xref: /openbmc/qemu/hw/i386/acpi-build.c (revision 660e8ec70065c8b1fd68b2cb137de16d831959f4)
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/acpi/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 /* #define DEBUG_ACPI_BUILD */
74 #ifdef DEBUG_ACPI_BUILD
75 #define ACPI_BUILD_DPRINTF(fmt, ...)        \
76     do {printf("ACPI_BUILD: " fmt, ## __VA_ARGS__); } while (0)
77 #else
78 #define ACPI_BUILD_DPRINTF(fmt, ...)
79 #endif
80 
81 typedef struct AcpiCpuInfo {
82     DECLARE_BITMAP(found_cpus, ACPI_CPU_HOTPLUG_ID_LIMIT);
83 } AcpiCpuInfo;
84 
85 typedef struct AcpiMcfgInfo {
86     uint64_t mcfg_base;
87     uint32_t mcfg_size;
88 } AcpiMcfgInfo;
89 
90 typedef struct AcpiPmInfo {
91     bool s3_disabled;
92     bool s4_disabled;
93     bool pcihp_bridge_en;
94     uint8_t s4_val;
95     uint16_t sci_int;
96     uint8_t acpi_enable_cmd;
97     uint8_t acpi_disable_cmd;
98     uint32_t gpe0_blk;
99     uint32_t gpe0_blk_len;
100     uint32_t io_base;
101     uint16_t cpu_hp_io_base;
102     uint16_t cpu_hp_io_len;
103     uint16_t mem_hp_io_base;
104     uint16_t mem_hp_io_len;
105     uint16_t pcihp_io_base;
106     uint16_t pcihp_io_len;
107 } AcpiPmInfo;
108 
109 typedef struct AcpiMiscInfo {
110     bool has_hpet;
111     bool has_tpm;
112     const unsigned char *dsdt_code;
113     unsigned dsdt_size;
114     uint16_t pvpanic_port;
115     uint16_t applesmc_io_base;
116 } AcpiMiscInfo;
117 
118 typedef struct AcpiBuildPciBusHotplugState {
119     GArray *device_table;
120     GArray *notify_table;
121     struct AcpiBuildPciBusHotplugState *parent;
122     bool pcihp_bridge_en;
123 } AcpiBuildPciBusHotplugState;
124 
125 static void acpi_get_dsdt(AcpiMiscInfo *info)
126 {
127     Object *piix = piix4_pm_find();
128     Object *lpc = ich9_lpc_find();
129     assert(!!piix != !!lpc);
130 
131     if (piix) {
132         info->dsdt_code = AcpiDsdtAmlCode;
133         info->dsdt_size = sizeof AcpiDsdtAmlCode;
134     }
135     if (lpc) {
136         info->dsdt_code = Q35AcpiDsdtAmlCode;
137         info->dsdt_size = sizeof Q35AcpiDsdtAmlCode;
138     }
139 }
140 
141 static
142 int acpi_add_cpu_info(Object *o, void *opaque)
143 {
144     AcpiCpuInfo *cpu = opaque;
145     uint64_t apic_id;
146 
147     if (object_dynamic_cast(o, TYPE_CPU)) {
148         apic_id = object_property_get_int(o, "apic-id", NULL);
149         assert(apic_id < ACPI_CPU_HOTPLUG_ID_LIMIT);
150 
151         set_bit(apic_id, cpu->found_cpus);
152     }
153 
154     object_child_foreach(o, acpi_add_cpu_info, opaque);
155     return 0;
156 }
157 
158 static void acpi_get_cpu_info(AcpiCpuInfo *cpu)
159 {
160     Object *root = object_get_root();
161 
162     memset(cpu->found_cpus, 0, sizeof cpu->found_cpus);
163     object_child_foreach(root, acpi_add_cpu_info, cpu);
164 }
165 
166 static void acpi_get_pm_info(AcpiPmInfo *pm)
167 {
168     Object *piix = piix4_pm_find();
169     Object *lpc = ich9_lpc_find();
170     Object *obj = NULL;
171     QObject *o;
172 
173     pm->pcihp_io_base = 0;
174     pm->pcihp_io_len = 0;
175     if (piix) {
176         obj = piix;
177         pm->cpu_hp_io_base = PIIX4_CPU_HOTPLUG_IO_BASE;
178         pm->pcihp_io_base =
179             object_property_get_int(obj, ACPI_PCIHP_IO_BASE_PROP, NULL);
180         pm->pcihp_io_len =
181             object_property_get_int(obj, ACPI_PCIHP_IO_LEN_PROP, NULL);
182     }
183     if (lpc) {
184         obj = lpc;
185         pm->cpu_hp_io_base = ICH9_CPU_HOTPLUG_IO_BASE;
186     }
187     assert(obj);
188 
189     pm->cpu_hp_io_len = ACPI_GPE_PROC_LEN;
190     pm->mem_hp_io_base = ACPI_MEMORY_HOTPLUG_BASE;
191     pm->mem_hp_io_len = ACPI_MEMORY_HOTPLUG_IO_LEN;
192 
193     /* Fill in optional s3/s4 related properties */
194     o = object_property_get_qobject(obj, ACPI_PM_PROP_S3_DISABLED, NULL);
195     if (o) {
196         pm->s3_disabled = qint_get_int(qobject_to_qint(o));
197     } else {
198         pm->s3_disabled = false;
199     }
200     qobject_decref(o);
201     o = object_property_get_qobject(obj, ACPI_PM_PROP_S4_DISABLED, NULL);
202     if (o) {
203         pm->s4_disabled = qint_get_int(qobject_to_qint(o));
204     } else {
205         pm->s4_disabled = false;
206     }
207     qobject_decref(o);
208     o = object_property_get_qobject(obj, ACPI_PM_PROP_S4_VAL, NULL);
209     if (o) {
210         pm->s4_val = qint_get_int(qobject_to_qint(o));
211     } else {
212         pm->s4_val = false;
213     }
214     qobject_decref(o);
215 
216     /* Fill in mandatory properties */
217     pm->sci_int = object_property_get_int(obj, ACPI_PM_PROP_SCI_INT, NULL);
218 
219     pm->acpi_enable_cmd = object_property_get_int(obj,
220                                                   ACPI_PM_PROP_ACPI_ENABLE_CMD,
221                                                   NULL);
222     pm->acpi_disable_cmd = object_property_get_int(obj,
223                                                   ACPI_PM_PROP_ACPI_DISABLE_CMD,
224                                                   NULL);
225     pm->io_base = object_property_get_int(obj, ACPI_PM_PROP_PM_IO_BASE,
226                                           NULL);
227     pm->gpe0_blk = object_property_get_int(obj, ACPI_PM_PROP_GPE0_BLK,
228                                            NULL);
229     pm->gpe0_blk_len = object_property_get_int(obj, ACPI_PM_PROP_GPE0_BLK_LEN,
230                                                NULL);
231     pm->pcihp_bridge_en =
232         object_property_get_bool(obj, "acpi-pci-hotplug-with-bridge-support",
233                                  NULL);
234 }
235 
236 static void acpi_get_misc_info(AcpiMiscInfo *info)
237 {
238     info->has_hpet = hpet_find();
239     info->has_tpm = tpm_find();
240     info->pvpanic_port = pvpanic_port();
241     info->applesmc_io_base = applesmc_port();
242 }
243 
244 static void acpi_get_pci_info(PcPciInfo *info)
245 {
246     Object *pci_host;
247     bool ambiguous;
248 
249     pci_host = object_resolve_path_type("", TYPE_PCI_HOST_BRIDGE, &ambiguous);
250     g_assert(!ambiguous);
251     g_assert(pci_host);
252 
253     info->w32.begin = object_property_get_int(pci_host,
254                                               PCI_HOST_PROP_PCI_HOLE_START,
255                                               NULL);
256     info->w32.end = object_property_get_int(pci_host,
257                                             PCI_HOST_PROP_PCI_HOLE_END,
258                                             NULL);
259     info->w64.begin = object_property_get_int(pci_host,
260                                               PCI_HOST_PROP_PCI_HOLE64_START,
261                                               NULL);
262     info->w64.end = object_property_get_int(pci_host,
263                                             PCI_HOST_PROP_PCI_HOLE64_END,
264                                             NULL);
265 }
266 
267 #define ACPI_PORT_SMI_CMD           0x00b2 /* TODO: this is APM_CNT_IOPORT */
268 
269 static void acpi_align_size(GArray *blob, unsigned align)
270 {
271     /* Align size to multiple of given size. This reduces the chance
272      * we need to change size in the future (breaking cross version migration).
273      */
274     g_array_set_size(blob, ROUND_UP(acpi_data_len(blob), align));
275 }
276 
277 /* FACS */
278 static void
279 build_facs(GArray *table_data, GArray *linker, PcGuestInfo *guest_info)
280 {
281     AcpiFacsDescriptorRev1 *facs = acpi_data_push(table_data, sizeof *facs);
282     memcpy(&facs->signature, "FACS", 4);
283     facs->length = cpu_to_le32(sizeof(*facs));
284 }
285 
286 /* Load chipset information in FADT */
287 static void fadt_setup(AcpiFadtDescriptorRev1 *fadt, AcpiPmInfo *pm)
288 {
289     fadt->model = 1;
290     fadt->reserved1 = 0;
291     fadt->sci_int = cpu_to_le16(pm->sci_int);
292     fadt->smi_cmd = cpu_to_le32(ACPI_PORT_SMI_CMD);
293     fadt->acpi_enable = pm->acpi_enable_cmd;
294     fadt->acpi_disable = pm->acpi_disable_cmd;
295     /* EVT, CNT, TMR offset matches hw/acpi/core.c */
296     fadt->pm1a_evt_blk = cpu_to_le32(pm->io_base);
297     fadt->pm1a_cnt_blk = cpu_to_le32(pm->io_base + 0x04);
298     fadt->pm_tmr_blk = cpu_to_le32(pm->io_base + 0x08);
299     fadt->gpe0_blk = cpu_to_le32(pm->gpe0_blk);
300     /* EVT, CNT, TMR length matches hw/acpi/core.c */
301     fadt->pm1_evt_len = 4;
302     fadt->pm1_cnt_len = 2;
303     fadt->pm_tmr_len = 4;
304     fadt->gpe0_blk_len = pm->gpe0_blk_len;
305     fadt->plvl2_lat = cpu_to_le16(0xfff); /* C2 state not supported */
306     fadt->plvl3_lat = cpu_to_le16(0xfff); /* C3 state not supported */
307     fadt->flags = cpu_to_le32((1 << ACPI_FADT_F_WBINVD) |
308                               (1 << ACPI_FADT_F_PROC_C1) |
309                               (1 << ACPI_FADT_F_SLP_BUTTON) |
310                               (1 << ACPI_FADT_F_RTC_S4));
311     fadt->flags |= cpu_to_le32(1 << ACPI_FADT_F_USE_PLATFORM_CLOCK);
312     /* APIC destination mode ("Flat Logical") has an upper limit of 8 CPUs
313      * For more than 8 CPUs, "Clustered Logical" mode has to be used
314      */
315     if (max_cpus > 8) {
316         fadt->flags |= cpu_to_le32(1 << ACPI_FADT_F_FORCE_APIC_CLUSTER_MODEL);
317     }
318 }
319 
320 
321 /* FADT */
322 static void
323 build_fadt(GArray *table_data, GArray *linker, AcpiPmInfo *pm,
324            unsigned facs, unsigned dsdt)
325 {
326     AcpiFadtDescriptorRev1 *fadt = acpi_data_push(table_data, sizeof(*fadt));
327 
328     fadt->firmware_ctrl = cpu_to_le32(facs);
329     /* FACS address to be filled by Guest linker */
330     bios_linker_loader_add_pointer(linker, ACPI_BUILD_TABLE_FILE,
331                                    ACPI_BUILD_TABLE_FILE,
332                                    table_data, &fadt->firmware_ctrl,
333                                    sizeof fadt->firmware_ctrl);
334 
335     fadt->dsdt = cpu_to_le32(dsdt);
336     /* DSDT address to be filled by Guest linker */
337     bios_linker_loader_add_pointer(linker, ACPI_BUILD_TABLE_FILE,
338                                    ACPI_BUILD_TABLE_FILE,
339                                    table_data, &fadt->dsdt,
340                                    sizeof fadt->dsdt);
341 
342     fadt_setup(fadt, pm);
343 
344     build_header(linker, table_data,
345                  (void *)fadt, "FACP", sizeof(*fadt), 1);
346 }
347 
348 static void
349 build_madt(GArray *table_data, GArray *linker, AcpiCpuInfo *cpu,
350            PcGuestInfo *guest_info)
351 {
352     int madt_start = table_data->len;
353 
354     AcpiMultipleApicTable *madt;
355     AcpiMadtIoApic *io_apic;
356     AcpiMadtIntsrcovr *intsrcovr;
357     AcpiMadtLocalNmi *local_nmi;
358     int i;
359 
360     madt = acpi_data_push(table_data, sizeof *madt);
361     madt->local_apic_address = cpu_to_le32(APIC_DEFAULT_ADDRESS);
362     madt->flags = cpu_to_le32(1);
363 
364     for (i = 0; i < guest_info->apic_id_limit; i++) {
365         AcpiMadtProcessorApic *apic = acpi_data_push(table_data, sizeof *apic);
366         apic->type = ACPI_APIC_PROCESSOR;
367         apic->length = sizeof(*apic);
368         apic->processor_id = i;
369         apic->local_apic_id = i;
370         if (test_bit(i, cpu->found_cpus)) {
371             apic->flags = cpu_to_le32(1);
372         } else {
373             apic->flags = cpu_to_le32(0);
374         }
375     }
376     io_apic = acpi_data_push(table_data, sizeof *io_apic);
377     io_apic->type = ACPI_APIC_IO;
378     io_apic->length = sizeof(*io_apic);
379 #define ACPI_BUILD_IOAPIC_ID 0x0
380     io_apic->io_apic_id = ACPI_BUILD_IOAPIC_ID;
381     io_apic->address = cpu_to_le32(IO_APIC_DEFAULT_ADDRESS);
382     io_apic->interrupt = cpu_to_le32(0);
383 
384     if (guest_info->apic_xrupt_override) {
385         intsrcovr = acpi_data_push(table_data, sizeof *intsrcovr);
386         intsrcovr->type   = ACPI_APIC_XRUPT_OVERRIDE;
387         intsrcovr->length = sizeof(*intsrcovr);
388         intsrcovr->source = 0;
389         intsrcovr->gsi    = cpu_to_le32(2);
390         intsrcovr->flags  = cpu_to_le16(0); /* conforms to bus specifications */
391     }
392     for (i = 1; i < 16; i++) {
393 #define ACPI_BUILD_PCI_IRQS ((1<<5) | (1<<9) | (1<<10) | (1<<11))
394         if (!(ACPI_BUILD_PCI_IRQS & (1 << i))) {
395             /* No need for a INT source override structure. */
396             continue;
397         }
398         intsrcovr = acpi_data_push(table_data, sizeof *intsrcovr);
399         intsrcovr->type   = ACPI_APIC_XRUPT_OVERRIDE;
400         intsrcovr->length = sizeof(*intsrcovr);
401         intsrcovr->source = i;
402         intsrcovr->gsi    = cpu_to_le32(i);
403         intsrcovr->flags  = cpu_to_le16(0xd); /* active high, level triggered */
404     }
405 
406     local_nmi = acpi_data_push(table_data, sizeof *local_nmi);
407     local_nmi->type         = ACPI_APIC_LOCAL_NMI;
408     local_nmi->length       = sizeof(*local_nmi);
409     local_nmi->processor_id = 0xff; /* all processors */
410     local_nmi->flags        = cpu_to_le16(0);
411     local_nmi->lint         = 1; /* ACPI_LINT1 */
412 
413     build_header(linker, table_data,
414                  (void *)(table_data->data + madt_start), "APIC",
415                  table_data->len - madt_start, 1);
416 }
417 
418 #include "hw/i386/ssdt-tpm.hex"
419 
420 /* Assign BSEL property to all buses.  In the future, this can be changed
421  * to only assign to buses that support hotplug.
422  */
423 static void *acpi_set_bsel(PCIBus *bus, void *opaque)
424 {
425     unsigned *bsel_alloc = opaque;
426     unsigned *bus_bsel;
427 
428     if (qbus_is_hotpluggable(BUS(bus))) {
429         bus_bsel = g_malloc(sizeof *bus_bsel);
430 
431         *bus_bsel = (*bsel_alloc)++;
432         object_property_add_uint32_ptr(OBJECT(bus), ACPI_PCIHP_PROP_BSEL,
433                                        bus_bsel, NULL);
434     }
435 
436     return bsel_alloc;
437 }
438 
439 static void acpi_set_pci_info(void)
440 {
441     PCIBus *bus = find_i440fx(); /* TODO: Q35 support */
442     unsigned bsel_alloc = 0;
443 
444     if (bus) {
445         /* Scan all PCI buses. Set property to enable acpi based hotplug. */
446         pci_for_each_bus_depth_first(bus, acpi_set_bsel, NULL, &bsel_alloc);
447     }
448 }
449 
450 static void build_append_pcihp_notify_entry(Aml *method, int slot)
451 {
452     Aml *if_ctx;
453     int32_t devfn = PCI_DEVFN(slot, 0);
454 
455     if_ctx = aml_if(aml_and(aml_arg(0), aml_int(0x1U << slot)));
456     aml_append(if_ctx, aml_notify(aml_name("S%.02X", devfn), aml_arg(1)));
457     aml_append(method, if_ctx);
458 }
459 
460 static void build_append_pci_bus_devices(Aml *parent_scope, PCIBus *bus,
461                                          bool pcihp_bridge_en)
462 {
463     Aml *dev, *notify_method, *method;
464     QObject *bsel;
465     PCIBus *sec;
466     int i;
467 
468     bsel = object_property_get_qobject(OBJECT(bus), ACPI_PCIHP_PROP_BSEL, NULL);
469     if (bsel) {
470         int64_t bsel_val = qint_get_int(qobject_to_qint(bsel));
471 
472         aml_append(parent_scope, aml_name_decl("BSEL", aml_int(bsel_val)));
473         notify_method = aml_method("DVNT", 2);
474     }
475 
476     for (i = 0; i < ARRAY_SIZE(bus->devices); i += PCI_FUNC_MAX) {
477         DeviceClass *dc;
478         PCIDeviceClass *pc;
479         PCIDevice *pdev = bus->devices[i];
480         int slot = PCI_SLOT(i);
481         bool hotplug_enabled_dev;
482         bool bridge_in_acpi;
483 
484         if (!pdev) {
485             if (bsel) { /* add hotplug slots for non present devices */
486                 dev = aml_device("S%.02X", PCI_DEVFN(slot, 0));
487                 aml_append(dev, aml_name_decl("_SUN", aml_int(slot)));
488                 aml_append(dev, aml_name_decl("_ADR", aml_int(slot << 16)));
489                 method = aml_method("_EJ0", 1);
490                 aml_append(method,
491                     aml_call2("PCEJ", aml_name("BSEL"), aml_name("_SUN"))
492                 );
493                 aml_append(dev, method);
494                 aml_append(parent_scope, dev);
495 
496                 build_append_pcihp_notify_entry(notify_method, slot);
497             }
498             continue;
499         }
500 
501         pc = PCI_DEVICE_GET_CLASS(pdev);
502         dc = DEVICE_GET_CLASS(pdev);
503 
504         /* When hotplug for bridges is enabled, bridges are
505          * described in ACPI separately (see build_pci_bus_end).
506          * In this case they aren't themselves hot-pluggable.
507          * Hotplugged bridges *are* hot-pluggable.
508          */
509         bridge_in_acpi = pc->is_bridge && pcihp_bridge_en &&
510             !DEVICE(pdev)->hotplugged;
511 
512         hotplug_enabled_dev = bsel && dc->hotpluggable && !bridge_in_acpi;
513 
514         if (pc->class_id == PCI_CLASS_BRIDGE_ISA) {
515             continue;
516         }
517 
518         /* start to compose PCI slot descriptor */
519         dev = aml_device("S%.02X", PCI_DEVFN(slot, 0));
520         aml_append(dev, aml_name_decl("_ADR", aml_int(slot << 16)));
521 
522         if (pc->class_id == PCI_CLASS_DISPLAY_VGA) {
523             /* add VGA specific AML methods */
524             int s3d;
525 
526             if (object_dynamic_cast(OBJECT(pdev), "qxl-vga")) {
527                 s3d = 3;
528             } else {
529                 s3d = 0;
530             }
531 
532             method = aml_method("_S1D", 0);
533             aml_append(method, aml_return(aml_int(0)));
534             aml_append(dev, method);
535 
536             method = aml_method("_S2D", 0);
537             aml_append(method, aml_return(aml_int(0)));
538             aml_append(dev, method);
539 
540             method = aml_method("_S3D", 0);
541             aml_append(method, aml_return(aml_int(s3d)));
542             aml_append(dev, method);
543         } else if (hotplug_enabled_dev) {
544             /* add _SUN/_EJ0 to make slot hotpluggable  */
545             aml_append(dev, aml_name_decl("_SUN", aml_int(slot)));
546 
547             method = aml_method("_EJ0", 1);
548             aml_append(method,
549                 aml_call2("PCEJ", aml_name("BSEL"), aml_name("_SUN"))
550             );
551             aml_append(dev, method);
552 
553             if (bsel) {
554                 build_append_pcihp_notify_entry(notify_method, slot);
555             }
556         } else if (bridge_in_acpi) {
557             /*
558              * device is coldplugged bridge,
559              * add child device descriptions into its scope
560              */
561             PCIBus *sec_bus = pci_bridge_get_sec_bus(PCI_BRIDGE(pdev));
562 
563             build_append_pci_bus_devices(dev, sec_bus, pcihp_bridge_en);
564         }
565         /* slot descriptor has been composed, add it into parent context */
566         aml_append(parent_scope, dev);
567     }
568 
569     if (bsel) {
570         aml_append(parent_scope, notify_method);
571     }
572 
573     /* Append PCNT method to notify about events on local and child buses.
574      * Add unconditionally for root since DSDT expects it.
575      */
576     method = aml_method("PCNT", 0);
577 
578     /* If bus supports hotplug select it and notify about local events */
579     if (bsel) {
580         int64_t bsel_val = qint_get_int(qobject_to_qint(bsel));
581         aml_append(method, aml_store(aml_int(bsel_val), aml_name("BNUM")));
582         aml_append(method,
583             aml_call2("DVNT", aml_name("PCIU"), aml_int(1) /* Device Check */)
584         );
585         aml_append(method,
586             aml_call2("DVNT", aml_name("PCID"), aml_int(3)/* Eject Request */)
587         );
588     }
589 
590     /* Notify about child bus events in any case */
591     if (pcihp_bridge_en) {
592         QLIST_FOREACH(sec, &bus->child, sibling) {
593             int32_t devfn = sec->parent_dev->devfn;
594 
595             aml_append(method, aml_name("^S%.02X.PCNT", devfn));
596         }
597     }
598     aml_append(parent_scope, method);
599 }
600 
601 static void
602 build_ssdt(GArray *table_data, GArray *linker,
603            AcpiCpuInfo *cpu, AcpiPmInfo *pm, AcpiMiscInfo *misc,
604            PcPciInfo *pci, PcGuestInfo *guest_info)
605 {
606     MachineState *machine = MACHINE(qdev_get_machine());
607     uint32_t nr_mem = machine->ram_slots;
608     unsigned acpi_cpus = guest_info->apic_id_limit;
609     Aml *ssdt, *sb_scope, *scope, *pkg, *dev, *method, *crs, *field, *ifctx;
610     int i;
611 
612     ssdt = init_aml_allocator();
613     /* The current AML generator can cover the APIC ID range [0..255],
614      * inclusive, for VCPU hotplug. */
615     QEMU_BUILD_BUG_ON(ACPI_CPU_HOTPLUG_ID_LIMIT > 256);
616     g_assert(acpi_cpus <= ACPI_CPU_HOTPLUG_ID_LIMIT);
617 
618     /* Reserve space for header */
619     acpi_data_push(ssdt->buf, sizeof(AcpiTableHeader));
620 
621     scope = aml_scope("\\_SB.PCI0");
622     /* build PCI0._CRS */
623     crs = aml_resource_template();
624     aml_append(crs,
625         aml_word_bus_number(aml_min_fixed, aml_max_fixed, aml_pos_decode,
626                             0x0000, 0x0000, 0x00FF, 0x0000, 0x0100));
627     aml_append(crs, aml_io(aml_decode16, 0x0CF8, 0x0CF8, 0x01, 0x08));
628 
629     aml_append(crs,
630         aml_word_io(aml_min_fixed, aml_max_fixed,
631                     aml_pos_decode, aml_entire_range,
632                     0x0000, 0x0000, 0x0CF7, 0x0000, 0x0CF8));
633     aml_append(crs,
634         aml_word_io(aml_min_fixed, aml_max_fixed,
635                     aml_pos_decode, aml_entire_range,
636                     0x0000, 0x0D00, 0xFFFF, 0x0000, 0xF300));
637     aml_append(crs,
638         aml_dword_memory(aml_pos_decode, aml_min_fixed, aml_max_fixed,
639                          aml_cacheable, aml_ReadWrite,
640                          0, 0x000A0000, 0x000BFFFF, 0, 0x00020000));
641     aml_append(crs,
642         aml_dword_memory(aml_pos_decode, aml_min_fixed, aml_max_fixed,
643                          aml_non_cacheable, aml_ReadWrite,
644                          0, pci->w32.begin, pci->w32.end - 1, 0,
645                          pci->w32.end - pci->w32.begin));
646     if (pci->w64.begin) {
647         aml_append(crs,
648             aml_qword_memory(aml_pos_decode, aml_min_fixed, aml_max_fixed,
649                              aml_cacheable, aml_ReadWrite,
650                              0, pci->w64.begin, pci->w64.end - 1, 0,
651                              pci->w64.end - pci->w64.begin));
652     }
653     aml_append(scope, aml_name_decl("_CRS", crs));
654 
655     /* reserve GPE0 block resources */
656     dev = aml_device("GPE0");
657     aml_append(dev, aml_name_decl("_HID", aml_string("PNP0A06")));
658     aml_append(dev, aml_name_decl("_UID", aml_string("GPE0 resources")));
659     /* device present, functioning, decoding, not shown in UI */
660     aml_append(dev, aml_name_decl("_STA", aml_int(0xB)));
661     crs = aml_resource_template();
662     aml_append(crs,
663         aml_io(aml_decode16, pm->gpe0_blk, pm->gpe0_blk, 1, pm->gpe0_blk_len)
664     );
665     aml_append(dev, aml_name_decl("_CRS", crs));
666     aml_append(scope, dev);
667 
668     /* reserve PCIHP resources */
669     if (pm->pcihp_io_len) {
670         dev = aml_device("PHPR");
671         aml_append(dev, aml_name_decl("_HID", aml_string("PNP0A06")));
672         aml_append(dev,
673             aml_name_decl("_UID", aml_string("PCI Hotplug resources")));
674         /* device present, functioning, decoding, not shown in UI */
675         aml_append(dev, aml_name_decl("_STA", aml_int(0xB)));
676         crs = aml_resource_template();
677         aml_append(crs,
678             aml_io(aml_decode16, pm->pcihp_io_base, pm->pcihp_io_base, 1,
679                    pm->pcihp_io_len)
680         );
681         aml_append(dev, aml_name_decl("_CRS", crs));
682         aml_append(scope, dev);
683     }
684     aml_append(ssdt, scope);
685 
686     /*  create S3_ / S4_ / S5_ packages if necessary */
687     scope = aml_scope("\\");
688     if (!pm->s3_disabled) {
689         pkg = aml_package(4);
690         aml_append(pkg, aml_int(1)); /* PM1a_CNT.SLP_TYP */
691         aml_append(pkg, aml_int(1)); /* PM1b_CNT.SLP_TYP, FIXME: not impl. */
692         aml_append(pkg, aml_int(0)); /* reserved */
693         aml_append(pkg, aml_int(0)); /* reserved */
694         aml_append(scope, aml_name_decl("_S3", pkg));
695     }
696 
697     if (!pm->s4_disabled) {
698         pkg = aml_package(4);
699         aml_append(pkg, aml_int(pm->s4_val)); /* PM1a_CNT.SLP_TYP */
700         /* PM1b_CNT.SLP_TYP, FIXME: not impl. */
701         aml_append(pkg, aml_int(pm->s4_val));
702         aml_append(pkg, aml_int(0)); /* reserved */
703         aml_append(pkg, aml_int(0)); /* reserved */
704         aml_append(scope, aml_name_decl("_S4", pkg));
705     }
706 
707     pkg = aml_package(4);
708     aml_append(pkg, aml_int(0)); /* PM1a_CNT.SLP_TYP */
709     aml_append(pkg, aml_int(0)); /* PM1b_CNT.SLP_TYP not impl. */
710     aml_append(pkg, aml_int(0)); /* reserved */
711     aml_append(pkg, aml_int(0)); /* reserved */
712     aml_append(scope, aml_name_decl("_S5", pkg));
713     aml_append(ssdt, scope);
714 
715     if (misc->applesmc_io_base) {
716         scope = aml_scope("\\_SB.PCI0.ISA");
717         dev = aml_device("SMC");
718 
719         aml_append(dev, aml_name_decl("_HID", aml_eisaid("APP0001")));
720         /* device present, functioning, decoding, not shown in UI */
721         aml_append(dev, aml_name_decl("_STA", aml_int(0xB)));
722 
723         crs = aml_resource_template();
724         aml_append(crs,
725             aml_io(aml_decode16, misc->applesmc_io_base, misc->applesmc_io_base,
726                    0x01, APPLESMC_MAX_DATA_LENGTH)
727         );
728         aml_append(crs, aml_irq_no_flags(6));
729         aml_append(dev, aml_name_decl("_CRS", crs));
730 
731         aml_append(scope, dev);
732         aml_append(ssdt, scope);
733     }
734 
735     if (misc->pvpanic_port) {
736         scope = aml_scope("\\_SB.PCI0.ISA");
737 
738         dev = aml_device("PEVR");
739         aml_append(dev, aml_name_decl("_HID", aml_string("QEMU0001")));
740 
741         crs = aml_resource_template();
742         aml_append(crs,
743             aml_io(aml_decode16, misc->pvpanic_port, misc->pvpanic_port, 1, 1)
744         );
745         aml_append(dev, aml_name_decl("_CRS", crs));
746 
747         aml_append(dev, aml_operation_region("PEOR", aml_system_io,
748                                               misc->pvpanic_port, 1));
749         field = aml_field("PEOR", aml_byte_acc, aml_preserve);
750         aml_append(field, aml_named_field("PEPT", 8));
751         aml_append(dev, field);
752 
753         method = aml_method("RDPT", 0);
754         aml_append(method, aml_store(aml_name("PEPT"), aml_local(0)));
755         aml_append(method, aml_return(aml_local(0)));
756         aml_append(dev, method);
757 
758         method = aml_method("WRPT", 1);
759         aml_append(method, aml_store(aml_arg(0), aml_name("PEPT")));
760         aml_append(dev, method);
761 
762         aml_append(scope, dev);
763         aml_append(ssdt, scope);
764     }
765 
766     sb_scope = aml_scope("_SB");
767     {
768         /* create PCI0.PRES device and its _CRS to reserve CPU hotplug MMIO */
769         dev = aml_device("PCI0." stringify(CPU_HOTPLUG_RESOURCE_DEVICE));
770         aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0A06")));
771         aml_append(dev,
772             aml_name_decl("_UID", aml_string("CPU Hotplug resources"))
773         );
774         /* device present, functioning, decoding, not shown in UI */
775         aml_append(dev, aml_name_decl("_STA", aml_int(0xB)));
776         crs = aml_resource_template();
777         aml_append(crs,
778             aml_io(aml_decode16, pm->cpu_hp_io_base, pm->cpu_hp_io_base, 1,
779                    pm->cpu_hp_io_len)
780         );
781         aml_append(dev, aml_name_decl("_CRS", crs));
782         aml_append(sb_scope, dev);
783         /* declare CPU hotplug MMIO region and PRS field to access it */
784         aml_append(sb_scope, aml_operation_region(
785             "PRST", aml_system_io, pm->cpu_hp_io_base, pm->cpu_hp_io_len));
786         field = aml_field("PRST", aml_byte_acc, aml_preserve);
787         aml_append(field, aml_named_field("PRS", 256));
788         aml_append(sb_scope, field);
789 
790         /* build Processor object for each processor */
791         for (i = 0; i < acpi_cpus; i++) {
792             dev = aml_processor(i, 0, 0, "CP%.02X", i);
793 
794             method = aml_method("_MAT", 0);
795             aml_append(method, aml_return(aml_call1("CPMA", aml_int(i))));
796             aml_append(dev, method);
797 
798             method = aml_method("_STA", 0);
799             aml_append(method, aml_return(aml_call1("CPST", aml_int(i))));
800             aml_append(dev, method);
801 
802             method = aml_method("_EJ0", 1);
803             aml_append(method,
804                 aml_return(aml_call2("CPEJ", aml_int(i), aml_arg(0)))
805             );
806             aml_append(dev, method);
807 
808             aml_append(sb_scope, dev);
809         }
810 
811         /* build this code:
812          *   Method(NTFY, 2) {If (LEqual(Arg0, 0x00)) {Notify(CP00, Arg1)} ...}
813          */
814         /* Arg0 = Processor ID = APIC ID */
815         method = aml_method("NTFY", 2);
816         for (i = 0; i < acpi_cpus; i++) {
817             ifctx = aml_if(aml_equal(aml_arg(0), aml_int(i)));
818             aml_append(ifctx,
819                 aml_notify(aml_name("CP%.02X", i), aml_arg(1))
820             );
821             aml_append(method, ifctx);
822         }
823         aml_append(sb_scope, method);
824 
825         /* build "Name(CPON, Package() { One, One, ..., Zero, Zero, ... })"
826          *
827          * Note: The ability to create variable-sized packages was first
828          * introduced in ACPI 2.0. ACPI 1.0 only allowed fixed-size packages
829          * ith up to 255 elements. Windows guests up to win2k8 fail when
830          * VarPackageOp is used.
831          */
832         pkg = acpi_cpus <= 255 ? aml_package(acpi_cpus) :
833                                  aml_varpackage(acpi_cpus);
834 
835         for (i = 0; i < acpi_cpus; i++) {
836             uint8_t b = test_bit(i, cpu->found_cpus) ? 0x01 : 0x00;
837             aml_append(pkg, aml_int(b));
838         }
839         aml_append(sb_scope, aml_name_decl("CPON", pkg));
840 
841         /* build memory devices */
842         assert(nr_mem <= ACPI_MAX_RAM_SLOTS);
843         scope = aml_scope("\\_SB.PCI0." stringify(MEMORY_HOTPLUG_DEVICE));
844         aml_append(scope,
845             aml_name_decl(stringify(MEMORY_SLOTS_NUMBER), aml_int(nr_mem))
846         );
847 
848         crs = aml_resource_template();
849         aml_append(crs,
850             aml_io(aml_decode16, pm->mem_hp_io_base, pm->mem_hp_io_base, 0,
851                    pm->mem_hp_io_len)
852         );
853         aml_append(scope, aml_name_decl("_CRS", crs));
854 
855         aml_append(scope, aml_operation_region(
856             stringify(MEMORY_HOTPLUG_IO_REGION), aml_system_io,
857             pm->mem_hp_io_base, pm->mem_hp_io_len)
858         );
859 
860         field = aml_field(stringify(MEMORY_HOTPLUG_IO_REGION), aml_dword_acc,
861                           aml_preserve);
862         aml_append(field, /* read only */
863             aml_named_field(stringify(MEMORY_SLOT_ADDR_LOW), 32));
864         aml_append(field, /* read only */
865             aml_named_field(stringify(MEMORY_SLOT_ADDR_HIGH), 32));
866         aml_append(field, /* read only */
867             aml_named_field(stringify(MEMORY_SLOT_SIZE_LOW), 32));
868         aml_append(field, /* read only */
869             aml_named_field(stringify(MEMORY_SLOT_SIZE_HIGH), 32));
870         aml_append(field, /* read only */
871             aml_named_field(stringify(MEMORY_SLOT_PROXIMITY), 32));
872         aml_append(scope, field);
873 
874         field = aml_field(stringify(MEMORY_HOTPLUG_IO_REGION), aml_byte_acc,
875                           aml_write_as_zeros);
876         aml_append(field, aml_reserved_field(160 /* bits, Offset(20) */));
877         aml_append(field, /* 1 if enabled, read only */
878             aml_named_field(stringify(MEMORY_SLOT_ENABLED), 1));
879         aml_append(field,
880             /*(read) 1 if has a insert event. (write) 1 to clear event */
881             aml_named_field(stringify(MEMORY_SLOT_INSERT_EVENT), 1));
882         aml_append(scope, field);
883 
884         field = aml_field(stringify(MEMORY_HOTPLUG_IO_REGION), aml_dword_acc,
885                           aml_preserve);
886         aml_append(field, /* DIMM selector, write only */
887             aml_named_field(stringify(MEMORY_SLOT_SLECTOR), 32));
888         aml_append(field, /* _OST event code, write only */
889             aml_named_field(stringify(MEMORY_SLOT_OST_EVENT), 32));
890         aml_append(field, /* _OST status code, write only */
891             aml_named_field(stringify(MEMORY_SLOT_OST_STATUS), 32));
892         aml_append(scope, field);
893 
894         aml_append(sb_scope, scope);
895 
896         for (i = 0; i < nr_mem; i++) {
897             #define BASEPATH "\\_SB.PCI0." stringify(MEMORY_HOTPLUG_DEVICE) "."
898             const char *s;
899 
900             dev = aml_device("MP%02X", i);
901             aml_append(dev, aml_name_decl("_UID", aml_string("0x%02X", i)));
902             aml_append(dev, aml_name_decl("_HID", aml_eisaid("PNP0C80")));
903 
904             method = aml_method("_CRS", 0);
905             s = BASEPATH stringify(MEMORY_SLOT_CRS_METHOD);
906             aml_append(method, aml_return(aml_call1(s, aml_name("_UID"))));
907             aml_append(dev, method);
908 
909             method = aml_method("_STA", 0);
910             s = BASEPATH stringify(MEMORY_SLOT_STATUS_METHOD);
911             aml_append(method, aml_return(aml_call1(s, aml_name("_UID"))));
912             aml_append(dev, method);
913 
914             method = aml_method("_PXM", 0);
915             s = BASEPATH stringify(MEMORY_SLOT_PROXIMITY_METHOD);
916             aml_append(method, aml_return(aml_call1(s, aml_name("_UID"))));
917             aml_append(dev, method);
918 
919             method = aml_method("_OST", 3);
920             s = BASEPATH stringify(MEMORY_SLOT_OST_METHOD);
921             aml_append(method, aml_return(aml_call4(
922                 s, aml_name("_UID"), aml_arg(0), aml_arg(1), aml_arg(2)
923             )));
924             aml_append(dev, method);
925 
926             aml_append(sb_scope, dev);
927         }
928 
929         /* build Method(MEMORY_SLOT_NOTIFY_METHOD, 2) {
930          *     If (LEqual(Arg0, 0x00)) {Notify(MP00, Arg1)} ... }
931          */
932         method = aml_method(stringify(MEMORY_SLOT_NOTIFY_METHOD), 2);
933         for (i = 0; i < nr_mem; i++) {
934             ifctx = aml_if(aml_equal(aml_arg(0), aml_int(i)));
935             aml_append(ifctx,
936                 aml_notify(aml_name("MP%.02X", i), aml_arg(1))
937             );
938             aml_append(method, ifctx);
939         }
940         aml_append(sb_scope, method);
941 
942         {
943             Object *pci_host;
944             PCIBus *bus = NULL;
945             bool ambiguous;
946 
947             pci_host = object_resolve_path_type("", TYPE_PCI_HOST_BRIDGE, &ambiguous);
948             if (!ambiguous && pci_host) {
949                 bus = PCI_HOST_BRIDGE(pci_host)->bus;
950             }
951 
952             if (bus) {
953                 Aml *scope = aml_scope("PCI0");
954                 /* Scan all PCI buses. Generate tables to support hotplug. */
955                 build_append_pci_bus_devices(scope, bus, pm->pcihp_bridge_en);
956                 aml_append(sb_scope, scope);
957             }
958         }
959         aml_append(ssdt, sb_scope);
960     }
961 
962     /* copy AML table into ACPI tables blob and patch header there */
963     g_array_append_vals(table_data, ssdt->buf->data, ssdt->buf->len);
964     build_header(linker, table_data,
965         (void *)(table_data->data + table_data->len - ssdt->buf->len),
966         "SSDT", ssdt->buf->len, 1);
967     free_aml_allocator();
968 }
969 
970 static void
971 build_hpet(GArray *table_data, GArray *linker)
972 {
973     Acpi20Hpet *hpet;
974 
975     hpet = acpi_data_push(table_data, sizeof(*hpet));
976     /* Note timer_block_id value must be kept in sync with value advertised by
977      * emulated hpet
978      */
979     hpet->timer_block_id = cpu_to_le32(0x8086a201);
980     hpet->addr.address = cpu_to_le64(HPET_BASE);
981     build_header(linker, table_data,
982                  (void *)hpet, "HPET", sizeof(*hpet), 1);
983 }
984 
985 static void
986 build_tpm_tcpa(GArray *table_data, GArray *linker, GArray *tcpalog)
987 {
988     Acpi20Tcpa *tcpa = acpi_data_push(table_data, sizeof *tcpa);
989     uint64_t log_area_start_address = acpi_data_len(tcpalog);
990 
991     tcpa->platform_class = cpu_to_le16(TPM_TCPA_ACPI_CLASS_CLIENT);
992     tcpa->log_area_minimum_length = cpu_to_le32(TPM_LOG_AREA_MINIMUM_SIZE);
993     tcpa->log_area_start_address = cpu_to_le64(log_area_start_address);
994 
995     bios_linker_loader_alloc(linker, ACPI_BUILD_TPMLOG_FILE, 1,
996                              false /* high memory */);
997 
998     /* log area start address to be filled by Guest linker */
999     bios_linker_loader_add_pointer(linker, ACPI_BUILD_TABLE_FILE,
1000                                    ACPI_BUILD_TPMLOG_FILE,
1001                                    table_data, &tcpa->log_area_start_address,
1002                                    sizeof(tcpa->log_area_start_address));
1003 
1004     build_header(linker, table_data,
1005                  (void *)tcpa, "TCPA", sizeof(*tcpa), 2);
1006 
1007     acpi_data_push(tcpalog, TPM_LOG_AREA_MINIMUM_SIZE);
1008 }
1009 
1010 static void
1011 build_tpm_ssdt(GArray *table_data, GArray *linker)
1012 {
1013     void *tpm_ptr;
1014 
1015     tpm_ptr = acpi_data_push(table_data, sizeof(ssdt_tpm_aml));
1016     memcpy(tpm_ptr, ssdt_tpm_aml, sizeof(ssdt_tpm_aml));
1017 }
1018 
1019 typedef enum {
1020     MEM_AFFINITY_NOFLAGS      = 0,
1021     MEM_AFFINITY_ENABLED      = (1 << 0),
1022     MEM_AFFINITY_HOTPLUGGABLE = (1 << 1),
1023     MEM_AFFINITY_NON_VOLATILE = (1 << 2),
1024 } MemoryAffinityFlags;
1025 
1026 static void
1027 acpi_build_srat_memory(AcpiSratMemoryAffinity *numamem, uint64_t base,
1028                        uint64_t len, int node, MemoryAffinityFlags flags)
1029 {
1030     numamem->type = ACPI_SRAT_MEMORY;
1031     numamem->length = sizeof(*numamem);
1032     memset(numamem->proximity, 0, 4);
1033     numamem->proximity[0] = node;
1034     numamem->flags = cpu_to_le32(flags);
1035     numamem->base_addr = cpu_to_le64(base);
1036     numamem->range_length = cpu_to_le64(len);
1037 }
1038 
1039 static void
1040 build_srat(GArray *table_data, GArray *linker, PcGuestInfo *guest_info)
1041 {
1042     AcpiSystemResourceAffinityTable *srat;
1043     AcpiSratProcessorAffinity *core;
1044     AcpiSratMemoryAffinity *numamem;
1045 
1046     int i;
1047     uint64_t curnode;
1048     int srat_start, numa_start, slots;
1049     uint64_t mem_len, mem_base, next_base;
1050     PCMachineState *pcms = PC_MACHINE(qdev_get_machine());
1051     ram_addr_t hotplugabble_address_space_size =
1052         object_property_get_int(OBJECT(pcms), PC_MACHINE_MEMHP_REGION_SIZE,
1053                                 NULL);
1054 
1055     srat_start = table_data->len;
1056 
1057     srat = acpi_data_push(table_data, sizeof *srat);
1058     srat->reserved1 = cpu_to_le32(1);
1059     core = (void *)(srat + 1);
1060 
1061     for (i = 0; i < guest_info->apic_id_limit; ++i) {
1062         core = acpi_data_push(table_data, sizeof *core);
1063         core->type = ACPI_SRAT_PROCESSOR;
1064         core->length = sizeof(*core);
1065         core->local_apic_id = i;
1066         curnode = guest_info->node_cpu[i];
1067         core->proximity_lo = curnode;
1068         memset(core->proximity_hi, 0, 3);
1069         core->local_sapic_eid = 0;
1070         core->flags = cpu_to_le32(1);
1071     }
1072 
1073 
1074     /* the memory map is a bit tricky, it contains at least one hole
1075      * from 640k-1M and possibly another one from 3.5G-4G.
1076      */
1077     next_base = 0;
1078     numa_start = table_data->len;
1079 
1080     numamem = acpi_data_push(table_data, sizeof *numamem);
1081     acpi_build_srat_memory(numamem, 0, 640*1024, 0, MEM_AFFINITY_ENABLED);
1082     next_base = 1024 * 1024;
1083     for (i = 1; i < guest_info->numa_nodes + 1; ++i) {
1084         mem_base = next_base;
1085         mem_len = guest_info->node_mem[i - 1];
1086         if (i == 1) {
1087             mem_len -= 1024 * 1024;
1088         }
1089         next_base = mem_base + mem_len;
1090 
1091         /* Cut out the ACPI_PCI hole */
1092         if (mem_base <= guest_info->ram_size_below_4g &&
1093             next_base > guest_info->ram_size_below_4g) {
1094             mem_len -= next_base - guest_info->ram_size_below_4g;
1095             if (mem_len > 0) {
1096                 numamem = acpi_data_push(table_data, sizeof *numamem);
1097                 acpi_build_srat_memory(numamem, mem_base, mem_len, i - 1,
1098                                        MEM_AFFINITY_ENABLED);
1099             }
1100             mem_base = 1ULL << 32;
1101             mem_len = next_base - guest_info->ram_size_below_4g;
1102             next_base += (1ULL << 32) - guest_info->ram_size_below_4g;
1103         }
1104         numamem = acpi_data_push(table_data, sizeof *numamem);
1105         acpi_build_srat_memory(numamem, mem_base, mem_len, i - 1,
1106                                MEM_AFFINITY_ENABLED);
1107     }
1108     slots = (table_data->len - numa_start) / sizeof *numamem;
1109     for (; slots < guest_info->numa_nodes + 2; slots++) {
1110         numamem = acpi_data_push(table_data, sizeof *numamem);
1111         acpi_build_srat_memory(numamem, 0, 0, 0, MEM_AFFINITY_NOFLAGS);
1112     }
1113 
1114     /*
1115      * Entry is required for Windows to enable memory hotplug in OS.
1116      * Memory devices may override proximity set by this entry,
1117      * providing _PXM method if necessary.
1118      */
1119     if (hotplugabble_address_space_size) {
1120         numamem = acpi_data_push(table_data, sizeof *numamem);
1121         acpi_build_srat_memory(numamem, pcms->hotplug_memory_base,
1122                                hotplugabble_address_space_size, 0,
1123                                MEM_AFFINITY_HOTPLUGGABLE |
1124                                MEM_AFFINITY_ENABLED);
1125     }
1126 
1127     build_header(linker, table_data,
1128                  (void *)(table_data->data + srat_start),
1129                  "SRAT",
1130                  table_data->len - srat_start, 1);
1131 }
1132 
1133 static void
1134 build_mcfg_q35(GArray *table_data, GArray *linker, AcpiMcfgInfo *info)
1135 {
1136     AcpiTableMcfg *mcfg;
1137     const char *sig;
1138     int len = sizeof(*mcfg) + 1 * sizeof(mcfg->allocation[0]);
1139 
1140     mcfg = acpi_data_push(table_data, len);
1141     mcfg->allocation[0].address = cpu_to_le64(info->mcfg_base);
1142     /* Only a single allocation so no need to play with segments */
1143     mcfg->allocation[0].pci_segment = cpu_to_le16(0);
1144     mcfg->allocation[0].start_bus_number = 0;
1145     mcfg->allocation[0].end_bus_number = PCIE_MMCFG_BUS(info->mcfg_size - 1);
1146 
1147     /* MCFG is used for ECAM which can be enabled or disabled by guest.
1148      * To avoid table size changes (which create migration issues),
1149      * always create the table even if there are no allocations,
1150      * but set the signature to a reserved value in this case.
1151      * ACPI spec requires OSPMs to ignore such tables.
1152      */
1153     if (info->mcfg_base == PCIE_BASE_ADDR_UNMAPPED) {
1154         /* Reserved signature: ignored by OSPM */
1155         sig = "QEMU";
1156     } else {
1157         sig = "MCFG";
1158     }
1159     build_header(linker, table_data, (void *)mcfg, sig, len, 1);
1160 }
1161 
1162 static void
1163 build_dmar_q35(GArray *table_data, GArray *linker)
1164 {
1165     int dmar_start = table_data->len;
1166 
1167     AcpiTableDmar *dmar;
1168     AcpiDmarHardwareUnit *drhd;
1169 
1170     dmar = acpi_data_push(table_data, sizeof(*dmar));
1171     dmar->host_address_width = VTD_HOST_ADDRESS_WIDTH - 1;
1172     dmar->flags = 0;    /* No intr_remap for now */
1173 
1174     /* DMAR Remapping Hardware Unit Definition structure */
1175     drhd = acpi_data_push(table_data, sizeof(*drhd));
1176     drhd->type = cpu_to_le16(ACPI_DMAR_TYPE_HARDWARE_UNIT);
1177     drhd->length = cpu_to_le16(sizeof(*drhd));   /* No device scope now */
1178     drhd->flags = ACPI_DMAR_INCLUDE_PCI_ALL;
1179     drhd->pci_segment = cpu_to_le16(0);
1180     drhd->address = cpu_to_le64(Q35_HOST_BRIDGE_IOMMU_ADDR);
1181 
1182     build_header(linker, table_data, (void *)(table_data->data + dmar_start),
1183                  "DMAR", table_data->len - dmar_start, 1);
1184 }
1185 
1186 static void
1187 build_dsdt(GArray *table_data, GArray *linker, AcpiMiscInfo *misc)
1188 {
1189     AcpiTableHeader *dsdt;
1190 
1191     assert(misc->dsdt_code && misc->dsdt_size);
1192 
1193     dsdt = acpi_data_push(table_data, misc->dsdt_size);
1194     memcpy(dsdt, misc->dsdt_code, misc->dsdt_size);
1195 
1196     memset(dsdt, 0, sizeof *dsdt);
1197     build_header(linker, table_data, dsdt, "DSDT",
1198                  misc->dsdt_size, 1);
1199 }
1200 
1201 /* Build final rsdt table */
1202 static void
1203 build_rsdt(GArray *table_data, GArray *linker, GArray *table_offsets)
1204 {
1205     AcpiRsdtDescriptorRev1 *rsdt;
1206     size_t rsdt_len;
1207     int i;
1208 
1209     rsdt_len = sizeof(*rsdt) + sizeof(uint32_t) * table_offsets->len;
1210     rsdt = acpi_data_push(table_data, rsdt_len);
1211     memcpy(rsdt->table_offset_entry, table_offsets->data,
1212            sizeof(uint32_t) * table_offsets->len);
1213     for (i = 0; i < table_offsets->len; ++i) {
1214         /* rsdt->table_offset_entry to be filled by Guest linker */
1215         bios_linker_loader_add_pointer(linker,
1216                                        ACPI_BUILD_TABLE_FILE,
1217                                        ACPI_BUILD_TABLE_FILE,
1218                                        table_data, &rsdt->table_offset_entry[i],
1219                                        sizeof(uint32_t));
1220     }
1221     build_header(linker, table_data,
1222                  (void *)rsdt, "RSDT", rsdt_len, 1);
1223 }
1224 
1225 static GArray *
1226 build_rsdp(GArray *rsdp_table, GArray *linker, unsigned rsdt)
1227 {
1228     AcpiRsdpDescriptor *rsdp = acpi_data_push(rsdp_table, sizeof *rsdp);
1229 
1230     bios_linker_loader_alloc(linker, ACPI_BUILD_RSDP_FILE, 16,
1231                              true /* fseg memory */);
1232 
1233     memcpy(&rsdp->signature, "RSD PTR ", 8);
1234     memcpy(rsdp->oem_id, ACPI_BUILD_APPNAME6, 6);
1235     rsdp->rsdt_physical_address = cpu_to_le32(rsdt);
1236     /* Address to be filled by Guest linker */
1237     bios_linker_loader_add_pointer(linker, ACPI_BUILD_RSDP_FILE,
1238                                    ACPI_BUILD_TABLE_FILE,
1239                                    rsdp_table, &rsdp->rsdt_physical_address,
1240                                    sizeof rsdp->rsdt_physical_address);
1241     rsdp->checksum = 0;
1242     /* Checksum to be filled by Guest linker */
1243     bios_linker_loader_add_checksum(linker, ACPI_BUILD_RSDP_FILE,
1244                                     rsdp, rsdp, sizeof *rsdp, &rsdp->checksum);
1245 
1246     return rsdp_table;
1247 }
1248 
1249 typedef
1250 struct AcpiBuildState {
1251     /* Copy of table in RAM (for patching). */
1252     ram_addr_t table_ram;
1253     /* Is table patched? */
1254     uint8_t patched;
1255     PcGuestInfo *guest_info;
1256     void *rsdp;
1257     ram_addr_t rsdp_ram;
1258     ram_addr_t linker_ram;
1259 } AcpiBuildState;
1260 
1261 static bool acpi_get_mcfg(AcpiMcfgInfo *mcfg)
1262 {
1263     Object *pci_host;
1264     QObject *o;
1265     bool ambiguous;
1266 
1267     pci_host = object_resolve_path_type("", TYPE_PCI_HOST_BRIDGE, &ambiguous);
1268     g_assert(!ambiguous);
1269     g_assert(pci_host);
1270 
1271     o = object_property_get_qobject(pci_host, PCIE_HOST_MCFG_BASE, NULL);
1272     if (!o) {
1273         return false;
1274     }
1275     mcfg->mcfg_base = qint_get_int(qobject_to_qint(o));
1276     qobject_decref(o);
1277 
1278     o = object_property_get_qobject(pci_host, PCIE_HOST_MCFG_SIZE, NULL);
1279     assert(o);
1280     mcfg->mcfg_size = qint_get_int(qobject_to_qint(o));
1281     qobject_decref(o);
1282     return true;
1283 }
1284 
1285 static bool acpi_has_iommu(void)
1286 {
1287     bool ambiguous;
1288     Object *intel_iommu;
1289 
1290     intel_iommu = object_resolve_path_type("", TYPE_INTEL_IOMMU_DEVICE,
1291                                            &ambiguous);
1292     return intel_iommu && !ambiguous;
1293 }
1294 
1295 static
1296 void acpi_build(PcGuestInfo *guest_info, AcpiBuildTables *tables)
1297 {
1298     GArray *table_offsets;
1299     unsigned facs, ssdt, dsdt, rsdt;
1300     AcpiCpuInfo cpu;
1301     AcpiPmInfo pm;
1302     AcpiMiscInfo misc;
1303     AcpiMcfgInfo mcfg;
1304     PcPciInfo pci;
1305     uint8_t *u;
1306     size_t aml_len = 0;
1307     GArray *tables_blob = tables->table_data;
1308 
1309     acpi_get_cpu_info(&cpu);
1310     acpi_get_pm_info(&pm);
1311     acpi_get_dsdt(&misc);
1312     acpi_get_misc_info(&misc);
1313     acpi_get_pci_info(&pci);
1314 
1315     table_offsets = g_array_new(false, true /* clear */,
1316                                         sizeof(uint32_t));
1317     ACPI_BUILD_DPRINTF("init ACPI tables\n");
1318 
1319     bios_linker_loader_alloc(tables->linker, ACPI_BUILD_TABLE_FILE,
1320                              64 /* Ensure FACS is aligned */,
1321                              false /* high memory */);
1322 
1323     /*
1324      * FACS is pointed to by FADT.
1325      * We place it first since it's the only table that has alignment
1326      * requirements.
1327      */
1328     facs = tables_blob->len;
1329     build_facs(tables_blob, tables->linker, guest_info);
1330 
1331     /* DSDT is pointed to by FADT */
1332     dsdt = tables_blob->len;
1333     build_dsdt(tables_blob, tables->linker, &misc);
1334 
1335     /* Count the size of the DSDT and SSDT, we will need it for legacy
1336      * sizing of ACPI tables.
1337      */
1338     aml_len += tables_blob->len - dsdt;
1339 
1340     /* ACPI tables pointed to by RSDT */
1341     acpi_add_table(table_offsets, tables_blob);
1342     build_fadt(tables_blob, tables->linker, &pm, facs, dsdt);
1343 
1344     ssdt = tables_blob->len;
1345     acpi_add_table(table_offsets, tables_blob);
1346     build_ssdt(tables_blob, tables->linker, &cpu, &pm, &misc, &pci,
1347                guest_info);
1348     aml_len += tables_blob->len - ssdt;
1349 
1350     acpi_add_table(table_offsets, tables_blob);
1351     build_madt(tables_blob, tables->linker, &cpu, guest_info);
1352 
1353     if (misc.has_hpet) {
1354         acpi_add_table(table_offsets, tables_blob);
1355         build_hpet(tables_blob, tables->linker);
1356     }
1357     if (misc.has_tpm) {
1358         acpi_add_table(table_offsets, tables_blob);
1359         build_tpm_tcpa(tables_blob, tables->linker, tables->tcpalog);
1360 
1361         acpi_add_table(table_offsets, tables_blob);
1362         build_tpm_ssdt(tables_blob, tables->linker);
1363     }
1364     if (guest_info->numa_nodes) {
1365         acpi_add_table(table_offsets, tables_blob);
1366         build_srat(tables_blob, tables->linker, guest_info);
1367     }
1368     if (acpi_get_mcfg(&mcfg)) {
1369         acpi_add_table(table_offsets, tables_blob);
1370         build_mcfg_q35(tables_blob, tables->linker, &mcfg);
1371     }
1372     if (acpi_has_iommu()) {
1373         acpi_add_table(table_offsets, tables_blob);
1374         build_dmar_q35(tables_blob, tables->linker);
1375     }
1376 
1377     /* Add tables supplied by user (if any) */
1378     for (u = acpi_table_first(); u; u = acpi_table_next(u)) {
1379         unsigned len = acpi_table_len(u);
1380 
1381         acpi_add_table(table_offsets, tables_blob);
1382         g_array_append_vals(tables_blob, u, len);
1383     }
1384 
1385     /* RSDT is pointed to by RSDP */
1386     rsdt = tables_blob->len;
1387     build_rsdt(tables_blob, tables->linker, table_offsets);
1388 
1389     /* RSDP is in FSEG memory, so allocate it separately */
1390     build_rsdp(tables->rsdp, tables->linker, rsdt);
1391 
1392     /* We'll expose it all to Guest so we want to reduce
1393      * chance of size changes.
1394      *
1395      * We used to align the tables to 4k, but of course this would
1396      * too simple to be enough.  4k turned out to be too small an
1397      * alignment very soon, and in fact it is almost impossible to
1398      * keep the table size stable for all (max_cpus, max_memory_slots)
1399      * combinations.  So the table size is always 64k for pc-i440fx-2.1
1400      * and we give an error if the table grows beyond that limit.
1401      *
1402      * We still have the problem of migrating from "-M pc-i440fx-2.0".  For
1403      * that, we exploit the fact that QEMU 2.1 generates _smaller_ tables
1404      * than 2.0 and we can always pad the smaller tables with zeros.  We can
1405      * then use the exact size of the 2.0 tables.
1406      *
1407      * All this is for PIIX4, since QEMU 2.0 didn't support Q35 migration.
1408      */
1409     if (guest_info->legacy_acpi_table_size) {
1410         /* Subtracting aml_len gives the size of fixed tables.  Then add the
1411          * size of the PIIX4 DSDT/SSDT in QEMU 2.0.
1412          */
1413         int legacy_aml_len =
1414             guest_info->legacy_acpi_table_size +
1415             ACPI_BUILD_LEGACY_CPU_AML_SIZE * max_cpus;
1416         int legacy_table_size =
1417             ROUND_UP(tables_blob->len - aml_len + legacy_aml_len,
1418                      ACPI_BUILD_ALIGN_SIZE);
1419         if (tables_blob->len > legacy_table_size) {
1420             /* Should happen only with PCI bridges and -M pc-i440fx-2.0.  */
1421             error_report("Warning: migration may not work.");
1422         }
1423         g_array_set_size(tables_blob, legacy_table_size);
1424     } else {
1425         /* Make sure we have a buffer in case we need to resize the tables. */
1426         if (tables_blob->len > ACPI_BUILD_TABLE_SIZE / 2) {
1427             /* As of QEMU 2.1, this fires with 160 VCPUs and 255 memory slots.  */
1428             error_report("Warning: ACPI tables are larger than 64k.");
1429             error_report("Warning: migration may not work.");
1430             error_report("Warning: please remove CPUs, NUMA nodes, "
1431                          "memory slots or PCI bridges.");
1432         }
1433         acpi_align_size(tables_blob, ACPI_BUILD_TABLE_SIZE);
1434     }
1435 
1436     acpi_align_size(tables->linker, ACPI_BUILD_ALIGN_SIZE);
1437 
1438     /* Cleanup memory that's no longer used. */
1439     g_array_free(table_offsets, true);
1440 }
1441 
1442 static void acpi_ram_update(ram_addr_t ram, GArray *data)
1443 {
1444     uint32_t size = acpi_data_len(data);
1445 
1446     /* Make sure RAM size is correct - in case it got changed e.g. by migration */
1447     qemu_ram_resize(ram, size, &error_abort);
1448 
1449     memcpy(qemu_get_ram_ptr(ram), data->data, size);
1450     cpu_physical_memory_set_dirty_range_nocode(ram, size);
1451 }
1452 
1453 static void acpi_build_update(void *build_opaque, uint32_t offset)
1454 {
1455     AcpiBuildState *build_state = build_opaque;
1456     AcpiBuildTables tables;
1457 
1458     /* No state to update or already patched? Nothing to do. */
1459     if (!build_state || build_state->patched) {
1460         return;
1461     }
1462     build_state->patched = 1;
1463 
1464     acpi_build_tables_init(&tables);
1465 
1466     acpi_build(build_state->guest_info, &tables);
1467 
1468     acpi_ram_update(build_state->table_ram, tables.table_data);
1469 
1470     if (build_state->rsdp) {
1471         memcpy(build_state->rsdp, tables.rsdp->data, acpi_data_len(tables.rsdp));
1472     } else {
1473         acpi_ram_update(build_state->rsdp_ram, tables.rsdp);
1474     }
1475 
1476     acpi_ram_update(build_state->linker_ram, tables.linker);
1477     acpi_build_tables_cleanup(&tables, true);
1478 }
1479 
1480 static void acpi_build_reset(void *build_opaque)
1481 {
1482     AcpiBuildState *build_state = build_opaque;
1483     build_state->patched = 0;
1484 }
1485 
1486 static ram_addr_t acpi_add_rom_blob(AcpiBuildState *build_state, GArray *blob,
1487                                const char *name, uint64_t max_size)
1488 {
1489     return rom_add_blob(name, blob->data, acpi_data_len(blob), max_size, -1,
1490                         name, acpi_build_update, build_state);
1491 }
1492 
1493 static const VMStateDescription vmstate_acpi_build = {
1494     .name = "acpi_build",
1495     .version_id = 1,
1496     .minimum_version_id = 1,
1497     .fields = (VMStateField[]) {
1498         VMSTATE_UINT8(patched, AcpiBuildState),
1499         VMSTATE_END_OF_LIST()
1500     },
1501 };
1502 
1503 void acpi_setup(PcGuestInfo *guest_info)
1504 {
1505     AcpiBuildTables tables;
1506     AcpiBuildState *build_state;
1507 
1508     if (!guest_info->fw_cfg) {
1509         ACPI_BUILD_DPRINTF("No fw cfg. Bailing out.\n");
1510         return;
1511     }
1512 
1513     if (!guest_info->has_acpi_build) {
1514         ACPI_BUILD_DPRINTF("ACPI build disabled. Bailing out.\n");
1515         return;
1516     }
1517 
1518     if (!acpi_enabled) {
1519         ACPI_BUILD_DPRINTF("ACPI disabled. Bailing out.\n");
1520         return;
1521     }
1522 
1523     build_state = g_malloc0(sizeof *build_state);
1524 
1525     build_state->guest_info = guest_info;
1526 
1527     acpi_set_pci_info();
1528 
1529     acpi_build_tables_init(&tables);
1530     acpi_build(build_state->guest_info, &tables);
1531 
1532     /* Now expose it all to Guest */
1533     build_state->table_ram = acpi_add_rom_blob(build_state, tables.table_data,
1534                                                ACPI_BUILD_TABLE_FILE,
1535                                                ACPI_BUILD_TABLE_MAX_SIZE);
1536     assert(build_state->table_ram != RAM_ADDR_MAX);
1537 
1538     build_state->linker_ram =
1539         acpi_add_rom_blob(build_state, tables.linker, "etc/table-loader", 0);
1540 
1541     fw_cfg_add_file(guest_info->fw_cfg, ACPI_BUILD_TPMLOG_FILE,
1542                     tables.tcpalog->data, acpi_data_len(tables.tcpalog));
1543 
1544     if (!guest_info->rsdp_in_ram) {
1545         /*
1546          * Keep for compatibility with old machine types.
1547          * Though RSDP is small, its contents isn't immutable, so
1548          * we'll update it along with the rest of tables on guest access.
1549          */
1550         uint32_t rsdp_size = acpi_data_len(tables.rsdp);
1551 
1552         build_state->rsdp = g_memdup(tables.rsdp->data, rsdp_size);
1553         fw_cfg_add_file_callback(guest_info->fw_cfg, ACPI_BUILD_RSDP_FILE,
1554                                  acpi_build_update, build_state,
1555                                  build_state->rsdp, rsdp_size);
1556         build_state->rsdp_ram = (ram_addr_t)-1;
1557     } else {
1558         build_state->rsdp = NULL;
1559         build_state->rsdp_ram = acpi_add_rom_blob(build_state, tables.rsdp,
1560                                                   ACPI_BUILD_RSDP_FILE, 0);
1561     }
1562 
1563     qemu_register_reset(acpi_build_reset, build_state);
1564     acpi_build_reset(build_state);
1565     vmstate_register(NULL, 0, &vmstate_acpi_build, build_state);
1566 
1567     /* Cleanup tables but don't free the memory: we track it
1568      * in build_state.
1569      */
1570     acpi_build_tables_cleanup(&tables, false);
1571 }
1572