xref: /qemu/hw/arm/virt.c (revision 69c22481)
1 /*
2  * ARM mach-virt emulation
3  *
4  * Copyright (c) 2013 Linaro Limited
5  *
6  * This program is free software; you can redistribute it and/or modify it
7  * under the terms and conditions of the GNU General Public License,
8  * version 2 or later, as published by the Free Software Foundation.
9  *
10  * This program is distributed in the hope it will be useful, but WITHOUT
11  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
12  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
13  * more details.
14  *
15  * You should have received a copy of the GNU General Public License along with
16  * this program.  If not, see <http://www.gnu.org/licenses/>.
17  *
18  * Emulate a virtual board which works by passing Linux all the information
19  * it needs about what devices are present via the device tree.
20  * There are some restrictions about what we can do here:
21  *  + we can only present devices whose Linux drivers will work based
22  *    purely on the device tree with no platform data at all
23  *  + we want to present a very stripped-down minimalist platform,
24  *    both because this reduces the security attack surface from the guest
25  *    and also because it reduces our exposure to being broken when
26  *    the kernel updates its device tree bindings and requires further
27  *    information in a device binding that we aren't providing.
28  * This is essentially the same approach kvmtool uses.
29  */
30 
31 #include "qemu/osdep.h"
32 #include "qemu/datadir.h"
33 #include "qemu/units.h"
34 #include "qemu/option.h"
35 #include "monitor/qdev.h"
36 #include "hw/sysbus.h"
37 #include "hw/arm/boot.h"
38 #include "hw/arm/primecell.h"
39 #include "hw/arm/virt.h"
40 #include "hw/block/flash.h"
41 #include "hw/vfio/vfio-calxeda-xgmac.h"
42 #include "hw/vfio/vfio-amd-xgbe.h"
43 #include "hw/display/ramfb.h"
44 #include "net/net.h"
45 #include "sysemu/device_tree.h"
46 #include "sysemu/numa.h"
47 #include "sysemu/runstate.h"
48 #include "sysemu/tpm.h"
49 #include "sysemu/tcg.h"
50 #include "sysemu/kvm.h"
51 #include "sysemu/hvf.h"
52 #include "sysemu/qtest.h"
53 #include "hw/loader.h"
54 #include "qapi/error.h"
55 #include "qemu/bitops.h"
56 #include "qemu/error-report.h"
57 #include "qemu/module.h"
58 #include "hw/pci-host/gpex.h"
59 #include "hw/virtio/virtio-pci.h"
60 #include "hw/core/sysbus-fdt.h"
61 #include "hw/platform-bus.h"
62 #include "hw/qdev-properties.h"
63 #include "hw/arm/fdt.h"
64 #include "hw/intc/arm_gic.h"
65 #include "hw/intc/arm_gicv3_common.h"
66 #include "hw/intc/arm_gicv3_its_common.h"
67 #include "hw/irq.h"
68 #include "kvm_arm.h"
69 #include "hw/firmware/smbios.h"
70 #include "qapi/visitor.h"
71 #include "qapi/qapi-visit-common.h"
72 #include "qapi/qmp/qlist.h"
73 #include "standard-headers/linux/input.h"
74 #include "hw/arm/smmuv3.h"
75 #include "hw/acpi/acpi.h"
76 #include "target/arm/internals.h"
77 #include "hw/mem/pc-dimm.h"
78 #include "hw/mem/nvdimm.h"
79 #include "hw/acpi/generic_event_device.h"
80 #include "hw/virtio/virtio-md-pci.h"
81 #include "hw/virtio/virtio-iommu.h"
82 #include "hw/char/pl011.h"
83 #include "qemu/guest-random.h"
84 
85 #define DEFINE_VIRT_MACHINE_LATEST(major, minor, latest) \
86     static void virt_##major##_##minor##_class_init(ObjectClass *oc, \
87                                                     void *data) \
88     { \
89         MachineClass *mc = MACHINE_CLASS(oc); \
90         virt_machine_##major##_##minor##_options(mc); \
91         mc->desc = "QEMU " # major "." # minor " ARM Virtual Machine"; \
92         if (latest) { \
93             mc->alias = "virt"; \
94         } \
95     } \
96     static const TypeInfo machvirt_##major##_##minor##_info = { \
97         .name = MACHINE_TYPE_NAME("virt-" # major "." # minor), \
98         .parent = TYPE_VIRT_MACHINE, \
99         .class_init = virt_##major##_##minor##_class_init, \
100     }; \
101     static void machvirt_machine_##major##_##minor##_init(void) \
102     { \
103         type_register_static(&machvirt_##major##_##minor##_info); \
104     } \
105     type_init(machvirt_machine_##major##_##minor##_init);
106 
107 #define DEFINE_VIRT_MACHINE_AS_LATEST(major, minor) \
108     DEFINE_VIRT_MACHINE_LATEST(major, minor, true)
109 #define DEFINE_VIRT_MACHINE(major, minor) \
110     DEFINE_VIRT_MACHINE_LATEST(major, minor, false)
111 
112 
113 /* Number of external interrupt lines to configure the GIC with */
114 #define NUM_IRQS 256
115 
116 #define PLATFORM_BUS_NUM_IRQS 64
117 
118 /* Legacy RAM limit in GB (< version 4.0) */
119 #define LEGACY_RAMLIMIT_GB 255
120 #define LEGACY_RAMLIMIT_BYTES (LEGACY_RAMLIMIT_GB * GiB)
121 
122 /* Addresses and sizes of our components.
123  * 0..128MB is space for a flash device so we can run bootrom code such as UEFI.
124  * 128MB..256MB is used for miscellaneous device I/O.
125  * 256MB..1GB is reserved for possible future PCI support (ie where the
126  * PCI memory window will go if we add a PCI host controller).
127  * 1GB and up is RAM (which may happily spill over into the
128  * high memory region beyond 4GB).
129  * This represents a compromise between how much RAM can be given to
130  * a 32 bit VM and leaving space for expansion and in particular for PCI.
131  * Note that devices should generally be placed at multiples of 0x10000,
132  * to accommodate guests using 64K pages.
133  */
134 static const MemMapEntry base_memmap[] = {
135     /* Space up to 0x8000000 is reserved for a boot ROM */
136     [VIRT_FLASH] =              {          0, 0x08000000 },
137     [VIRT_CPUPERIPHS] =         { 0x08000000, 0x00020000 },
138     /* GIC distributor and CPU interfaces sit inside the CPU peripheral space */
139     [VIRT_GIC_DIST] =           { 0x08000000, 0x00010000 },
140     [VIRT_GIC_CPU] =            { 0x08010000, 0x00010000 },
141     [VIRT_GIC_V2M] =            { 0x08020000, 0x00001000 },
142     [VIRT_GIC_HYP] =            { 0x08030000, 0x00010000 },
143     [VIRT_GIC_VCPU] =           { 0x08040000, 0x00010000 },
144     /* The space in between here is reserved for GICv3 CPU/vCPU/HYP */
145     [VIRT_GIC_ITS] =            { 0x08080000, 0x00020000 },
146     /* This redistributor space allows up to 2*64kB*123 CPUs */
147     [VIRT_GIC_REDIST] =         { 0x080A0000, 0x00F60000 },
148     [VIRT_UART] =               { 0x09000000, 0x00001000 },
149     [VIRT_RTC] =                { 0x09010000, 0x00001000 },
150     [VIRT_FW_CFG] =             { 0x09020000, 0x00000018 },
151     [VIRT_GPIO] =               { 0x09030000, 0x00001000 },
152     [VIRT_SECURE_UART] =        { 0x09040000, 0x00001000 },
153     [VIRT_SMMU] =               { 0x09050000, 0x00020000 },
154     [VIRT_PCDIMM_ACPI] =        { 0x09070000, MEMORY_HOTPLUG_IO_LEN },
155     [VIRT_ACPI_GED] =           { 0x09080000, ACPI_GED_EVT_SEL_LEN },
156     [VIRT_NVDIMM_ACPI] =        { 0x09090000, NVDIMM_ACPI_IO_LEN},
157     [VIRT_PVTIME] =             { 0x090a0000, 0x00010000 },
158     [VIRT_SECURE_GPIO] =        { 0x090b0000, 0x00001000 },
159     [VIRT_MMIO] =               { 0x0a000000, 0x00000200 },
160     /* ...repeating for a total of NUM_VIRTIO_TRANSPORTS, each of that size */
161     [VIRT_PLATFORM_BUS] =       { 0x0c000000, 0x02000000 },
162     [VIRT_SECURE_MEM] =         { 0x0e000000, 0x01000000 },
163     [VIRT_PCIE_MMIO] =          { 0x10000000, 0x2eff0000 },
164     [VIRT_PCIE_PIO] =           { 0x3eff0000, 0x00010000 },
165     [VIRT_PCIE_ECAM] =          { 0x3f000000, 0x01000000 },
166     /* Actual RAM size depends on initial RAM and device memory settings */
167     [VIRT_MEM] =                { GiB, LEGACY_RAMLIMIT_BYTES },
168 };
169 
170 /*
171  * Highmem IO Regions: This memory map is floating, located after the RAM.
172  * Each MemMapEntry base (GPA) will be dynamically computed, depending on the
173  * top of the RAM, so that its base get the same alignment as the size,
174  * ie. a 512GiB entry will be aligned on a 512GiB boundary. If there is
175  * less than 256GiB of RAM, the floating area starts at the 256GiB mark.
176  * Note the extended_memmap is sized so that it eventually also includes the
177  * base_memmap entries (VIRT_HIGH_GIC_REDIST2 index is greater than the last
178  * index of base_memmap).
179  *
180  * The memory map for these Highmem IO Regions can be in legacy or compact
181  * layout, depending on 'compact-highmem' property. With legacy layout, the
182  * PA space for one specific region is always reserved, even if the region
183  * has been disabled or doesn't fit into the PA space. However, the PA space
184  * for the region won't be reserved in these circumstances with compact layout.
185  */
186 static MemMapEntry extended_memmap[] = {
187     /* Additional 64 MB redist region (can contain up to 512 redistributors) */
188     [VIRT_HIGH_GIC_REDIST2] =   { 0x0, 64 * MiB },
189     [VIRT_HIGH_PCIE_ECAM] =     { 0x0, 256 * MiB },
190     /* Second PCIe window */
191     [VIRT_HIGH_PCIE_MMIO] =     { 0x0, 512 * GiB },
192 };
193 
194 static const int a15irqmap[] = {
195     [VIRT_UART] = 1,
196     [VIRT_RTC] = 2,
197     [VIRT_PCIE] = 3, /* ... to 6 */
198     [VIRT_GPIO] = 7,
199     [VIRT_SECURE_UART] = 8,
200     [VIRT_ACPI_GED] = 9,
201     [VIRT_MMIO] = 16, /* ...to 16 + NUM_VIRTIO_TRANSPORTS - 1 */
202     [VIRT_GIC_V2M] = 48, /* ...to 48 + NUM_GICV2M_SPIS - 1 */
203     [VIRT_SMMU] = 74,    /* ...to 74 + NUM_SMMU_IRQS - 1 */
204     [VIRT_PLATFORM_BUS] = 112, /* ...to 112 + PLATFORM_BUS_NUM_IRQS -1 */
205 };
206 
207 static const char *valid_cpus[] = {
208 #ifdef CONFIG_TCG
209     ARM_CPU_TYPE_NAME("cortex-a7"),
210     ARM_CPU_TYPE_NAME("cortex-a15"),
211     ARM_CPU_TYPE_NAME("cortex-a35"),
212     ARM_CPU_TYPE_NAME("cortex-a55"),
213     ARM_CPU_TYPE_NAME("cortex-a72"),
214     ARM_CPU_TYPE_NAME("cortex-a76"),
215     ARM_CPU_TYPE_NAME("cortex-a710"),
216     ARM_CPU_TYPE_NAME("a64fx"),
217     ARM_CPU_TYPE_NAME("neoverse-n1"),
218     ARM_CPU_TYPE_NAME("neoverse-v1"),
219     ARM_CPU_TYPE_NAME("neoverse-n2"),
220 #endif
221     ARM_CPU_TYPE_NAME("cortex-a53"),
222     ARM_CPU_TYPE_NAME("cortex-a57"),
223     ARM_CPU_TYPE_NAME("host"),
224     ARM_CPU_TYPE_NAME("max"),
225 };
226 
227 static bool cpu_type_valid(const char *cpu)
228 {
229     int i;
230 
231     for (i = 0; i < ARRAY_SIZE(valid_cpus); i++) {
232         if (strcmp(cpu, valid_cpus[i]) == 0) {
233             return true;
234         }
235     }
236     return false;
237 }
238 
239 static void create_randomness(MachineState *ms, const char *node)
240 {
241     struct {
242         uint64_t kaslr;
243         uint8_t rng[32];
244     } seed;
245 
246     if (qemu_guest_getrandom(&seed, sizeof(seed), NULL)) {
247         return;
248     }
249     qemu_fdt_setprop_u64(ms->fdt, node, "kaslr-seed", seed.kaslr);
250     qemu_fdt_setprop(ms->fdt, node, "rng-seed", seed.rng, sizeof(seed.rng));
251 }
252 
253 static void create_fdt(VirtMachineState *vms)
254 {
255     MachineState *ms = MACHINE(vms);
256     int nb_numa_nodes = ms->numa_state->num_nodes;
257     void *fdt = create_device_tree(&vms->fdt_size);
258 
259     if (!fdt) {
260         error_report("create_device_tree() failed");
261         exit(1);
262     }
263 
264     ms->fdt = fdt;
265 
266     /* Header */
267     qemu_fdt_setprop_string(fdt, "/", "compatible", "linux,dummy-virt");
268     qemu_fdt_setprop_cell(fdt, "/", "#address-cells", 0x2);
269     qemu_fdt_setprop_cell(fdt, "/", "#size-cells", 0x2);
270     qemu_fdt_setprop_string(fdt, "/", "model", "linux,dummy-virt");
271 
272     /* /chosen must exist for load_dtb to fill in necessary properties later */
273     qemu_fdt_add_subnode(fdt, "/chosen");
274     if (vms->dtb_randomness) {
275         create_randomness(ms, "/chosen");
276     }
277 
278     if (vms->secure) {
279         qemu_fdt_add_subnode(fdt, "/secure-chosen");
280         if (vms->dtb_randomness) {
281             create_randomness(ms, "/secure-chosen");
282         }
283     }
284 
285     /* Clock node, for the benefit of the UART. The kernel device tree
286      * binding documentation claims the PL011 node clock properties are
287      * optional but in practice if you omit them the kernel refuses to
288      * probe for the device.
289      */
290     vms->clock_phandle = qemu_fdt_alloc_phandle(fdt);
291     qemu_fdt_add_subnode(fdt, "/apb-pclk");
292     qemu_fdt_setprop_string(fdt, "/apb-pclk", "compatible", "fixed-clock");
293     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "#clock-cells", 0x0);
294     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "clock-frequency", 24000000);
295     qemu_fdt_setprop_string(fdt, "/apb-pclk", "clock-output-names",
296                                 "clk24mhz");
297     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "phandle", vms->clock_phandle);
298 
299     if (nb_numa_nodes > 0 && ms->numa_state->have_numa_distance) {
300         int size = nb_numa_nodes * nb_numa_nodes * 3 * sizeof(uint32_t);
301         uint32_t *matrix = g_malloc0(size);
302         int idx, i, j;
303 
304         for (i = 0; i < nb_numa_nodes; i++) {
305             for (j = 0; j < nb_numa_nodes; j++) {
306                 idx = (i * nb_numa_nodes + j) * 3;
307                 matrix[idx + 0] = cpu_to_be32(i);
308                 matrix[idx + 1] = cpu_to_be32(j);
309                 matrix[idx + 2] =
310                     cpu_to_be32(ms->numa_state->nodes[i].distance[j]);
311             }
312         }
313 
314         qemu_fdt_add_subnode(fdt, "/distance-map");
315         qemu_fdt_setprop_string(fdt, "/distance-map", "compatible",
316                                 "numa-distance-map-v1");
317         qemu_fdt_setprop(fdt, "/distance-map", "distance-matrix",
318                          matrix, size);
319         g_free(matrix);
320     }
321 }
322 
323 static void fdt_add_timer_nodes(const VirtMachineState *vms)
324 {
325     /* On real hardware these interrupts are level-triggered.
326      * On KVM they were edge-triggered before host kernel version 4.4,
327      * and level-triggered afterwards.
328      * On emulated QEMU they are level-triggered.
329      *
330      * Getting the DTB info about them wrong is awkward for some
331      * guest kernels:
332      *  pre-4.8 ignore the DT and leave the interrupt configured
333      *   with whatever the GIC reset value (or the bootloader) left it at
334      *  4.8 before rc6 honour the incorrect data by programming it back
335      *   into the GIC, causing problems
336      *  4.8rc6 and later ignore the DT and always write "level triggered"
337      *   into the GIC
338      *
339      * For backwards-compatibility, virt-2.8 and earlier will continue
340      * to say these are edge-triggered, but later machines will report
341      * the correct information.
342      */
343     ARMCPU *armcpu;
344     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
345     uint32_t irqflags = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
346     MachineState *ms = MACHINE(vms);
347 
348     if (vmc->claim_edge_triggered_timers) {
349         irqflags = GIC_FDT_IRQ_FLAGS_EDGE_LO_HI;
350     }
351 
352     if (vms->gic_version == VIRT_GIC_VERSION_2) {
353         irqflags = deposit32(irqflags, GIC_FDT_IRQ_PPI_CPU_START,
354                              GIC_FDT_IRQ_PPI_CPU_WIDTH,
355                              (1 << MACHINE(vms)->smp.cpus) - 1);
356     }
357 
358     qemu_fdt_add_subnode(ms->fdt, "/timer");
359 
360     armcpu = ARM_CPU(qemu_get_cpu(0));
361     if (arm_feature(&armcpu->env, ARM_FEATURE_V8)) {
362         const char compat[] = "arm,armv8-timer\0arm,armv7-timer";
363         qemu_fdt_setprop(ms->fdt, "/timer", "compatible",
364                          compat, sizeof(compat));
365     } else {
366         qemu_fdt_setprop_string(ms->fdt, "/timer", "compatible",
367                                 "arm,armv7-timer");
368     }
369     qemu_fdt_setprop(ms->fdt, "/timer", "always-on", NULL, 0);
370     qemu_fdt_setprop_cells(ms->fdt, "/timer", "interrupts",
371                            GIC_FDT_IRQ_TYPE_PPI,
372                            INTID_TO_PPI(ARCH_TIMER_S_EL1_IRQ), irqflags,
373                            GIC_FDT_IRQ_TYPE_PPI,
374                            INTID_TO_PPI(ARCH_TIMER_NS_EL1_IRQ), irqflags,
375                            GIC_FDT_IRQ_TYPE_PPI,
376                            INTID_TO_PPI(ARCH_TIMER_VIRT_IRQ), irqflags,
377                            GIC_FDT_IRQ_TYPE_PPI,
378                            INTID_TO_PPI(ARCH_TIMER_NS_EL2_IRQ), irqflags);
379 }
380 
381 static void fdt_add_cpu_nodes(const VirtMachineState *vms)
382 {
383     int cpu;
384     int addr_cells = 1;
385     const MachineState *ms = MACHINE(vms);
386     const VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
387     int smp_cpus = ms->smp.cpus;
388 
389     /*
390      * See Linux Documentation/devicetree/bindings/arm/cpus.yaml
391      * On ARM v8 64-bit systems value should be set to 2,
392      * that corresponds to the MPIDR_EL1 register size.
393      * If MPIDR_EL1[63:32] value is equal to 0 on all CPUs
394      * in the system, #address-cells can be set to 1, since
395      * MPIDR_EL1[63:32] bits are not used for CPUs
396      * identification.
397      *
398      * Here we actually don't know whether our system is 32- or 64-bit one.
399      * The simplest way to go is to examine affinity IDs of all our CPUs. If
400      * at least one of them has Aff3 populated, we set #address-cells to 2.
401      */
402     for (cpu = 0; cpu < smp_cpus; cpu++) {
403         ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(cpu));
404 
405         if (armcpu->mp_affinity & ARM_AFF3_MASK) {
406             addr_cells = 2;
407             break;
408         }
409     }
410 
411     qemu_fdt_add_subnode(ms->fdt, "/cpus");
412     qemu_fdt_setprop_cell(ms->fdt, "/cpus", "#address-cells", addr_cells);
413     qemu_fdt_setprop_cell(ms->fdt, "/cpus", "#size-cells", 0x0);
414 
415     for (cpu = smp_cpus - 1; cpu >= 0; cpu--) {
416         char *nodename = g_strdup_printf("/cpus/cpu@%d", cpu);
417         ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(cpu));
418         CPUState *cs = CPU(armcpu);
419 
420         qemu_fdt_add_subnode(ms->fdt, nodename);
421         qemu_fdt_setprop_string(ms->fdt, nodename, "device_type", "cpu");
422         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
423                                     armcpu->dtb_compatible);
424 
425         if (vms->psci_conduit != QEMU_PSCI_CONDUIT_DISABLED && smp_cpus > 1) {
426             qemu_fdt_setprop_string(ms->fdt, nodename,
427                                         "enable-method", "psci");
428         }
429 
430         if (addr_cells == 2) {
431             qemu_fdt_setprop_u64(ms->fdt, nodename, "reg",
432                                  armcpu->mp_affinity);
433         } else {
434             qemu_fdt_setprop_cell(ms->fdt, nodename, "reg",
435                                   armcpu->mp_affinity);
436         }
437 
438         if (ms->possible_cpus->cpus[cs->cpu_index].props.has_node_id) {
439             qemu_fdt_setprop_cell(ms->fdt, nodename, "numa-node-id",
440                 ms->possible_cpus->cpus[cs->cpu_index].props.node_id);
441         }
442 
443         if (!vmc->no_cpu_topology) {
444             qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle",
445                                   qemu_fdt_alloc_phandle(ms->fdt));
446         }
447 
448         g_free(nodename);
449     }
450 
451     if (!vmc->no_cpu_topology) {
452         /*
453          * Add vCPU topology description through fdt node cpu-map.
454          *
455          * See Linux Documentation/devicetree/bindings/cpu/cpu-topology.txt
456          * In a SMP system, the hierarchy of CPUs can be defined through
457          * four entities that are used to describe the layout of CPUs in
458          * the system: socket/cluster/core/thread.
459          *
460          * A socket node represents the boundary of system physical package
461          * and its child nodes must be one or more cluster nodes. A system
462          * can contain several layers of clustering within a single physical
463          * package and cluster nodes can be contained in parent cluster nodes.
464          *
465          * Note: currently we only support one layer of clustering within
466          * each physical package.
467          */
468         qemu_fdt_add_subnode(ms->fdt, "/cpus/cpu-map");
469 
470         for (cpu = smp_cpus - 1; cpu >= 0; cpu--) {
471             char *cpu_path = g_strdup_printf("/cpus/cpu@%d", cpu);
472             char *map_path;
473 
474             if (ms->smp.threads > 1) {
475                 map_path = g_strdup_printf(
476                     "/cpus/cpu-map/socket%d/cluster%d/core%d/thread%d",
477                     cpu / (ms->smp.clusters * ms->smp.cores * ms->smp.threads),
478                     (cpu / (ms->smp.cores * ms->smp.threads)) % ms->smp.clusters,
479                     (cpu / ms->smp.threads) % ms->smp.cores,
480                     cpu % ms->smp.threads);
481             } else {
482                 map_path = g_strdup_printf(
483                     "/cpus/cpu-map/socket%d/cluster%d/core%d",
484                     cpu / (ms->smp.clusters * ms->smp.cores),
485                     (cpu / ms->smp.cores) % ms->smp.clusters,
486                     cpu % ms->smp.cores);
487             }
488             qemu_fdt_add_path(ms->fdt, map_path);
489             qemu_fdt_setprop_phandle(ms->fdt, map_path, "cpu", cpu_path);
490 
491             g_free(map_path);
492             g_free(cpu_path);
493         }
494     }
495 }
496 
497 static void fdt_add_its_gic_node(VirtMachineState *vms)
498 {
499     char *nodename;
500     MachineState *ms = MACHINE(vms);
501 
502     vms->msi_phandle = qemu_fdt_alloc_phandle(ms->fdt);
503     nodename = g_strdup_printf("/intc/its@%" PRIx64,
504                                vms->memmap[VIRT_GIC_ITS].base);
505     qemu_fdt_add_subnode(ms->fdt, nodename);
506     qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
507                             "arm,gic-v3-its");
508     qemu_fdt_setprop(ms->fdt, nodename, "msi-controller", NULL, 0);
509     qemu_fdt_setprop_cell(ms->fdt, nodename, "#msi-cells", 1);
510     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
511                                  2, vms->memmap[VIRT_GIC_ITS].base,
512                                  2, vms->memmap[VIRT_GIC_ITS].size);
513     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", vms->msi_phandle);
514     g_free(nodename);
515 }
516 
517 static void fdt_add_v2m_gic_node(VirtMachineState *vms)
518 {
519     MachineState *ms = MACHINE(vms);
520     char *nodename;
521 
522     nodename = g_strdup_printf("/intc/v2m@%" PRIx64,
523                                vms->memmap[VIRT_GIC_V2M].base);
524     vms->msi_phandle = qemu_fdt_alloc_phandle(ms->fdt);
525     qemu_fdt_add_subnode(ms->fdt, nodename);
526     qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
527                             "arm,gic-v2m-frame");
528     qemu_fdt_setprop(ms->fdt, nodename, "msi-controller", NULL, 0);
529     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
530                                  2, vms->memmap[VIRT_GIC_V2M].base,
531                                  2, vms->memmap[VIRT_GIC_V2M].size);
532     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", vms->msi_phandle);
533     g_free(nodename);
534 }
535 
536 static void fdt_add_gic_node(VirtMachineState *vms)
537 {
538     MachineState *ms = MACHINE(vms);
539     char *nodename;
540 
541     vms->gic_phandle = qemu_fdt_alloc_phandle(ms->fdt);
542     qemu_fdt_setprop_cell(ms->fdt, "/", "interrupt-parent", vms->gic_phandle);
543 
544     nodename = g_strdup_printf("/intc@%" PRIx64,
545                                vms->memmap[VIRT_GIC_DIST].base);
546     qemu_fdt_add_subnode(ms->fdt, nodename);
547     qemu_fdt_setprop_cell(ms->fdt, nodename, "#interrupt-cells", 3);
548     qemu_fdt_setprop(ms->fdt, nodename, "interrupt-controller", NULL, 0);
549     qemu_fdt_setprop_cell(ms->fdt, nodename, "#address-cells", 0x2);
550     qemu_fdt_setprop_cell(ms->fdt, nodename, "#size-cells", 0x2);
551     qemu_fdt_setprop(ms->fdt, nodename, "ranges", NULL, 0);
552     if (vms->gic_version != VIRT_GIC_VERSION_2) {
553         int nb_redist_regions = virt_gicv3_redist_region_count(vms);
554 
555         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
556                                 "arm,gic-v3");
557 
558         qemu_fdt_setprop_cell(ms->fdt, nodename,
559                               "#redistributor-regions", nb_redist_regions);
560 
561         if (nb_redist_regions == 1) {
562             qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
563                                          2, vms->memmap[VIRT_GIC_DIST].base,
564                                          2, vms->memmap[VIRT_GIC_DIST].size,
565                                          2, vms->memmap[VIRT_GIC_REDIST].base,
566                                          2, vms->memmap[VIRT_GIC_REDIST].size);
567         } else {
568             qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
569                                  2, vms->memmap[VIRT_GIC_DIST].base,
570                                  2, vms->memmap[VIRT_GIC_DIST].size,
571                                  2, vms->memmap[VIRT_GIC_REDIST].base,
572                                  2, vms->memmap[VIRT_GIC_REDIST].size,
573                                  2, vms->memmap[VIRT_HIGH_GIC_REDIST2].base,
574                                  2, vms->memmap[VIRT_HIGH_GIC_REDIST2].size);
575         }
576 
577         if (vms->virt) {
578             qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
579                                    GIC_FDT_IRQ_TYPE_PPI,
580                                    INTID_TO_PPI(ARCH_GIC_MAINT_IRQ),
581                                    GIC_FDT_IRQ_FLAGS_LEVEL_HI);
582         }
583     } else {
584         /* 'cortex-a15-gic' means 'GIC v2' */
585         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
586                                 "arm,cortex-a15-gic");
587         if (!vms->virt) {
588             qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
589                                          2, vms->memmap[VIRT_GIC_DIST].base,
590                                          2, vms->memmap[VIRT_GIC_DIST].size,
591                                          2, vms->memmap[VIRT_GIC_CPU].base,
592                                          2, vms->memmap[VIRT_GIC_CPU].size);
593         } else {
594             qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
595                                          2, vms->memmap[VIRT_GIC_DIST].base,
596                                          2, vms->memmap[VIRT_GIC_DIST].size,
597                                          2, vms->memmap[VIRT_GIC_CPU].base,
598                                          2, vms->memmap[VIRT_GIC_CPU].size,
599                                          2, vms->memmap[VIRT_GIC_HYP].base,
600                                          2, vms->memmap[VIRT_GIC_HYP].size,
601                                          2, vms->memmap[VIRT_GIC_VCPU].base,
602                                          2, vms->memmap[VIRT_GIC_VCPU].size);
603             qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
604                                    GIC_FDT_IRQ_TYPE_PPI,
605                                    INTID_TO_PPI(ARCH_GIC_MAINT_IRQ),
606                                    GIC_FDT_IRQ_FLAGS_LEVEL_HI);
607         }
608     }
609 
610     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", vms->gic_phandle);
611     g_free(nodename);
612 }
613 
614 static void fdt_add_pmu_nodes(const VirtMachineState *vms)
615 {
616     ARMCPU *armcpu = ARM_CPU(first_cpu);
617     uint32_t irqflags = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
618     MachineState *ms = MACHINE(vms);
619 
620     if (!arm_feature(&armcpu->env, ARM_FEATURE_PMU)) {
621         assert(!object_property_get_bool(OBJECT(armcpu), "pmu", NULL));
622         return;
623     }
624 
625     if (vms->gic_version == VIRT_GIC_VERSION_2) {
626         irqflags = deposit32(irqflags, GIC_FDT_IRQ_PPI_CPU_START,
627                              GIC_FDT_IRQ_PPI_CPU_WIDTH,
628                              (1 << MACHINE(vms)->smp.cpus) - 1);
629     }
630 
631     qemu_fdt_add_subnode(ms->fdt, "/pmu");
632     if (arm_feature(&armcpu->env, ARM_FEATURE_V8)) {
633         const char compat[] = "arm,armv8-pmuv3";
634         qemu_fdt_setprop(ms->fdt, "/pmu", "compatible",
635                          compat, sizeof(compat));
636         qemu_fdt_setprop_cells(ms->fdt, "/pmu", "interrupts",
637                                GIC_FDT_IRQ_TYPE_PPI,
638                                INTID_TO_PPI(VIRTUAL_PMU_IRQ), irqflags);
639     }
640 }
641 
642 static inline DeviceState *create_acpi_ged(VirtMachineState *vms)
643 {
644     DeviceState *dev;
645     MachineState *ms = MACHINE(vms);
646     int irq = vms->irqmap[VIRT_ACPI_GED];
647     uint32_t event = ACPI_GED_PWR_DOWN_EVT;
648 
649     if (ms->ram_slots) {
650         event |= ACPI_GED_MEM_HOTPLUG_EVT;
651     }
652 
653     if (ms->nvdimms_state->is_enabled) {
654         event |= ACPI_GED_NVDIMM_HOTPLUG_EVT;
655     }
656 
657     dev = qdev_new(TYPE_ACPI_GED);
658     qdev_prop_set_uint32(dev, "ged-event", event);
659     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
660 
661     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_ACPI_GED].base);
662     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 1, vms->memmap[VIRT_PCDIMM_ACPI].base);
663     sysbus_connect_irq(SYS_BUS_DEVICE(dev), 0, qdev_get_gpio_in(vms->gic, irq));
664 
665     return dev;
666 }
667 
668 static void create_its(VirtMachineState *vms)
669 {
670     const char *itsclass = its_class_name();
671     DeviceState *dev;
672 
673     if (!strcmp(itsclass, "arm-gicv3-its")) {
674         if (!vms->tcg_its) {
675             itsclass = NULL;
676         }
677     }
678 
679     if (!itsclass) {
680         /* Do nothing if not supported */
681         return;
682     }
683 
684     dev = qdev_new(itsclass);
685 
686     object_property_set_link(OBJECT(dev), "parent-gicv3", OBJECT(vms->gic),
687                              &error_abort);
688     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
689     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_GIC_ITS].base);
690 
691     fdt_add_its_gic_node(vms);
692     vms->msi_controller = VIRT_MSI_CTRL_ITS;
693 }
694 
695 static void create_v2m(VirtMachineState *vms)
696 {
697     int i;
698     int irq = vms->irqmap[VIRT_GIC_V2M];
699     DeviceState *dev;
700 
701     dev = qdev_new("arm-gicv2m");
702     qdev_prop_set_uint32(dev, "base-spi", irq);
703     qdev_prop_set_uint32(dev, "num-spi", NUM_GICV2M_SPIS);
704     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
705     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_GIC_V2M].base);
706 
707     for (i = 0; i < NUM_GICV2M_SPIS; i++) {
708         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
709                            qdev_get_gpio_in(vms->gic, irq + i));
710     }
711 
712     fdt_add_v2m_gic_node(vms);
713     vms->msi_controller = VIRT_MSI_CTRL_GICV2M;
714 }
715 
716 static void create_gic(VirtMachineState *vms, MemoryRegion *mem)
717 {
718     MachineState *ms = MACHINE(vms);
719     /* We create a standalone GIC */
720     SysBusDevice *gicbusdev;
721     const char *gictype;
722     int i;
723     unsigned int smp_cpus = ms->smp.cpus;
724     uint32_t nb_redist_regions = 0;
725     int revision;
726 
727     if (vms->gic_version == VIRT_GIC_VERSION_2) {
728         gictype = gic_class_name();
729     } else {
730         gictype = gicv3_class_name();
731     }
732 
733     switch (vms->gic_version) {
734     case VIRT_GIC_VERSION_2:
735         revision = 2;
736         break;
737     case VIRT_GIC_VERSION_3:
738         revision = 3;
739         break;
740     case VIRT_GIC_VERSION_4:
741         revision = 4;
742         break;
743     default:
744         g_assert_not_reached();
745     }
746     vms->gic = qdev_new(gictype);
747     qdev_prop_set_uint32(vms->gic, "revision", revision);
748     qdev_prop_set_uint32(vms->gic, "num-cpu", smp_cpus);
749     /* Note that the num-irq property counts both internal and external
750      * interrupts; there are always 32 of the former (mandated by GIC spec).
751      */
752     qdev_prop_set_uint32(vms->gic, "num-irq", NUM_IRQS + 32);
753     if (!kvm_irqchip_in_kernel()) {
754         qdev_prop_set_bit(vms->gic, "has-security-extensions", vms->secure);
755     }
756 
757     if (vms->gic_version != VIRT_GIC_VERSION_2) {
758         QList *redist_region_count;
759         uint32_t redist0_capacity = virt_redist_capacity(vms, VIRT_GIC_REDIST);
760         uint32_t redist0_count = MIN(smp_cpus, redist0_capacity);
761 
762         nb_redist_regions = virt_gicv3_redist_region_count(vms);
763 
764         redist_region_count = qlist_new();
765         qlist_append_int(redist_region_count, redist0_count);
766         if (nb_redist_regions == 2) {
767             uint32_t redist1_capacity =
768                 virt_redist_capacity(vms, VIRT_HIGH_GIC_REDIST2);
769 
770             qlist_append_int(redist_region_count,
771                 MIN(smp_cpus - redist0_count, redist1_capacity));
772         }
773         qdev_prop_set_array(vms->gic, "redist-region-count",
774                             redist_region_count);
775 
776         if (!kvm_irqchip_in_kernel()) {
777             if (vms->tcg_its) {
778                 object_property_set_link(OBJECT(vms->gic), "sysmem",
779                                          OBJECT(mem), &error_fatal);
780                 qdev_prop_set_bit(vms->gic, "has-lpi", true);
781             }
782         }
783     } else {
784         if (!kvm_irqchip_in_kernel()) {
785             qdev_prop_set_bit(vms->gic, "has-virtualization-extensions",
786                               vms->virt);
787         }
788     }
789     gicbusdev = SYS_BUS_DEVICE(vms->gic);
790     sysbus_realize_and_unref(gicbusdev, &error_fatal);
791     sysbus_mmio_map(gicbusdev, 0, vms->memmap[VIRT_GIC_DIST].base);
792     if (vms->gic_version != VIRT_GIC_VERSION_2) {
793         sysbus_mmio_map(gicbusdev, 1, vms->memmap[VIRT_GIC_REDIST].base);
794         if (nb_redist_regions == 2) {
795             sysbus_mmio_map(gicbusdev, 2,
796                             vms->memmap[VIRT_HIGH_GIC_REDIST2].base);
797         }
798     } else {
799         sysbus_mmio_map(gicbusdev, 1, vms->memmap[VIRT_GIC_CPU].base);
800         if (vms->virt) {
801             sysbus_mmio_map(gicbusdev, 2, vms->memmap[VIRT_GIC_HYP].base);
802             sysbus_mmio_map(gicbusdev, 3, vms->memmap[VIRT_GIC_VCPU].base);
803         }
804     }
805 
806     /* Wire the outputs from each CPU's generic timer and the GICv3
807      * maintenance interrupt signal to the appropriate GIC PPI inputs,
808      * and the GIC's IRQ/FIQ/VIRQ/VFIQ interrupt outputs to the CPU's inputs.
809      */
810     for (i = 0; i < smp_cpus; i++) {
811         DeviceState *cpudev = DEVICE(qemu_get_cpu(i));
812         int intidbase = NUM_IRQS + i * GIC_INTERNAL;
813         /* Mapping from the output timer irq lines from the CPU to the
814          * GIC PPI inputs we use for the virt board.
815          */
816         const int timer_irq[] = {
817             [GTIMER_PHYS] = ARCH_TIMER_NS_EL1_IRQ,
818             [GTIMER_VIRT] = ARCH_TIMER_VIRT_IRQ,
819             [GTIMER_HYP]  = ARCH_TIMER_NS_EL2_IRQ,
820             [GTIMER_SEC]  = ARCH_TIMER_S_EL1_IRQ,
821         };
822 
823         for (unsigned irq = 0; irq < ARRAY_SIZE(timer_irq); irq++) {
824             qdev_connect_gpio_out(cpudev, irq,
825                                   qdev_get_gpio_in(vms->gic,
826                                                    intidbase + timer_irq[irq]));
827         }
828 
829         if (vms->gic_version != VIRT_GIC_VERSION_2) {
830             qemu_irq irq = qdev_get_gpio_in(vms->gic,
831                                             intidbase + ARCH_GIC_MAINT_IRQ);
832             qdev_connect_gpio_out_named(cpudev, "gicv3-maintenance-interrupt",
833                                         0, irq);
834         } else if (vms->virt) {
835             qemu_irq irq = qdev_get_gpio_in(vms->gic,
836                                             intidbase + ARCH_GIC_MAINT_IRQ);
837             sysbus_connect_irq(gicbusdev, i + 4 * smp_cpus, irq);
838         }
839 
840         qdev_connect_gpio_out_named(cpudev, "pmu-interrupt", 0,
841                                     qdev_get_gpio_in(vms->gic, intidbase
842                                                      + VIRTUAL_PMU_IRQ));
843 
844         sysbus_connect_irq(gicbusdev, i, qdev_get_gpio_in(cpudev, ARM_CPU_IRQ));
845         sysbus_connect_irq(gicbusdev, i + smp_cpus,
846                            qdev_get_gpio_in(cpudev, ARM_CPU_FIQ));
847         sysbus_connect_irq(gicbusdev, i + 2 * smp_cpus,
848                            qdev_get_gpio_in(cpudev, ARM_CPU_VIRQ));
849         sysbus_connect_irq(gicbusdev, i + 3 * smp_cpus,
850                            qdev_get_gpio_in(cpudev, ARM_CPU_VFIQ));
851     }
852 
853     fdt_add_gic_node(vms);
854 
855     if (vms->gic_version != VIRT_GIC_VERSION_2 && vms->its) {
856         create_its(vms);
857     } else if (vms->gic_version == VIRT_GIC_VERSION_2) {
858         create_v2m(vms);
859     }
860 }
861 
862 static void create_uart(const VirtMachineState *vms, int uart,
863                         MemoryRegion *mem, Chardev *chr)
864 {
865     char *nodename;
866     hwaddr base = vms->memmap[uart].base;
867     hwaddr size = vms->memmap[uart].size;
868     int irq = vms->irqmap[uart];
869     const char compat[] = "arm,pl011\0arm,primecell";
870     const char clocknames[] = "uartclk\0apb_pclk";
871     DeviceState *dev = qdev_new(TYPE_PL011);
872     SysBusDevice *s = SYS_BUS_DEVICE(dev);
873     MachineState *ms = MACHINE(vms);
874 
875     qdev_prop_set_chr(dev, "chardev", chr);
876     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
877     memory_region_add_subregion(mem, base,
878                                 sysbus_mmio_get_region(s, 0));
879     sysbus_connect_irq(s, 0, qdev_get_gpio_in(vms->gic, irq));
880 
881     nodename = g_strdup_printf("/pl011@%" PRIx64, base);
882     qemu_fdt_add_subnode(ms->fdt, nodename);
883     /* Note that we can't use setprop_string because of the embedded NUL */
884     qemu_fdt_setprop(ms->fdt, nodename, "compatible",
885                          compat, sizeof(compat));
886     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
887                                      2, base, 2, size);
888     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
889                                GIC_FDT_IRQ_TYPE_SPI, irq,
890                                GIC_FDT_IRQ_FLAGS_LEVEL_HI);
891     qemu_fdt_setprop_cells(ms->fdt, nodename, "clocks",
892                                vms->clock_phandle, vms->clock_phandle);
893     qemu_fdt_setprop(ms->fdt, nodename, "clock-names",
894                          clocknames, sizeof(clocknames));
895 
896     if (uart == VIRT_UART) {
897         qemu_fdt_setprop_string(ms->fdt, "/chosen", "stdout-path", nodename);
898     } else {
899         /* Mark as not usable by the normal world */
900         qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
901         qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
902 
903         qemu_fdt_setprop_string(ms->fdt, "/secure-chosen", "stdout-path",
904                                 nodename);
905     }
906 
907     g_free(nodename);
908 }
909 
910 static void create_rtc(const VirtMachineState *vms)
911 {
912     char *nodename;
913     hwaddr base = vms->memmap[VIRT_RTC].base;
914     hwaddr size = vms->memmap[VIRT_RTC].size;
915     int irq = vms->irqmap[VIRT_RTC];
916     const char compat[] = "arm,pl031\0arm,primecell";
917     MachineState *ms = MACHINE(vms);
918 
919     sysbus_create_simple("pl031", base, qdev_get_gpio_in(vms->gic, irq));
920 
921     nodename = g_strdup_printf("/pl031@%" PRIx64, base);
922     qemu_fdt_add_subnode(ms->fdt, nodename);
923     qemu_fdt_setprop(ms->fdt, nodename, "compatible", compat, sizeof(compat));
924     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
925                                  2, base, 2, size);
926     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
927                            GIC_FDT_IRQ_TYPE_SPI, irq,
928                            GIC_FDT_IRQ_FLAGS_LEVEL_HI);
929     qemu_fdt_setprop_cell(ms->fdt, nodename, "clocks", vms->clock_phandle);
930     qemu_fdt_setprop_string(ms->fdt, nodename, "clock-names", "apb_pclk");
931     g_free(nodename);
932 }
933 
934 static DeviceState *gpio_key_dev;
935 static void virt_powerdown_req(Notifier *n, void *opaque)
936 {
937     VirtMachineState *s = container_of(n, VirtMachineState, powerdown_notifier);
938 
939     if (s->acpi_dev) {
940         acpi_send_event(s->acpi_dev, ACPI_POWER_DOWN_STATUS);
941     } else {
942         /* use gpio Pin 3 for power button event */
943         qemu_set_irq(qdev_get_gpio_in(gpio_key_dev, 0), 1);
944     }
945 }
946 
947 static void create_gpio_keys(char *fdt, DeviceState *pl061_dev,
948                              uint32_t phandle)
949 {
950     gpio_key_dev = sysbus_create_simple("gpio-key", -1,
951                                         qdev_get_gpio_in(pl061_dev, 3));
952 
953     qemu_fdt_add_subnode(fdt, "/gpio-keys");
954     qemu_fdt_setprop_string(fdt, "/gpio-keys", "compatible", "gpio-keys");
955 
956     qemu_fdt_add_subnode(fdt, "/gpio-keys/poweroff");
957     qemu_fdt_setprop_string(fdt, "/gpio-keys/poweroff",
958                             "label", "GPIO Key Poweroff");
959     qemu_fdt_setprop_cell(fdt, "/gpio-keys/poweroff", "linux,code",
960                           KEY_POWER);
961     qemu_fdt_setprop_cells(fdt, "/gpio-keys/poweroff",
962                            "gpios", phandle, 3, 0);
963 }
964 
965 #define SECURE_GPIO_POWEROFF 0
966 #define SECURE_GPIO_RESET    1
967 
968 static void create_secure_gpio_pwr(char *fdt, DeviceState *pl061_dev,
969                                    uint32_t phandle)
970 {
971     DeviceState *gpio_pwr_dev;
972 
973     /* gpio-pwr */
974     gpio_pwr_dev = sysbus_create_simple("gpio-pwr", -1, NULL);
975 
976     /* connect secure pl061 to gpio-pwr */
977     qdev_connect_gpio_out(pl061_dev, SECURE_GPIO_RESET,
978                           qdev_get_gpio_in_named(gpio_pwr_dev, "reset", 0));
979     qdev_connect_gpio_out(pl061_dev, SECURE_GPIO_POWEROFF,
980                           qdev_get_gpio_in_named(gpio_pwr_dev, "shutdown", 0));
981 
982     qemu_fdt_add_subnode(fdt, "/gpio-poweroff");
983     qemu_fdt_setprop_string(fdt, "/gpio-poweroff", "compatible",
984                             "gpio-poweroff");
985     qemu_fdt_setprop_cells(fdt, "/gpio-poweroff",
986                            "gpios", phandle, SECURE_GPIO_POWEROFF, 0);
987     qemu_fdt_setprop_string(fdt, "/gpio-poweroff", "status", "disabled");
988     qemu_fdt_setprop_string(fdt, "/gpio-poweroff", "secure-status",
989                             "okay");
990 
991     qemu_fdt_add_subnode(fdt, "/gpio-restart");
992     qemu_fdt_setprop_string(fdt, "/gpio-restart", "compatible",
993                             "gpio-restart");
994     qemu_fdt_setprop_cells(fdt, "/gpio-restart",
995                            "gpios", phandle, SECURE_GPIO_RESET, 0);
996     qemu_fdt_setprop_string(fdt, "/gpio-restart", "status", "disabled");
997     qemu_fdt_setprop_string(fdt, "/gpio-restart", "secure-status",
998                             "okay");
999 }
1000 
1001 static void create_gpio_devices(const VirtMachineState *vms, int gpio,
1002                                 MemoryRegion *mem)
1003 {
1004     char *nodename;
1005     DeviceState *pl061_dev;
1006     hwaddr base = vms->memmap[gpio].base;
1007     hwaddr size = vms->memmap[gpio].size;
1008     int irq = vms->irqmap[gpio];
1009     const char compat[] = "arm,pl061\0arm,primecell";
1010     SysBusDevice *s;
1011     MachineState *ms = MACHINE(vms);
1012 
1013     pl061_dev = qdev_new("pl061");
1014     /* Pull lines down to 0 if not driven by the PL061 */
1015     qdev_prop_set_uint32(pl061_dev, "pullups", 0);
1016     qdev_prop_set_uint32(pl061_dev, "pulldowns", 0xff);
1017     s = SYS_BUS_DEVICE(pl061_dev);
1018     sysbus_realize_and_unref(s, &error_fatal);
1019     memory_region_add_subregion(mem, base, sysbus_mmio_get_region(s, 0));
1020     sysbus_connect_irq(s, 0, qdev_get_gpio_in(vms->gic, irq));
1021 
1022     uint32_t phandle = qemu_fdt_alloc_phandle(ms->fdt);
1023     nodename = g_strdup_printf("/pl061@%" PRIx64, base);
1024     qemu_fdt_add_subnode(ms->fdt, nodename);
1025     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1026                                  2, base, 2, size);
1027     qemu_fdt_setprop(ms->fdt, nodename, "compatible", compat, sizeof(compat));
1028     qemu_fdt_setprop_cell(ms->fdt, nodename, "#gpio-cells", 2);
1029     qemu_fdt_setprop(ms->fdt, nodename, "gpio-controller", NULL, 0);
1030     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
1031                            GIC_FDT_IRQ_TYPE_SPI, irq,
1032                            GIC_FDT_IRQ_FLAGS_LEVEL_HI);
1033     qemu_fdt_setprop_cell(ms->fdt, nodename, "clocks", vms->clock_phandle);
1034     qemu_fdt_setprop_string(ms->fdt, nodename, "clock-names", "apb_pclk");
1035     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", phandle);
1036 
1037     if (gpio != VIRT_GPIO) {
1038         /* Mark as not usable by the normal world */
1039         qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
1040         qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
1041     }
1042     g_free(nodename);
1043 
1044     /* Child gpio devices */
1045     if (gpio == VIRT_GPIO) {
1046         create_gpio_keys(ms->fdt, pl061_dev, phandle);
1047     } else {
1048         create_secure_gpio_pwr(ms->fdt, pl061_dev, phandle);
1049     }
1050 }
1051 
1052 static void create_virtio_devices(const VirtMachineState *vms)
1053 {
1054     int i;
1055     hwaddr size = vms->memmap[VIRT_MMIO].size;
1056     MachineState *ms = MACHINE(vms);
1057 
1058     /* We create the transports in forwards order. Since qbus_realize()
1059      * prepends (not appends) new child buses, the incrementing loop below will
1060      * create a list of virtio-mmio buses with decreasing base addresses.
1061      *
1062      * When a -device option is processed from the command line,
1063      * qbus_find_recursive() picks the next free virtio-mmio bus in forwards
1064      * order. The upshot is that -device options in increasing command line
1065      * order are mapped to virtio-mmio buses with decreasing base addresses.
1066      *
1067      * When this code was originally written, that arrangement ensured that the
1068      * guest Linux kernel would give the lowest "name" (/dev/vda, eth0, etc) to
1069      * the first -device on the command line. (The end-to-end order is a
1070      * function of this loop, qbus_realize(), qbus_find_recursive(), and the
1071      * guest kernel's name-to-address assignment strategy.)
1072      *
1073      * Meanwhile, the kernel's traversal seems to have been reversed; see eg.
1074      * the message, if not necessarily the code, of commit 70161ff336.
1075      * Therefore the loop now establishes the inverse of the original intent.
1076      *
1077      * Unfortunately, we can't counteract the kernel change by reversing the
1078      * loop; it would break existing command lines.
1079      *
1080      * In any case, the kernel makes no guarantee about the stability of
1081      * enumeration order of virtio devices (as demonstrated by it changing
1082      * between kernel versions). For reliable and stable identification
1083      * of disks users must use UUIDs or similar mechanisms.
1084      */
1085     for (i = 0; i < NUM_VIRTIO_TRANSPORTS; i++) {
1086         int irq = vms->irqmap[VIRT_MMIO] + i;
1087         hwaddr base = vms->memmap[VIRT_MMIO].base + i * size;
1088 
1089         sysbus_create_simple("virtio-mmio", base,
1090                              qdev_get_gpio_in(vms->gic, irq));
1091     }
1092 
1093     /* We add dtb nodes in reverse order so that they appear in the finished
1094      * device tree lowest address first.
1095      *
1096      * Note that this mapping is independent of the loop above. The previous
1097      * loop influences virtio device to virtio transport assignment, whereas
1098      * this loop controls how virtio transports are laid out in the dtb.
1099      */
1100     for (i = NUM_VIRTIO_TRANSPORTS - 1; i >= 0; i--) {
1101         char *nodename;
1102         int irq = vms->irqmap[VIRT_MMIO] + i;
1103         hwaddr base = vms->memmap[VIRT_MMIO].base + i * size;
1104 
1105         nodename = g_strdup_printf("/virtio_mmio@%" PRIx64, base);
1106         qemu_fdt_add_subnode(ms->fdt, nodename);
1107         qemu_fdt_setprop_string(ms->fdt, nodename,
1108                                 "compatible", "virtio,mmio");
1109         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1110                                      2, base, 2, size);
1111         qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
1112                                GIC_FDT_IRQ_TYPE_SPI, irq,
1113                                GIC_FDT_IRQ_FLAGS_EDGE_LO_HI);
1114         qemu_fdt_setprop(ms->fdt, nodename, "dma-coherent", NULL, 0);
1115         g_free(nodename);
1116     }
1117 }
1118 
1119 #define VIRT_FLASH_SECTOR_SIZE (256 * KiB)
1120 
1121 static PFlashCFI01 *virt_flash_create1(VirtMachineState *vms,
1122                                         const char *name,
1123                                         const char *alias_prop_name)
1124 {
1125     /*
1126      * Create a single flash device.  We use the same parameters as
1127      * the flash devices on the Versatile Express board.
1128      */
1129     DeviceState *dev = qdev_new(TYPE_PFLASH_CFI01);
1130 
1131     qdev_prop_set_uint64(dev, "sector-length", VIRT_FLASH_SECTOR_SIZE);
1132     qdev_prop_set_uint8(dev, "width", 4);
1133     qdev_prop_set_uint8(dev, "device-width", 2);
1134     qdev_prop_set_bit(dev, "big-endian", false);
1135     qdev_prop_set_uint16(dev, "id0", 0x89);
1136     qdev_prop_set_uint16(dev, "id1", 0x18);
1137     qdev_prop_set_uint16(dev, "id2", 0x00);
1138     qdev_prop_set_uint16(dev, "id3", 0x00);
1139     qdev_prop_set_string(dev, "name", name);
1140     object_property_add_child(OBJECT(vms), name, OBJECT(dev));
1141     object_property_add_alias(OBJECT(vms), alias_prop_name,
1142                               OBJECT(dev), "drive");
1143     return PFLASH_CFI01(dev);
1144 }
1145 
1146 static void virt_flash_create(VirtMachineState *vms)
1147 {
1148     vms->flash[0] = virt_flash_create1(vms, "virt.flash0", "pflash0");
1149     vms->flash[1] = virt_flash_create1(vms, "virt.flash1", "pflash1");
1150 }
1151 
1152 static void virt_flash_map1(PFlashCFI01 *flash,
1153                             hwaddr base, hwaddr size,
1154                             MemoryRegion *sysmem)
1155 {
1156     DeviceState *dev = DEVICE(flash);
1157 
1158     assert(QEMU_IS_ALIGNED(size, VIRT_FLASH_SECTOR_SIZE));
1159     assert(size / VIRT_FLASH_SECTOR_SIZE <= UINT32_MAX);
1160     qdev_prop_set_uint32(dev, "num-blocks", size / VIRT_FLASH_SECTOR_SIZE);
1161     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1162 
1163     memory_region_add_subregion(sysmem, base,
1164                                 sysbus_mmio_get_region(SYS_BUS_DEVICE(dev),
1165                                                        0));
1166 }
1167 
1168 static void virt_flash_map(VirtMachineState *vms,
1169                            MemoryRegion *sysmem,
1170                            MemoryRegion *secure_sysmem)
1171 {
1172     /*
1173      * Map two flash devices to fill the VIRT_FLASH space in the memmap.
1174      * sysmem is the system memory space. secure_sysmem is the secure view
1175      * of the system, and the first flash device should be made visible only
1176      * there. The second flash device is visible to both secure and nonsecure.
1177      * If sysmem == secure_sysmem this means there is no separate Secure
1178      * address space and both flash devices are generally visible.
1179      */
1180     hwaddr flashsize = vms->memmap[VIRT_FLASH].size / 2;
1181     hwaddr flashbase = vms->memmap[VIRT_FLASH].base;
1182 
1183     virt_flash_map1(vms->flash[0], flashbase, flashsize,
1184                     secure_sysmem);
1185     virt_flash_map1(vms->flash[1], flashbase + flashsize, flashsize,
1186                     sysmem);
1187 }
1188 
1189 static void virt_flash_fdt(VirtMachineState *vms,
1190                            MemoryRegion *sysmem,
1191                            MemoryRegion *secure_sysmem)
1192 {
1193     hwaddr flashsize = vms->memmap[VIRT_FLASH].size / 2;
1194     hwaddr flashbase = vms->memmap[VIRT_FLASH].base;
1195     MachineState *ms = MACHINE(vms);
1196     char *nodename;
1197 
1198     if (sysmem == secure_sysmem) {
1199         /* Report both flash devices as a single node in the DT */
1200         nodename = g_strdup_printf("/flash@%" PRIx64, flashbase);
1201         qemu_fdt_add_subnode(ms->fdt, nodename);
1202         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible", "cfi-flash");
1203         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1204                                      2, flashbase, 2, flashsize,
1205                                      2, flashbase + flashsize, 2, flashsize);
1206         qemu_fdt_setprop_cell(ms->fdt, nodename, "bank-width", 4);
1207         g_free(nodename);
1208     } else {
1209         /*
1210          * Report the devices as separate nodes so we can mark one as
1211          * only visible to the secure world.
1212          */
1213         nodename = g_strdup_printf("/secflash@%" PRIx64, flashbase);
1214         qemu_fdt_add_subnode(ms->fdt, nodename);
1215         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible", "cfi-flash");
1216         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1217                                      2, flashbase, 2, flashsize);
1218         qemu_fdt_setprop_cell(ms->fdt, nodename, "bank-width", 4);
1219         qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
1220         qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
1221         g_free(nodename);
1222 
1223         nodename = g_strdup_printf("/flash@%" PRIx64, flashbase + flashsize);
1224         qemu_fdt_add_subnode(ms->fdt, nodename);
1225         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible", "cfi-flash");
1226         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1227                                      2, flashbase + flashsize, 2, flashsize);
1228         qemu_fdt_setprop_cell(ms->fdt, nodename, "bank-width", 4);
1229         g_free(nodename);
1230     }
1231 }
1232 
1233 static bool virt_firmware_init(VirtMachineState *vms,
1234                                MemoryRegion *sysmem,
1235                                MemoryRegion *secure_sysmem)
1236 {
1237     int i;
1238     const char *bios_name;
1239     BlockBackend *pflash_blk0;
1240 
1241     /* Map legacy -drive if=pflash to machine properties */
1242     for (i = 0; i < ARRAY_SIZE(vms->flash); i++) {
1243         pflash_cfi01_legacy_drive(vms->flash[i],
1244                                   drive_get(IF_PFLASH, 0, i));
1245     }
1246 
1247     virt_flash_map(vms, sysmem, secure_sysmem);
1248 
1249     pflash_blk0 = pflash_cfi01_get_blk(vms->flash[0]);
1250 
1251     bios_name = MACHINE(vms)->firmware;
1252     if (bios_name) {
1253         char *fname;
1254         MemoryRegion *mr;
1255         int image_size;
1256 
1257         if (pflash_blk0) {
1258             error_report("The contents of the first flash device may be "
1259                          "specified with -bios or with -drive if=pflash... "
1260                          "but you cannot use both options at once");
1261             exit(1);
1262         }
1263 
1264         /* Fall back to -bios */
1265 
1266         fname = qemu_find_file(QEMU_FILE_TYPE_BIOS, bios_name);
1267         if (!fname) {
1268             error_report("Could not find ROM image '%s'", bios_name);
1269             exit(1);
1270         }
1271         mr = sysbus_mmio_get_region(SYS_BUS_DEVICE(vms->flash[0]), 0);
1272         image_size = load_image_mr(fname, mr);
1273         g_free(fname);
1274         if (image_size < 0) {
1275             error_report("Could not load ROM image '%s'", bios_name);
1276             exit(1);
1277         }
1278     }
1279 
1280     return pflash_blk0 || bios_name;
1281 }
1282 
1283 static FWCfgState *create_fw_cfg(const VirtMachineState *vms, AddressSpace *as)
1284 {
1285     MachineState *ms = MACHINE(vms);
1286     hwaddr base = vms->memmap[VIRT_FW_CFG].base;
1287     hwaddr size = vms->memmap[VIRT_FW_CFG].size;
1288     FWCfgState *fw_cfg;
1289     char *nodename;
1290 
1291     fw_cfg = fw_cfg_init_mem_wide(base + 8, base, 8, base + 16, as);
1292     fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, (uint16_t)ms->smp.cpus);
1293 
1294     nodename = g_strdup_printf("/fw-cfg@%" PRIx64, base);
1295     qemu_fdt_add_subnode(ms->fdt, nodename);
1296     qemu_fdt_setprop_string(ms->fdt, nodename,
1297                             "compatible", "qemu,fw-cfg-mmio");
1298     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1299                                  2, base, 2, size);
1300     qemu_fdt_setprop(ms->fdt, nodename, "dma-coherent", NULL, 0);
1301     g_free(nodename);
1302     return fw_cfg;
1303 }
1304 
1305 static void create_pcie_irq_map(const MachineState *ms,
1306                                 uint32_t gic_phandle,
1307                                 int first_irq, const char *nodename)
1308 {
1309     int devfn, pin;
1310     uint32_t full_irq_map[4 * 4 * 10] = { 0 };
1311     uint32_t *irq_map = full_irq_map;
1312 
1313     for (devfn = 0; devfn <= 0x18; devfn += 0x8) {
1314         for (pin = 0; pin < 4; pin++) {
1315             int irq_type = GIC_FDT_IRQ_TYPE_SPI;
1316             int irq_nr = first_irq + ((pin + PCI_SLOT(devfn)) % PCI_NUM_PINS);
1317             int irq_level = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
1318             int i;
1319 
1320             uint32_t map[] = {
1321                 devfn << 8, 0, 0,                           /* devfn */
1322                 pin + 1,                                    /* PCI pin */
1323                 gic_phandle, 0, 0, irq_type, irq_nr, irq_level }; /* GIC irq */
1324 
1325             /* Convert map to big endian */
1326             for (i = 0; i < 10; i++) {
1327                 irq_map[i] = cpu_to_be32(map[i]);
1328             }
1329             irq_map += 10;
1330         }
1331     }
1332 
1333     qemu_fdt_setprop(ms->fdt, nodename, "interrupt-map",
1334                      full_irq_map, sizeof(full_irq_map));
1335 
1336     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupt-map-mask",
1337                            cpu_to_be16(PCI_DEVFN(3, 0)), /* Slot 3 */
1338                            0, 0,
1339                            0x7           /* PCI irq */);
1340 }
1341 
1342 static void create_smmu(const VirtMachineState *vms,
1343                         PCIBus *bus)
1344 {
1345     char *node;
1346     const char compat[] = "arm,smmu-v3";
1347     int irq =  vms->irqmap[VIRT_SMMU];
1348     int i;
1349     hwaddr base = vms->memmap[VIRT_SMMU].base;
1350     hwaddr size = vms->memmap[VIRT_SMMU].size;
1351     const char irq_names[] = "eventq\0priq\0cmdq-sync\0gerror";
1352     DeviceState *dev;
1353     MachineState *ms = MACHINE(vms);
1354 
1355     if (vms->iommu != VIRT_IOMMU_SMMUV3 || !vms->iommu_phandle) {
1356         return;
1357     }
1358 
1359     dev = qdev_new(TYPE_ARM_SMMUV3);
1360 
1361     object_property_set_link(OBJECT(dev), "primary-bus", OBJECT(bus),
1362                              &error_abort);
1363     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1364     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, base);
1365     for (i = 0; i < NUM_SMMU_IRQS; i++) {
1366         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
1367                            qdev_get_gpio_in(vms->gic, irq + i));
1368     }
1369 
1370     node = g_strdup_printf("/smmuv3@%" PRIx64, base);
1371     qemu_fdt_add_subnode(ms->fdt, node);
1372     qemu_fdt_setprop(ms->fdt, node, "compatible", compat, sizeof(compat));
1373     qemu_fdt_setprop_sized_cells(ms->fdt, node, "reg", 2, base, 2, size);
1374 
1375     qemu_fdt_setprop_cells(ms->fdt, node, "interrupts",
1376             GIC_FDT_IRQ_TYPE_SPI, irq    , GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1377             GIC_FDT_IRQ_TYPE_SPI, irq + 1, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1378             GIC_FDT_IRQ_TYPE_SPI, irq + 2, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1379             GIC_FDT_IRQ_TYPE_SPI, irq + 3, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI);
1380 
1381     qemu_fdt_setprop(ms->fdt, node, "interrupt-names", irq_names,
1382                      sizeof(irq_names));
1383 
1384     qemu_fdt_setprop(ms->fdt, node, "dma-coherent", NULL, 0);
1385 
1386     qemu_fdt_setprop_cell(ms->fdt, node, "#iommu-cells", 1);
1387 
1388     qemu_fdt_setprop_cell(ms->fdt, node, "phandle", vms->iommu_phandle);
1389     g_free(node);
1390 }
1391 
1392 static void create_virtio_iommu_dt_bindings(VirtMachineState *vms)
1393 {
1394     const char compat[] = "virtio,pci-iommu\0pci1af4,1057";
1395     uint16_t bdf = vms->virtio_iommu_bdf;
1396     MachineState *ms = MACHINE(vms);
1397     char *node;
1398 
1399     vms->iommu_phandle = qemu_fdt_alloc_phandle(ms->fdt);
1400 
1401     node = g_strdup_printf("%s/virtio_iommu@%x,%x", vms->pciehb_nodename,
1402                            PCI_SLOT(bdf), PCI_FUNC(bdf));
1403     qemu_fdt_add_subnode(ms->fdt, node);
1404     qemu_fdt_setprop(ms->fdt, node, "compatible", compat, sizeof(compat));
1405     qemu_fdt_setprop_sized_cells(ms->fdt, node, "reg",
1406                                  1, bdf << 8, 1, 0, 1, 0,
1407                                  1, 0, 1, 0);
1408 
1409     qemu_fdt_setprop_cell(ms->fdt, node, "#iommu-cells", 1);
1410     qemu_fdt_setprop_cell(ms->fdt, node, "phandle", vms->iommu_phandle);
1411     g_free(node);
1412 
1413     qemu_fdt_setprop_cells(ms->fdt, vms->pciehb_nodename, "iommu-map",
1414                            0x0, vms->iommu_phandle, 0x0, bdf,
1415                            bdf + 1, vms->iommu_phandle, bdf + 1, 0xffff - bdf);
1416 }
1417 
1418 static void create_pcie(VirtMachineState *vms)
1419 {
1420     hwaddr base_mmio = vms->memmap[VIRT_PCIE_MMIO].base;
1421     hwaddr size_mmio = vms->memmap[VIRT_PCIE_MMIO].size;
1422     hwaddr base_mmio_high = vms->memmap[VIRT_HIGH_PCIE_MMIO].base;
1423     hwaddr size_mmio_high = vms->memmap[VIRT_HIGH_PCIE_MMIO].size;
1424     hwaddr base_pio = vms->memmap[VIRT_PCIE_PIO].base;
1425     hwaddr size_pio = vms->memmap[VIRT_PCIE_PIO].size;
1426     hwaddr base_ecam, size_ecam;
1427     hwaddr base = base_mmio;
1428     int nr_pcie_buses;
1429     int irq = vms->irqmap[VIRT_PCIE];
1430     MemoryRegion *mmio_alias;
1431     MemoryRegion *mmio_reg;
1432     MemoryRegion *ecam_alias;
1433     MemoryRegion *ecam_reg;
1434     DeviceState *dev;
1435     char *nodename;
1436     int i, ecam_id;
1437     PCIHostState *pci;
1438     MachineState *ms = MACHINE(vms);
1439     MachineClass *mc = MACHINE_GET_CLASS(ms);
1440 
1441     dev = qdev_new(TYPE_GPEX_HOST);
1442     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1443 
1444     ecam_id = VIRT_ECAM_ID(vms->highmem_ecam);
1445     base_ecam = vms->memmap[ecam_id].base;
1446     size_ecam = vms->memmap[ecam_id].size;
1447     nr_pcie_buses = size_ecam / PCIE_MMCFG_SIZE_MIN;
1448     /* Map only the first size_ecam bytes of ECAM space */
1449     ecam_alias = g_new0(MemoryRegion, 1);
1450     ecam_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 0);
1451     memory_region_init_alias(ecam_alias, OBJECT(dev), "pcie-ecam",
1452                              ecam_reg, 0, size_ecam);
1453     memory_region_add_subregion(get_system_memory(), base_ecam, ecam_alias);
1454 
1455     /* Map the MMIO window into system address space so as to expose
1456      * the section of PCI MMIO space which starts at the same base address
1457      * (ie 1:1 mapping for that part of PCI MMIO space visible through
1458      * the window).
1459      */
1460     mmio_alias = g_new0(MemoryRegion, 1);
1461     mmio_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 1);
1462     memory_region_init_alias(mmio_alias, OBJECT(dev), "pcie-mmio",
1463                              mmio_reg, base_mmio, size_mmio);
1464     memory_region_add_subregion(get_system_memory(), base_mmio, mmio_alias);
1465 
1466     if (vms->highmem_mmio) {
1467         /* Map high MMIO space */
1468         MemoryRegion *high_mmio_alias = g_new0(MemoryRegion, 1);
1469 
1470         memory_region_init_alias(high_mmio_alias, OBJECT(dev), "pcie-mmio-high",
1471                                  mmio_reg, base_mmio_high, size_mmio_high);
1472         memory_region_add_subregion(get_system_memory(), base_mmio_high,
1473                                     high_mmio_alias);
1474     }
1475 
1476     /* Map IO port space */
1477     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 2, base_pio);
1478 
1479     for (i = 0; i < GPEX_NUM_IRQS; i++) {
1480         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
1481                            qdev_get_gpio_in(vms->gic, irq + i));
1482         gpex_set_irq_num(GPEX_HOST(dev), i, irq + i);
1483     }
1484 
1485     pci = PCI_HOST_BRIDGE(dev);
1486     pci->bypass_iommu = vms->default_bus_bypass_iommu;
1487     vms->bus = pci->bus;
1488     if (vms->bus) {
1489         for (i = 0; i < nb_nics; i++) {
1490             pci_nic_init_nofail(&nd_table[i], pci->bus, mc->default_nic, NULL);
1491         }
1492     }
1493 
1494     nodename = vms->pciehb_nodename = g_strdup_printf("/pcie@%" PRIx64, base);
1495     qemu_fdt_add_subnode(ms->fdt, nodename);
1496     qemu_fdt_setprop_string(ms->fdt, nodename,
1497                             "compatible", "pci-host-ecam-generic");
1498     qemu_fdt_setprop_string(ms->fdt, nodename, "device_type", "pci");
1499     qemu_fdt_setprop_cell(ms->fdt, nodename, "#address-cells", 3);
1500     qemu_fdt_setprop_cell(ms->fdt, nodename, "#size-cells", 2);
1501     qemu_fdt_setprop_cell(ms->fdt, nodename, "linux,pci-domain", 0);
1502     qemu_fdt_setprop_cells(ms->fdt, nodename, "bus-range", 0,
1503                            nr_pcie_buses - 1);
1504     qemu_fdt_setprop(ms->fdt, nodename, "dma-coherent", NULL, 0);
1505 
1506     if (vms->msi_phandle) {
1507         qemu_fdt_setprop_cells(ms->fdt, nodename, "msi-map",
1508                                0, vms->msi_phandle, 0, 0x10000);
1509     }
1510 
1511     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1512                                  2, base_ecam, 2, size_ecam);
1513 
1514     if (vms->highmem_mmio) {
1515         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "ranges",
1516                                      1, FDT_PCI_RANGE_IOPORT, 2, 0,
1517                                      2, base_pio, 2, size_pio,
1518                                      1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
1519                                      2, base_mmio, 2, size_mmio,
1520                                      1, FDT_PCI_RANGE_MMIO_64BIT,
1521                                      2, base_mmio_high,
1522                                      2, base_mmio_high, 2, size_mmio_high);
1523     } else {
1524         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "ranges",
1525                                      1, FDT_PCI_RANGE_IOPORT, 2, 0,
1526                                      2, base_pio, 2, size_pio,
1527                                      1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
1528                                      2, base_mmio, 2, size_mmio);
1529     }
1530 
1531     qemu_fdt_setprop_cell(ms->fdt, nodename, "#interrupt-cells", 1);
1532     create_pcie_irq_map(ms, vms->gic_phandle, irq, nodename);
1533 
1534     if (vms->iommu) {
1535         vms->iommu_phandle = qemu_fdt_alloc_phandle(ms->fdt);
1536 
1537         switch (vms->iommu) {
1538         case VIRT_IOMMU_SMMUV3:
1539             create_smmu(vms, vms->bus);
1540             qemu_fdt_setprop_cells(ms->fdt, nodename, "iommu-map",
1541                                    0x0, vms->iommu_phandle, 0x0, 0x10000);
1542             break;
1543         default:
1544             g_assert_not_reached();
1545         }
1546     }
1547 }
1548 
1549 static void create_platform_bus(VirtMachineState *vms)
1550 {
1551     DeviceState *dev;
1552     SysBusDevice *s;
1553     int i;
1554     MemoryRegion *sysmem = get_system_memory();
1555 
1556     dev = qdev_new(TYPE_PLATFORM_BUS_DEVICE);
1557     dev->id = g_strdup(TYPE_PLATFORM_BUS_DEVICE);
1558     qdev_prop_set_uint32(dev, "num_irqs", PLATFORM_BUS_NUM_IRQS);
1559     qdev_prop_set_uint32(dev, "mmio_size", vms->memmap[VIRT_PLATFORM_BUS].size);
1560     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1561     vms->platform_bus_dev = dev;
1562 
1563     s = SYS_BUS_DEVICE(dev);
1564     for (i = 0; i < PLATFORM_BUS_NUM_IRQS; i++) {
1565         int irq = vms->irqmap[VIRT_PLATFORM_BUS] + i;
1566         sysbus_connect_irq(s, i, qdev_get_gpio_in(vms->gic, irq));
1567     }
1568 
1569     memory_region_add_subregion(sysmem,
1570                                 vms->memmap[VIRT_PLATFORM_BUS].base,
1571                                 sysbus_mmio_get_region(s, 0));
1572 }
1573 
1574 static void create_tag_ram(MemoryRegion *tag_sysmem,
1575                            hwaddr base, hwaddr size,
1576                            const char *name)
1577 {
1578     MemoryRegion *tagram = g_new(MemoryRegion, 1);
1579 
1580     memory_region_init_ram(tagram, NULL, name, size / 32, &error_fatal);
1581     memory_region_add_subregion(tag_sysmem, base / 32, tagram);
1582 }
1583 
1584 static void create_secure_ram(VirtMachineState *vms,
1585                               MemoryRegion *secure_sysmem,
1586                               MemoryRegion *secure_tag_sysmem)
1587 {
1588     MemoryRegion *secram = g_new(MemoryRegion, 1);
1589     char *nodename;
1590     hwaddr base = vms->memmap[VIRT_SECURE_MEM].base;
1591     hwaddr size = vms->memmap[VIRT_SECURE_MEM].size;
1592     MachineState *ms = MACHINE(vms);
1593 
1594     memory_region_init_ram(secram, NULL, "virt.secure-ram", size,
1595                            &error_fatal);
1596     memory_region_add_subregion(secure_sysmem, base, secram);
1597 
1598     nodename = g_strdup_printf("/secram@%" PRIx64, base);
1599     qemu_fdt_add_subnode(ms->fdt, nodename);
1600     qemu_fdt_setprop_string(ms->fdt, nodename, "device_type", "memory");
1601     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg", 2, base, 2, size);
1602     qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
1603     qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
1604 
1605     if (secure_tag_sysmem) {
1606         create_tag_ram(secure_tag_sysmem, base, size, "mach-virt.secure-tag");
1607     }
1608 
1609     g_free(nodename);
1610 }
1611 
1612 static void *machvirt_dtb(const struct arm_boot_info *binfo, int *fdt_size)
1613 {
1614     const VirtMachineState *board = container_of(binfo, VirtMachineState,
1615                                                  bootinfo);
1616     MachineState *ms = MACHINE(board);
1617 
1618 
1619     *fdt_size = board->fdt_size;
1620     return ms->fdt;
1621 }
1622 
1623 static void virt_build_smbios(VirtMachineState *vms)
1624 {
1625     MachineClass *mc = MACHINE_GET_CLASS(vms);
1626     MachineState *ms = MACHINE(vms);
1627     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
1628     uint8_t *smbios_tables, *smbios_anchor;
1629     size_t smbios_tables_len, smbios_anchor_len;
1630     struct smbios_phys_mem_area mem_array;
1631     const char *product = "QEMU Virtual Machine";
1632 
1633     if (kvm_enabled()) {
1634         product = "KVM Virtual Machine";
1635     }
1636 
1637     smbios_set_defaults("QEMU", product,
1638                         vmc->smbios_old_sys_ver ? "1.0" : mc->name, false,
1639                         true, SMBIOS_ENTRY_POINT_TYPE_64);
1640 
1641     /* build the array of physical mem area from base_memmap */
1642     mem_array.address = vms->memmap[VIRT_MEM].base;
1643     mem_array.length = ms->ram_size;
1644 
1645     smbios_get_tables(ms, &mem_array, 1,
1646                       &smbios_tables, &smbios_tables_len,
1647                       &smbios_anchor, &smbios_anchor_len,
1648                       &error_fatal);
1649 
1650     if (smbios_anchor) {
1651         fw_cfg_add_file(vms->fw_cfg, "etc/smbios/smbios-tables",
1652                         smbios_tables, smbios_tables_len);
1653         fw_cfg_add_file(vms->fw_cfg, "etc/smbios/smbios-anchor",
1654                         smbios_anchor, smbios_anchor_len);
1655     }
1656 }
1657 
1658 static
1659 void virt_machine_done(Notifier *notifier, void *data)
1660 {
1661     VirtMachineState *vms = container_of(notifier, VirtMachineState,
1662                                          machine_done);
1663     MachineState *ms = MACHINE(vms);
1664     ARMCPU *cpu = ARM_CPU(first_cpu);
1665     struct arm_boot_info *info = &vms->bootinfo;
1666     AddressSpace *as = arm_boot_address_space(cpu, info);
1667 
1668     /*
1669      * If the user provided a dtb, we assume the dynamic sysbus nodes
1670      * already are integrated there. This corresponds to a use case where
1671      * the dynamic sysbus nodes are complex and their generation is not yet
1672      * supported. In that case the user can take charge of the guest dt
1673      * while qemu takes charge of the qom stuff.
1674      */
1675     if (info->dtb_filename == NULL) {
1676         platform_bus_add_all_fdt_nodes(ms->fdt, "/intc",
1677                                        vms->memmap[VIRT_PLATFORM_BUS].base,
1678                                        vms->memmap[VIRT_PLATFORM_BUS].size,
1679                                        vms->irqmap[VIRT_PLATFORM_BUS]);
1680     }
1681     if (arm_load_dtb(info->dtb_start, info, info->dtb_limit, as, ms) < 0) {
1682         exit(1);
1683     }
1684 
1685     fw_cfg_add_extra_pci_roots(vms->bus, vms->fw_cfg);
1686 
1687     virt_acpi_setup(vms);
1688     virt_build_smbios(vms);
1689 }
1690 
1691 static uint64_t virt_cpu_mp_affinity(VirtMachineState *vms, int idx)
1692 {
1693     uint8_t clustersz = ARM_DEFAULT_CPUS_PER_CLUSTER;
1694     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
1695 
1696     if (!vmc->disallow_affinity_adjustment) {
1697         /* Adjust MPIDR like 64-bit KVM hosts, which incorporate the
1698          * GIC's target-list limitations. 32-bit KVM hosts currently
1699          * always create clusters of 4 CPUs, but that is expected to
1700          * change when they gain support for gicv3. When KVM is enabled
1701          * it will override the changes we make here, therefore our
1702          * purposes are to make TCG consistent (with 64-bit KVM hosts)
1703          * and to improve SGI efficiency.
1704          */
1705         if (vms->gic_version == VIRT_GIC_VERSION_2) {
1706             clustersz = GIC_TARGETLIST_BITS;
1707         } else {
1708             clustersz = GICV3_TARGETLIST_BITS;
1709         }
1710     }
1711     return arm_cpu_mp_affinity(idx, clustersz);
1712 }
1713 
1714 static inline bool *virt_get_high_memmap_enabled(VirtMachineState *vms,
1715                                                  int index)
1716 {
1717     bool *enabled_array[] = {
1718         &vms->highmem_redists,
1719         &vms->highmem_ecam,
1720         &vms->highmem_mmio,
1721     };
1722 
1723     assert(ARRAY_SIZE(extended_memmap) - VIRT_LOWMEMMAP_LAST ==
1724            ARRAY_SIZE(enabled_array));
1725     assert(index - VIRT_LOWMEMMAP_LAST < ARRAY_SIZE(enabled_array));
1726 
1727     return enabled_array[index - VIRT_LOWMEMMAP_LAST];
1728 }
1729 
1730 static void virt_set_high_memmap(VirtMachineState *vms,
1731                                  hwaddr base, int pa_bits)
1732 {
1733     hwaddr region_base, region_size;
1734     bool *region_enabled, fits;
1735     int i;
1736 
1737     for (i = VIRT_LOWMEMMAP_LAST; i < ARRAY_SIZE(extended_memmap); i++) {
1738         region_enabled = virt_get_high_memmap_enabled(vms, i);
1739         region_base = ROUND_UP(base, extended_memmap[i].size);
1740         region_size = extended_memmap[i].size;
1741 
1742         vms->memmap[i].base = region_base;
1743         vms->memmap[i].size = region_size;
1744 
1745         /*
1746          * Check each device to see if it fits in the PA space,
1747          * moving highest_gpa as we go. For compatibility, move
1748          * highest_gpa for disabled fitting devices as well, if
1749          * the compact layout has been disabled.
1750          *
1751          * For each device that doesn't fit, disable it.
1752          */
1753         fits = (region_base + region_size) <= BIT_ULL(pa_bits);
1754         *region_enabled &= fits;
1755         if (vms->highmem_compact && !*region_enabled) {
1756             continue;
1757         }
1758 
1759         base = region_base + region_size;
1760         if (fits) {
1761             vms->highest_gpa = base - 1;
1762         }
1763     }
1764 }
1765 
1766 static void virt_set_memmap(VirtMachineState *vms, int pa_bits)
1767 {
1768     MachineState *ms = MACHINE(vms);
1769     hwaddr base, device_memory_base, device_memory_size, memtop;
1770     int i;
1771 
1772     vms->memmap = extended_memmap;
1773 
1774     for (i = 0; i < ARRAY_SIZE(base_memmap); i++) {
1775         vms->memmap[i] = base_memmap[i];
1776     }
1777 
1778     if (ms->ram_slots > ACPI_MAX_RAM_SLOTS) {
1779         error_report("unsupported number of memory slots: %"PRIu64,
1780                      ms->ram_slots);
1781         exit(EXIT_FAILURE);
1782     }
1783 
1784     /*
1785      * !highmem is exactly the same as limiting the PA space to 32bit,
1786      * irrespective of the underlying capabilities of the HW.
1787      */
1788     if (!vms->highmem) {
1789         pa_bits = 32;
1790     }
1791 
1792     /*
1793      * We compute the base of the high IO region depending on the
1794      * amount of initial and device memory. The device memory start/size
1795      * is aligned on 1GiB. We never put the high IO region below 256GiB
1796      * so that if maxram_size is < 255GiB we keep the legacy memory map.
1797      * The device region size assumes 1GiB page max alignment per slot.
1798      */
1799     device_memory_base =
1800         ROUND_UP(vms->memmap[VIRT_MEM].base + ms->ram_size, GiB);
1801     device_memory_size = ms->maxram_size - ms->ram_size + ms->ram_slots * GiB;
1802 
1803     /* Base address of the high IO region */
1804     memtop = base = device_memory_base + ROUND_UP(device_memory_size, GiB);
1805     if (memtop > BIT_ULL(pa_bits)) {
1806 	    error_report("Addressing limited to %d bits, but memory exceeds it by %llu bytes\n",
1807 			 pa_bits, memtop - BIT_ULL(pa_bits));
1808         exit(EXIT_FAILURE);
1809     }
1810     if (base < device_memory_base) {
1811         error_report("maxmem/slots too huge");
1812         exit(EXIT_FAILURE);
1813     }
1814     if (base < vms->memmap[VIRT_MEM].base + LEGACY_RAMLIMIT_BYTES) {
1815         base = vms->memmap[VIRT_MEM].base + LEGACY_RAMLIMIT_BYTES;
1816     }
1817 
1818     /* We know for sure that at least the memory fits in the PA space */
1819     vms->highest_gpa = memtop - 1;
1820 
1821     virt_set_high_memmap(vms, base, pa_bits);
1822 
1823     if (device_memory_size > 0) {
1824         machine_memory_devices_init(ms, device_memory_base, device_memory_size);
1825     }
1826 }
1827 
1828 static VirtGICType finalize_gic_version_do(const char *accel_name,
1829                                            VirtGICType gic_version,
1830                                            int gics_supported,
1831                                            unsigned int max_cpus)
1832 {
1833     /* Convert host/max/nosel to GIC version number */
1834     switch (gic_version) {
1835     case VIRT_GIC_VERSION_HOST:
1836         if (!kvm_enabled()) {
1837             error_report("gic-version=host requires KVM");
1838             exit(1);
1839         }
1840 
1841         /* For KVM, gic-version=host means gic-version=max */
1842         return finalize_gic_version_do(accel_name, VIRT_GIC_VERSION_MAX,
1843                                        gics_supported, max_cpus);
1844     case VIRT_GIC_VERSION_MAX:
1845         if (gics_supported & VIRT_GIC_VERSION_4_MASK) {
1846             gic_version = VIRT_GIC_VERSION_4;
1847         } else if (gics_supported & VIRT_GIC_VERSION_3_MASK) {
1848             gic_version = VIRT_GIC_VERSION_3;
1849         } else {
1850             gic_version = VIRT_GIC_VERSION_2;
1851         }
1852         break;
1853     case VIRT_GIC_VERSION_NOSEL:
1854         if ((gics_supported & VIRT_GIC_VERSION_2_MASK) &&
1855             max_cpus <= GIC_NCPU) {
1856             gic_version = VIRT_GIC_VERSION_2;
1857         } else if (gics_supported & VIRT_GIC_VERSION_3_MASK) {
1858             /*
1859              * in case the host does not support v2 emulation or
1860              * the end-user requested more than 8 VCPUs we now default
1861              * to v3. In any case defaulting to v2 would be broken.
1862              */
1863             gic_version = VIRT_GIC_VERSION_3;
1864         } else if (max_cpus > GIC_NCPU) {
1865             error_report("%s only supports GICv2 emulation but more than 8 "
1866                          "vcpus are requested", accel_name);
1867             exit(1);
1868         }
1869         break;
1870     case VIRT_GIC_VERSION_2:
1871     case VIRT_GIC_VERSION_3:
1872     case VIRT_GIC_VERSION_4:
1873         break;
1874     }
1875 
1876     /* Check chosen version is effectively supported */
1877     switch (gic_version) {
1878     case VIRT_GIC_VERSION_2:
1879         if (!(gics_supported & VIRT_GIC_VERSION_2_MASK)) {
1880             error_report("%s does not support GICv2 emulation", accel_name);
1881             exit(1);
1882         }
1883         break;
1884     case VIRT_GIC_VERSION_3:
1885         if (!(gics_supported & VIRT_GIC_VERSION_3_MASK)) {
1886             error_report("%s does not support GICv3 emulation", accel_name);
1887             exit(1);
1888         }
1889         break;
1890     case VIRT_GIC_VERSION_4:
1891         if (!(gics_supported & VIRT_GIC_VERSION_4_MASK)) {
1892             error_report("%s does not support GICv4 emulation, is virtualization=on?",
1893                          accel_name);
1894             exit(1);
1895         }
1896         break;
1897     default:
1898         error_report("logic error in finalize_gic_version");
1899         exit(1);
1900         break;
1901     }
1902 
1903     return gic_version;
1904 }
1905 
1906 /*
1907  * finalize_gic_version - Determines the final gic_version
1908  * according to the gic-version property
1909  *
1910  * Default GIC type is v2
1911  */
1912 static void finalize_gic_version(VirtMachineState *vms)
1913 {
1914     const char *accel_name = current_accel_name();
1915     unsigned int max_cpus = MACHINE(vms)->smp.max_cpus;
1916     int gics_supported = 0;
1917 
1918     /* Determine which GIC versions the current environment supports */
1919     if (kvm_enabled() && kvm_irqchip_in_kernel()) {
1920         int probe_bitmap = kvm_arm_vgic_probe();
1921 
1922         if (!probe_bitmap) {
1923             error_report("Unable to determine GIC version supported by host");
1924             exit(1);
1925         }
1926 
1927         if (probe_bitmap & KVM_ARM_VGIC_V2) {
1928             gics_supported |= VIRT_GIC_VERSION_2_MASK;
1929         }
1930         if (probe_bitmap & KVM_ARM_VGIC_V3) {
1931             gics_supported |= VIRT_GIC_VERSION_3_MASK;
1932         }
1933     } else if (kvm_enabled() && !kvm_irqchip_in_kernel()) {
1934         /* KVM w/o kernel irqchip can only deal with GICv2 */
1935         gics_supported |= VIRT_GIC_VERSION_2_MASK;
1936         accel_name = "KVM with kernel-irqchip=off";
1937     } else if (tcg_enabled() || hvf_enabled() || qtest_enabled())  {
1938         gics_supported |= VIRT_GIC_VERSION_2_MASK;
1939         if (module_object_class_by_name("arm-gicv3")) {
1940             gics_supported |= VIRT_GIC_VERSION_3_MASK;
1941             if (vms->virt) {
1942                 /* GICv4 only makes sense if CPU has EL2 */
1943                 gics_supported |= VIRT_GIC_VERSION_4_MASK;
1944             }
1945         }
1946     } else {
1947         error_report("Unsupported accelerator, can not determine GIC support");
1948         exit(1);
1949     }
1950 
1951     /*
1952      * Then convert helpers like host/max to concrete GIC versions and ensure
1953      * the desired version is supported
1954      */
1955     vms->gic_version = finalize_gic_version_do(accel_name, vms->gic_version,
1956                                                gics_supported, max_cpus);
1957 }
1958 
1959 /*
1960  * virt_cpu_post_init() must be called after the CPUs have
1961  * been realized and the GIC has been created.
1962  */
1963 static void virt_cpu_post_init(VirtMachineState *vms, MemoryRegion *sysmem)
1964 {
1965     int max_cpus = MACHINE(vms)->smp.max_cpus;
1966     bool aarch64, pmu, steal_time;
1967     CPUState *cpu;
1968 
1969     aarch64 = object_property_get_bool(OBJECT(first_cpu), "aarch64", NULL);
1970     pmu = object_property_get_bool(OBJECT(first_cpu), "pmu", NULL);
1971     steal_time = object_property_get_bool(OBJECT(first_cpu),
1972                                           "kvm-steal-time", NULL);
1973 
1974     if (kvm_enabled()) {
1975         hwaddr pvtime_reg_base = vms->memmap[VIRT_PVTIME].base;
1976         hwaddr pvtime_reg_size = vms->memmap[VIRT_PVTIME].size;
1977 
1978         if (steal_time) {
1979             MemoryRegion *pvtime = g_new(MemoryRegion, 1);
1980             hwaddr pvtime_size = max_cpus * PVTIME_SIZE_PER_CPU;
1981 
1982             /* The memory region size must be a multiple of host page size. */
1983             pvtime_size = REAL_HOST_PAGE_ALIGN(pvtime_size);
1984 
1985             if (pvtime_size > pvtime_reg_size) {
1986                 error_report("pvtime requires a %" HWADDR_PRId
1987                              " byte memory region for %d CPUs,"
1988                              " but only %" HWADDR_PRId " has been reserved",
1989                              pvtime_size, max_cpus, pvtime_reg_size);
1990                 exit(1);
1991             }
1992 
1993             memory_region_init_ram(pvtime, NULL, "pvtime", pvtime_size, NULL);
1994             memory_region_add_subregion(sysmem, pvtime_reg_base, pvtime);
1995         }
1996 
1997         CPU_FOREACH(cpu) {
1998             if (pmu) {
1999                 assert(arm_feature(&ARM_CPU(cpu)->env, ARM_FEATURE_PMU));
2000                 if (kvm_irqchip_in_kernel()) {
2001                     kvm_arm_pmu_set_irq(cpu, VIRTUAL_PMU_IRQ);
2002                 }
2003                 kvm_arm_pmu_init(cpu);
2004             }
2005             if (steal_time) {
2006                 kvm_arm_pvtime_init(cpu, pvtime_reg_base +
2007                                          cpu->cpu_index * PVTIME_SIZE_PER_CPU);
2008             }
2009         }
2010     } else {
2011         if (aarch64 && vms->highmem) {
2012             int requested_pa_size = 64 - clz64(vms->highest_gpa);
2013             int pamax = arm_pamax(ARM_CPU(first_cpu));
2014 
2015             if (pamax < requested_pa_size) {
2016                 error_report("VCPU supports less PA bits (%d) than "
2017                              "requested by the memory map (%d)",
2018                              pamax, requested_pa_size);
2019                 exit(1);
2020             }
2021         }
2022     }
2023 }
2024 
2025 static void machvirt_init(MachineState *machine)
2026 {
2027     VirtMachineState *vms = VIRT_MACHINE(machine);
2028     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(machine);
2029     MachineClass *mc = MACHINE_GET_CLASS(machine);
2030     const CPUArchIdList *possible_cpus;
2031     MemoryRegion *sysmem = get_system_memory();
2032     MemoryRegion *secure_sysmem = NULL;
2033     MemoryRegion *tag_sysmem = NULL;
2034     MemoryRegion *secure_tag_sysmem = NULL;
2035     int n, virt_max_cpus;
2036     bool firmware_loaded;
2037     bool aarch64 = true;
2038     bool has_ged = !vmc->no_ged;
2039     unsigned int smp_cpus = machine->smp.cpus;
2040     unsigned int max_cpus = machine->smp.max_cpus;
2041 
2042     if (!cpu_type_valid(machine->cpu_type)) {
2043         error_report("mach-virt: CPU type %s not supported", machine->cpu_type);
2044         exit(1);
2045     }
2046 
2047     possible_cpus = mc->possible_cpu_arch_ids(machine);
2048 
2049     /*
2050      * In accelerated mode, the memory map is computed earlier in kvm_type()
2051      * to create a VM with the right number of IPA bits.
2052      */
2053     if (!vms->memmap) {
2054         Object *cpuobj;
2055         ARMCPU *armcpu;
2056         int pa_bits;
2057 
2058         /*
2059          * Instantiate a temporary CPU object to find out about what
2060          * we are about to deal with. Once this is done, get rid of
2061          * the object.
2062          */
2063         cpuobj = object_new(possible_cpus->cpus[0].type);
2064         armcpu = ARM_CPU(cpuobj);
2065 
2066         pa_bits = arm_pamax(armcpu);
2067 
2068         object_unref(cpuobj);
2069 
2070         virt_set_memmap(vms, pa_bits);
2071     }
2072 
2073     /* We can probe only here because during property set
2074      * KVM is not available yet
2075      */
2076     finalize_gic_version(vms);
2077 
2078     if (vms->secure) {
2079         /*
2080          * The Secure view of the world is the same as the NonSecure,
2081          * but with a few extra devices. Create it as a container region
2082          * containing the system memory at low priority; any secure-only
2083          * devices go in at higher priority and take precedence.
2084          */
2085         secure_sysmem = g_new(MemoryRegion, 1);
2086         memory_region_init(secure_sysmem, OBJECT(machine), "secure-memory",
2087                            UINT64_MAX);
2088         memory_region_add_subregion_overlap(secure_sysmem, 0, sysmem, -1);
2089     }
2090 
2091     firmware_loaded = virt_firmware_init(vms, sysmem,
2092                                          secure_sysmem ?: sysmem);
2093 
2094     /* If we have an EL3 boot ROM then the assumption is that it will
2095      * implement PSCI itself, so disable QEMU's internal implementation
2096      * so it doesn't get in the way. Instead of starting secondary
2097      * CPUs in PSCI powerdown state we will start them all running and
2098      * let the boot ROM sort them out.
2099      * The usual case is that we do use QEMU's PSCI implementation;
2100      * if the guest has EL2 then we will use SMC as the conduit,
2101      * and otherwise we will use HVC (for backwards compatibility and
2102      * because if we're using KVM then we must use HVC).
2103      */
2104     if (vms->secure && firmware_loaded) {
2105         vms->psci_conduit = QEMU_PSCI_CONDUIT_DISABLED;
2106     } else if (vms->virt) {
2107         vms->psci_conduit = QEMU_PSCI_CONDUIT_SMC;
2108     } else {
2109         vms->psci_conduit = QEMU_PSCI_CONDUIT_HVC;
2110     }
2111 
2112     /*
2113      * The maximum number of CPUs depends on the GIC version, or on how
2114      * many redistributors we can fit into the memory map (which in turn
2115      * depends on whether this is a GICv3 or v4).
2116      */
2117     if (vms->gic_version == VIRT_GIC_VERSION_2) {
2118         virt_max_cpus = GIC_NCPU;
2119     } else {
2120         virt_max_cpus = virt_redist_capacity(vms, VIRT_GIC_REDIST);
2121         if (vms->highmem_redists) {
2122             virt_max_cpus += virt_redist_capacity(vms, VIRT_HIGH_GIC_REDIST2);
2123         }
2124     }
2125 
2126     if (max_cpus > virt_max_cpus) {
2127         error_report("Number of SMP CPUs requested (%d) exceeds max CPUs "
2128                      "supported by machine 'mach-virt' (%d)",
2129                      max_cpus, virt_max_cpus);
2130         if (vms->gic_version != VIRT_GIC_VERSION_2 && !vms->highmem_redists) {
2131             error_printf("Try 'highmem-redists=on' for more CPUs\n");
2132         }
2133 
2134         exit(1);
2135     }
2136 
2137     if (vms->secure && (kvm_enabled() || hvf_enabled())) {
2138         error_report("mach-virt: %s does not support providing "
2139                      "Security extensions (TrustZone) to the guest CPU",
2140                      current_accel_name());
2141         exit(1);
2142     }
2143 
2144     if (vms->virt && (kvm_enabled() || hvf_enabled())) {
2145         error_report("mach-virt: %s does not support providing "
2146                      "Virtualization extensions to the guest CPU",
2147                      current_accel_name());
2148         exit(1);
2149     }
2150 
2151     if (vms->mte && (kvm_enabled() || hvf_enabled())) {
2152         error_report("mach-virt: %s does not support providing "
2153                      "MTE to the guest CPU",
2154                      current_accel_name());
2155         exit(1);
2156     }
2157 
2158     create_fdt(vms);
2159 
2160     assert(possible_cpus->len == max_cpus);
2161     for (n = 0; n < possible_cpus->len; n++) {
2162         Object *cpuobj;
2163         CPUState *cs;
2164 
2165         if (n >= smp_cpus) {
2166             break;
2167         }
2168 
2169         cpuobj = object_new(possible_cpus->cpus[n].type);
2170         object_property_set_int(cpuobj, "mp-affinity",
2171                                 possible_cpus->cpus[n].arch_id, NULL);
2172 
2173         cs = CPU(cpuobj);
2174         cs->cpu_index = n;
2175 
2176         numa_cpu_pre_plug(&possible_cpus->cpus[cs->cpu_index], DEVICE(cpuobj),
2177                           &error_fatal);
2178 
2179         aarch64 &= object_property_get_bool(cpuobj, "aarch64", NULL);
2180 
2181         if (!vms->secure) {
2182             object_property_set_bool(cpuobj, "has_el3", false, NULL);
2183         }
2184 
2185         if (!vms->virt && object_property_find(cpuobj, "has_el2")) {
2186             object_property_set_bool(cpuobj, "has_el2", false, NULL);
2187         }
2188 
2189         if (vmc->kvm_no_adjvtime &&
2190             object_property_find(cpuobj, "kvm-no-adjvtime")) {
2191             object_property_set_bool(cpuobj, "kvm-no-adjvtime", true, NULL);
2192         }
2193 
2194         if (vmc->no_kvm_steal_time &&
2195             object_property_find(cpuobj, "kvm-steal-time")) {
2196             object_property_set_bool(cpuobj, "kvm-steal-time", false, NULL);
2197         }
2198 
2199         if (vmc->no_pmu && object_property_find(cpuobj, "pmu")) {
2200             object_property_set_bool(cpuobj, "pmu", false, NULL);
2201         }
2202 
2203         if (vmc->no_tcg_lpa2 && object_property_find(cpuobj, "lpa2")) {
2204             object_property_set_bool(cpuobj, "lpa2", false, NULL);
2205         }
2206 
2207         if (object_property_find(cpuobj, "reset-cbar")) {
2208             object_property_set_int(cpuobj, "reset-cbar",
2209                                     vms->memmap[VIRT_CPUPERIPHS].base,
2210                                     &error_abort);
2211         }
2212 
2213         object_property_set_link(cpuobj, "memory", OBJECT(sysmem),
2214                                  &error_abort);
2215         if (vms->secure) {
2216             object_property_set_link(cpuobj, "secure-memory",
2217                                      OBJECT(secure_sysmem), &error_abort);
2218         }
2219 
2220         if (vms->mte) {
2221             /* Create the memory region only once, but link to all cpus. */
2222             if (!tag_sysmem) {
2223                 /*
2224                  * The property exists only if MemTag is supported.
2225                  * If it is, we must allocate the ram to back that up.
2226                  */
2227                 if (!object_property_find(cpuobj, "tag-memory")) {
2228                     error_report("MTE requested, but not supported "
2229                                  "by the guest CPU");
2230                     exit(1);
2231                 }
2232 
2233                 tag_sysmem = g_new(MemoryRegion, 1);
2234                 memory_region_init(tag_sysmem, OBJECT(machine),
2235                                    "tag-memory", UINT64_MAX / 32);
2236 
2237                 if (vms->secure) {
2238                     secure_tag_sysmem = g_new(MemoryRegion, 1);
2239                     memory_region_init(secure_tag_sysmem, OBJECT(machine),
2240                                        "secure-tag-memory", UINT64_MAX / 32);
2241 
2242                     /* As with ram, secure-tag takes precedence over tag.  */
2243                     memory_region_add_subregion_overlap(secure_tag_sysmem, 0,
2244                                                         tag_sysmem, -1);
2245                 }
2246             }
2247 
2248             object_property_set_link(cpuobj, "tag-memory", OBJECT(tag_sysmem),
2249                                      &error_abort);
2250             if (vms->secure) {
2251                 object_property_set_link(cpuobj, "secure-tag-memory",
2252                                          OBJECT(secure_tag_sysmem),
2253                                          &error_abort);
2254             }
2255         }
2256 
2257         qdev_realize(DEVICE(cpuobj), NULL, &error_fatal);
2258         object_unref(cpuobj);
2259     }
2260     fdt_add_timer_nodes(vms);
2261     fdt_add_cpu_nodes(vms);
2262 
2263     memory_region_add_subregion(sysmem, vms->memmap[VIRT_MEM].base,
2264                                 machine->ram);
2265 
2266     virt_flash_fdt(vms, sysmem, secure_sysmem ?: sysmem);
2267 
2268     create_gic(vms, sysmem);
2269 
2270     virt_cpu_post_init(vms, sysmem);
2271 
2272     fdt_add_pmu_nodes(vms);
2273 
2274     create_uart(vms, VIRT_UART, sysmem, serial_hd(0));
2275 
2276     if (vms->secure) {
2277         create_secure_ram(vms, secure_sysmem, secure_tag_sysmem);
2278         create_uart(vms, VIRT_SECURE_UART, secure_sysmem, serial_hd(1));
2279     }
2280 
2281     if (tag_sysmem) {
2282         create_tag_ram(tag_sysmem, vms->memmap[VIRT_MEM].base,
2283                        machine->ram_size, "mach-virt.tag");
2284     }
2285 
2286     vms->highmem_ecam &= (!firmware_loaded || aarch64);
2287 
2288     create_rtc(vms);
2289 
2290     create_pcie(vms);
2291 
2292     if (has_ged && aarch64 && firmware_loaded && virt_is_acpi_enabled(vms)) {
2293         vms->acpi_dev = create_acpi_ged(vms);
2294     } else {
2295         create_gpio_devices(vms, VIRT_GPIO, sysmem);
2296     }
2297 
2298     if (vms->secure && !vmc->no_secure_gpio) {
2299         create_gpio_devices(vms, VIRT_SECURE_GPIO, secure_sysmem);
2300     }
2301 
2302      /* connect powerdown request */
2303      vms->powerdown_notifier.notify = virt_powerdown_req;
2304      qemu_register_powerdown_notifier(&vms->powerdown_notifier);
2305 
2306     /* Create mmio transports, so the user can create virtio backends
2307      * (which will be automatically plugged in to the transports). If
2308      * no backend is created the transport will just sit harmlessly idle.
2309      */
2310     create_virtio_devices(vms);
2311 
2312     vms->fw_cfg = create_fw_cfg(vms, &address_space_memory);
2313     rom_set_fw(vms->fw_cfg);
2314 
2315     create_platform_bus(vms);
2316 
2317     if (machine->nvdimms_state->is_enabled) {
2318         const struct AcpiGenericAddress arm_virt_nvdimm_acpi_dsmio = {
2319             .space_id = AML_AS_SYSTEM_MEMORY,
2320             .address = vms->memmap[VIRT_NVDIMM_ACPI].base,
2321             .bit_width = NVDIMM_ACPI_IO_LEN << 3
2322         };
2323 
2324         nvdimm_init_acpi_state(machine->nvdimms_state, sysmem,
2325                                arm_virt_nvdimm_acpi_dsmio,
2326                                vms->fw_cfg, OBJECT(vms));
2327     }
2328 
2329     vms->bootinfo.ram_size = machine->ram_size;
2330     vms->bootinfo.board_id = -1;
2331     vms->bootinfo.loader_start = vms->memmap[VIRT_MEM].base;
2332     vms->bootinfo.get_dtb = machvirt_dtb;
2333     vms->bootinfo.skip_dtb_autoload = true;
2334     vms->bootinfo.firmware_loaded = firmware_loaded;
2335     vms->bootinfo.psci_conduit = vms->psci_conduit;
2336     arm_load_kernel(ARM_CPU(first_cpu), machine, &vms->bootinfo);
2337 
2338     vms->machine_done.notify = virt_machine_done;
2339     qemu_add_machine_init_done_notifier(&vms->machine_done);
2340 }
2341 
2342 static bool virt_get_secure(Object *obj, Error **errp)
2343 {
2344     VirtMachineState *vms = VIRT_MACHINE(obj);
2345 
2346     return vms->secure;
2347 }
2348 
2349 static void virt_set_secure(Object *obj, bool value, Error **errp)
2350 {
2351     VirtMachineState *vms = VIRT_MACHINE(obj);
2352 
2353     vms->secure = value;
2354 }
2355 
2356 static bool virt_get_virt(Object *obj, Error **errp)
2357 {
2358     VirtMachineState *vms = VIRT_MACHINE(obj);
2359 
2360     return vms->virt;
2361 }
2362 
2363 static void virt_set_virt(Object *obj, bool value, Error **errp)
2364 {
2365     VirtMachineState *vms = VIRT_MACHINE(obj);
2366 
2367     vms->virt = value;
2368 }
2369 
2370 static bool virt_get_highmem(Object *obj, Error **errp)
2371 {
2372     VirtMachineState *vms = VIRT_MACHINE(obj);
2373 
2374     return vms->highmem;
2375 }
2376 
2377 static void virt_set_highmem(Object *obj, bool value, Error **errp)
2378 {
2379     VirtMachineState *vms = VIRT_MACHINE(obj);
2380 
2381     vms->highmem = value;
2382 }
2383 
2384 static bool virt_get_compact_highmem(Object *obj, Error **errp)
2385 {
2386     VirtMachineState *vms = VIRT_MACHINE(obj);
2387 
2388     return vms->highmem_compact;
2389 }
2390 
2391 static void virt_set_compact_highmem(Object *obj, bool value, Error **errp)
2392 {
2393     VirtMachineState *vms = VIRT_MACHINE(obj);
2394 
2395     vms->highmem_compact = value;
2396 }
2397 
2398 static bool virt_get_highmem_redists(Object *obj, Error **errp)
2399 {
2400     VirtMachineState *vms = VIRT_MACHINE(obj);
2401 
2402     return vms->highmem_redists;
2403 }
2404 
2405 static void virt_set_highmem_redists(Object *obj, bool value, Error **errp)
2406 {
2407     VirtMachineState *vms = VIRT_MACHINE(obj);
2408 
2409     vms->highmem_redists = value;
2410 }
2411 
2412 static bool virt_get_highmem_ecam(Object *obj, Error **errp)
2413 {
2414     VirtMachineState *vms = VIRT_MACHINE(obj);
2415 
2416     return vms->highmem_ecam;
2417 }
2418 
2419 static void virt_set_highmem_ecam(Object *obj, bool value, Error **errp)
2420 {
2421     VirtMachineState *vms = VIRT_MACHINE(obj);
2422 
2423     vms->highmem_ecam = value;
2424 }
2425 
2426 static bool virt_get_highmem_mmio(Object *obj, Error **errp)
2427 {
2428     VirtMachineState *vms = VIRT_MACHINE(obj);
2429 
2430     return vms->highmem_mmio;
2431 }
2432 
2433 static void virt_set_highmem_mmio(Object *obj, bool value, Error **errp)
2434 {
2435     VirtMachineState *vms = VIRT_MACHINE(obj);
2436 
2437     vms->highmem_mmio = value;
2438 }
2439 
2440 
2441 static bool virt_get_its(Object *obj, Error **errp)
2442 {
2443     VirtMachineState *vms = VIRT_MACHINE(obj);
2444 
2445     return vms->its;
2446 }
2447 
2448 static void virt_set_its(Object *obj, bool value, Error **errp)
2449 {
2450     VirtMachineState *vms = VIRT_MACHINE(obj);
2451 
2452     vms->its = value;
2453 }
2454 
2455 static bool virt_get_dtb_randomness(Object *obj, Error **errp)
2456 {
2457     VirtMachineState *vms = VIRT_MACHINE(obj);
2458 
2459     return vms->dtb_randomness;
2460 }
2461 
2462 static void virt_set_dtb_randomness(Object *obj, bool value, Error **errp)
2463 {
2464     VirtMachineState *vms = VIRT_MACHINE(obj);
2465 
2466     vms->dtb_randomness = value;
2467 }
2468 
2469 static char *virt_get_oem_id(Object *obj, Error **errp)
2470 {
2471     VirtMachineState *vms = VIRT_MACHINE(obj);
2472 
2473     return g_strdup(vms->oem_id);
2474 }
2475 
2476 static void virt_set_oem_id(Object *obj, const char *value, Error **errp)
2477 {
2478     VirtMachineState *vms = VIRT_MACHINE(obj);
2479     size_t len = strlen(value);
2480 
2481     if (len > 6) {
2482         error_setg(errp,
2483                    "User specified oem-id value is bigger than 6 bytes in size");
2484         return;
2485     }
2486 
2487     strncpy(vms->oem_id, value, 6);
2488 }
2489 
2490 static char *virt_get_oem_table_id(Object *obj, Error **errp)
2491 {
2492     VirtMachineState *vms = VIRT_MACHINE(obj);
2493 
2494     return g_strdup(vms->oem_table_id);
2495 }
2496 
2497 static void virt_set_oem_table_id(Object *obj, const char *value,
2498                                   Error **errp)
2499 {
2500     VirtMachineState *vms = VIRT_MACHINE(obj);
2501     size_t len = strlen(value);
2502 
2503     if (len > 8) {
2504         error_setg(errp,
2505                    "User specified oem-table-id value is bigger than 8 bytes in size");
2506         return;
2507     }
2508     strncpy(vms->oem_table_id, value, 8);
2509 }
2510 
2511 
2512 bool virt_is_acpi_enabled(VirtMachineState *vms)
2513 {
2514     if (vms->acpi == ON_OFF_AUTO_OFF) {
2515         return false;
2516     }
2517     return true;
2518 }
2519 
2520 static void virt_get_acpi(Object *obj, Visitor *v, const char *name,
2521                           void *opaque, Error **errp)
2522 {
2523     VirtMachineState *vms = VIRT_MACHINE(obj);
2524     OnOffAuto acpi = vms->acpi;
2525 
2526     visit_type_OnOffAuto(v, name, &acpi, errp);
2527 }
2528 
2529 static void virt_set_acpi(Object *obj, Visitor *v, const char *name,
2530                           void *opaque, Error **errp)
2531 {
2532     VirtMachineState *vms = VIRT_MACHINE(obj);
2533 
2534     visit_type_OnOffAuto(v, name, &vms->acpi, errp);
2535 }
2536 
2537 static bool virt_get_ras(Object *obj, Error **errp)
2538 {
2539     VirtMachineState *vms = VIRT_MACHINE(obj);
2540 
2541     return vms->ras;
2542 }
2543 
2544 static void virt_set_ras(Object *obj, bool value, Error **errp)
2545 {
2546     VirtMachineState *vms = VIRT_MACHINE(obj);
2547 
2548     vms->ras = value;
2549 }
2550 
2551 static bool virt_get_mte(Object *obj, Error **errp)
2552 {
2553     VirtMachineState *vms = VIRT_MACHINE(obj);
2554 
2555     return vms->mte;
2556 }
2557 
2558 static void virt_set_mte(Object *obj, bool value, Error **errp)
2559 {
2560     VirtMachineState *vms = VIRT_MACHINE(obj);
2561 
2562     vms->mte = value;
2563 }
2564 
2565 static char *virt_get_gic_version(Object *obj, Error **errp)
2566 {
2567     VirtMachineState *vms = VIRT_MACHINE(obj);
2568     const char *val;
2569 
2570     switch (vms->gic_version) {
2571     case VIRT_GIC_VERSION_4:
2572         val = "4";
2573         break;
2574     case VIRT_GIC_VERSION_3:
2575         val = "3";
2576         break;
2577     default:
2578         val = "2";
2579         break;
2580     }
2581     return g_strdup(val);
2582 }
2583 
2584 static void virt_set_gic_version(Object *obj, const char *value, Error **errp)
2585 {
2586     VirtMachineState *vms = VIRT_MACHINE(obj);
2587 
2588     if (!strcmp(value, "4")) {
2589         vms->gic_version = VIRT_GIC_VERSION_4;
2590     } else if (!strcmp(value, "3")) {
2591         vms->gic_version = VIRT_GIC_VERSION_3;
2592     } else if (!strcmp(value, "2")) {
2593         vms->gic_version = VIRT_GIC_VERSION_2;
2594     } else if (!strcmp(value, "host")) {
2595         vms->gic_version = VIRT_GIC_VERSION_HOST; /* Will probe later */
2596     } else if (!strcmp(value, "max")) {
2597         vms->gic_version = VIRT_GIC_VERSION_MAX; /* Will probe later */
2598     } else {
2599         error_setg(errp, "Invalid gic-version value");
2600         error_append_hint(errp, "Valid values are 3, 2, host, max.\n");
2601     }
2602 }
2603 
2604 static char *virt_get_iommu(Object *obj, Error **errp)
2605 {
2606     VirtMachineState *vms = VIRT_MACHINE(obj);
2607 
2608     switch (vms->iommu) {
2609     case VIRT_IOMMU_NONE:
2610         return g_strdup("none");
2611     case VIRT_IOMMU_SMMUV3:
2612         return g_strdup("smmuv3");
2613     default:
2614         g_assert_not_reached();
2615     }
2616 }
2617 
2618 static void virt_set_iommu(Object *obj, const char *value, Error **errp)
2619 {
2620     VirtMachineState *vms = VIRT_MACHINE(obj);
2621 
2622     if (!strcmp(value, "smmuv3")) {
2623         vms->iommu = VIRT_IOMMU_SMMUV3;
2624     } else if (!strcmp(value, "none")) {
2625         vms->iommu = VIRT_IOMMU_NONE;
2626     } else {
2627         error_setg(errp, "Invalid iommu value");
2628         error_append_hint(errp, "Valid values are none, smmuv3.\n");
2629     }
2630 }
2631 
2632 static bool virt_get_default_bus_bypass_iommu(Object *obj, Error **errp)
2633 {
2634     VirtMachineState *vms = VIRT_MACHINE(obj);
2635 
2636     return vms->default_bus_bypass_iommu;
2637 }
2638 
2639 static void virt_set_default_bus_bypass_iommu(Object *obj, bool value,
2640                                               Error **errp)
2641 {
2642     VirtMachineState *vms = VIRT_MACHINE(obj);
2643 
2644     vms->default_bus_bypass_iommu = value;
2645 }
2646 
2647 static CpuInstanceProperties
2648 virt_cpu_index_to_props(MachineState *ms, unsigned cpu_index)
2649 {
2650     MachineClass *mc = MACHINE_GET_CLASS(ms);
2651     const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(ms);
2652 
2653     assert(cpu_index < possible_cpus->len);
2654     return possible_cpus->cpus[cpu_index].props;
2655 }
2656 
2657 static int64_t virt_get_default_cpu_node_id(const MachineState *ms, int idx)
2658 {
2659     int64_t socket_id = ms->possible_cpus->cpus[idx].props.socket_id;
2660 
2661     return socket_id % ms->numa_state->num_nodes;
2662 }
2663 
2664 static const CPUArchIdList *virt_possible_cpu_arch_ids(MachineState *ms)
2665 {
2666     int n;
2667     unsigned int max_cpus = ms->smp.max_cpus;
2668     VirtMachineState *vms = VIRT_MACHINE(ms);
2669     MachineClass *mc = MACHINE_GET_CLASS(vms);
2670 
2671     if (ms->possible_cpus) {
2672         assert(ms->possible_cpus->len == max_cpus);
2673         return ms->possible_cpus;
2674     }
2675 
2676     ms->possible_cpus = g_malloc0(sizeof(CPUArchIdList) +
2677                                   sizeof(CPUArchId) * max_cpus);
2678     ms->possible_cpus->len = max_cpus;
2679     for (n = 0; n < ms->possible_cpus->len; n++) {
2680         ms->possible_cpus->cpus[n].type = ms->cpu_type;
2681         ms->possible_cpus->cpus[n].arch_id =
2682             virt_cpu_mp_affinity(vms, n);
2683 
2684         assert(!mc->smp_props.dies_supported);
2685         ms->possible_cpus->cpus[n].props.has_socket_id = true;
2686         ms->possible_cpus->cpus[n].props.socket_id =
2687             n / (ms->smp.clusters * ms->smp.cores * ms->smp.threads);
2688         ms->possible_cpus->cpus[n].props.has_cluster_id = true;
2689         ms->possible_cpus->cpus[n].props.cluster_id =
2690             (n / (ms->smp.cores * ms->smp.threads)) % ms->smp.clusters;
2691         ms->possible_cpus->cpus[n].props.has_core_id = true;
2692         ms->possible_cpus->cpus[n].props.core_id =
2693             (n / ms->smp.threads) % ms->smp.cores;
2694         ms->possible_cpus->cpus[n].props.has_thread_id = true;
2695         ms->possible_cpus->cpus[n].props.thread_id =
2696             n % ms->smp.threads;
2697     }
2698     return ms->possible_cpus;
2699 }
2700 
2701 static void virt_memory_pre_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
2702                                  Error **errp)
2703 {
2704     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2705     const MachineState *ms = MACHINE(hotplug_dev);
2706     const bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2707 
2708     if (!vms->acpi_dev) {
2709         error_setg(errp,
2710                    "memory hotplug is not enabled: missing acpi-ged device");
2711         return;
2712     }
2713 
2714     if (vms->mte) {
2715         error_setg(errp, "memory hotplug is not enabled: MTE is enabled");
2716         return;
2717     }
2718 
2719     if (is_nvdimm && !ms->nvdimms_state->is_enabled) {
2720         error_setg(errp, "nvdimm is not enabled: add 'nvdimm=on' to '-M'");
2721         return;
2722     }
2723 
2724     pc_dimm_pre_plug(PC_DIMM(dev), MACHINE(hotplug_dev), NULL, errp);
2725 }
2726 
2727 static void virt_memory_plug(HotplugHandler *hotplug_dev,
2728                              DeviceState *dev, Error **errp)
2729 {
2730     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2731     MachineState *ms = MACHINE(hotplug_dev);
2732     bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2733 
2734     pc_dimm_plug(PC_DIMM(dev), MACHINE(vms));
2735 
2736     if (is_nvdimm) {
2737         nvdimm_plug(ms->nvdimms_state);
2738     }
2739 
2740     hotplug_handler_plug(HOTPLUG_HANDLER(vms->acpi_dev),
2741                          dev, &error_abort);
2742 }
2743 
2744 static void virt_machine_device_pre_plug_cb(HotplugHandler *hotplug_dev,
2745                                             DeviceState *dev, Error **errp)
2746 {
2747     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2748 
2749     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2750         virt_memory_pre_plug(hotplug_dev, dev, errp);
2751     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI)) {
2752         virtio_md_pci_pre_plug(VIRTIO_MD_PCI(dev), MACHINE(hotplug_dev), errp);
2753     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2754         hwaddr db_start = 0, db_end = 0;
2755         QList *reserved_regions;
2756         char *resv_prop_str;
2757 
2758         if (vms->iommu != VIRT_IOMMU_NONE) {
2759             error_setg(errp, "virt machine does not support multiple IOMMUs");
2760             return;
2761         }
2762 
2763         switch (vms->msi_controller) {
2764         case VIRT_MSI_CTRL_NONE:
2765             return;
2766         case VIRT_MSI_CTRL_ITS:
2767             /* GITS_TRANSLATER page */
2768             db_start = base_memmap[VIRT_GIC_ITS].base + 0x10000;
2769             db_end = base_memmap[VIRT_GIC_ITS].base +
2770                      base_memmap[VIRT_GIC_ITS].size - 1;
2771             break;
2772         case VIRT_MSI_CTRL_GICV2M:
2773             /* MSI_SETSPI_NS page */
2774             db_start = base_memmap[VIRT_GIC_V2M].base;
2775             db_end = db_start + base_memmap[VIRT_GIC_V2M].size - 1;
2776             break;
2777         }
2778         resv_prop_str = g_strdup_printf("0x%"PRIx64":0x%"PRIx64":%u",
2779                                         db_start, db_end,
2780                                         VIRTIO_IOMMU_RESV_MEM_T_MSI);
2781 
2782         reserved_regions = qlist_new();
2783         qlist_append_str(reserved_regions, resv_prop_str);
2784         qdev_prop_set_array(dev, "reserved-regions", reserved_regions);
2785         g_free(resv_prop_str);
2786     }
2787 }
2788 
2789 static void virt_machine_device_plug_cb(HotplugHandler *hotplug_dev,
2790                                         DeviceState *dev, Error **errp)
2791 {
2792     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2793 
2794     if (vms->platform_bus_dev) {
2795         MachineClass *mc = MACHINE_GET_CLASS(vms);
2796 
2797         if (device_is_dynamic_sysbus(mc, dev)) {
2798             platform_bus_link_device(PLATFORM_BUS_DEVICE(vms->platform_bus_dev),
2799                                      SYS_BUS_DEVICE(dev));
2800         }
2801     }
2802 
2803     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2804         virt_memory_plug(hotplug_dev, dev, errp);
2805     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI)) {
2806         virtio_md_pci_plug(VIRTIO_MD_PCI(dev), MACHINE(hotplug_dev), errp);
2807     }
2808 
2809     if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2810         PCIDevice *pdev = PCI_DEVICE(dev);
2811 
2812         vms->iommu = VIRT_IOMMU_VIRTIO;
2813         vms->virtio_iommu_bdf = pci_get_bdf(pdev);
2814         create_virtio_iommu_dt_bindings(vms);
2815     }
2816 }
2817 
2818 static void virt_dimm_unplug_request(HotplugHandler *hotplug_dev,
2819                                      DeviceState *dev, Error **errp)
2820 {
2821     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2822 
2823     if (!vms->acpi_dev) {
2824         error_setg(errp,
2825                    "memory hotplug is not enabled: missing acpi-ged device");
2826         return;
2827     }
2828 
2829     if (object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM)) {
2830         error_setg(errp, "nvdimm device hot unplug is not supported yet.");
2831         return;
2832     }
2833 
2834     hotplug_handler_unplug_request(HOTPLUG_HANDLER(vms->acpi_dev), dev,
2835                                    errp);
2836 }
2837 
2838 static void virt_dimm_unplug(HotplugHandler *hotplug_dev,
2839                              DeviceState *dev, Error **errp)
2840 {
2841     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2842     Error *local_err = NULL;
2843 
2844     hotplug_handler_unplug(HOTPLUG_HANDLER(vms->acpi_dev), dev, &local_err);
2845     if (local_err) {
2846         goto out;
2847     }
2848 
2849     pc_dimm_unplug(PC_DIMM(dev), MACHINE(vms));
2850     qdev_unrealize(dev);
2851 
2852 out:
2853     error_propagate(errp, local_err);
2854 }
2855 
2856 static void virt_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
2857                                           DeviceState *dev, Error **errp)
2858 {
2859     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2860         virt_dimm_unplug_request(hotplug_dev, dev, errp);
2861     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI)) {
2862         virtio_md_pci_unplug_request(VIRTIO_MD_PCI(dev), MACHINE(hotplug_dev),
2863                                      errp);
2864     } else {
2865         error_setg(errp, "device unplug request for unsupported device"
2866                    " type: %s", object_get_typename(OBJECT(dev)));
2867     }
2868 }
2869 
2870 static void virt_machine_device_unplug_cb(HotplugHandler *hotplug_dev,
2871                                           DeviceState *dev, Error **errp)
2872 {
2873     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2874         virt_dimm_unplug(hotplug_dev, dev, errp);
2875     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI)) {
2876         virtio_md_pci_unplug(VIRTIO_MD_PCI(dev), MACHINE(hotplug_dev), errp);
2877     } else {
2878         error_setg(errp, "virt: device unplug for unsupported device"
2879                    " type: %s", object_get_typename(OBJECT(dev)));
2880     }
2881 }
2882 
2883 static HotplugHandler *virt_machine_get_hotplug_handler(MachineState *machine,
2884                                                         DeviceState *dev)
2885 {
2886     MachineClass *mc = MACHINE_GET_CLASS(machine);
2887 
2888     if (device_is_dynamic_sysbus(mc, dev) ||
2889         object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) ||
2890         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI) ||
2891         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2892         return HOTPLUG_HANDLER(machine);
2893     }
2894     return NULL;
2895 }
2896 
2897 /*
2898  * for arm64 kvm_type [7-0] encodes the requested number of bits
2899  * in the IPA address space
2900  */
2901 static int virt_kvm_type(MachineState *ms, const char *type_str)
2902 {
2903     VirtMachineState *vms = VIRT_MACHINE(ms);
2904     int max_vm_pa_size, requested_pa_size;
2905     bool fixed_ipa;
2906 
2907     max_vm_pa_size = kvm_arm_get_max_vm_ipa_size(ms, &fixed_ipa);
2908 
2909     /* we freeze the memory map to compute the highest gpa */
2910     virt_set_memmap(vms, max_vm_pa_size);
2911 
2912     requested_pa_size = 64 - clz64(vms->highest_gpa);
2913 
2914     /*
2915      * KVM requires the IPA size to be at least 32 bits.
2916      */
2917     if (requested_pa_size < 32) {
2918         requested_pa_size = 32;
2919     }
2920 
2921     if (requested_pa_size > max_vm_pa_size) {
2922         error_report("-m and ,maxmem option values "
2923                      "require an IPA range (%d bits) larger than "
2924                      "the one supported by the host (%d bits)",
2925                      requested_pa_size, max_vm_pa_size);
2926         return -1;
2927     }
2928     /*
2929      * We return the requested PA log size, unless KVM only supports
2930      * the implicit legacy 40b IPA setting, in which case the kvm_type
2931      * must be 0.
2932      */
2933     return fixed_ipa ? 0 : requested_pa_size;
2934 }
2935 
2936 static void virt_machine_class_init(ObjectClass *oc, void *data)
2937 {
2938     MachineClass *mc = MACHINE_CLASS(oc);
2939     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
2940 
2941     mc->init = machvirt_init;
2942     /* Start with max_cpus set to 512, which is the maximum supported by KVM.
2943      * The value may be reduced later when we have more information about the
2944      * configuration of the particular instance.
2945      */
2946     mc->max_cpus = 512;
2947     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_CALXEDA_XGMAC);
2948     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_AMD_XGBE);
2949     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_RAMFB_DEVICE);
2950     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_PLATFORM);
2951 #ifdef CONFIG_TPM
2952     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_TPM_TIS_SYSBUS);
2953 #endif
2954     mc->block_default_type = IF_VIRTIO;
2955     mc->no_cdrom = 1;
2956     mc->pci_allow_0_address = true;
2957     /* We know we will never create a pre-ARMv7 CPU which needs 1K pages */
2958     mc->minimum_page_bits = 12;
2959     mc->possible_cpu_arch_ids = virt_possible_cpu_arch_ids;
2960     mc->cpu_index_to_instance_props = virt_cpu_index_to_props;
2961 #ifdef CONFIG_TCG
2962     mc->default_cpu_type = ARM_CPU_TYPE_NAME("cortex-a15");
2963 #else
2964     mc->default_cpu_type = ARM_CPU_TYPE_NAME("max");
2965 #endif
2966     mc->get_default_cpu_node_id = virt_get_default_cpu_node_id;
2967     mc->kvm_type = virt_kvm_type;
2968     assert(!mc->get_hotplug_handler);
2969     mc->get_hotplug_handler = virt_machine_get_hotplug_handler;
2970     hc->pre_plug = virt_machine_device_pre_plug_cb;
2971     hc->plug = virt_machine_device_plug_cb;
2972     hc->unplug_request = virt_machine_device_unplug_request_cb;
2973     hc->unplug = virt_machine_device_unplug_cb;
2974     mc->nvdimm_supported = true;
2975     mc->smp_props.clusters_supported = true;
2976     mc->auto_enable_numa_with_memhp = true;
2977     mc->auto_enable_numa_with_memdev = true;
2978     /* platform instead of architectural choice */
2979     mc->cpu_cluster_has_numa_boundary = true;
2980     mc->default_ram_id = "mach-virt.ram";
2981     mc->default_nic = "virtio-net-pci";
2982 
2983     object_class_property_add(oc, "acpi", "OnOffAuto",
2984         virt_get_acpi, virt_set_acpi,
2985         NULL, NULL);
2986     object_class_property_set_description(oc, "acpi",
2987         "Enable ACPI");
2988     object_class_property_add_bool(oc, "secure", virt_get_secure,
2989                                    virt_set_secure);
2990     object_class_property_set_description(oc, "secure",
2991                                                 "Set on/off to enable/disable the ARM "
2992                                                 "Security Extensions (TrustZone)");
2993 
2994     object_class_property_add_bool(oc, "virtualization", virt_get_virt,
2995                                    virt_set_virt);
2996     object_class_property_set_description(oc, "virtualization",
2997                                           "Set on/off to enable/disable emulating a "
2998                                           "guest CPU which implements the ARM "
2999                                           "Virtualization Extensions");
3000 
3001     object_class_property_add_bool(oc, "highmem", virt_get_highmem,
3002                                    virt_set_highmem);
3003     object_class_property_set_description(oc, "highmem",
3004                                           "Set on/off to enable/disable using "
3005                                           "physical address space above 32 bits");
3006 
3007     object_class_property_add_bool(oc, "compact-highmem",
3008                                    virt_get_compact_highmem,
3009                                    virt_set_compact_highmem);
3010     object_class_property_set_description(oc, "compact-highmem",
3011                                           "Set on/off to enable/disable compact "
3012                                           "layout for high memory regions");
3013 
3014     object_class_property_add_bool(oc, "highmem-redists",
3015                                    virt_get_highmem_redists,
3016                                    virt_set_highmem_redists);
3017     object_class_property_set_description(oc, "highmem-redists",
3018                                           "Set on/off to enable/disable high "
3019                                           "memory region for GICv3 or GICv4 "
3020                                           "redistributor");
3021 
3022     object_class_property_add_bool(oc, "highmem-ecam",
3023                                    virt_get_highmem_ecam,
3024                                    virt_set_highmem_ecam);
3025     object_class_property_set_description(oc, "highmem-ecam",
3026                                           "Set on/off to enable/disable high "
3027                                           "memory region for PCI ECAM");
3028 
3029     object_class_property_add_bool(oc, "highmem-mmio",
3030                                    virt_get_highmem_mmio,
3031                                    virt_set_highmem_mmio);
3032     object_class_property_set_description(oc, "highmem-mmio",
3033                                           "Set on/off to enable/disable high "
3034                                           "memory region for PCI MMIO");
3035 
3036     object_class_property_add_str(oc, "gic-version", virt_get_gic_version,
3037                                   virt_set_gic_version);
3038     object_class_property_set_description(oc, "gic-version",
3039                                           "Set GIC version. "
3040                                           "Valid values are 2, 3, 4, host and max");
3041 
3042     object_class_property_add_str(oc, "iommu", virt_get_iommu, virt_set_iommu);
3043     object_class_property_set_description(oc, "iommu",
3044                                           "Set the IOMMU type. "
3045                                           "Valid values are none and smmuv3");
3046 
3047     object_class_property_add_bool(oc, "default-bus-bypass-iommu",
3048                                    virt_get_default_bus_bypass_iommu,
3049                                    virt_set_default_bus_bypass_iommu);
3050     object_class_property_set_description(oc, "default-bus-bypass-iommu",
3051                                           "Set on/off to enable/disable "
3052                                           "bypass_iommu for default root bus");
3053 
3054     object_class_property_add_bool(oc, "ras", virt_get_ras,
3055                                    virt_set_ras);
3056     object_class_property_set_description(oc, "ras",
3057                                           "Set on/off to enable/disable reporting host memory errors "
3058                                           "to a KVM guest using ACPI and guest external abort exceptions");
3059 
3060     object_class_property_add_bool(oc, "mte", virt_get_mte, virt_set_mte);
3061     object_class_property_set_description(oc, "mte",
3062                                           "Set on/off to enable/disable emulating a "
3063                                           "guest CPU which implements the ARM "
3064                                           "Memory Tagging Extension");
3065 
3066     object_class_property_add_bool(oc, "its", virt_get_its,
3067                                    virt_set_its);
3068     object_class_property_set_description(oc, "its",
3069                                           "Set on/off to enable/disable "
3070                                           "ITS instantiation");
3071 
3072     object_class_property_add_bool(oc, "dtb-randomness",
3073                                    virt_get_dtb_randomness,
3074                                    virt_set_dtb_randomness);
3075     object_class_property_set_description(oc, "dtb-randomness",
3076                                           "Set off to disable passing random or "
3077                                           "non-deterministic dtb nodes to guest");
3078 
3079     object_class_property_add_bool(oc, "dtb-kaslr-seed",
3080                                    virt_get_dtb_randomness,
3081                                    virt_set_dtb_randomness);
3082     object_class_property_set_description(oc, "dtb-kaslr-seed",
3083                                           "Deprecated synonym of dtb-randomness");
3084 
3085     object_class_property_add_str(oc, "x-oem-id",
3086                                   virt_get_oem_id,
3087                                   virt_set_oem_id);
3088     object_class_property_set_description(oc, "x-oem-id",
3089                                           "Override the default value of field OEMID "
3090                                           "in ACPI table header."
3091                                           "The string may be up to 6 bytes in size");
3092 
3093 
3094     object_class_property_add_str(oc, "x-oem-table-id",
3095                                   virt_get_oem_table_id,
3096                                   virt_set_oem_table_id);
3097     object_class_property_set_description(oc, "x-oem-table-id",
3098                                           "Override the default value of field OEM Table ID "
3099                                           "in ACPI table header."
3100                                           "The string may be up to 8 bytes in size");
3101 
3102 }
3103 
3104 static void virt_instance_init(Object *obj)
3105 {
3106     VirtMachineState *vms = VIRT_MACHINE(obj);
3107     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
3108 
3109     /* EL3 is disabled by default on virt: this makes us consistent
3110      * between KVM and TCG for this board, and it also allows us to
3111      * boot UEFI blobs which assume no TrustZone support.
3112      */
3113     vms->secure = false;
3114 
3115     /* EL2 is also disabled by default, for similar reasons */
3116     vms->virt = false;
3117 
3118     /* High memory is enabled by default */
3119     vms->highmem = true;
3120     vms->highmem_compact = !vmc->no_highmem_compact;
3121     vms->gic_version = VIRT_GIC_VERSION_NOSEL;
3122 
3123     vms->highmem_ecam = !vmc->no_highmem_ecam;
3124     vms->highmem_mmio = true;
3125     vms->highmem_redists = true;
3126 
3127     if (vmc->no_its) {
3128         vms->its = false;
3129     } else {
3130         /* Default allows ITS instantiation */
3131         vms->its = true;
3132 
3133         if (vmc->no_tcg_its) {
3134             vms->tcg_its = false;
3135         } else {
3136             vms->tcg_its = true;
3137         }
3138     }
3139 
3140     /* Default disallows iommu instantiation */
3141     vms->iommu = VIRT_IOMMU_NONE;
3142 
3143     /* The default root bus is attached to iommu by default */
3144     vms->default_bus_bypass_iommu = false;
3145 
3146     /* Default disallows RAS instantiation */
3147     vms->ras = false;
3148 
3149     /* MTE is disabled by default.  */
3150     vms->mte = false;
3151 
3152     /* Supply kaslr-seed and rng-seed by default */
3153     vms->dtb_randomness = true;
3154 
3155     vms->irqmap = a15irqmap;
3156 
3157     virt_flash_create(vms);
3158 
3159     vms->oem_id = g_strndup(ACPI_BUILD_APPNAME6, 6);
3160     vms->oem_table_id = g_strndup(ACPI_BUILD_APPNAME8, 8);
3161 }
3162 
3163 static const TypeInfo virt_machine_info = {
3164     .name          = TYPE_VIRT_MACHINE,
3165     .parent        = TYPE_MACHINE,
3166     .abstract      = true,
3167     .instance_size = sizeof(VirtMachineState),
3168     .class_size    = sizeof(VirtMachineClass),
3169     .class_init    = virt_machine_class_init,
3170     .instance_init = virt_instance_init,
3171     .interfaces = (InterfaceInfo[]) {
3172          { TYPE_HOTPLUG_HANDLER },
3173          { }
3174     },
3175 };
3176 
3177 static void machvirt_machine_init(void)
3178 {
3179     type_register_static(&virt_machine_info);
3180 }
3181 type_init(machvirt_machine_init);
3182 
3183 static void virt_machine_8_2_options(MachineClass *mc)
3184 {
3185 }
3186 DEFINE_VIRT_MACHINE_AS_LATEST(8, 2)
3187 
3188 static void virt_machine_8_1_options(MachineClass *mc)
3189 {
3190     virt_machine_8_2_options(mc);
3191     compat_props_add(mc->compat_props, hw_compat_8_1, hw_compat_8_1_len);
3192 }
3193 DEFINE_VIRT_MACHINE(8, 1)
3194 
3195 static void virt_machine_8_0_options(MachineClass *mc)
3196 {
3197     virt_machine_8_1_options(mc);
3198     compat_props_add(mc->compat_props, hw_compat_8_0, hw_compat_8_0_len);
3199 }
3200 DEFINE_VIRT_MACHINE(8, 0)
3201 
3202 static void virt_machine_7_2_options(MachineClass *mc)
3203 {
3204     virt_machine_8_0_options(mc);
3205     compat_props_add(mc->compat_props, hw_compat_7_2, hw_compat_7_2_len);
3206 }
3207 DEFINE_VIRT_MACHINE(7, 2)
3208 
3209 static void virt_machine_7_1_options(MachineClass *mc)
3210 {
3211     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3212 
3213     virt_machine_7_2_options(mc);
3214     compat_props_add(mc->compat_props, hw_compat_7_1, hw_compat_7_1_len);
3215     /* Compact layout for high memory regions was introduced with 7.2 */
3216     vmc->no_highmem_compact = true;
3217 }
3218 DEFINE_VIRT_MACHINE(7, 1)
3219 
3220 static void virt_machine_7_0_options(MachineClass *mc)
3221 {
3222     virt_machine_7_1_options(mc);
3223     compat_props_add(mc->compat_props, hw_compat_7_0, hw_compat_7_0_len);
3224 }
3225 DEFINE_VIRT_MACHINE(7, 0)
3226 
3227 static void virt_machine_6_2_options(MachineClass *mc)
3228 {
3229     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3230 
3231     virt_machine_7_0_options(mc);
3232     compat_props_add(mc->compat_props, hw_compat_6_2, hw_compat_6_2_len);
3233     vmc->no_tcg_lpa2 = true;
3234 }
3235 DEFINE_VIRT_MACHINE(6, 2)
3236 
3237 static void virt_machine_6_1_options(MachineClass *mc)
3238 {
3239     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3240 
3241     virt_machine_6_2_options(mc);
3242     compat_props_add(mc->compat_props, hw_compat_6_1, hw_compat_6_1_len);
3243     mc->smp_props.prefer_sockets = true;
3244     vmc->no_cpu_topology = true;
3245 
3246     /* qemu ITS was introduced with 6.2 */
3247     vmc->no_tcg_its = true;
3248 }
3249 DEFINE_VIRT_MACHINE(6, 1)
3250 
3251 static void virt_machine_6_0_options(MachineClass *mc)
3252 {
3253     virt_machine_6_1_options(mc);
3254     compat_props_add(mc->compat_props, hw_compat_6_0, hw_compat_6_0_len);
3255 }
3256 DEFINE_VIRT_MACHINE(6, 0)
3257 
3258 static void virt_machine_5_2_options(MachineClass *mc)
3259 {
3260     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3261 
3262     virt_machine_6_0_options(mc);
3263     compat_props_add(mc->compat_props, hw_compat_5_2, hw_compat_5_2_len);
3264     vmc->no_secure_gpio = true;
3265 }
3266 DEFINE_VIRT_MACHINE(5, 2)
3267 
3268 static void virt_machine_5_1_options(MachineClass *mc)
3269 {
3270     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3271 
3272     virt_machine_5_2_options(mc);
3273     compat_props_add(mc->compat_props, hw_compat_5_1, hw_compat_5_1_len);
3274     vmc->no_kvm_steal_time = true;
3275 }
3276 DEFINE_VIRT_MACHINE(5, 1)
3277 
3278 static void virt_machine_5_0_options(MachineClass *mc)
3279 {
3280     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3281 
3282     virt_machine_5_1_options(mc);
3283     compat_props_add(mc->compat_props, hw_compat_5_0, hw_compat_5_0_len);
3284     mc->numa_mem_supported = true;
3285     vmc->acpi_expose_flash = true;
3286     mc->auto_enable_numa_with_memdev = false;
3287 }
3288 DEFINE_VIRT_MACHINE(5, 0)
3289 
3290 static void virt_machine_4_2_options(MachineClass *mc)
3291 {
3292     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3293 
3294     virt_machine_5_0_options(mc);
3295     compat_props_add(mc->compat_props, hw_compat_4_2, hw_compat_4_2_len);
3296     vmc->kvm_no_adjvtime = true;
3297 }
3298 DEFINE_VIRT_MACHINE(4, 2)
3299 
3300 static void virt_machine_4_1_options(MachineClass *mc)
3301 {
3302     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3303 
3304     virt_machine_4_2_options(mc);
3305     compat_props_add(mc->compat_props, hw_compat_4_1, hw_compat_4_1_len);
3306     vmc->no_ged = true;
3307     mc->auto_enable_numa_with_memhp = false;
3308 }
3309 DEFINE_VIRT_MACHINE(4, 1)
3310 
3311 static void virt_machine_4_0_options(MachineClass *mc)
3312 {
3313     virt_machine_4_1_options(mc);
3314     compat_props_add(mc->compat_props, hw_compat_4_0, hw_compat_4_0_len);
3315 }
3316 DEFINE_VIRT_MACHINE(4, 0)
3317 
3318 static void virt_machine_3_1_options(MachineClass *mc)
3319 {
3320     virt_machine_4_0_options(mc);
3321     compat_props_add(mc->compat_props, hw_compat_3_1, hw_compat_3_1_len);
3322 }
3323 DEFINE_VIRT_MACHINE(3, 1)
3324 
3325 static void virt_machine_3_0_options(MachineClass *mc)
3326 {
3327     virt_machine_3_1_options(mc);
3328     compat_props_add(mc->compat_props, hw_compat_3_0, hw_compat_3_0_len);
3329 }
3330 DEFINE_VIRT_MACHINE(3, 0)
3331 
3332 static void virt_machine_2_12_options(MachineClass *mc)
3333 {
3334     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3335 
3336     virt_machine_3_0_options(mc);
3337     compat_props_add(mc->compat_props, hw_compat_2_12, hw_compat_2_12_len);
3338     vmc->no_highmem_ecam = true;
3339     mc->max_cpus = 255;
3340 }
3341 DEFINE_VIRT_MACHINE(2, 12)
3342 
3343 static void virt_machine_2_11_options(MachineClass *mc)
3344 {
3345     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3346 
3347     virt_machine_2_12_options(mc);
3348     compat_props_add(mc->compat_props, hw_compat_2_11, hw_compat_2_11_len);
3349     vmc->smbios_old_sys_ver = true;
3350 }
3351 DEFINE_VIRT_MACHINE(2, 11)
3352 
3353 static void virt_machine_2_10_options(MachineClass *mc)
3354 {
3355     virt_machine_2_11_options(mc);
3356     compat_props_add(mc->compat_props, hw_compat_2_10, hw_compat_2_10_len);
3357     /* before 2.11 we never faulted accesses to bad addresses */
3358     mc->ignore_memory_transaction_failures = true;
3359 }
3360 DEFINE_VIRT_MACHINE(2, 10)
3361 
3362 static void virt_machine_2_9_options(MachineClass *mc)
3363 {
3364     virt_machine_2_10_options(mc);
3365     compat_props_add(mc->compat_props, hw_compat_2_9, hw_compat_2_9_len);
3366 }
3367 DEFINE_VIRT_MACHINE(2, 9)
3368 
3369 static void virt_machine_2_8_options(MachineClass *mc)
3370 {
3371     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3372 
3373     virt_machine_2_9_options(mc);
3374     compat_props_add(mc->compat_props, hw_compat_2_8, hw_compat_2_8_len);
3375     /* For 2.8 and earlier we falsely claimed in the DT that
3376      * our timers were edge-triggered, not level-triggered.
3377      */
3378     vmc->claim_edge_triggered_timers = true;
3379 }
3380 DEFINE_VIRT_MACHINE(2, 8)
3381 
3382 static void virt_machine_2_7_options(MachineClass *mc)
3383 {
3384     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3385 
3386     virt_machine_2_8_options(mc);
3387     compat_props_add(mc->compat_props, hw_compat_2_7, hw_compat_2_7_len);
3388     /* ITS was introduced with 2.8 */
3389     vmc->no_its = true;
3390     /* Stick with 1K pages for migration compatibility */
3391     mc->minimum_page_bits = 0;
3392 }
3393 DEFINE_VIRT_MACHINE(2, 7)
3394 
3395 static void virt_machine_2_6_options(MachineClass *mc)
3396 {
3397     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3398 
3399     virt_machine_2_7_options(mc);
3400     compat_props_add(mc->compat_props, hw_compat_2_6, hw_compat_2_6_len);
3401     vmc->disallow_affinity_adjustment = true;
3402     /* Disable PMU for 2.6 as PMU support was first introduced in 2.7 */
3403     vmc->no_pmu = true;
3404 }
3405 DEFINE_VIRT_MACHINE(2, 6)
3406