2 * QEMU fw_cfg helpers (X86 specific)
4 * Copyright (c) 2019 Red Hat, Inc.
7 * Philippe Mathieu-Daudé <philmd@redhat.com>
9 * SPDX-License-Identifier: GPL-2.0-or-later
11 * This work is licensed under the terms of the GNU GPL, version 2 or later.
12 * See the COPYING file in the top-level directory.
15 #include "qemu/osdep.h"
16 #include "sysemu/numa.h"
17 #include "hw/acpi/acpi.h"
18 #include "hw/acpi/aml-build.h"
19 #include "hw/firmware/smbios.h"
20 #include "hw/i386/fw_cfg.h"
21 #include "hw/timer/hpet.h"
22 #include "hw/nvram/fw_cfg.h"
23 #include "e820_memory_layout.h"
24 #include "kvm/kvm_i386.h"
25 #include "qapi/error.h"
26 #include CONFIG_DEVICES
27 #include "target/i386/cpu.h"
29 struct hpet_fw_config hpet_cfg
= {.count
= UINT8_MAX
};
31 const char *fw_cfg_arch_key_name(uint16_t key
)
36 } fw_cfg_arch_wellknown_keys
[] = {
37 {FW_CFG_ACPI_TABLES
, "acpi_tables"},
38 {FW_CFG_SMBIOS_ENTRIES
, "smbios_entries"},
39 {FW_CFG_IRQ0_OVERRIDE
, "irq0_override"},
40 {FW_CFG_HPET
, "hpet"},
43 for (size_t i
= 0; i
< ARRAY_SIZE(fw_cfg_arch_wellknown_keys
); i
++) {
44 if (fw_cfg_arch_wellknown_keys
[i
].key
== key
) {
45 return fw_cfg_arch_wellknown_keys
[i
].name
;
51 /* Add etc/e820 late, once all regions should be present */
52 void fw_cfg_add_e820(FWCfgState
*fw_cfg
)
54 struct e820_entry
*table
;
55 int nr_e820
= e820_get_table(&table
);
57 fw_cfg_add_file(fw_cfg
, "etc/e820", table
, nr_e820
* sizeof(*table
));
60 void fw_cfg_build_smbios(PCMachineState
*pcms
, FWCfgState
*fw_cfg
,
61 SmbiosEntryPointType ep_type
)
64 uint8_t *smbios_tables
, *smbios_anchor
;
65 size_t smbios_tables_len
, smbios_anchor_len
;
66 struct smbios_phys_mem_area
*mem_array
;
67 unsigned i
, array_count
;
68 MachineState
*ms
= MACHINE(pcms
);
69 PCMachineClass
*pcmc
= PC_MACHINE_GET_CLASS(pcms
);
70 MachineClass
*mc
= MACHINE_GET_CLASS(pcms
);
71 X86CPU
*cpu
= X86_CPU(ms
->possible_cpus
->cpus
[0].cpu
);
74 if (pcmc
->smbios_defaults
) {
75 /* These values are guest ABI, do not change */
76 smbios_set_defaults("QEMU", mc
->desc
, mc
->name
);
79 /* tell smbios about cpuid version and features */
80 smbios_set_cpuid(cpu
->env
.cpuid_version
, cpu
->env
.features
[FEAT_1_EDX
]);
82 if (pcmc
->smbios_legacy_mode
) {
83 smbios_tables
= smbios_get_table_legacy(&smbios_tables_len
,
85 fw_cfg_add_bytes(fw_cfg
, FW_CFG_SMBIOS_ENTRIES
,
86 smbios_tables
, smbios_tables_len
);
90 /* build the array of physical mem area from e820 table */
91 nr_e820
= e820_get_table(NULL
);
92 mem_array
= g_malloc0(sizeof(*mem_array
) * nr_e820
);
93 for (i
= 0, array_count
= 0; i
< nr_e820
; i
++) {
96 if (e820_get_entry(i
, E820_RAM
, &addr
, &len
)) {
97 mem_array
[array_count
].address
= addr
;
98 mem_array
[array_count
].length
= len
;
102 smbios_get_tables(ms
, ep_type
, mem_array
, array_count
,
103 &smbios_tables
, &smbios_tables_len
,
104 &smbios_anchor
, &smbios_anchor_len
,
109 fw_cfg_add_file(fw_cfg
, "etc/smbios/smbios-tables",
110 smbios_tables
, smbios_tables_len
);
111 fw_cfg_add_file(fw_cfg
, "etc/smbios/smbios-anchor",
112 smbios_anchor
, smbios_anchor_len
);
117 FWCfgState
*fw_cfg_arch_create(MachineState
*ms
,
119 uint16_t apic_id_limit
)
122 uint64_t *numa_fw_cfg
;
124 MachineClass
*mc
= MACHINE_GET_CLASS(ms
);
125 const CPUArchIdList
*cpus
= mc
->possible_cpu_arch_ids(ms
);
126 int nb_numa_nodes
= ms
->numa_state
->num_nodes
;
128 fw_cfg
= fw_cfg_init_io_dma(FW_CFG_IO_BASE
, FW_CFG_IO_BASE
+ 4,
129 &address_space_memory
);
130 fw_cfg_add_i16(fw_cfg
, FW_CFG_NB_CPUS
, boot_cpus
);
132 /* FW_CFG_MAX_CPUS is a bit confusing/problematic on x86:
134 * For machine types prior to 1.8, SeaBIOS needs FW_CFG_MAX_CPUS for
135 * building MPTable, ACPI MADT, ACPI CPU hotplug and ACPI SRAT table,
136 * that tables are based on xAPIC ID and QEMU<->SeaBIOS interface
137 * for CPU hotplug also uses APIC ID and not "CPU index".
138 * This means that FW_CFG_MAX_CPUS is not the "maximum number of CPUs",
139 * but the "limit to the APIC ID values SeaBIOS may see".
141 * So for compatibility reasons with old BIOSes we are stuck with
142 * "etc/max-cpus" actually being apic_id_limit
144 fw_cfg_add_i16(fw_cfg
, FW_CFG_MAX_CPUS
, apic_id_limit
);
145 fw_cfg_add_i64(fw_cfg
, FW_CFG_RAM_SIZE
, ms
->ram_size
);
147 fw_cfg_add_bytes(fw_cfg
, FW_CFG_ACPI_TABLES
,
148 acpi_tables
, acpi_tables_len
);
150 fw_cfg_add_i32(fw_cfg
, FW_CFG_IRQ0_OVERRIDE
, 1);
152 fw_cfg_add_bytes(fw_cfg
, FW_CFG_HPET
, &hpet_cfg
, sizeof(hpet_cfg
));
153 /* allocate memory for the NUMA channel: one (64bit) word for the number
154 * of nodes, one word for each VCPU->node and one word for each node to
155 * hold the amount of memory.
157 numa_fw_cfg
= g_new0(uint64_t, 1 + apic_id_limit
+ nb_numa_nodes
);
158 numa_fw_cfg
[0] = cpu_to_le64(nb_numa_nodes
);
159 for (i
= 0; i
< cpus
->len
; i
++) {
160 unsigned int apic_id
= cpus
->cpus
[i
].arch_id
;
161 assert(apic_id
< apic_id_limit
);
162 numa_fw_cfg
[apic_id
+ 1] = cpu_to_le64(cpus
->cpus
[i
].props
.node_id
);
164 for (i
= 0; i
< nb_numa_nodes
; i
++) {
165 numa_fw_cfg
[apic_id_limit
+ 1 + i
] =
166 cpu_to_le64(ms
->numa_state
->nodes
[i
].node_mem
);
168 fw_cfg_add_bytes(fw_cfg
, FW_CFG_NUMA
, numa_fw_cfg
,
169 (1 + apic_id_limit
+ nb_numa_nodes
) *
170 sizeof(*numa_fw_cfg
));
175 void fw_cfg_build_feature_control(MachineState
*ms
, FWCfgState
*fw_cfg
)
177 X86CPU
*cpu
= X86_CPU(ms
->possible_cpus
->cpus
[0].cpu
);
178 CPUX86State
*env
= &cpu
->env
;
179 uint32_t unused
, ebx
, ecx
, edx
;
180 uint64_t feature_control_bits
= 0;
183 cpu_x86_cpuid(env
, 1, 0, &unused
, &unused
, &ecx
, &edx
);
184 if (ecx
& CPUID_EXT_VMX
) {
185 feature_control_bits
|= FEATURE_CONTROL_VMXON_ENABLED_OUTSIDE_SMX
;
188 if ((edx
& (CPUID_EXT2_MCE
| CPUID_EXT2_MCA
)) ==
189 (CPUID_EXT2_MCE
| CPUID_EXT2_MCA
) &&
190 (env
->mcg_cap
& MCG_LMCE_P
)) {
191 feature_control_bits
|= FEATURE_CONTROL_LMCE
;
194 if (env
->cpuid_level
>= 7) {
195 cpu_x86_cpuid(env
, 0x7, 0, &unused
, &ebx
, &ecx
, &unused
);
196 if (ebx
& CPUID_7_0_EBX_SGX
) {
197 feature_control_bits
|= FEATURE_CONTROL_SGX
;
199 if (ecx
& CPUID_7_0_ECX_SGX_LC
) {
200 feature_control_bits
|= FEATURE_CONTROL_SGX_LC
;
204 if (!feature_control_bits
) {
208 val
= g_malloc(sizeof(*val
));
209 *val
= cpu_to_le64(feature_control_bits
| FEATURE_CONTROL_LOCKED
);
210 fw_cfg_add_file(fw_cfg
, "etc/msr_feature_control", val
, sizeof(*val
));
214 void fw_cfg_add_acpi_dsdt(Aml
*scope
, FWCfgState
*fw_cfg
)
217 * when using port i/o, the 8-bit data register *always* overlaps
218 * with half of the 16-bit control register. Hence, the total size
219 * of the i/o region used is FW_CFG_CTL_SIZE; when using DMA, the
220 * DMA control register is located at FW_CFG_DMA_IO_BASE + 4
222 Object
*obj
= OBJECT(fw_cfg
);
223 uint8_t io_size
= object_property_get_bool(obj
, "dma_enabled", NULL
) ?
224 ROUND_UP(FW_CFG_CTL_SIZE
, 4) + sizeof(dma_addr_t
) :
226 Aml
*dev
= aml_device("FWCF");
227 Aml
*crs
= aml_resource_template();
229 aml_append(dev
, aml_name_decl("_HID", aml_string("QEMU0002")));
231 /* device present, functioning, decoding, not shown in UI */
232 aml_append(dev
, aml_name_decl("_STA", aml_int(0xB)));
235 aml_io(AML_DECODE16
, FW_CFG_IO_BASE
, FW_CFG_IO_BASE
, 0x01, io_size
));
237 aml_append(dev
, aml_name_decl("_CRS", crs
));
238 aml_append(scope
, dev
);