_qint(o));
1355     qobject_decref(o);
1356     return true;
1357 }
1358 
1359 static bool acpi_has_iommu(void)
1360 {
1361     bool ambiguous;
1362     Object *intel_iommu;
1363 
1364     intel_iommu = object_resolve_path_type("", TYPE_INTEL_IOMMU_DEVICE,
1365                                            &ambiguous);
1366     return intel_iommu && !ambiguous;
1367 }
1368 
1369 static
1370 void acpi_build(PcGuestInfo *guest_info, AcpiBuildTables *tables)
1371 {
1372     GArray *table_offsets;
1373     unsigned facs, ssdt, dsdt, rsdt;
1374     AcpiCpuInfo cpu;
1375     AcpiPmInfo pm;
1376     AcpiMiscInfo misc;
1377     AcpiMcfgInfo mcfg;
1378     PcPciInfo pci;
1379     uint8_t *u;
1380     size_t aml_len = 0;
1381     GArray *tables_blob = tables->table_data;
1382 
1383     acpi_get_cpu_info(&cpu);
1384     acpi_get_pm_info(&pm);
1385     acpi_get_dsdt(&misc);
1386     acpi_get_misc_info(&misc);
1387     acpi_get_pci_info(&pci);
1388 
1389     table_offsets = g_array_new(false, true /* clear */,
1390                                         sizeof(uint32_t));
1391     ACPI_BUILD_DPRINTF("init ACPI tables\n");
1392 
1393     bios_linker_loader_alloc(tables->linker, ACPI_BUILD_TABLE_FILE,
1394                              64 /* Ensure FACS is aligned */,
1395                              false /* high memory */);
1396 
1397     /*
1398      * FACS is pointed to by FADT.
1399      * We place it first since it's the only table that has alignment
1400      * requirements.
1401      */
1402     facs = tables_blob->len;
1403     build_facs(tables_blob, tables->linker, guest_info);
1404 
1405     /* DSDT is pointed to by FADT */
1406     dsdt = tables_blob->len;
1407     build_dsdt(tables_blob, tables->linker, &misc);
1408 
1409     /* Count the size of the DSDT and SSDT, we will need it for legacy
1410      * sizing of ACPI tables.
1411      */
1412     aml_len += tables_blob->len - dsdt;
1413 
1414     /* ACPI tables pointed to by RSDT */
1415     acpi_add_table(table_offsets, tables_blob);
1416     build_fadt(tables_blob, tables->linker, &pm, facs, dsdt);
1417 
1418     ssdt = tables_blob->len;
1419     acpi_add_table(table_offsets, tables_blob);
1420     build_ssdt(tables_blob, tables->linker, &cpu, &pm, &misc, &pci,
1421                guest_info);
1422     aml_len += tables_blob->len - ssdt;
1423 
1424     acpi_add_table(table_offsets, tables_blob);
1425     build_madt(tables_blob, tables->linker, &cpu, guest_info);
1426 
1427     if (misc.has_hpet) {
1428         acpi_add_table(table_offsets, tables_blob);
1429         build_hpet(tables_blob, tables->linker);
1430     }
1431     if (misc.has_tpm) {
1432         acpi_add_table(table_offsets, tables_blob);
1433         build_tpm_tcpa(tables_blob, tables->linker, tables->tcpalog);
1434 
1435         acpi_add_table(table_offsets, tables_blob);
1436         build_tpm_ssdt(tables_blob, tables->linker);
1437     }
1438     if (guest_info->numa_nodes) {
1439         acpi_add_table(table_offsets, tables_blob);
1440         build_srat(tables_blob, tables->linker, guest_info);
1441     }
1442     if (acpi_get_mcfg(&mcfg)) {
1443         acpi_add_table(table_offsets, tables_blob);
1444         build_mcfg_q35(tables_blob, tables->linker, &mcfg);
1445     }
1446     if (acpi_has_iommu()) {
1447         acpi_add_table(table_offsets, tables_blob);
1448         build_dmar_q35(tables_blob, tables->linker);
1449     }
1450 
1451     /* Add tables supplied by user (if any) */
1452     for (u = acpi_table_first(); u; u = acpi_table_next(u)) {
1453         unsigned len = acpi_table_len(u);
1454 
1455         acpi_add_table(table_offsets, tables_blob);
1456         g_array_append_vals(tables_blob, u, len);
1457     }
1458 
1459     /* RSDT is pointed to by RSDP */
1460     rsdt = tables_blob->len;
1461     build_rsdt(tables_blob, tables->linker, table_offsets);
1462 
1463     /* RSDP is in FSEG memory, so allocate it separately */
1464     build_rsdp(tables->rsdp, tables->linker, rsdt);
1465 
1466     /* We'll expose it all to Guest so we want to reduce
1467      * chance of size changes.
1468      *
1469      * We used to align the tables to 4k, but of course this would
1470      * too simple to be enough.  4k turned out to be too small an
1471      * alignment very soon, and in fact it is almost impossible to
1472      * keep the table size stable for all (max_cpus, max_memory_slots)
1473      * combinations.  So the table size is always 64k for pc-i440fx-2.1
1474      * and we give an error if the table grows beyond that limit.
1475      *
1476      * We still have the problem of migrating from "-M pc-i440fx-2.0".  For
1477      * that, we exploit the fact that QEMU 2.1 generates _smaller_ tables
1478      * than 2.0 and we can always pad the smaller tables with zeros.  We can
1479      * then use the exact size of the 2.0 tables.
1480      *
1481      * All this is for PIIX4, since QEMU 2.0 didn't support Q35 migration.
1482      */
1483     if (guest_info->legacy_acpi_table_size) {
1484         /* Subtracting aml_len gives the size of fixed tables.  Then add the
1485          * size of the PIIX4 DSDT/SSDT in QEMU 2.0.
1486          */
1487         int legacy_aml_len =
1488             guest_info->legacy_acpi_table_size +
1489             ACPI_BUILD_LEGACY_CPU_AML_SIZE * max_cpus;
1490         int legacy_table_size =
1491             ROUND_UP(tables_blob->len - aml_len + legacy_aml_len,
1492                      ACPI_BUILD_ALIGN_SIZE);
1493         if (tables_blob->len > legacy_table_size) {
1494             /* Should happen only with PCI bridges and -M pc-i440fx-2.0.  */
1495             error_report("Warning: migration may not work.");
1496         }
1497         g_array_set_size(tables_blob, legacy_table_size);
1498     } else {
1499         /* Make sure we have a buffer in case we need to resize the tables. */
1500         if (tables_blob->len > ACPI_BUILD_TABLE_SIZE / 2) {
1501             /* As of QEMU 2.1, this fires with 160 VCPUs and 255 memory slots.  */
1502             error_report("Warning: ACPI tables are larger than 64k.");
1503             error_report("Warning: migration may not work.");
1504             error_report("Warning: please remove CPUs, NUMA nodes, "
1505                          "memory slots or PCI bridges.");
1506         }
1507         acpi_align_size(tables_blob, ACPI_BUILD_TABLE_SIZE);
1508     }
1509 
1510     acpi_align_size(tables->linker, ACPI_BUILD_ALIGN_SIZE);
1511 
1512     /* Cleanup memory that's no longer used. */
1513     g_array_free(table_offsets, true);
1514 }
1515 
1516 static void acpi_ram_update(ram_addr_t ram, GArray *data)
1517 {
1518     uint32_t size = acpi_data_len(data);
1519 
1520     /* Make sure RAM size is correct - in case it got changed e.g. by migration */
1521     qemu_ram_resize(ram, size, &error_abort);
1522 
1523     memcpy(qemu_get_ram_ptr(ram), data->data, size);
1524     cpu_physical_memory_set_dirty_range_nocode(ram, size);
1525 }
1526 
1527 static void acpi_build_update(void *build_opaque, uint32_t offset)
1528 {
1529     AcpiBuildState *build_state = build_opaque;
1530     AcpiBuildTables tables;
1531 
1532     /* No state to update or already patched? Nothing to do. */
1533     if (!build_state || build_state->patched) {
1534         return;
1535     }
1536     build_state->patched = 1;
1537 
1538     acpi_build_tables_init(&tables);
1539 
1540     acpi_build(build_state->guest_info, &tables);
1541 
1542     acpi_ram_update(build_state->table_ram, tables.table_data);
1543 
1544     if (build_state->rsdp) {
1545         memcpy(build_state->rsdp, tables.rsdp->data, acpi_data_len(tables.rsdp));
1546     } else {
1547         acpi_ram_update(build_state->rsdp_ram, tables.rsdp);
1548     }
1549 
1550     acpi_ram_update(build_state->linker_ram, tables.linker);
1551     acpi_build_tables_cleanup(&tables, true);
1552 }
1553 
1554 static void acpi_build_reset(void *build_opaque)
1555 {
1556     AcpiBuildState *build_state = build_opaque;
1557     build_state->patched = 0;
1558 }
1559 
1560 static ram_addr_t acpi_add_rom_blob(AcpiBuildState *build_state, GArray *blob,
1561                                const char *name, uint64_t max_size)
1562 {
1563     return rom_add_blob(name, blob->data, acpi_data_len(blob), max_size, -1,
1564                         name, acpi_build_update, build_state);
1565 }
1566 
1567 static const VMStateDescription vmstate_acpi_build = {
1568     .name = "acpi_build",
1569     .version_id = 1,
1570     .minimum_version_id = 1,
1571     .fields = (VMStateField[]) {
1572         VMSTATE_UINT8(patched, AcpiBuildState),
1573         VMSTATE_END_OF_LIST()
1574     },
1575 };
1576 
1577 void acpi_setup(PcGuestInfo *guest_info)
1578 {
1579     AcpiBuildTables tables;
1580     AcpiBuildState *build_state;
1581 
1582     if (!guest_info->fw_cfg) {
1583         ACPI_BUILD_DPRINTF("No fw cfg. Bailing out.\n");
1584         return;
1585     }
1586 
1587     if (!guest_info->has_acpi_build) {
1588         ACPI_BUILD_DPRINTF("ACPI build disabled. Bailing out.\n");
1589         return;
1590     }
1591 
1592     if (!acpi_enabled) {
1593         ACPI_BUILD_DPRINTF("ACPI disabled. Bailing out.\n");
1594         return;
1595     }
1596 
1597     build_state = g_malloc0(sizeof *build_state);
1598 
1599     build_state->guest_info = guest_info;
1600 
1601     acpi_set_pci_info();
1602 
1603     acpi_build_tables_init(&tables);
1604     acpi_build(build_state->guest_info, &tables);
1605 
1606     /* Now expose it all to Guest */
1607     build_state->table_ram = acpi_add_rom_blob(build_state, tables.table_data,
1608                                                ACPI_BUILD_TABLE_FILE,
1609                                                ACPI_BUILD_TABLE_MAX_SIZE);
1610     assert(build_state->table_ram != RAM_ADDR_MAX);
1611 
1612     build_state->linker_ram =
1613         acpi_add_rom_blob(build_state, tables.linker, "etc/table-loader", 0);
1614 
1615     fw_cfg_add_file(guest_info->fw_cfg, ACPI_BUILD_TPMLOG_FILE,
1616                     tables.tcpalog->data, acpi_data_len(tables.tcpalog));
1617 
1618     if (!guest_info->rsdp_in_ram) {
1619         /*
1620          * Keep for compatibility with old machine types.
1621          * Though RSDP is small, its contents isn't immutable, so
1622          * we'll update it along with the rest of tables on guest access.
1623          */
1624         uint32_t rsdp_size = acpi_data_len(tables.rsdp);
1625 
1626         build_state->rsdp = g_memdup(tables.rsdp->data, rsdp_size);
1627         fw_cfg_add_file_callback(guest_info->fw_cfg, ACPI_BUILD_RSDP_FILE,
1628                                  acpi_build_update, build_state,
1629                                  build_state->rsdp, rsdp_size);
1630         build_state->rsdp_ram = (ram_addr_t)-1;
1631     } else {
1632         build_state->rsdp = NULL;
1633         build_state->rsdp_ram = acpi_add_rom_blob(build_state, tables.rsdp,
1634                                                   ACPI_BUILD_RSDP_FILE, 0);
1635     }
1636 
1637     qemu_register_reset(acpi_build_reset, build_state);
1638     acpi_build_reset(build_state);
1639     vmstate_register(NULL, 0, &vmstate_acpi_build, build_state);
1640 
1641     /* Cleanup tables but don't free the memory: we track it
1642      * in build_state.
1643      */
1644     acpi_build_tables_cleanup(&tables, false);
1645 }
1646