xref: /openbmc/qemu/hw/i386/acpi-build.c (revision b8eb5512fd8a115f164edbbe897cdf8884920ccb)
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 
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 /* Reserve RAM space for tables: add another order of magnitude. */
73 #define ACPI_BUILD_TABLE_MAX_SIZE         0x200000
74 
75 /* #define DEBUG_ACPI_BUILD */
76 #ifdef DEBUG_ACPI_BUILD
77 #define ACPI_BUILD_DPRINTF(fmt, ...)        \
78     do {printf("ACPI_BUILD: " fmt, ## __VA_ARGS__); } while (0)
79 #else
80 #define ACPI_BUILD_DPRINTF(fmt, ...)
81 #endif
82 
83 typedef struct AcpiCpuInfo {
84     DECLARE_BITMAP(found_cpus, ACPI_CPU_HOTPLUG_ID_LIMIT);
85 } AcpiCpuInfo;
86 
87 typedef struct AcpiMcfgInfo {
88     uint64_t mcfg_base;
89     uint32_t mcfg_size;
90 } AcpiMcfgInfo;
91 
92 typedef struct AcpiPmInfo {
93     bool s3_disabled;
94     bool s4_disabled;
95     bool pcihp_bridge_en;
96     uint8_t s4_val;
97     uint16_t sci_int;
98     uint8_t acpi_enable_cmd;
99     uint8_t acpi_disable_cmd;
100     uint32_t gpe0_blk;
101     uint32_t gpe0_blk_len;
102     uint32_t io_base;
103     uint16_t cpu_hp_io_base;
104     uint16_t cpu_hp_io_len;
105     uint16_t mem_hp_io_base;
106     uint16_t mem_hp_io_len;
107     uint16_t pcihp_io_base;
108     uint16_t pcihp_io_len;
109 } AcpiPmInfo;
110 
111 typedef struct AcpiMiscInfo {
112     bool has_hpet;
113     bool has_tpm;
114     const unsigned char *dsdt_code;
115     unsigned dsdt_size;
116     uint16_t pvpanic_port;
117     uint16_t applesmc_io_base;
118 } AcpiMiscInfo;
119 
120 typedef struct AcpiBuildPciBusHotplugState {
121     GArray *device_table;
122     GArray *notify_table;
123     struct AcpiBuildPciBusHotplugState *parent;
124     bool pcihp_bridge_en;
125 } AcpiBuildPciBusHotplugState;
126 
127 static void acpi_get_dsdt(AcpiMiscInfo *info)
128 {
129     Object *piix = piix4_pm_find();
130     Object *lpc = ich9_lpc_find();
131     assert(!!piix != !!lpc);
132 
133     if (piix) {
134         info->dsdt_code = AcpiDsdtAmlCode;
135         info->dsdt_size = sizeof AcpiDsdtAmlCode;
136     }
137     if (lpc) {
138         info->dsdt_code = Q35AcpiDsdtAmlCode;
139         info->dsdt_size = sizeof Q35AcpiDsdtAmlCode;
140     }
141 }
142 
143 static
144 int acpi_add_cpu_info(Object *o, void *opaque)
145 {
146     AcpiCpuInfo *cpu = opaque;
147     uint64_t apic_id;
148 
149     if (object_dynamic_cast(o, TYPE_CPU)) {
150         apic_id = object_property_get_int(o, "apic-id", NULL);
151         assert(apic_id < ACPI_CPU_HOTPLUG_ID_LIMIT);
152 
153         set_bit(apic_id, cpu->found_cpus);
154     }
155 
156     object_child_foreach(o, acpi_add_cpu_info, opaque);
157     return 0;
158 }
159 
160 static void acpi_get_cpu_info(AcpiCpuInfo *cpu)
161 {
162     Object *root = object_get_root();
163 
164     memset(cpu->found_cpus, 0, sizeof cpu->found_cpus);
165     object_child_foreach(root, acpi_add_cpu_info, cpu);
166 }
167 
168 static void acpi_get_pm_info(AcpiPmInfo *pm)
169 {
170     Object *piix = piix4_pm_find();
171     Object *lpc = ich9_lpc_find();
172     Object *obj = NULL;
173     QObject *o;
174 
175     pm->pcihp_io_base = 0;
176     pm->pcihp_io_len = 0;
177     if (piix) {
178         obj = piix;
179         pm->cpu_hp_io_base = PIIX4_CPU_HOTPLUG_IO_BASE;
180         pm->pcihp_io_base =
181             object_property_get_int(obj, ACPI_PCIHP_IO_BASE_PROP, NULL);
182         pm->pcihp_io_len =
183             object_property_get_int(obj, ACPI_PCIHP_IO_LEN_PROP, NULL);
184     }
185     if (lpc) {
186         obj = lpc;
187         pm->cpu_hp_io_base = ICH9_CPU_HOTPLUG_IO_BASE;
188     }
189     assert(obj);
190 
191     pm->cpu_hp_io_len = ACPI_GPE_PROC_LEN;
192     pm->mem_hp_io_base = ACPI_MEMORY_HOTPLUG_BASE;
193     pm->mem_hp_io_len = ACPI_MEMORY_HOTPLUG_IO_LEN;
194 
195     /* Fill in optional s3/s4 related properties */
196     o = object_property_get_qobject(obj, ACPI_PM_PROP_S3_DISABLED, NULL);
197     if (o) {
198         pm->s3_disabled = qint_get_int(qobject_to_qint(o));
199     } else {
200         pm->s3_disabled = false;
201     }
202     qobject_decref(o);
203     o = object_property_get_qobject(obj, ACPI_PM_PROP_S4_DISABLED, NULL);
204     if (o) {
205         pm->s4_disabled = qint_get_int(qobject_to_qint(o));
206     } else {
207         pm->s4_disabled = false;
208     }
209     qobject_decref(o);
210     o = object_property_get_qobject(obj, ACPI_PM_PROP_S4_VAL, NULL);
211     if (o) {
212         pm->s4_val = qint_get_int(qobject_to_qint(o));
213     } else {
214         pm->s4_val = false;
215     }
216     qobject_decref(o);
217 
218     /* Fill in mandatory properties */
219     pm->sci_int = object_property_get_int(obj, ACPI_PM_PROP_SCI_INT, NULL);
220 
221     pm->acpi_enable_cmd = object_property_get_int(obj,
222                                                   ACPI_PM_PROP_ACPI_ENABLE_CMD,
223                                                   NULL);
224     pm->acpi_disable_cmd = object_property_get_int(obj,
225                                                   ACPI_PM_PROP_ACPI_DISABLE_CMD,
226                                                   NULL);
227     pm->io_base = object_property_get_int(obj, ACPI_PM_PROP_PM_IO_BASE,
228                                           NULL);
229     pm->gpe0_blk = object_property_get_int(obj, ACPI_PM_PROP_GPE0_BLK,
230                                            NULL);
231     pm->gpe0_blk_len = object_property_get_int(obj, ACPI_PM_PROP_GPE0_BLK_LEN,
232                                                NULL);
233     pm->pcihp_bridge_en =
234         object_property_get_bool(obj, "acpi-pci-hotplug-with-bridge-support",
235                                  NULL);
236 }
237 
238 static void acpi_get_misc_info(AcpiMiscInfo *info)
239 {
240     info->has_hpet = hpet_find();
241     info->has_tpm = tpm_find();
242     info->pvpanic_port = pvpanic_port();
243     info->applesmc_io_base = applesmc_port();
244 }
245 
246 static void acpi_get_pci_info(PcPciInfo *info)
247 {
248     Object *pci_host;
249     bool ambiguous;
250 
251     pci_host = object_resolve_path_type("", TYPE_PCI_HOST_BRIDGE, &ambiguous);
252     g_assert(!ambiguous);
253     g_assert(pci_host);
254 
255     info->w32.begin = object_property_get_int(pci_host,
256                                               PCI_HOST_PROP_PCI_HOLE_START,
257                                               NULL);
258     info->w32.end = object_property_get_int(pci_host,
259                                             PCI_HOST_PROP_PCI_HOLE_END,
260                                             NULL);
261     info->w64.begin = object_property_get_int(pci_host,
262                                               PCI_HOST_PROP_PCI_HOLE64_START,
263                                               NULL);
264     info->w64.end = object_property_get_int(pci_host,
265                                             PCI_HOST_PROP_PCI_HOLE64_END,
266                                             NULL);
267 }
268 
269 #define ACPI_BUILD_APPNAME  "Bochs"
270 #define ACPI_BUILD_APPNAME6 "BOCHS "
271 #define ACPI_BUILD_APPNAME4 "BXPC"
272 
273 #define ACPI_BUILD_TABLE_FILE "etc/acpi/tables"
274 #define ACPI_BUILD_RSDP_FILE "etc/acpi/rsdp"
275 #define ACPI_BUILD_TPMLOG_FILE "etc/tpm/log"
276 
277 static void
278 build_header(GArray *linker, GArray *table_data,
279              AcpiTableHeader *h, const char *sig, int len, uint8_t rev)
280 {
281     memcpy(&h->signature, sig, 4);
282     h->length = cpu_to_le32(len);
283     h->revision = rev;
284     memcpy(h->oem_id, ACPI_BUILD_APPNAME6, 6);
285     memcpy(h->oem_table_id, ACPI_BUILD_APPNAME4, 4);
286     memcpy(h->oem_table_id + 4, sig, 4);
287     h->oem_revision = cpu_to_le32(1);
288     memcpy(h->asl_compiler_id, ACPI_BUILD_APPNAME4, 4);
289     h->asl_compiler_revision = cpu_to_le32(1);
290     h->checksum = 0;
291     /* Checksum to be filled in by Guest linker */
292     bios_linker_loader_add_checksum(linker, ACPI_BUILD_TABLE_FILE,
293                                     table_data->data, h, len, &h->checksum);
294 }
295 
296 /* End here */
297 #define ACPI_PORT_SMI_CMD           0x00b2 /* TODO: this is APM_CNT_IOPORT */
298 
299 static inline void *acpi_data_push(GArray *table_data, unsigned size)
300 {
301     unsigned off = table_data->len;
302     g_array_set_size(table_data, off + size);
303     return table_data->data + off;
304 }
305 
306 static unsigned acpi_data_len(GArray *table)
307 {
308 #if GLIB_CHECK_VERSION(2, 22, 0)
309     assert(g_array_get_element_size(table) == 1);
310 #endif
311     return table->len;
312 }
313 
314 static void acpi_align_size(GArray *blob, unsigned align)
315 {
316     /* Align size to multiple of given size. This reduces the chance
317      * we need to change size in the future (breaking cross version migration).
318      */
319     g_array_set_size(blob, ROUND_UP(acpi_data_len(blob), align));
320 }
321 
322 static inline void acpi_add_table(GArray *table_offsets, GArray *table_data)
323 {
324     uint32_t offset = cpu_to_le32(table_data->len);
325     g_array_append_val(table_offsets, offset);
326 }
327 
328 /* FACS */
329 static void
330 build_facs(GArray *table_data, GArray *linker, PcGuestInfo *guest_info)
331 {
332     AcpiFacsDescriptorRev1 *facs = acpi_data_push(table_data, sizeof *facs);
333     memcpy(&facs->signature, "FACS", 4);
334     facs->length = cpu_to_le32(sizeof(*facs));
335 }
336 
337 /* Load chipset information in FADT */
338 static void fadt_setup(AcpiFadtDescriptorRev1 *fadt, AcpiPmInfo *pm)
339 {
340     fadt->model = 1;
341     fadt->reserved1 = 0;
342     fadt->sci_int = cpu_to_le16(pm->sci_int);
343     fadt->smi_cmd = cpu_to_le32(ACPI_PORT_SMI_CMD);
344     fadt->acpi_enable = pm->acpi_enable_cmd;
345     fadt->acpi_disable = pm->acpi_disable_cmd;
346     /* EVT, CNT, TMR offset matches hw/acpi/core.c */
347     fadt->pm1a_evt_blk = cpu_to_le32(pm->io_base);
348     fadt->pm1a_cnt_blk = cpu_to_le32(pm->io_base + 0x04);
349     fadt->pm_tmr_blk = cpu_to_le32(pm->io_base + 0x08);
350     fadt->gpe0_blk = cpu_to_le32(pm->gpe0_blk);
351     /* EVT, CNT, TMR length matches hw/acpi/core.c */
352     fadt->pm1_evt_len = 4;
353     fadt->pm1_cnt_len = 2;
354     fadt->pm_tmr_len = 4;
355     fadt->gpe0_blk_len = pm->gpe0_blk_len;
356     fadt->plvl2_lat = cpu_to_le16(0xfff); /* C2 state not supported */
357     fadt->plvl3_lat = cpu_to_le16(0xfff); /* C3 state not supported */
358     fadt->flags = cpu_to_le32((1 << ACPI_FADT_F_WBINVD) |
359                               (1 << ACPI_FADT_F_PROC_C1) |
360                               (1 << ACPI_FADT_F_SLP_BUTTON) |
361                               (1 << ACPI_FADT_F_RTC_S4));
362     fadt->flags |= cpu_to_le32(1 << ACPI_FADT_F_USE_PLATFORM_CLOCK);
363     /* APIC destination mode ("Flat Logical") has an upper limit of 8 CPUs
364      * For more than 8 CPUs, "Clustered Logical" mode has to be used
365      */
366     if (max_cpus > 8) {
367         fadt->flags |= cpu_to_le32(1 << ACPI_FADT_F_FORCE_APIC_CLUSTER_MODEL);
368     }
369 }
370 
371 
372 /* FADT */
373 static void
374 build_fadt(GArray *table_data, GArray *linker, AcpiPmInfo *pm,
375            unsigned facs, unsigned dsdt)
376 {
377     AcpiFadtDescriptorRev1 *fadt = acpi_data_push(table_data, sizeof(*fadt));
378 
379     fadt->firmware_ctrl = cpu_to_le32(facs);
380     /* FACS address to be filled by Guest linker */
381     bios_linker_loader_add_pointer(linker, ACPI_BUILD_TABLE_FILE,
382                                    ACPI_BUILD_TABLE_FILE,
383                                    table_data, &fadt->firmware_ctrl,
384                                    sizeof fadt->firmware_ctrl);
385 
386     fadt->dsdt = cpu_to_le32(dsdt);
387     /* DSDT address to be filled by Guest linker */
388     bios_linker_loader_add_pointer(linker, ACPI_BUILD_TABLE_FILE,
389                                    ACPI_BUILD_TABLE_FILE,
390                                    table_data, &fadt->dsdt,
391                                    sizeof fadt->dsdt);
392 
393     fadt_setup(fadt, pm);
394 
395     build_header(linker, table_data,
396                  (void *)fadt, "FACP", sizeof(*fadt), 1);
397 }
398 
399 static void
400 build_madt(GArray *table_data, GArray *linker, AcpiCpuInfo *cpu,
401            PcGuestInfo *guest_info)
402 {
403     int madt_start = table_data->len;
404 
405     AcpiMultipleApicTable *madt;
406     AcpiMadtIoApic *io_apic;
407     AcpiMadtIntsrcovr *intsrcovr;
408     AcpiMadtLocalNmi *local_nmi;
409     int i;
410 
411     madt = acpi_data_push(table_data, sizeof *madt);
412     madt->local_apic_address = cpu_to_le32(APIC_DEFAULT_ADDRESS);
413     madt->flags = cpu_to_le32(1);
414 
415     for (i = 0; i < guest_info->apic_id_limit; i++) {
416         AcpiMadtProcessorApic *apic = acpi_data_push(table_data, sizeof *apic);
417         apic->type = ACPI_APIC_PROCESSOR;
418         apic->length = sizeof(*apic);
419         apic->processor_id = i;
420         apic->local_apic_id = i;
421         if (test_bit(i, cpu->found_cpus)) {
422             apic->flags = cpu_to_le32(1);
423         } else {
424             apic->flags = cpu_to_le32(0);
425         }
426     }
427     io_apic = acpi_data_push(table_data, sizeof *io_apic);
428     io_apic->type = ACPI_APIC_IO;
429     io_apic->length = sizeof(*io_apic);
430 #define ACPI_BUILD_IOAPIC_ID 0x0
431     io_apic->io_apic_id = ACPI_BUILD_IOAPIC_ID;
432     io_apic->address = cpu_to_le32(IO_APIC_DEFAULT_ADDRESS);
433     io_apic->interrupt = cpu_to_le32(0);
434 
435     if (guest_info->apic_xrupt_override) {
436         intsrcovr = acpi_data_push(table_data, sizeof *intsrcovr);
437         intsrcovr->type   = ACPI_APIC_XRUPT_OVERRIDE;
438         intsrcovr->length = sizeof(*intsrcovr);
439         intsrcovr->source = 0;
440         intsrcovr->gsi    = cpu_to_le32(2);
441         intsrcovr->flags  = cpu_to_le16(0); /* conforms to bus specifications */
442     }
443     for (i = 1; i < 16; i++) {
444 #define ACPI_BUILD_PCI_IRQS ((1<<5) | (1<<9) | (1<<10) | (1<<11))
445         if (!(ACPI_BUILD_PCI_IRQS & (1 << i))) {
446             /* No need for a INT source override structure. */
447             continue;
448         }
449         intsrcovr = acpi_data_push(table_data, sizeof *intsrcovr);
450         intsrcovr->type   = ACPI_APIC_XRUPT_OVERRIDE;
451         intsrcovr->length = sizeof(*intsrcovr);
452         intsrcovr->source = i;
453         intsrcovr->gsi    = cpu_to_le32(i);
454         intsrcovr->flags  = cpu_to_le16(0xd); /* active high, level triggered */
455     }
456 
457     local_nmi = acpi_data_push(table_data, sizeof *local_nmi);
458     local_nmi->type         = ACPI_APIC_LOCAL_NMI;
459     local_nmi->length       = sizeof(*local_nmi);
460     local_nmi->processor_id = 0xff; /* all processors */
461     local_nmi->flags        = cpu_to_le16(0);
462     local_nmi->lint         = 1; /* ACPI_LINT1 */
463 
464     build_header(linker, table_data,
465                  (void *)(table_data->data + madt_start), "APIC",
466                  table_data->len - madt_start, 1);
467 }
468 
469 #include "hw/i386/ssdt-tpm.hex"
470 
471 /* Assign BSEL property to all buses.  In the future, this can be changed
472  * to only assign to buses that support hotplug.
473  */
474 static void *acpi_set_bsel(PCIBus *bus, void *opaque)
475 {
476     unsigned *bsel_alloc = opaque;
477     unsigned *bus_bsel;
478 
479     if (qbus_is_hotpluggable(BUS(bus))) {
480         bus_bsel = g_malloc(sizeof *bus_bsel);
481 
482         *bus_bsel = (*bsel_alloc)++;
483         object_property_add_uint32_ptr(OBJECT(bus), ACPI_PCIHP_PROP_BSEL,
484                                        bus_bsel, NULL);
485     }
486 
487     return bsel_alloc;
488 }
489 
490 static void acpi_set_pci_info(void)
491 {
492     PCIBus *bus = find_i440fx(); /* TODO: Q35 support */
493     unsigned bsel_alloc = 0;
494 
495     if (bus) {
496         /* Scan all PCI buses. Set property to enable acpi based hotplug. */
497         pci_for_each_bus_depth_first(bus, acpi_set_bsel, NULL, &bsel_alloc);
498     }
499 }
500 
501 static void build_append_pcihp_notify_entry(Aml *method, int slot)
502 {
503     Aml *if_ctx;
504     int32_t devfn = PCI_DEVFN(slot, 0);
505 
506     if_ctx = aml_if(aml_and(aml_arg(0), aml_int(0x1U << slot)));
507     aml_append(if_ctx, aml_notify(aml_name("S%.02X", devfn), aml_arg(1)));
508     aml_append(method, if_ctx);
509 }
510 
511 static void build_append_pci_bus_devices(Aml *parent_scope, PCIBus *bus,
512                                          bool pcihp_bridge_en)
513 {
514     Aml *dev, *notify_method, *method;
515     QObject *bsel;
516     PCIBus *sec;
517     int i;
518 
519     bsel = object_property_get_qobject(OBJECT(bus), ACPI_PCIHP_PROP_BSEL, NULL);
520     if (bsel) {
521         int64_t bsel_val = qint_get_int(qobject_to_qint(bsel));
522 
523         aml_append(parent_scope, aml_name_decl("BSEL", aml_int(bsel_val)));
524         notify_method = aml_method("DVNT", 2);
525     }
526 
527     for (i = 0; i < ARRAY_SIZE(bus->devices); i += PCI_FUNC_MAX) {
528         DeviceClass *dc;
529         PCIDeviceClass *pc;
530         PCIDevice *pdev = bus->devices[i];
531         int slot = PCI_SLOT(i);
532         bool hotplug_enabled_dev;
533         bool bridge_in_acpi;
534 
535         if (!pdev) {
536             if (bsel) { /* add hotplug slots for non present devices */
537                 dev = aml_device("S%.02X", PCI_DEVFN(slot, 0));
538                 aml_append(dev, aml_name_decl("_SUN", aml_int(slot)));
539                 aml_append(dev, aml_name_decl("_ADR", aml_int(slot << 16)));
540                 method = aml_method("_EJ0", 1);
541                 aml_append(method,
542                     aml_call2("PCEJ", aml_name("BSEL"), aml_name("_SUN"))
543                 );
544                 aml_append(dev, method);
545                 aml_append(parent_scope, dev);
546 
547                 build_append_pcihp_notify_entry(notify_method, slot);
548             }
549             continue;
550         }
551 
552         pc = PCI_DEVICE_GET_CLASS(pdev);
553         dc = DEVICE_GET_CLASS(pdev);
554 
555         /* When hotplug for bridges is enabled, bridges are
556          * described in ACPI separately (see build_pci_bus_end).
557          * In this case they aren't themselves hot-pluggable.
558          * Hotplugged bridges *are* hot-pluggable.
559          */
560         bridge_in_acpi = pc->is_bridge && pcihp_bridge_en &&
561             !DEVICE(pdev)->hotplugged;
562 
563         hotplug_enabled_dev = bsel && dc->hotpluggable && !bridge_in_acpi;
564 
565         if (pc->class_id == PCI_CLASS_BRIDGE_ISA) {
566             continue;
567         }
568 
569         /* start to compose PCI slot descriptor */
570         dev = aml_device("S%.02X", PCI_DEVFN(slot, 0));
571         aml_append(dev, aml_name_decl("_ADR", aml_int(slot << 16)));
572 
573         if (pc->class_id == PCI_CLASS_DISPLAY_VGA) {
574             /* add VGA specific AML methods */
575             int s3d;
576 
577             if (object_dynamic_cast(OBJECT(pdev), "qxl-vga")) {
578                 s3d = 3;
579             } else {
580                 s3d = 0;
581             }
582 
583             method = aml_method("_S1D", 0);
584             aml_append(method, aml_return(aml_int(0)));
585             aml_append(dev, method);
586 
587             method = aml_method("_S2D", 0);
588             aml_append(method, aml_return(aml_int(0)));
589             aml_append(dev, method);
590 
591             method = aml_method("_S3D", 0);
592             aml_append(method, aml_return(aml_int(s3d)));
593             aml_append(dev, method);
594         } else if (hotplug_enabled_dev) {
595             /* add _SUN/_EJ0 to make slot hotpluggable  */
596             aml_append(dev, aml_name_decl("_SUN", aml_int(slot)));
597 
598             method = aml_method("_EJ0", 1);
599             aml_append(method,
600                 aml_call2("PCEJ", aml_name("BSEL"), aml_name("_SUN"))
601             );
602             aml_append(dev, method);
603 
604             if (bsel) {
605                 build_append_pcihp_notify_entry(notify_method, slot);
606             }
607         } else if (bridge_in_acpi) {
608             /*
609              * device is coldplugged bridge,
610              * add child device descriptions into its scope
611              */
612             PCIBus *sec_bus = pci_bridge_get_sec_bus(PCI_BRIDGE(pdev));
613 
614             build_append_pci_bus_devices(dev, sec_bus, pcihp_bridge_en);
615         }
616         /* slot descriptor has been composed, add it into parent context */
617         aml_append(parent_scope, dev);
618     }
619 
620     if (bsel) {
621         aml_append(parent_scope, notify_method);
622     }
623 
624     /* Append PCNT method to notify about events on local and child buses.
625      * Add unconditionally for root since DSDT expects it.
626      */
627     method = aml_method("PCNT", 0);
628 
629     /* If bus supports hotplug select it and notify about local events */
630     if (bsel) {
631         int64_t bsel_val = qint_get_int(qobject_to_qint(bsel));
632         aml_append(method, aml_store(aml_int(bsel_val), aml_name("BNUM")));
633         aml_append(method,
634             aml_call2("DVNT", aml_name("PCIU"), aml_int(1) /* Device Check */)
635         );
636         aml_append(method,
637             aml_call2("DVNT", aml_name("PCID"), aml_int(3)/* Eject Request */)
638         );
639     }
640 
641     /* Notify about child bus events in any case */
642     if (pcihp_bridge_en) {
643         QLIST_FOREACH(sec, &bus->child, sibling) {
644             int32_t devfn = sec->parent_dev->devfn;
645 
646             aml_append(method, aml_name("^S%.02X.PCNT", devfn));
647         }
648     }
649     aml_append(parent_scope, method);
650 }
651 
652 static void
653 build_ssdt(GArray *table_data, GArray *linker,
654            AcpiCpuInfo *cpu, AcpiPmInfo *pm, AcpiMiscInfo *misc,
655            PcPciInfo *pci, PcGuestInfo *guest_info)
656 {
657     MachineState *machine = MACHINE(qdev_get_machine());
658     uint32_t nr_mem = machine->ram_slots;
659     unsigned acpi_cpus = guest_info->apic_id_limit;
660     Aml *ssdt, *sb_scope, *scope, *pkg, *dev, *method, *crs, *field, *ifctx;
661     int i;
662 
663     ssdt = init_aml_allocator();
664     /* The current AML generator can cover the APIC ID range [0..255],
665      * inclusive, for VCPU hotplug. */
666     QEMU_BUILD_BUG_ON(ACPI_CPU_HOTPLUG_ID_LIMIT > 256);
667     g_assert(acpi_cpus <= ACPI_CPU_HOTPLUG_ID_LIMIT);
668 
669     /* Reserve space for header */
670     acpi_data_push(ssdt->buf, sizeof(AcpiTableHeader));
671 
672     scope = aml_scope("\\_SB.PCI0");
673     /* build PCI0._CRS */
674     crs = aml_resource_template();
675     aml_append(crs,
676         aml_word_bus_number(aml_min_fixed, aml_max_fixed, aml_pos_decode,
677                             0x0000, 0x0000, 0x00FF, 0x0000, 0x0100));
678     aml_append(crs, aml_io(aml_decode16, 0x0CF8, 0x0CF8, 0x01, 0x08));
679 
680     aml_append(crs,
681         aml_word_io(aml_min_fixed, aml_max_fixed,
682                     aml_pos_decode, aml_entire_range,
683                     0x0000, 0x0000, 0x0CF7, 0x0000, 0x0CF8));
684     aml_append(crs,
685         aml_word_io(aml_min_fixed, aml_max_fixed,
686                     aml_pos_decode, aml_entire_range,
687                     0x0000, 0x0D00, 0xFFFF, 0x0000, 0xF300));
688     aml_append(crs,
689         aml_dword_memory(aml_pos_decode, aml_min_fixed, aml_max_fixed,
690                          aml_cacheable, aml_ReadWrite,
691                          0, 0x000A0000, 0x000BFFFF, 0, 0x00020000));
692     aml_append(crs,
693         aml_dword_memory(aml_pos_decode, aml_min_fixed, aml_max_fixed,
694                          aml_non_cacheable, aml_ReadWrite,
695                          0, pci->w32.begin, pci->w32.end - 1, 0,
696                          pci->w32.end - pci->w32.begin));
697     if (pci->w64.begin) {
698         aml_append(crs,
699             aml_qword_memory(aml_pos_decode, aml_min_fixed, aml_max_fixed,
700                              aml_cacheable, aml_ReadWrite,
701                              0, pci->w64.begin, pci->w64.end - 1, 0,
702                              pci->w64.end - pci->w64.begin));
703     }
704     aml_append(scope, aml_name_decl("_CRS", crs));
705 
706     /* reserve GPE0 block resources */
707     dev = aml_device("GPE0");
708     aml_append(dev, aml_name_decl("_HID", aml_string("PNP0A06")));
709     aml_append(dev, aml_name_decl("_UID", aml_string("GPE0 resources")));
710     /* device present, functioning, decoding, not shown in UI */
711     aml_append(dev, aml_name_decl("_STA", aml_int(0xB)));
712     crs = aml_resource_template();
713     aml_append(crs,
714         aml_io(aml_decode16, pm->gpe0_blk, pm->gpe0_blk, 1, pm->gpe0_blk_len)
715     );
716     aml_append(dev, aml_name_decl("_CRS", crs));
717     aml_append(scope, dev);
718 
719     /* reserve PCIHP resources */
720     if (pm->pcihp_io_len) {
721         dev = aml_device("PHPR");
722         aml_append(dev, aml_name_decl("_HID", aml_string("PNP0A06")));
723         aml_append(dev,
724             aml_name_decl("_UID", aml_string("PCI Hotplug resources")));
725         /* device present, functioning, decoding, not shown in UI */
726         aml_append(dev, aml_name_decl("_STA", aml_int(0xB)));
727         crs = aml_resource_template();
728         aml_append(crs,
729             aml_io(aml_decode16, pm->pcihp_io_base, pm->pcihp_io_base, 1,
730                    pm->pcihp_io_len)
731         );
732         aml_append(dev, aml_name_decl("_CRS", crs));
733         aml_append(scope, dev);
734     }
735     aml_append(ssdt, scope);
736 
737     /*  create S3_ / S4_ / S5_ packages if necessary */
738     scope = aml_scope("\\");
739     if (!pm->s3_disabled) {
740         pkg = aml_package(4);
741         aml_append(pkg, aml_int(1)); /* PM1a_CNT.SLP_TYP */
742         aml_append(pkg, aml_int(1)); /* PM1b_CNT.SLP_TYP, FIXME: not impl. */
743         aml_append(pkg, aml_int(0)); /* reserved */
744         aml_append(pkg, aml_int(0)); /* reserved */
745         aml_append(scope, aml_name_decl("_S3", pkg));
746     }
747 
748     if (!pm->s4_disabled) {
749         pkg = aml_package(4);
750         aml_append(pkg, aml_int(pm->s4_val)); /* PM1a_CNT.SLP_TYP */
751         /* PM1b_CNT.SLP_TYP, FIXME: not impl. */
752         aml_append(pkg, aml_int(pm->s4_val));
753         aml_append(pkg, aml_int(0)); /* reserved */
754         aml_append(pkg, aml_int(0)); /* reserved */
755         aml_append(scope, aml_name_decl("_S4", pkg));
756     }
757 
758     pkg = aml_package(4);
759     aml_append(pkg, aml_int(0)); /* PM1a_CNT.SLP_TYP */
760     aml_append(pkg, aml_int(0)); /* PM1b_CNT.SLP_TYP not impl. */
761     aml_append(pkg, aml_int(0)); /* reserved */
762     aml_append(pkg, aml_int(0)); /* reserved */
763     aml_append(scope, aml_name_decl("_S5", pkg));
764     aml_append(ssdt, scope);
765 
766     if (misc->applesmc_io_base) {
767         scope = aml_scope("\\_SB.PCI0.ISA");
768         dev = aml_device("SMC");
769 
770         aml_append(dev, aml_name_decl("_HID", aml_eisaid("APP0001")));
771         /* device present, functioning, decoding, not shown in UI */
772         aml_append(dev, aml_name_decl("_STA", aml_int(0xB)));
773 
774         crs = aml_resource_template();
775         aml_append(crs,
776             aml_io(aml_decode16, misc->applesmc_io_base, misc->applesmc_io_base,
777                    0x01, APPLESMC_MAX_DATA_LENGTH)
778         );
779         aml_append(crs, aml_irq_no_flags(6));
780         aml_append(dev, aml_name_decl("_CRS", crs));
781 
782         aml_append(scope, dev);
783         aml_append(ssdt, scope);
784     }
785 
786     if (misc->pvpanic_port) {
787         scope = aml_scope("\\_SB.PCI0.ISA");
788 
789         dev = aml_device("PEVR");
790         aml_append(dev, aml_name_decl("_HID", aml_string("QEMU0001")));
791 
792         crs = aml_resource_template();
793         aml_append(crs,
794             aml_io(aml_decode16, misc->pvpanic_port, misc->pvpanic_port, 1, 1)
795         );
796         aml_append(dev, aml_name_decl("_CRS", crs));
797 
798         aml_append(dev, aml_operation_region("PEOR", aml_system_io,
799                                               misc->pvpanic_port, 1));
800         field = aml_field("PEOR", aml_byte_acc);
801         aml_append(field, aml_named_field("PEPT", 8));
802         aml_append(dev, field);
803 
804         method = aml_method("RDPT", 0);
805         aml_append(method, aml_store(aml_name("PEPT"), aml_local(0)));
806         aml_append(method, aml_return(aml_local(0)));
807         aml_append(dev, method);
808 
809         method = aml_method("WRPT", 1);
810         aml_append(method, aml_store(aml_arg(0), aml_name("PEPT")));
811         aml_append(dev, method);
812 
813         aml_append(scope, dev);
814         aml_append(ssdt, scope);
815     }
816 
817     sb_scope = aml_scope("_SB");
818     {
819         /* create PCI0.PRES device and its _CRS to reserve CPU hotplug MMIO */
820         dev = aml_device("PCI0." stringify(CPU_HOTPLUG_RESOURCE_DEVICE));
821         aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0A06")));
822         aml_append(dev,
823             aml_name_decl("_UID", aml_string("CPU Hotplug resources"))
824         );
825         /* device present, functioning, decoding, not shown in UI */
826         aml_append(dev, aml_name_decl("_STA", aml_int(0xB)));
827         crs = aml_resource_template();
828         aml_append(crs,
829             aml_io(aml_decode16, pm->cpu_hp_io_base, pm->cpu_hp_io_base, 1,
830                    pm->cpu_hp_io_len)
831         );
832         aml_append(dev, aml_name_decl("_CRS", crs));
833         aml_append(sb_scope, dev);
834         /* declare CPU hotplug MMIO region and PRS field to access it */
835         aml_append(sb_scope, aml_operation_region(
836             "PRST", aml_system_io, pm->cpu_hp_io_base, pm->cpu_hp_io_len));
837         field = aml_field("PRST", aml_byte_acc);
838         aml_append(field, aml_named_field("PRS", 256));
839         aml_append(sb_scope, field);
840 
841         /* build Processor object for each processor */
842         for (i = 0; i < acpi_cpus; i++) {
843             dev = aml_processor(i, 0, 0, "CP%.02X", i);
844 
845             method = aml_method("_MAT", 0);
846             aml_append(method, aml_return(aml_call1("CPMA", aml_int(i))));
847             aml_append(dev, method);
848 
849             method = aml_method("_STA", 0);
850             aml_append(method, aml_return(aml_call1("CPST", aml_int(i))));
851             aml_append(dev, method);
852 
853             method = aml_method("_EJ0", 1);
854             aml_append(method,
855                 aml_return(aml_call2("CPEJ", aml_int(i), aml_arg(0)))
856             );
857             aml_append(dev, method);
858 
859             aml_append(sb_scope, dev);
860         }
861 
862         /* build this code:
863          *   Method(NTFY, 2) {If (LEqual(Arg0, 0x00)) {Notify(CP00, Arg1)} ...}
864          */
865         /* Arg0 = Processor ID = APIC ID */
866         method = aml_method("NTFY", 2);
867         for (i = 0; i < acpi_cpus; i++) {
868             ifctx = aml_if(aml_equal(aml_arg(0), aml_int(i)));
869             aml_append(ifctx,
870                 aml_notify(aml_name("CP%.02X", i), aml_arg(1))
871             );
872             aml_append(method, ifctx);
873         }
874         aml_append(sb_scope, method);
875 
876         /* build "Name(CPON, Package() { One, One, ..., Zero, Zero, ... })"
877          *
878          * Note: The ability to create variable-sized packages was first
879          * introduced in ACPI 2.0. ACPI 1.0 only allowed fixed-size packages
880          * ith up to 255 elements. Windows guests up to win2k8 fail when
881          * VarPackageOp is used.
882          */
883         pkg = acpi_cpus <= 255 ? aml_package(acpi_cpus) :
884                                  aml_varpackage(acpi_cpus);
885 
886         for (i = 0; i < acpi_cpus; i++) {
887             uint8_t b = test_bit(i, cpu->found_cpus) ? 0x01 : 0x00;
888             aml_append(pkg, aml_int(b));
889         }
890         aml_append(sb_scope, aml_name_decl("CPON", pkg));
891 
892         /* build memory devices */
893         assert(nr_mem <= ACPI_MAX_RAM_SLOTS);
894         scope = aml_scope("\\_SB.PCI0." stringify(MEMORY_HOTPLUG_DEVICE));
895         aml_append(scope,
896             aml_name_decl(stringify(MEMORY_SLOTS_NUMBER), aml_int(nr_mem))
897         );
898 
899         crs = aml_resource_template();
900         aml_append(crs,
901             aml_io(aml_decode16, pm->mem_hp_io_base, pm->mem_hp_io_base, 0,
902                    pm->mem_hp_io_len)
903         );
904         aml_append(scope, aml_name_decl("_CRS", crs));
905 
906         aml_append(scope, aml_operation_region(
907             stringify(MEMORY_HOTPLUG_IO_REGION), aml_system_io,
908             pm->mem_hp_io_base, pm->mem_hp_io_len)
909         );
910 
911         field = aml_field(stringify(MEMORY_HOTPLUG_IO_REGION), aml_dword_acc);
912         aml_append(field, /* read only */
913             aml_named_field(stringify(MEMORY_SLOT_ADDR_LOW), 32));
914         aml_append(field, /* read only */
915             aml_named_field(stringify(MEMORY_SLOT_ADDR_HIGH), 32));
916         aml_append(field, /* read only */
917             aml_named_field(stringify(MEMORY_SLOT_SIZE_LOW), 32));
918         aml_append(field, /* read only */
919             aml_named_field(stringify(MEMORY_SLOT_SIZE_HIGH), 32));
920         aml_append(field, /* read only */
921             aml_named_field(stringify(MEMORY_SLOT_PROXIMITY), 32));
922         aml_append(scope, field);
923 
924         field = aml_field(stringify(MEMORY_HOTPLUG_IO_REGION), aml_byte_acc);
925         aml_append(field, aml_reserved_field(160 /* bits, Offset(20) */));
926         aml_append(field, /* 1 if enabled, read only */
927             aml_named_field(stringify(MEMORY_SLOT_ENABLED), 1));
928         aml_append(field,
929             /*(read) 1 if has a insert event. (write) 1 to clear event */
930             aml_named_field(stringify(MEMORY_SLOT_INSERT_EVENT), 1));
931         aml_append(scope, field);
932 
933         field = aml_field(stringify(MEMORY_HOTPLUG_IO_REGION), aml_dword_acc);
934         aml_append(field, /* DIMM selector, write only */
935             aml_named_field(stringify(MEMORY_SLOT_SLECTOR), 32));
936         aml_append(field, /* _OST event code, write only */
937             aml_named_field(stringify(MEMORY_SLOT_OST_EVENT), 32));
938         aml_append(field, /* _OST status code, write only */
939             aml_named_field(stringify(MEMORY_SLOT_OST_STATUS), 32));
940         aml_append(scope, field);
941 
942         aml_append(sb_scope, scope);
943 
944         for (i = 0; i < nr_mem; i++) {
945             #define BASEPATH "\\_SB.PCI0." stringify(MEMORY_HOTPLUG_DEVICE) "."
946             const char *s;
947 
948             dev = aml_device("MP%02X", i);
949             aml_append(dev, aml_name_decl("_UID", aml_string("0x%02X", i)));
950             aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0C80")));
951 
952             method = aml_method("_CRS", 0);
953             s = BASEPATH stringify(MEMORY_SLOT_CRS_METHOD);
954             aml_append(method, aml_return(aml_call1(s, aml_name("_UID"))));
955             aml_append(dev, method);
956 
957             method = aml_method("_STA", 0);
958             s = BASEPATH stringify(MEMORY_SLOT_STATUS_METHOD);
959             aml_append(method, aml_return(aml_call1(s, aml_name("_UID"))));
960             aml_append(dev, method);
961 
962             method = aml_method("_PXM", 0);
963             s = BASEPATH stringify(MEMORY_SLOT_PROXIMITY_METHOD);
964             aml_append(method, aml_return(aml_call1(s, aml_name("_UID"))));
965             aml_append(dev, method);
966 
967             method = aml_method("_OST", 3);
968             s = BASEPATH stringify(MEMORY_SLOT_OST_METHOD);
969             aml_append(method, aml_return(aml_call4(
970                 s, aml_name("_UID"), aml_arg(0), aml_arg(1), aml_arg(2)
971             )));
972             aml_append(dev, method);
973 
974             aml_append(sb_scope, dev);
975         }
976 
977         /* build Method(MEMORY_SLOT_NOTIFY_METHOD, 2) {
978          *     If (LEqual(Arg0, 0x00)) {Notify(MP00, Arg1)} ...
979          */
980         method = aml_method(stringify(MEMORY_SLOT_NOTIFY_METHOD), 2);
981         for (i = 0; i < nr_mem; i++) {
982             ifctx = aml_if(aml_equal(aml_arg(0), aml_int(i)));
983             aml_append(ifctx,
984                 aml_notify(aml_name("MP%.02X", i), aml_arg(1))
985             );
986             aml_append(method, ifctx);
987         }
988         aml_append(sb_scope, method);
989 
990         {
991             Object *pci_host;
992             PCIBus *bus = NULL;
993             bool ambiguous;
994 
995             pci_host = object_resolve_path_type("", TYPE_PCI_HOST_BRIDGE, &ambiguous);
996             if (!ambiguous && pci_host) {
997                 bus = PCI_HOST_BRIDGE(pci_host)->bus;
998             }
999 
1000             if (bus) {
1001                 Aml *scope = aml_scope("PCI0");
1002                 /* Scan all PCI buses. Generate tables to support hotplug. */
1003                 build_append_pci_bus_devices(scope, bus, pm->pcihp_bridge_en);
1004                 aml_append(sb_scope, scope);
1005             }
1006         }
1007         aml_append(ssdt, sb_scope);
1008     }
1009 
1010     /* copy AML table into ACPI tables blob and patch header there */
1011     g_array_append_vals(table_data, ssdt->buf->data, ssdt->buf->len);
1012     build_header(linker, table_data,
1013         (void *)(table_data->data + table_data->len - ssdt->buf->len),
1014         "SSDT", ssdt->buf->len, 1);
1015     free_aml_allocator();
1016 }
1017 
1018 static void
1019 build_hpet(GArray *table_data, GArray *linker)
1020 {
1021     Acpi20Hpet *hpet;
1022 
1023     hpet = acpi_data_push(table_data, sizeof(*hpet));
1024     /* Note timer_block_id value must be kept in sync with value advertised by
1025      * emulated hpet
1026      */
1027     hpet->timer_block_id = cpu_to_le32(0x8086a201);
1028     hpet->addr.address = cpu_to_le64(HPET_BASE);
1029     build_header(linker, table_data,
1030                  (void *)hpet, "HPET", sizeof(*hpet), 1);
1031 }
1032 
1033 static void
1034 build_tpm_tcpa(GArray *table_data, GArray *linker, GArray *tcpalog)
1035 {
1036     Acpi20Tcpa *tcpa = acpi_data_push(table_data, sizeof *tcpa);
1037     uint64_t log_area_start_address = acpi_data_len(tcpalog);
1038 
1039     tcpa->platform_class = cpu_to_le16(TPM_TCPA_ACPI_CLASS_CLIENT);
1040     tcpa->log_area_minimum_length = cpu_to_le32(TPM_LOG_AREA_MINIMUM_SIZE);
1041     tcpa->log_area_start_address = cpu_to_le64(log_area_start_address);
1042 
1043     bios_linker_loader_alloc(linker, ACPI_BUILD_TPMLOG_FILE, 1,
1044                              false /* high memory */);
1045 
1046     /* log area start address to be filled by Guest linker */
1047     bios_linker_loader_add_pointer(linker, ACPI_BUILD_TABLE_FILE,
1048                                    ACPI_BUILD_TPMLOG_FILE,
1049                                    table_data, &tcpa->log_area_start_address,
1050                                    sizeof(tcpa->log_area_start_address));
1051 
1052     build_header(linker, table_data,
1053                  (void *)tcpa, "TCPA", sizeof(*tcpa), 2);
1054 
1055     acpi_data_push(tcpalog, TPM_LOG_AREA_MINIMUM_SIZE);
1056 }
1057 
1058 static void
1059 build_tpm_ssdt(GArray *table_data, GArray *linker)
1060 {
1061     void *tpm_ptr;
1062 
1063     tpm_ptr = acpi_data_push(table_data, sizeof(ssdt_tpm_aml));
1064     memcpy(tpm_ptr, ssdt_tpm_aml, sizeof(ssdt_tpm_aml));
1065 }
1066 
1067 typedef enum {
1068     MEM_AFFINITY_NOFLAGS      = 0,
1069     MEM_AFFINITY_ENABLED      = (1 << 0),
1070     MEM_AFFINITY_HOTPLUGGABLE = (1 << 1),
1071     MEM_AFFINITY_NON_VOLATILE = (1 << 2),
1072 } MemoryAffinityFlags;
1073 
1074 static void
1075 acpi_build_srat_memory(AcpiSratMemoryAffinity *numamem, uint64_t base,
1076                        uint64_t len, int node, MemoryAffinityFlags flags)
1077 {
1078     numamem->type = ACPI_SRAT_MEMORY;
1079     numamem->length = sizeof(*numamem);
1080     memset(numamem->proximity, 0, 4);
1081     numamem->proximity[0] = node;
1082     numamem->flags = cpu_to_le32(flags);
1083     numamem->base_addr = cpu_to_le64(base);
1084     numamem->range_length = cpu_to_le64(len);
1085 }
1086 
1087 static void
1088 build_srat(GArray *table_data, GArray *linker, PcGuestInfo *guest_info)
1089 {
1090     AcpiSystemResourceAffinityTable *srat;
1091     AcpiSratProcessorAffinity *core;
1092     AcpiSratMemoryAffinity *numamem;
1093 
1094     int i;
1095     uint64_t curnode;
1096     int srat_start, numa_start, slots;
1097     uint64_t mem_len, mem_base, next_base;
1098     PCMachineState *pcms = PC_MACHINE(qdev_get_machine());
1099     ram_addr_t hotplugabble_address_space_size =
1100         object_property_get_int(OBJECT(pcms), PC_MACHINE_MEMHP_REGION_SIZE,
1101                                 NULL);
1102 
1103     srat_start = table_data->len;
1104 
1105     srat = acpi_data_push(table_data, sizeof *srat);
1106     srat->reserved1 = cpu_to_le32(1);
1107     core = (void *)(srat + 1);
1108 
1109     for (i = 0; i < guest_info->apic_id_limit; ++i) {
1110         core = acpi_data_push(table_data, sizeof *core);
1111         core->type = ACPI_SRAT_PROCESSOR;
1112         core->length = sizeof(*core);
1113         core->local_apic_id = i;
1114         curnode = guest_info->node_cpu[i];
1115         core->proximity_lo = curnode;
1116         memset(core->proximity_hi, 0, 3);
1117         core->local_sapic_eid = 0;
1118         core->flags = cpu_to_le32(1);
1119     }
1120 
1121 
1122     /* the memory map is a bit tricky, it contains at least one hole
1123      * from 640k-1M and possibly another one from 3.5G-4G.
1124      */
1125     next_base = 0;
1126     numa_start = table_data->len;
1127 
1128     numamem = acpi_data_push(table_data, sizeof *numamem);
1129     acpi_build_srat_memory(numamem, 0, 640*1024, 0, MEM_AFFINITY_ENABLED);
1130     next_base = 1024 * 1024;
1131     for (i = 1; i < guest_info->numa_nodes + 1; ++i) {
1132         mem_base = next_base;
1133         mem_len = guest_info->node_mem[i - 1];
1134         if (i == 1) {
1135             mem_len -= 1024 * 1024;
1136         }
1137         next_base = mem_base + mem_len;
1138 
1139         /* Cut out the ACPI_PCI hole */
1140         if (mem_base <= guest_info->ram_size_below_4g &&
1141             next_base > guest_info->ram_size_below_4g) {
1142             mem_len -= next_base - guest_info->ram_size_below_4g;
1143             if (mem_len > 0) {
1144                 numamem = acpi_data_push(table_data, sizeof *numamem);
1145                 acpi_build_srat_memory(numamem, mem_base, mem_len, i - 1,
1146                                        MEM_AFFINITY_ENABLED);
1147             }
1148             mem_base = 1ULL << 32;
1149             mem_len = next_base - guest_info->ram_size_below_4g;
1150             next_base += (1ULL << 32) - guest_info->ram_size_below_4g;
1151         }
1152         numamem = acpi_data_push(table_data, sizeof *numamem);
1153         acpi_build_srat_memory(numamem, mem_base, mem_len, i - 1,
1154                                MEM_AFFINITY_ENABLED);
1155     }
1156     slots = (table_data->len - numa_start) / sizeof *numamem;
1157     for (; slots < guest_info->numa_nodes + 2; slots++) {
1158         numamem = acpi_data_push(table_data, sizeof *numamem);
1159         acpi_build_srat_memory(numamem, 0, 0, 0, MEM_AFFINITY_NOFLAGS);
1160     }
1161 
1162     /*
1163      * Entry is required for Windows to enable memory hotplug in OS.
1164      * Memory devices may override proximity set by this entry,
1165      * providing _PXM method if necessary.
1166      */
1167     if (hotplugabble_address_space_size) {
1168         numamem = acpi_data_push(table_data, sizeof *numamem);
1169         acpi_build_srat_memory(numamem, pcms->hotplug_memory_base,
1170                                hotplugabble_address_space_size, 0,
1171                                MEM_AFFINITY_HOTPLUGGABLE |
1172                                MEM_AFFINITY_ENABLED);
1173     }
1174 
1175     build_header(linker, table_data,
1176                  (void *)(table_data->data + srat_start),
1177                  "SRAT",
1178                  table_data->len - srat_start, 1);
1179 }
1180 
1181 static void
1182 build_mcfg_q35(GArray *table_data, GArray *linker, AcpiMcfgInfo *info)
1183 {
1184     AcpiTableMcfg *mcfg;
1185     const char *sig;
1186     int len = sizeof(*mcfg) + 1 * sizeof(mcfg->allocation[0]);
1187 
1188     mcfg = acpi_data_push(table_data, len);
1189     mcfg->allocation[0].address = cpu_to_le64(info->mcfg_base);
1190     /* Only a single allocation so no need to play with segments */
1191     mcfg->allocation[0].pci_segment = cpu_to_le16(0);
1192     mcfg->allocation[0].start_bus_number = 0;
1193     mcfg->allocation[0].end_bus_number = PCIE_MMCFG_BUS(info->mcfg_size - 1);
1194 
1195     /* MCFG is used for ECAM which can be enabled or disabled by guest.
1196      * To avoid table size changes (which create migration issues),
1197      * always create the table even if there are no allocations,
1198      * but set the signature to a reserved value in this case.
1199      * ACPI spec requires OSPMs to ignore such tables.
1200      */
1201     if (info->mcfg_base == PCIE_BASE_ADDR_UNMAPPED) {
1202         /* Reserved signature: ignored by OSPM */
1203         sig = "QEMU";
1204     } else {
1205         sig = "MCFG";
1206     }
1207     build_header(linker, table_data, (void *)mcfg, sig, len, 1);
1208 }
1209 
1210 static void
1211 build_dmar_q35(GArray *table_data, GArray *linker)
1212 {
1213     int dmar_start = table_data->len;
1214 
1215     AcpiTableDmar *dmar;
1216     AcpiDmarHardwareUnit *drhd;
1217 
1218     dmar = acpi_data_push(table_data, sizeof(*dmar));
1219     dmar->host_address_width = VTD_HOST_ADDRESS_WIDTH - 1;
1220     dmar->flags = 0;    /* No intr_remap for now */
1221 
1222     /* DMAR Remapping Hardware Unit Definition structure */
1223     drhd = acpi_data_push(table_data, sizeof(*drhd));
1224     drhd->type = cpu_to_le16(ACPI_DMAR_TYPE_HARDWARE_UNIT);
1225     drhd->length = cpu_to_le16(sizeof(*drhd));   /* No device scope now */
1226     drhd->flags = ACPI_DMAR_INCLUDE_PCI_ALL;
1227     drhd->pci_segment = cpu_to_le16(0);
1228     drhd->address = cpu_to_le64(Q35_HOST_BRIDGE_IOMMU_ADDR);
1229 
1230     build_header(linker, table_data, (void *)(table_data->data + dmar_start),
1231                  "DMAR", table_data->len - dmar_start, 1);
1232 }
1233 
1234 static void
1235 build_dsdt(GArray *table_data, GArray *linker, AcpiMiscInfo *misc)
1236 {
1237     AcpiTableHeader *dsdt;
1238 
1239     assert(misc->dsdt_code && misc->dsdt_size);
1240 
1241     dsdt = acpi_data_push(table_data, misc->dsdt_size);
1242     memcpy(dsdt, misc->dsdt_code, misc->dsdt_size);
1243 
1244     memset(dsdt, 0, sizeof *dsdt);
1245     build_header(linker, table_data, dsdt, "DSDT",
1246                  misc->dsdt_size, 1);
1247 }
1248 
1249 /* Build final rsdt table */
1250 static void
1251 build_rsdt(GArray *table_data, GArray *linker, GArray *table_offsets)
1252 {
1253     AcpiRsdtDescriptorRev1 *rsdt;
1254     size_t rsdt_len;
1255     int i;
1256 
1257     rsdt_len = sizeof(*rsdt) + sizeof(uint32_t) * table_offsets->len;
1258     rsdt = acpi_data_push(table_data, rsdt_len);
1259     memcpy(rsdt->table_offset_entry, table_offsets->data,
1260            sizeof(uint32_t) * table_offsets->len);
1261     for (i = 0; i < table_offsets->len; ++i) {
1262         /* rsdt->table_offset_entry to be filled by Guest linker */
1263         bios_linker_loader_add_pointer(linker,
1264                                        ACPI_BUILD_TABLE_FILE,
1265                                        ACPI_BUILD_TABLE_FILE,
1266                                        table_data, &rsdt->table_offset_entry[i],
1267                                        sizeof(uint32_t));
1268     }
1269     build_header(linker, table_data,
1270                  (void *)rsdt, "RSDT", rsdt_len, 1);
1271 }
1272 
1273 static GArray *
1274 build_rsdp(GArray *rsdp_table, GArray *linker, unsigned rsdt)
1275 {
1276     AcpiRsdpDescriptor *rsdp = acpi_data_push(rsdp_table, sizeof *rsdp);
1277 
1278     bios_linker_loader_alloc(linker, ACPI_BUILD_RSDP_FILE, 16,
1279                              true /* fseg memory */);
1280 
1281     memcpy(&rsdp->signature, "RSD PTR ", 8);
1282     memcpy(rsdp->oem_id, ACPI_BUILD_APPNAME6, 6);
1283     rsdp->rsdt_physical_address = cpu_to_le32(rsdt);
1284     /* Address to be filled by Guest linker */
1285     bios_linker_loader_add_pointer(linker, ACPI_BUILD_RSDP_FILE,
1286                                    ACPI_BUILD_TABLE_FILE,
1287                                    rsdp_table, &rsdp->rsdt_physical_address,
1288                                    sizeof rsdp->rsdt_physical_address);
1289     rsdp->checksum = 0;
1290     /* Checksum to be filled by Guest linker */
1291     bios_linker_loader_add_checksum(linker, ACPI_BUILD_RSDP_FILE,
1292                                     rsdp, rsdp, sizeof *rsdp, &rsdp->checksum);
1293 
1294     return rsdp_table;
1295 }
1296 
1297 typedef
1298 struct AcpiBuildTables {
1299     GArray *table_data;
1300     GArray *rsdp;
1301     GArray *tcpalog;
1302     GArray *linker;
1303 } AcpiBuildTables;
1304 
1305 static inline void acpi_build_tables_init(AcpiBuildTables *tables)
1306 {
1307     tables->rsdp = g_array_new(false, true /* clear */, 1);
1308     tables->table_data = g_array_new(false, true /* clear */, 1);
1309     tables->tcpalog = g_array_new(false, true /* clear */, 1);
1310     tables->linker = bios_linker_loader_init();
1311 }
1312 
1313 static inline void acpi_build_tables_cleanup(AcpiBuildTables *tables, bool mfre)
1314 {
1315     void *linker_data = bios_linker_loader_cleanup(tables->linker);
1316     g_free(linker_data);
1317     g_array_free(tables->rsdp, true);
1318     g_array_free(tables->table_data, true);
1319     g_array_free(tables->tcpalog, mfre);
1320 }
1321 
1322 typedef
1323 struct AcpiBuildState {
1324     /* Copy of table in RAM (for patching). */
1325     MemoryRegion *table_mr;
1326     /* Is table patched? */
1327     uint8_t patched;
1328     PcGuestInfo *guest_info;
1329     void *rsdp;
1330     MemoryRegion *rsdp_mr;
1331     MemoryRegion *linker_mr;
1332 } AcpiBuildState;
1333 
1334 static bool acpi_get_mcfg(AcpiMcfgInfo *mcfg)
1335 {
1336     Object *pci_host;
1337     QObject *o;
1338     bool ambiguous;
1339 
1340     pci_host = object_resolve_path_type("", TYPE_PCI_HOST_BRIDGE, &ambiguous);
1341     g_assert(!ambiguous);
1342     g_assert(pci_host);
1343 
1344     o = object_property_get_qobject(pci_host, PCIE_HOST_MCFG_BASE, NULL);
1345     if (!o) {
1346         return false;
1347     }
1348     mcfg->mcfg_base = qint_get_int(qobject_to_qint(o));
1349     qobject_decref(o);
1350 
1351     o = object_property_get_qobject(pci_host, PCIE_HOST_MCFG_SIZE, NULL);
1352     assert(o);
1353     mcfg->mcfg_size = qint_get_int(qobject_to_qint(o));
1354     qobject_decref(o);
1355     return true;
1356 }
1357 
1358 static bool acpi_has_iommu(void)
1359 {
1360     bool ambiguous;
1361     Object *intel_iommu;
1362 
1363     intel_iommu = object_resolve_path_type("", TYPE_INTEL_IOMMU_DEVICE,
1364                                            &ambiguous);
1365     return intel_iommu && !ambiguous;
1366 }
1367 
1368 static
1369 void acpi_build(PcGuestInfo *guest_info, AcpiBuildTables *tables)
1370 {
1371     GArray *table_offsets;
1372     unsigned facs, ssdt, dsdt, rsdt;
1373     AcpiCpuInfo cpu;
1374     AcpiPmInfo pm;
1375     AcpiMiscInfo misc;
1376     AcpiMcfgInfo mcfg;
1377     PcPciInfo pci;
1378     uint8_t *u;
1379     size_t aml_len = 0;
1380     GArray *tables_blob = tables->table_data;
1381 
1382     acpi_get_cpu_info(&cpu);
1383     acpi_get_pm_info(&pm);
1384     acpi_get_dsdt(&misc);
1385     acpi_get_misc_info(&misc);
1386     acpi_get_pci_info(&pci);
1387 
1388     table_offsets = g_array_new(false, true /* clear */,
1389                                         sizeof(uint32_t));
1390     ACPI_BUILD_DPRINTF("init ACPI tables\n");
1391 
1392     bios_linker_loader_alloc(tables->linker, ACPI_BUILD_TABLE_FILE,
1393                              64 /* Ensure FACS is aligned */,
1394                              false /* high memory */);
1395 
1396     /*
1397      * FACS is pointed to by FADT.
1398      * We place it first since it's the only table that has alignment
1399      * requirements.
1400      */
1401     facs = tables_blob->len;
1402     build_facs(tables_blob, tables->linker, guest_info);
1403 
1404     /* DSDT is pointed to by FADT */
1405     dsdt = tables_blob->len;
1406     build_dsdt(tables_blob, tables->linker, &misc);
1407 
1408     /* Count the size of the DSDT and SSDT, we will need it for legacy
1409      * sizing of ACPI tables.
1410      */
1411     aml_len += tables_blob->len - dsdt;
1412 
1413     /* ACPI tables pointed to by RSDT */
1414     acpi_add_table(table_offsets, tables_blob);
1415     build_fadt(tables_blob, tables->linker, &pm, facs, dsdt);
1416 
1417     ssdt = tables_blob->len;
1418     acpi_add_table(table_offsets, tables_blob);
1419     build_ssdt(tables_blob, tables->linker, &cpu, &pm, &misc, &pci,
1420                guest_info);
1421     aml_len += tables_blob->len - ssdt;
1422 
1423     acpi_add_table(table_offsets, tables_blob);
1424     build_madt(tables_blob, tables->linker, &cpu, guest_info);
1425 
1426     if (misc.has_hpet) {
1427         acpi_add_table(table_offsets, tables_blob);
1428         build_hpet(tables_blob, tables->linker);
1429     }
1430     if (misc.has_tpm) {
1431         acpi_add_table(table_offsets, tables_blob);
1432         build_tpm_tcpa(tables_blob, tables->linker, tables->tcpalog);
1433 
1434         acpi_add_table(table_offsets, tables_blob);
1435         build_tpm_ssdt(tables_blob, tables->linker);
1436     }
1437     if (guest_info->numa_nodes) {
1438         acpi_add_table(table_offsets, tables_blob);
1439         build_srat(tables_blob, tables->linker, guest_info);
1440     }
1441     if (acpi_get_mcfg(&mcfg)) {
1442         acpi_add_table(table_offsets, tables_blob);
1443         build_mcfg_q35(tables_blob, tables->linker, &mcfg);
1444     }
1445     if (acpi_has_iommu()) {
1446         acpi_add_table(table_offsets, tables_blob);
1447         build_dmar_q35(tables_blob, tables->linker);
1448     }
1449 
1450     /* Add tables supplied by user (if any) */
1451     for (u = acpi_table_first(); u; u = acpi_table_next(u)) {
1452         unsigned len = acpi_table_len(u);
1453 
1454         acpi_add_table(table_offsets, tables_blob);
1455         g_array_append_vals(tables_blob, u, len);
1456     }
1457 
1458     /* RSDT is pointed to by RSDP */
1459     rsdt = tables_blob->len;
1460     build_rsdt(tables_blob, tables->linker, table_offsets);
1461 
1462     /* RSDP is in FSEG memory, so allocate it separately */
1463     build_rsdp(tables->rsdp, tables->linker, rsdt);
1464 
1465     /* We'll expose it all to Guest so we want to reduce
1466      * chance of size changes.
1467      *
1468      * We used to align the tables to 4k, but of course this would
1469      * too simple to be enough.  4k turned out to be too small an
1470      * alignment very soon, and in fact it is almost impossible to
1471      * keep the table size stable for all (max_cpus, max_memory_slots)
1472      * combinations.  So the table size is always 64k for pc-i440fx-2.1
1473      * and we give an error if the table grows beyond that limit.
1474      *
1475      * We still have the problem of migrating from "-M pc-i440fx-2.0".  For
1476      * that, we exploit the fact that QEMU 2.1 generates _smaller_ tables
1477      * than 2.0 and we can always pad the smaller tables with zeros.  We can
1478      * then use the exact size of the 2.0 tables.
1479      *
1480      * All this is for PIIX4, since QEMU 2.0 didn't support Q35 migration.
1481      */
1482     if (guest_info->legacy_acpi_table_size) {
1483         /* Subtracting aml_len gives the size of fixed tables.  Then add the
1484          * size of the PIIX4 DSDT/SSDT in QEMU 2.0.
1485          */
1486         int legacy_aml_len =
1487             guest_info->legacy_acpi_table_size +
1488             ACPI_BUILD_LEGACY_CPU_AML_SIZE * max_cpus;
1489         int legacy_table_size =
1490             ROUND_UP(tables_blob->len - aml_len + legacy_aml_len,
1491                      ACPI_BUILD_ALIGN_SIZE);
1492         if (tables_blob->len > legacy_table_size) {
1493             /* Should happen only with PCI bridges and -M pc-i440fx-2.0.  */
1494             error_report("Warning: migration may not work.");
1495         }
1496         g_array_set_size(tables_blob, legacy_table_size);
1497     } else {
1498         /* Make sure we have a buffer in case we need to resize the tables. */
1499         if (tables_blob->len > ACPI_BUILD_TABLE_SIZE / 2) {
1500             /* As of QEMU 2.1, this fires with 160 VCPUs and 255 memory slots.  */
1501             error_report("Warning: ACPI tables are larger than 64k.");
1502             error_report("Warning: migration may not work.");
1503             error_report("Warning: please remove CPUs, NUMA nodes, "
1504                          "memory slots or PCI bridges.");
1505         }
1506         acpi_align_size(tables_blob, ACPI_BUILD_TABLE_SIZE);
1507     }
1508 
1509     acpi_align_size(tables->linker, ACPI_BUILD_ALIGN_SIZE);
1510 
1511     /* Cleanup memory that's no longer used. */
1512     g_array_free(table_offsets, true);
1513 }
1514 
1515 static void acpi_ram_update(MemoryRegion *mr, GArray *data)
1516 {
1517     uint32_t size = acpi_data_len(data);
1518 
1519     /* Make sure RAM size is correct - in case it got changed e.g. by migration */
1520     memory_region_ram_resize(mr, size, &error_abort);
1521 
1522     memcpy(memory_region_get_ram_ptr(mr), data->data, size);
1523     memory_region_set_dirty(mr, 0, size);
1524 }
1525 
1526 static void acpi_build_update(void *build_opaque, uint32_t offset)
1527 {
1528     AcpiBuildState *build_state = build_opaque;
1529     AcpiBuildTables tables;
1530 
1531     /* No state to update or already patched? Nothing to do. */
1532     if (!build_state || build_state->patched) {
1533         return;
1534     }
1535     build_state->patched = 1;
1536 
1537     acpi_build_tables_init(&tables);
1538 
1539     acpi_build(build_state->guest_info, &tables);
1540 
1541     acpi_ram_update(build_state->table_mr, tables.table_data);
1542 
1543     if (build_state->rsdp) {
1544         memcpy(build_state->rsdp, tables.rsdp->data, acpi_data_len(tables.rsdp));
1545     } else {
1546         acpi_ram_update(build_state->rsdp_mr, tables.rsdp);
1547     }
1548 
1549     acpi_ram_update(build_state->linker_mr, tables.linker);
1550     acpi_build_tables_cleanup(&tables, true);
1551 }
1552 
1553 static void acpi_build_reset(void *build_opaque)
1554 {
1555     AcpiBuildState *build_state = build_opaque;
1556     build_state->patched = 0;
1557 }
1558 
1559 static MemoryRegion *acpi_add_rom_blob(AcpiBuildState *build_state,
1560                                        GArray *blob, const char *name,
1561                                        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_mr = 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_mr != NULL);
1611 
1612     build_state->linker_mr =
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_mr = NULL;
1631     } else {
1632         build_state->rsdp = NULL;
1633         build_state->rsdp_mr = 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