xref: /openbmc/qemu/hw/arm/virt.c (revision c79aa350)
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 "qapi/error.h"
37 #include "hw/sysbus.h"
38 #include "hw/arm/boot.h"
39 #include "hw/arm/primecell.h"
40 #include "hw/arm/virt.h"
41 #include "hw/block/flash.h"
42 #include "hw/vfio/vfio-calxeda-xgmac.h"
43 #include "hw/vfio/vfio-amd-xgbe.h"
44 #include "hw/display/ramfb.h"
45 #include "net/net.h"
46 #include "sysemu/device_tree.h"
47 #include "sysemu/numa.h"
48 #include "sysemu/runstate.h"
49 #include "sysemu/tpm.h"
50 #include "sysemu/kvm.h"
51 #include "sysemu/hvf.h"
52 #include "hw/loader.h"
53 #include "qapi/error.h"
54 #include "qemu/bitops.h"
55 #include "qemu/error-report.h"
56 #include "qemu/module.h"
57 #include "hw/pci-host/gpex.h"
58 #include "hw/virtio/virtio-pci.h"
59 #include "hw/core/sysbus-fdt.h"
60 #include "hw/platform-bus.h"
61 #include "hw/qdev-properties.h"
62 #include "hw/arm/fdt.h"
63 #include "hw/intc/arm_gic.h"
64 #include "hw/intc/arm_gicv3_common.h"
65 #include "hw/irq.h"
66 #include "kvm_arm.h"
67 #include "hw/firmware/smbios.h"
68 #include "qapi/visitor.h"
69 #include "qapi/qapi-visit-common.h"
70 #include "standard-headers/linux/input.h"
71 #include "hw/arm/smmuv3.h"
72 #include "hw/acpi/acpi.h"
73 #include "target/arm/internals.h"
74 #include "hw/mem/memory-device.h"
75 #include "hw/mem/pc-dimm.h"
76 #include "hw/mem/nvdimm.h"
77 #include "hw/acpi/generic_event_device.h"
78 #include "hw/virtio/virtio-mem-pci.h"
79 #include "hw/virtio/virtio-iommu.h"
80 #include "hw/char/pl011.h"
81 #include "qemu/guest-random.h"
82 
83 #define DEFINE_VIRT_MACHINE_LATEST(major, minor, latest) \
84     static void virt_##major##_##minor##_class_init(ObjectClass *oc, \
85                                                     void *data) \
86     { \
87         MachineClass *mc = MACHINE_CLASS(oc); \
88         virt_machine_##major##_##minor##_options(mc); \
89         mc->desc = "QEMU " # major "." # minor " ARM Virtual Machine"; \
90         if (latest) { \
91             mc->alias = "virt"; \
92         } \
93     } \
94     static const TypeInfo machvirt_##major##_##minor##_info = { \
95         .name = MACHINE_TYPE_NAME("virt-" # major "." # minor), \
96         .parent = TYPE_VIRT_MACHINE, \
97         .class_init = virt_##major##_##minor##_class_init, \
98     }; \
99     static void machvirt_machine_##major##_##minor##_init(void) \
100     { \
101         type_register_static(&machvirt_##major##_##minor##_info); \
102     } \
103     type_init(machvirt_machine_##major##_##minor##_init);
104 
105 #define DEFINE_VIRT_MACHINE_AS_LATEST(major, minor) \
106     DEFINE_VIRT_MACHINE_LATEST(major, minor, true)
107 #define DEFINE_VIRT_MACHINE(major, minor) \
108     DEFINE_VIRT_MACHINE_LATEST(major, minor, false)
109 
110 
111 /* Number of external interrupt lines to configure the GIC with */
112 #define NUM_IRQS 256
113 
114 #define PLATFORM_BUS_NUM_IRQS 64
115 
116 /* Legacy RAM limit in GB (< version 4.0) */
117 #define LEGACY_RAMLIMIT_GB 255
118 #define LEGACY_RAMLIMIT_BYTES (LEGACY_RAMLIMIT_GB * GiB)
119 
120 /* Addresses and sizes of our components.
121  * 0..128MB is space for a flash device so we can run bootrom code such as UEFI.
122  * 128MB..256MB is used for miscellaneous device I/O.
123  * 256MB..1GB is reserved for possible future PCI support (ie where the
124  * PCI memory window will go if we add a PCI host controller).
125  * 1GB and up is RAM (which may happily spill over into the
126  * high memory region beyond 4GB).
127  * This represents a compromise between how much RAM can be given to
128  * a 32 bit VM and leaving space for expansion and in particular for PCI.
129  * Note that devices should generally be placed at multiples of 0x10000,
130  * to accommodate guests using 64K pages.
131  */
132 static const MemMapEntry base_memmap[] = {
133     /* Space up to 0x8000000 is reserved for a boot ROM */
134     [VIRT_FLASH] =              {          0, 0x08000000 },
135     [VIRT_CPUPERIPHS] =         { 0x08000000, 0x00020000 },
136     /* GIC distributor and CPU interfaces sit inside the CPU peripheral space */
137     [VIRT_GIC_DIST] =           { 0x08000000, 0x00010000 },
138     [VIRT_GIC_CPU] =            { 0x08010000, 0x00010000 },
139     [VIRT_GIC_V2M] =            { 0x08020000, 0x00001000 },
140     [VIRT_GIC_HYP] =            { 0x08030000, 0x00010000 },
141     [VIRT_GIC_VCPU] =           { 0x08040000, 0x00010000 },
142     /* The space in between here is reserved for GICv3 CPU/vCPU/HYP */
143     [VIRT_GIC_ITS] =            { 0x08080000, 0x00020000 },
144     /* This redistributor space allows up to 2*64kB*123 CPUs */
145     [VIRT_GIC_REDIST] =         { 0x080A0000, 0x00F60000 },
146     [VIRT_UART] =               { 0x09000000, 0x00001000 },
147     [VIRT_RTC] =                { 0x09010000, 0x00001000 },
148     [VIRT_FW_CFG] =             { 0x09020000, 0x00000018 },
149     [VIRT_GPIO] =               { 0x09030000, 0x00001000 },
150     [VIRT_SECURE_UART] =        { 0x09040000, 0x00001000 },
151     [VIRT_SMMU] =               { 0x09050000, 0x00020000 },
152     [VIRT_PCDIMM_ACPI] =        { 0x09070000, MEMORY_HOTPLUG_IO_LEN },
153     [VIRT_ACPI_GED] =           { 0x09080000, ACPI_GED_EVT_SEL_LEN },
154     [VIRT_NVDIMM_ACPI] =        { 0x09090000, NVDIMM_ACPI_IO_LEN},
155     [VIRT_PVTIME] =             { 0x090a0000, 0x00010000 },
156     [VIRT_SECURE_GPIO] =        { 0x090b0000, 0x00001000 },
157     [VIRT_MMIO] =               { 0x0a000000, 0x00000200 },
158     /* ...repeating for a total of NUM_VIRTIO_TRANSPORTS, each of that size */
159     [VIRT_PLATFORM_BUS] =       { 0x0c000000, 0x02000000 },
160     [VIRT_SECURE_MEM] =         { 0x0e000000, 0x01000000 },
161     [VIRT_PCIE_MMIO] =          { 0x10000000, 0x2eff0000 },
162     [VIRT_PCIE_PIO] =           { 0x3eff0000, 0x00010000 },
163     [VIRT_PCIE_ECAM] =          { 0x3f000000, 0x01000000 },
164     /* Actual RAM size depends on initial RAM and device memory settings */
165     [VIRT_MEM] =                { GiB, LEGACY_RAMLIMIT_BYTES },
166 };
167 
168 /*
169  * Highmem IO Regions: This memory map is floating, located after the RAM.
170  * Each MemMapEntry base (GPA) will be dynamically computed, depending on the
171  * top of the RAM, so that its base get the same alignment as the size,
172  * ie. a 512GiB entry will be aligned on a 512GiB boundary. If there is
173  * less than 256GiB of RAM, the floating area starts at the 256GiB mark.
174  * Note the extended_memmap is sized so that it eventually also includes the
175  * base_memmap entries (VIRT_HIGH_GIC_REDIST2 index is greater than the last
176  * index of base_memmap).
177  *
178  * The memory map for these Highmem IO Regions can be in legacy or compact
179  * layout, depending on 'compact-highmem' property. With legacy layout, the
180  * PA space for one specific region is always reserved, even if the region
181  * has been disabled or doesn't fit into the PA space. However, the PA space
182  * for the region won't be reserved in these circumstances with compact layout.
183  */
184 static MemMapEntry extended_memmap[] = {
185     /* Additional 64 MB redist region (can contain up to 512 redistributors) */
186     [VIRT_HIGH_GIC_REDIST2] =   { 0x0, 64 * MiB },
187     [VIRT_HIGH_PCIE_ECAM] =     { 0x0, 256 * MiB },
188     /* Second PCIe window */
189     [VIRT_HIGH_PCIE_MMIO] =     { 0x0, 512 * GiB },
190 };
191 
192 static const int a15irqmap[] = {
193     [VIRT_UART] = 1,
194     [VIRT_RTC] = 2,
195     [VIRT_PCIE] = 3, /* ... to 6 */
196     [VIRT_GPIO] = 7,
197     [VIRT_SECURE_UART] = 8,
198     [VIRT_ACPI_GED] = 9,
199     [VIRT_MMIO] = 16, /* ...to 16 + NUM_VIRTIO_TRANSPORTS - 1 */
200     [VIRT_GIC_V2M] = 48, /* ...to 48 + NUM_GICV2M_SPIS - 1 */
201     [VIRT_SMMU] = 74,    /* ...to 74 + NUM_SMMU_IRQS - 1 */
202     [VIRT_PLATFORM_BUS] = 112, /* ...to 112 + PLATFORM_BUS_NUM_IRQS -1 */
203 };
204 
205 static const char *valid_cpus[] = {
206     ARM_CPU_TYPE_NAME("cortex-a7"),
207     ARM_CPU_TYPE_NAME("cortex-a15"),
208     ARM_CPU_TYPE_NAME("cortex-a35"),
209     ARM_CPU_TYPE_NAME("cortex-a53"),
210     ARM_CPU_TYPE_NAME("cortex-a55"),
211     ARM_CPU_TYPE_NAME("cortex-a57"),
212     ARM_CPU_TYPE_NAME("cortex-a72"),
213     ARM_CPU_TYPE_NAME("cortex-a76"),
214     ARM_CPU_TYPE_NAME("a64fx"),
215     ARM_CPU_TYPE_NAME("neoverse-n1"),
216     ARM_CPU_TYPE_NAME("host"),
217     ARM_CPU_TYPE_NAME("max"),
218 };
219 
220 static bool cpu_type_valid(const char *cpu)
221 {
222     int i;
223 
224     for (i = 0; i < ARRAY_SIZE(valid_cpus); i++) {
225         if (strcmp(cpu, valid_cpus[i]) == 0) {
226             return true;
227         }
228     }
229     return false;
230 }
231 
232 static void create_randomness(MachineState *ms, const char *node)
233 {
234     struct {
235         uint64_t kaslr;
236         uint8_t rng[32];
237     } seed;
238 
239     if (qemu_guest_getrandom(&seed, sizeof(seed), NULL)) {
240         return;
241     }
242     qemu_fdt_setprop_u64(ms->fdt, node, "kaslr-seed", seed.kaslr);
243     qemu_fdt_setprop(ms->fdt, node, "rng-seed", seed.rng, sizeof(seed.rng));
244 }
245 
246 static void create_fdt(VirtMachineState *vms)
247 {
248     MachineState *ms = MACHINE(vms);
249     int nb_numa_nodes = ms->numa_state->num_nodes;
250     void *fdt = create_device_tree(&vms->fdt_size);
251 
252     if (!fdt) {
253         error_report("create_device_tree() failed");
254         exit(1);
255     }
256 
257     ms->fdt = fdt;
258 
259     /* Header */
260     qemu_fdt_setprop_string(fdt, "/", "compatible", "linux,dummy-virt");
261     qemu_fdt_setprop_cell(fdt, "/", "#address-cells", 0x2);
262     qemu_fdt_setprop_cell(fdt, "/", "#size-cells", 0x2);
263     qemu_fdt_setprop_string(fdt, "/", "model", "linux,dummy-virt");
264 
265     /* /chosen must exist for load_dtb to fill in necessary properties later */
266     qemu_fdt_add_subnode(fdt, "/chosen");
267     if (vms->dtb_randomness) {
268         create_randomness(ms, "/chosen");
269     }
270 
271     if (vms->secure) {
272         qemu_fdt_add_subnode(fdt, "/secure-chosen");
273         if (vms->dtb_randomness) {
274             create_randomness(ms, "/secure-chosen");
275         }
276     }
277 
278     /* Clock node, for the benefit of the UART. The kernel device tree
279      * binding documentation claims the PL011 node clock properties are
280      * optional but in practice if you omit them the kernel refuses to
281      * probe for the device.
282      */
283     vms->clock_phandle = qemu_fdt_alloc_phandle(fdt);
284     qemu_fdt_add_subnode(fdt, "/apb-pclk");
285     qemu_fdt_setprop_string(fdt, "/apb-pclk", "compatible", "fixed-clock");
286     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "#clock-cells", 0x0);
287     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "clock-frequency", 24000000);
288     qemu_fdt_setprop_string(fdt, "/apb-pclk", "clock-output-names",
289                                 "clk24mhz");
290     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "phandle", vms->clock_phandle);
291 
292     if (nb_numa_nodes > 0 && ms->numa_state->have_numa_distance) {
293         int size = nb_numa_nodes * nb_numa_nodes * 3 * sizeof(uint32_t);
294         uint32_t *matrix = g_malloc0(size);
295         int idx, i, j;
296 
297         for (i = 0; i < nb_numa_nodes; i++) {
298             for (j = 0; j < nb_numa_nodes; j++) {
299                 idx = (i * nb_numa_nodes + j) * 3;
300                 matrix[idx + 0] = cpu_to_be32(i);
301                 matrix[idx + 1] = cpu_to_be32(j);
302                 matrix[idx + 2] =
303                     cpu_to_be32(ms->numa_state->nodes[i].distance[j]);
304             }
305         }
306 
307         qemu_fdt_add_subnode(fdt, "/distance-map");
308         qemu_fdt_setprop_string(fdt, "/distance-map", "compatible",
309                                 "numa-distance-map-v1");
310         qemu_fdt_setprop(fdt, "/distance-map", "distance-matrix",
311                          matrix, size);
312         g_free(matrix);
313     }
314 }
315 
316 static void fdt_add_timer_nodes(const VirtMachineState *vms)
317 {
318     /* On real hardware these interrupts are level-triggered.
319      * On KVM they were edge-triggered before host kernel version 4.4,
320      * and level-triggered afterwards.
321      * On emulated QEMU they are level-triggered.
322      *
323      * Getting the DTB info about them wrong is awkward for some
324      * guest kernels:
325      *  pre-4.8 ignore the DT and leave the interrupt configured
326      *   with whatever the GIC reset value (or the bootloader) left it at
327      *  4.8 before rc6 honour the incorrect data by programming it back
328      *   into the GIC, causing problems
329      *  4.8rc6 and later ignore the DT and always write "level triggered"
330      *   into the GIC
331      *
332      * For backwards-compatibility, virt-2.8 and earlier will continue
333      * to say these are edge-triggered, but later machines will report
334      * the correct information.
335      */
336     ARMCPU *armcpu;
337     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
338     uint32_t irqflags = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
339     MachineState *ms = MACHINE(vms);
340 
341     if (vmc->claim_edge_triggered_timers) {
342         irqflags = GIC_FDT_IRQ_FLAGS_EDGE_LO_HI;
343     }
344 
345     if (vms->gic_version == VIRT_GIC_VERSION_2) {
346         irqflags = deposit32(irqflags, GIC_FDT_IRQ_PPI_CPU_START,
347                              GIC_FDT_IRQ_PPI_CPU_WIDTH,
348                              (1 << MACHINE(vms)->smp.cpus) - 1);
349     }
350 
351     qemu_fdt_add_subnode(ms->fdt, "/timer");
352 
353     armcpu = ARM_CPU(qemu_get_cpu(0));
354     if (arm_feature(&armcpu->env, ARM_FEATURE_V8)) {
355         const char compat[] = "arm,armv8-timer\0arm,armv7-timer";
356         qemu_fdt_setprop(ms->fdt, "/timer", "compatible",
357                          compat, sizeof(compat));
358     } else {
359         qemu_fdt_setprop_string(ms->fdt, "/timer", "compatible",
360                                 "arm,armv7-timer");
361     }
362     qemu_fdt_setprop(ms->fdt, "/timer", "always-on", NULL, 0);
363     qemu_fdt_setprop_cells(ms->fdt, "/timer", "interrupts",
364                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_S_EL1_IRQ, irqflags,
365                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_NS_EL1_IRQ, irqflags,
366                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_VIRT_IRQ, irqflags,
367                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_NS_EL2_IRQ, irqflags);
368 }
369 
370 static void fdt_add_cpu_nodes(const VirtMachineState *vms)
371 {
372     int cpu;
373     int addr_cells = 1;
374     const MachineState *ms = MACHINE(vms);
375     const VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
376     int smp_cpus = ms->smp.cpus;
377 
378     /*
379      * See Linux Documentation/devicetree/bindings/arm/cpus.yaml
380      * On ARM v8 64-bit systems value should be set to 2,
381      * that corresponds to the MPIDR_EL1 register size.
382      * If MPIDR_EL1[63:32] value is equal to 0 on all CPUs
383      * in the system, #address-cells can be set to 1, since
384      * MPIDR_EL1[63:32] bits are not used for CPUs
385      * identification.
386      *
387      * Here we actually don't know whether our system is 32- or 64-bit one.
388      * The simplest way to go is to examine affinity IDs of all our CPUs. If
389      * at least one of them has Aff3 populated, we set #address-cells to 2.
390      */
391     for (cpu = 0; cpu < smp_cpus; cpu++) {
392         ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(cpu));
393 
394         if (armcpu->mp_affinity & ARM_AFF3_MASK) {
395             addr_cells = 2;
396             break;
397         }
398     }
399 
400     qemu_fdt_add_subnode(ms->fdt, "/cpus");
401     qemu_fdt_setprop_cell(ms->fdt, "/cpus", "#address-cells", addr_cells);
402     qemu_fdt_setprop_cell(ms->fdt, "/cpus", "#size-cells", 0x0);
403 
404     for (cpu = smp_cpus - 1; cpu >= 0; cpu--) {
405         char *nodename = g_strdup_printf("/cpus/cpu@%d", cpu);
406         ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(cpu));
407         CPUState *cs = CPU(armcpu);
408 
409         qemu_fdt_add_subnode(ms->fdt, nodename);
410         qemu_fdt_setprop_string(ms->fdt, nodename, "device_type", "cpu");
411         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
412                                     armcpu->dtb_compatible);
413 
414         if (vms->psci_conduit != QEMU_PSCI_CONDUIT_DISABLED && smp_cpus > 1) {
415             qemu_fdt_setprop_string(ms->fdt, nodename,
416                                         "enable-method", "psci");
417         }
418 
419         if (addr_cells == 2) {
420             qemu_fdt_setprop_u64(ms->fdt, nodename, "reg",
421                                  armcpu->mp_affinity);
422         } else {
423             qemu_fdt_setprop_cell(ms->fdt, nodename, "reg",
424                                   armcpu->mp_affinity);
425         }
426 
427         if (ms->possible_cpus->cpus[cs->cpu_index].props.has_node_id) {
428             qemu_fdt_setprop_cell(ms->fdt, nodename, "numa-node-id",
429                 ms->possible_cpus->cpus[cs->cpu_index].props.node_id);
430         }
431 
432         if (!vmc->no_cpu_topology) {
433             qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle",
434                                   qemu_fdt_alloc_phandle(ms->fdt));
435         }
436 
437         g_free(nodename);
438     }
439 
440     if (!vmc->no_cpu_topology) {
441         /*
442          * Add vCPU topology description through fdt node cpu-map.
443          *
444          * See Linux Documentation/devicetree/bindings/cpu/cpu-topology.txt
445          * In a SMP system, the hierarchy of CPUs can be defined through
446          * four entities that are used to describe the layout of CPUs in
447          * the system: socket/cluster/core/thread.
448          *
449          * A socket node represents the boundary of system physical package
450          * and its child nodes must be one or more cluster nodes. A system
451          * can contain several layers of clustering within a single physical
452          * package and cluster nodes can be contained in parent cluster nodes.
453          *
454          * Note: currently we only support one layer of clustering within
455          * each physical package.
456          */
457         qemu_fdt_add_subnode(ms->fdt, "/cpus/cpu-map");
458 
459         for (cpu = smp_cpus - 1; cpu >= 0; cpu--) {
460             char *cpu_path = g_strdup_printf("/cpus/cpu@%d", cpu);
461             char *map_path;
462 
463             if (ms->smp.threads > 1) {
464                 map_path = g_strdup_printf(
465                     "/cpus/cpu-map/socket%d/cluster%d/core%d/thread%d",
466                     cpu / (ms->smp.clusters * ms->smp.cores * ms->smp.threads),
467                     (cpu / (ms->smp.cores * ms->smp.threads)) % ms->smp.clusters,
468                     (cpu / ms->smp.threads) % ms->smp.cores,
469                     cpu % ms->smp.threads);
470             } else {
471                 map_path = g_strdup_printf(
472                     "/cpus/cpu-map/socket%d/cluster%d/core%d",
473                     cpu / (ms->smp.clusters * ms->smp.cores),
474                     (cpu / ms->smp.cores) % ms->smp.clusters,
475                     cpu % ms->smp.cores);
476             }
477             qemu_fdt_add_path(ms->fdt, map_path);
478             qemu_fdt_setprop_phandle(ms->fdt, map_path, "cpu", cpu_path);
479 
480             g_free(map_path);
481             g_free(cpu_path);
482         }
483     }
484 }
485 
486 static void fdt_add_its_gic_node(VirtMachineState *vms)
487 {
488     char *nodename;
489     MachineState *ms = MACHINE(vms);
490 
491     vms->msi_phandle = qemu_fdt_alloc_phandle(ms->fdt);
492     nodename = g_strdup_printf("/intc/its@%" PRIx64,
493                                vms->memmap[VIRT_GIC_ITS].base);
494     qemu_fdt_add_subnode(ms->fdt, nodename);
495     qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
496                             "arm,gic-v3-its");
497     qemu_fdt_setprop(ms->fdt, nodename, "msi-controller", NULL, 0);
498     qemu_fdt_setprop_cell(ms->fdt, nodename, "#msi-cells", 1);
499     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
500                                  2, vms->memmap[VIRT_GIC_ITS].base,
501                                  2, vms->memmap[VIRT_GIC_ITS].size);
502     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", vms->msi_phandle);
503     g_free(nodename);
504 }
505 
506 static void fdt_add_v2m_gic_node(VirtMachineState *vms)
507 {
508     MachineState *ms = MACHINE(vms);
509     char *nodename;
510 
511     nodename = g_strdup_printf("/intc/v2m@%" PRIx64,
512                                vms->memmap[VIRT_GIC_V2M].base);
513     vms->msi_phandle = qemu_fdt_alloc_phandle(ms->fdt);
514     qemu_fdt_add_subnode(ms->fdt, nodename);
515     qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
516                             "arm,gic-v2m-frame");
517     qemu_fdt_setprop(ms->fdt, nodename, "msi-controller", NULL, 0);
518     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
519                                  2, vms->memmap[VIRT_GIC_V2M].base,
520                                  2, vms->memmap[VIRT_GIC_V2M].size);
521     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", vms->msi_phandle);
522     g_free(nodename);
523 }
524 
525 static void fdt_add_gic_node(VirtMachineState *vms)
526 {
527     MachineState *ms = MACHINE(vms);
528     char *nodename;
529 
530     vms->gic_phandle = qemu_fdt_alloc_phandle(ms->fdt);
531     qemu_fdt_setprop_cell(ms->fdt, "/", "interrupt-parent", vms->gic_phandle);
532 
533     nodename = g_strdup_printf("/intc@%" PRIx64,
534                                vms->memmap[VIRT_GIC_DIST].base);
535     qemu_fdt_add_subnode(ms->fdt, nodename);
536     qemu_fdt_setprop_cell(ms->fdt, nodename, "#interrupt-cells", 3);
537     qemu_fdt_setprop(ms->fdt, nodename, "interrupt-controller", NULL, 0);
538     qemu_fdt_setprop_cell(ms->fdt, nodename, "#address-cells", 0x2);
539     qemu_fdt_setprop_cell(ms->fdt, nodename, "#size-cells", 0x2);
540     qemu_fdt_setprop(ms->fdt, nodename, "ranges", NULL, 0);
541     if (vms->gic_version != VIRT_GIC_VERSION_2) {
542         int nb_redist_regions = virt_gicv3_redist_region_count(vms);
543 
544         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
545                                 "arm,gic-v3");
546 
547         qemu_fdt_setprop_cell(ms->fdt, nodename,
548                               "#redistributor-regions", nb_redist_regions);
549 
550         if (nb_redist_regions == 1) {
551             qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
552                                          2, vms->memmap[VIRT_GIC_DIST].base,
553                                          2, vms->memmap[VIRT_GIC_DIST].size,
554                                          2, vms->memmap[VIRT_GIC_REDIST].base,
555                                          2, vms->memmap[VIRT_GIC_REDIST].size);
556         } else {
557             qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
558                                  2, vms->memmap[VIRT_GIC_DIST].base,
559                                  2, vms->memmap[VIRT_GIC_DIST].size,
560                                  2, vms->memmap[VIRT_GIC_REDIST].base,
561                                  2, vms->memmap[VIRT_GIC_REDIST].size,
562                                  2, vms->memmap[VIRT_HIGH_GIC_REDIST2].base,
563                                  2, vms->memmap[VIRT_HIGH_GIC_REDIST2].size);
564         }
565 
566         if (vms->virt) {
567             qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
568                                    GIC_FDT_IRQ_TYPE_PPI, ARCH_GIC_MAINT_IRQ,
569                                    GIC_FDT_IRQ_FLAGS_LEVEL_HI);
570         }
571     } else {
572         /* 'cortex-a15-gic' means 'GIC v2' */
573         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
574                                 "arm,cortex-a15-gic");
575         if (!vms->virt) {
576             qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
577                                          2, vms->memmap[VIRT_GIC_DIST].base,
578                                          2, vms->memmap[VIRT_GIC_DIST].size,
579                                          2, vms->memmap[VIRT_GIC_CPU].base,
580                                          2, vms->memmap[VIRT_GIC_CPU].size);
581         } else {
582             qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
583                                          2, vms->memmap[VIRT_GIC_DIST].base,
584                                          2, vms->memmap[VIRT_GIC_DIST].size,
585                                          2, vms->memmap[VIRT_GIC_CPU].base,
586                                          2, vms->memmap[VIRT_GIC_CPU].size,
587                                          2, vms->memmap[VIRT_GIC_HYP].base,
588                                          2, vms->memmap[VIRT_GIC_HYP].size,
589                                          2, vms->memmap[VIRT_GIC_VCPU].base,
590                                          2, vms->memmap[VIRT_GIC_VCPU].size);
591             qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
592                                    GIC_FDT_IRQ_TYPE_PPI, ARCH_GIC_MAINT_IRQ,
593                                    GIC_FDT_IRQ_FLAGS_LEVEL_HI);
594         }
595     }
596 
597     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", vms->gic_phandle);
598     g_free(nodename);
599 }
600 
601 static void fdt_add_pmu_nodes(const VirtMachineState *vms)
602 {
603     ARMCPU *armcpu = ARM_CPU(first_cpu);
604     uint32_t irqflags = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
605     MachineState *ms = MACHINE(vms);
606 
607     if (!arm_feature(&armcpu->env, ARM_FEATURE_PMU)) {
608         assert(!object_property_get_bool(OBJECT(armcpu), "pmu", NULL));
609         return;
610     }
611 
612     if (vms->gic_version == VIRT_GIC_VERSION_2) {
613         irqflags = deposit32(irqflags, GIC_FDT_IRQ_PPI_CPU_START,
614                              GIC_FDT_IRQ_PPI_CPU_WIDTH,
615                              (1 << MACHINE(vms)->smp.cpus) - 1);
616     }
617 
618     qemu_fdt_add_subnode(ms->fdt, "/pmu");
619     if (arm_feature(&armcpu->env, ARM_FEATURE_V8)) {
620         const char compat[] = "arm,armv8-pmuv3";
621         qemu_fdt_setprop(ms->fdt, "/pmu", "compatible",
622                          compat, sizeof(compat));
623         qemu_fdt_setprop_cells(ms->fdt, "/pmu", "interrupts",
624                                GIC_FDT_IRQ_TYPE_PPI, VIRTUAL_PMU_IRQ, irqflags);
625     }
626 }
627 
628 static inline DeviceState *create_acpi_ged(VirtMachineState *vms)
629 {
630     DeviceState *dev;
631     MachineState *ms = MACHINE(vms);
632     int irq = vms->irqmap[VIRT_ACPI_GED];
633     uint32_t event = ACPI_GED_PWR_DOWN_EVT;
634 
635     if (ms->ram_slots) {
636         event |= ACPI_GED_MEM_HOTPLUG_EVT;
637     }
638 
639     if (ms->nvdimms_state->is_enabled) {
640         event |= ACPI_GED_NVDIMM_HOTPLUG_EVT;
641     }
642 
643     dev = qdev_new(TYPE_ACPI_GED);
644     qdev_prop_set_uint32(dev, "ged-event", event);
645 
646     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_ACPI_GED].base);
647     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 1, vms->memmap[VIRT_PCDIMM_ACPI].base);
648     sysbus_connect_irq(SYS_BUS_DEVICE(dev), 0, qdev_get_gpio_in(vms->gic, irq));
649 
650     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
651 
652     return dev;
653 }
654 
655 static void create_its(VirtMachineState *vms)
656 {
657     const char *itsclass = its_class_name();
658     DeviceState *dev;
659 
660     if (!strcmp(itsclass, "arm-gicv3-its")) {
661         if (!vms->tcg_its) {
662             itsclass = NULL;
663         }
664     }
665 
666     if (!itsclass) {
667         /* Do nothing if not supported */
668         return;
669     }
670 
671     dev = qdev_new(itsclass);
672 
673     object_property_set_link(OBJECT(dev), "parent-gicv3", OBJECT(vms->gic),
674                              &error_abort);
675     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
676     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_GIC_ITS].base);
677 
678     fdt_add_its_gic_node(vms);
679     vms->msi_controller = VIRT_MSI_CTRL_ITS;
680 }
681 
682 static void create_v2m(VirtMachineState *vms)
683 {
684     int i;
685     int irq = vms->irqmap[VIRT_GIC_V2M];
686     DeviceState *dev;
687 
688     dev = qdev_new("arm-gicv2m");
689     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_GIC_V2M].base);
690     qdev_prop_set_uint32(dev, "base-spi", irq);
691     qdev_prop_set_uint32(dev, "num-spi", NUM_GICV2M_SPIS);
692     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
693 
694     for (i = 0; i < NUM_GICV2M_SPIS; i++) {
695         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
696                            qdev_get_gpio_in(vms->gic, irq + i));
697     }
698 
699     fdt_add_v2m_gic_node(vms);
700     vms->msi_controller = VIRT_MSI_CTRL_GICV2M;
701 }
702 
703 static void create_gic(VirtMachineState *vms, MemoryRegion *mem)
704 {
705     MachineState *ms = MACHINE(vms);
706     /* We create a standalone GIC */
707     SysBusDevice *gicbusdev;
708     const char *gictype;
709     int i;
710     unsigned int smp_cpus = ms->smp.cpus;
711     uint32_t nb_redist_regions = 0;
712     int revision;
713 
714     if (vms->gic_version == VIRT_GIC_VERSION_2) {
715         gictype = gic_class_name();
716     } else {
717         gictype = gicv3_class_name();
718     }
719 
720     switch (vms->gic_version) {
721     case VIRT_GIC_VERSION_2:
722         revision = 2;
723         break;
724     case VIRT_GIC_VERSION_3:
725         revision = 3;
726         break;
727     case VIRT_GIC_VERSION_4:
728         revision = 4;
729         break;
730     default:
731         g_assert_not_reached();
732     }
733     vms->gic = qdev_new(gictype);
734     qdev_prop_set_uint32(vms->gic, "revision", revision);
735     qdev_prop_set_uint32(vms->gic, "num-cpu", smp_cpus);
736     /* Note that the num-irq property counts both internal and external
737      * interrupts; there are always 32 of the former (mandated by GIC spec).
738      */
739     qdev_prop_set_uint32(vms->gic, "num-irq", NUM_IRQS + 32);
740     if (!kvm_irqchip_in_kernel()) {
741         qdev_prop_set_bit(vms->gic, "has-security-extensions", vms->secure);
742     }
743 
744     if (vms->gic_version != VIRT_GIC_VERSION_2) {
745         uint32_t redist0_capacity = virt_redist_capacity(vms, VIRT_GIC_REDIST);
746         uint32_t redist0_count = MIN(smp_cpus, redist0_capacity);
747 
748         nb_redist_regions = virt_gicv3_redist_region_count(vms);
749 
750         qdev_prop_set_uint32(vms->gic, "len-redist-region-count",
751                              nb_redist_regions);
752         qdev_prop_set_uint32(vms->gic, "redist-region-count[0]", redist0_count);
753 
754         if (!kvm_irqchip_in_kernel()) {
755             if (vms->tcg_its) {
756                 object_property_set_link(OBJECT(vms->gic), "sysmem",
757                                          OBJECT(mem), &error_fatal);
758                 qdev_prop_set_bit(vms->gic, "has-lpi", true);
759             }
760         }
761 
762         if (nb_redist_regions == 2) {
763             uint32_t redist1_capacity =
764                 virt_redist_capacity(vms, VIRT_HIGH_GIC_REDIST2);
765 
766             qdev_prop_set_uint32(vms->gic, "redist-region-count[1]",
767                 MIN(smp_cpus - redist0_count, redist1_capacity));
768         }
769     } else {
770         if (!kvm_irqchip_in_kernel()) {
771             qdev_prop_set_bit(vms->gic, "has-virtualization-extensions",
772                               vms->virt);
773         }
774     }
775     gicbusdev = SYS_BUS_DEVICE(vms->gic);
776     sysbus_realize_and_unref(gicbusdev, &error_fatal);
777     sysbus_mmio_map(gicbusdev, 0, vms->memmap[VIRT_GIC_DIST].base);
778     if (vms->gic_version != VIRT_GIC_VERSION_2) {
779         sysbus_mmio_map(gicbusdev, 1, vms->memmap[VIRT_GIC_REDIST].base);
780         if (nb_redist_regions == 2) {
781             sysbus_mmio_map(gicbusdev, 2,
782                             vms->memmap[VIRT_HIGH_GIC_REDIST2].base);
783         }
784     } else {
785         sysbus_mmio_map(gicbusdev, 1, vms->memmap[VIRT_GIC_CPU].base);
786         if (vms->virt) {
787             sysbus_mmio_map(gicbusdev, 2, vms->memmap[VIRT_GIC_HYP].base);
788             sysbus_mmio_map(gicbusdev, 3, vms->memmap[VIRT_GIC_VCPU].base);
789         }
790     }
791 
792     /* Wire the outputs from each CPU's generic timer and the GICv3
793      * maintenance interrupt signal to the appropriate GIC PPI inputs,
794      * and the GIC's IRQ/FIQ/VIRQ/VFIQ interrupt outputs to the CPU's inputs.
795      */
796     for (i = 0; i < smp_cpus; i++) {
797         DeviceState *cpudev = DEVICE(qemu_get_cpu(i));
798         int ppibase = NUM_IRQS + i * GIC_INTERNAL + GIC_NR_SGIS;
799         int irq;
800         /* Mapping from the output timer irq lines from the CPU to the
801          * GIC PPI inputs we use for the virt board.
802          */
803         const int timer_irq[] = {
804             [GTIMER_PHYS] = ARCH_TIMER_NS_EL1_IRQ,
805             [GTIMER_VIRT] = ARCH_TIMER_VIRT_IRQ,
806             [GTIMER_HYP]  = ARCH_TIMER_NS_EL2_IRQ,
807             [GTIMER_SEC]  = ARCH_TIMER_S_EL1_IRQ,
808         };
809 
810         for (irq = 0; irq < ARRAY_SIZE(timer_irq); irq++) {
811             qdev_connect_gpio_out(cpudev, irq,
812                                   qdev_get_gpio_in(vms->gic,
813                                                    ppibase + timer_irq[irq]));
814         }
815 
816         if (vms->gic_version != VIRT_GIC_VERSION_2) {
817             qemu_irq irq = qdev_get_gpio_in(vms->gic,
818                                             ppibase + ARCH_GIC_MAINT_IRQ);
819             qdev_connect_gpio_out_named(cpudev, "gicv3-maintenance-interrupt",
820                                         0, irq);
821         } else if (vms->virt) {
822             qemu_irq irq = qdev_get_gpio_in(vms->gic,
823                                             ppibase + ARCH_GIC_MAINT_IRQ);
824             sysbus_connect_irq(gicbusdev, i + 4 * smp_cpus, irq);
825         }
826 
827         qdev_connect_gpio_out_named(cpudev, "pmu-interrupt", 0,
828                                     qdev_get_gpio_in(vms->gic, ppibase
829                                                      + VIRTUAL_PMU_IRQ));
830 
831         sysbus_connect_irq(gicbusdev, i, qdev_get_gpio_in(cpudev, ARM_CPU_IRQ));
832         sysbus_connect_irq(gicbusdev, i + smp_cpus,
833                            qdev_get_gpio_in(cpudev, ARM_CPU_FIQ));
834         sysbus_connect_irq(gicbusdev, i + 2 * smp_cpus,
835                            qdev_get_gpio_in(cpudev, ARM_CPU_VIRQ));
836         sysbus_connect_irq(gicbusdev, i + 3 * smp_cpus,
837                            qdev_get_gpio_in(cpudev, ARM_CPU_VFIQ));
838     }
839 
840     fdt_add_gic_node(vms);
841 
842     if (vms->gic_version != VIRT_GIC_VERSION_2 && vms->its) {
843         create_its(vms);
844     } else if (vms->gic_version == VIRT_GIC_VERSION_2) {
845         create_v2m(vms);
846     }
847 }
848 
849 static void create_uart(const VirtMachineState *vms, int uart,
850                         MemoryRegion *mem, Chardev *chr)
851 {
852     char *nodename;
853     hwaddr base = vms->memmap[uart].base;
854     hwaddr size = vms->memmap[uart].size;
855     int irq = vms->irqmap[uart];
856     const char compat[] = "arm,pl011\0arm,primecell";
857     const char clocknames[] = "uartclk\0apb_pclk";
858     DeviceState *dev = qdev_new(TYPE_PL011);
859     SysBusDevice *s = SYS_BUS_DEVICE(dev);
860     MachineState *ms = MACHINE(vms);
861 
862     qdev_prop_set_chr(dev, "chardev", chr);
863     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
864     memory_region_add_subregion(mem, base,
865                                 sysbus_mmio_get_region(s, 0));
866     sysbus_connect_irq(s, 0, qdev_get_gpio_in(vms->gic, irq));
867 
868     nodename = g_strdup_printf("/pl011@%" PRIx64, base);
869     qemu_fdt_add_subnode(ms->fdt, nodename);
870     /* Note that we can't use setprop_string because of the embedded NUL */
871     qemu_fdt_setprop(ms->fdt, nodename, "compatible",
872                          compat, sizeof(compat));
873     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
874                                      2, base, 2, size);
875     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
876                                GIC_FDT_IRQ_TYPE_SPI, irq,
877                                GIC_FDT_IRQ_FLAGS_LEVEL_HI);
878     qemu_fdt_setprop_cells(ms->fdt, nodename, "clocks",
879                                vms->clock_phandle, vms->clock_phandle);
880     qemu_fdt_setprop(ms->fdt, nodename, "clock-names",
881                          clocknames, sizeof(clocknames));
882 
883     if (uart == VIRT_UART) {
884         qemu_fdt_setprop_string(ms->fdt, "/chosen", "stdout-path", nodename);
885     } else {
886         /* Mark as not usable by the normal world */
887         qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
888         qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
889 
890         qemu_fdt_setprop_string(ms->fdt, "/secure-chosen", "stdout-path",
891                                 nodename);
892     }
893 
894     g_free(nodename);
895 }
896 
897 static void create_rtc(const VirtMachineState *vms)
898 {
899     char *nodename;
900     hwaddr base = vms->memmap[VIRT_RTC].base;
901     hwaddr size = vms->memmap[VIRT_RTC].size;
902     int irq = vms->irqmap[VIRT_RTC];
903     const char compat[] = "arm,pl031\0arm,primecell";
904     MachineState *ms = MACHINE(vms);
905 
906     sysbus_create_simple("pl031", base, qdev_get_gpio_in(vms->gic, irq));
907 
908     nodename = g_strdup_printf("/pl031@%" PRIx64, base);
909     qemu_fdt_add_subnode(ms->fdt, nodename);
910     qemu_fdt_setprop(ms->fdt, nodename, "compatible", compat, sizeof(compat));
911     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
912                                  2, base, 2, size);
913     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
914                            GIC_FDT_IRQ_TYPE_SPI, irq,
915                            GIC_FDT_IRQ_FLAGS_LEVEL_HI);
916     qemu_fdt_setprop_cell(ms->fdt, nodename, "clocks", vms->clock_phandle);
917     qemu_fdt_setprop_string(ms->fdt, nodename, "clock-names", "apb_pclk");
918     g_free(nodename);
919 }
920 
921 static DeviceState *gpio_key_dev;
922 static void virt_powerdown_req(Notifier *n, void *opaque)
923 {
924     VirtMachineState *s = container_of(n, VirtMachineState, powerdown_notifier);
925 
926     if (s->acpi_dev) {
927         acpi_send_event(s->acpi_dev, ACPI_POWER_DOWN_STATUS);
928     } else {
929         /* use gpio Pin 3 for power button event */
930         qemu_set_irq(qdev_get_gpio_in(gpio_key_dev, 0), 1);
931     }
932 }
933 
934 static void create_gpio_keys(char *fdt, DeviceState *pl061_dev,
935                              uint32_t phandle)
936 {
937     gpio_key_dev = sysbus_create_simple("gpio-key", -1,
938                                         qdev_get_gpio_in(pl061_dev, 3));
939 
940     qemu_fdt_add_subnode(fdt, "/gpio-keys");
941     qemu_fdt_setprop_string(fdt, "/gpio-keys", "compatible", "gpio-keys");
942 
943     qemu_fdt_add_subnode(fdt, "/gpio-keys/poweroff");
944     qemu_fdt_setprop_string(fdt, "/gpio-keys/poweroff",
945                             "label", "GPIO Key Poweroff");
946     qemu_fdt_setprop_cell(fdt, "/gpio-keys/poweroff", "linux,code",
947                           KEY_POWER);
948     qemu_fdt_setprop_cells(fdt, "/gpio-keys/poweroff",
949                            "gpios", phandle, 3, 0);
950 }
951 
952 #define SECURE_GPIO_POWEROFF 0
953 #define SECURE_GPIO_RESET    1
954 
955 static void create_secure_gpio_pwr(char *fdt, DeviceState *pl061_dev,
956                                    uint32_t phandle)
957 {
958     DeviceState *gpio_pwr_dev;
959 
960     /* gpio-pwr */
961     gpio_pwr_dev = sysbus_create_simple("gpio-pwr", -1, NULL);
962 
963     /* connect secure pl061 to gpio-pwr */
964     qdev_connect_gpio_out(pl061_dev, SECURE_GPIO_RESET,
965                           qdev_get_gpio_in_named(gpio_pwr_dev, "reset", 0));
966     qdev_connect_gpio_out(pl061_dev, SECURE_GPIO_POWEROFF,
967                           qdev_get_gpio_in_named(gpio_pwr_dev, "shutdown", 0));
968 
969     qemu_fdt_add_subnode(fdt, "/gpio-poweroff");
970     qemu_fdt_setprop_string(fdt, "/gpio-poweroff", "compatible",
971                             "gpio-poweroff");
972     qemu_fdt_setprop_cells(fdt, "/gpio-poweroff",
973                            "gpios", phandle, SECURE_GPIO_POWEROFF, 0);
974     qemu_fdt_setprop_string(fdt, "/gpio-poweroff", "status", "disabled");
975     qemu_fdt_setprop_string(fdt, "/gpio-poweroff", "secure-status",
976                             "okay");
977 
978     qemu_fdt_add_subnode(fdt, "/gpio-restart");
979     qemu_fdt_setprop_string(fdt, "/gpio-restart", "compatible",
980                             "gpio-restart");
981     qemu_fdt_setprop_cells(fdt, "/gpio-restart",
982                            "gpios", phandle, SECURE_GPIO_RESET, 0);
983     qemu_fdt_setprop_string(fdt, "/gpio-restart", "status", "disabled");
984     qemu_fdt_setprop_string(fdt, "/gpio-restart", "secure-status",
985                             "okay");
986 }
987 
988 static void create_gpio_devices(const VirtMachineState *vms, int gpio,
989                                 MemoryRegion *mem)
990 {
991     char *nodename;
992     DeviceState *pl061_dev;
993     hwaddr base = vms->memmap[gpio].base;
994     hwaddr size = vms->memmap[gpio].size;
995     int irq = vms->irqmap[gpio];
996     const char compat[] = "arm,pl061\0arm,primecell";
997     SysBusDevice *s;
998     MachineState *ms = MACHINE(vms);
999 
1000     pl061_dev = qdev_new("pl061");
1001     /* Pull lines down to 0 if not driven by the PL061 */
1002     qdev_prop_set_uint32(pl061_dev, "pullups", 0);
1003     qdev_prop_set_uint32(pl061_dev, "pulldowns", 0xff);
1004     s = SYS_BUS_DEVICE(pl061_dev);
1005     sysbus_realize_and_unref(s, &error_fatal);
1006     memory_region_add_subregion(mem, base, sysbus_mmio_get_region(s, 0));
1007     sysbus_connect_irq(s, 0, qdev_get_gpio_in(vms->gic, irq));
1008 
1009     uint32_t phandle = qemu_fdt_alloc_phandle(ms->fdt);
1010     nodename = g_strdup_printf("/pl061@%" PRIx64, base);
1011     qemu_fdt_add_subnode(ms->fdt, nodename);
1012     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1013                                  2, base, 2, size);
1014     qemu_fdt_setprop(ms->fdt, nodename, "compatible", compat, sizeof(compat));
1015     qemu_fdt_setprop_cell(ms->fdt, nodename, "#gpio-cells", 2);
1016     qemu_fdt_setprop(ms->fdt, nodename, "gpio-controller", NULL, 0);
1017     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
1018                            GIC_FDT_IRQ_TYPE_SPI, irq,
1019                            GIC_FDT_IRQ_FLAGS_LEVEL_HI);
1020     qemu_fdt_setprop_cell(ms->fdt, nodename, "clocks", vms->clock_phandle);
1021     qemu_fdt_setprop_string(ms->fdt, nodename, "clock-names", "apb_pclk");
1022     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", phandle);
1023 
1024     if (gpio != VIRT_GPIO) {
1025         /* Mark as not usable by the normal world */
1026         qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
1027         qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
1028     }
1029     g_free(nodename);
1030 
1031     /* Child gpio devices */
1032     if (gpio == VIRT_GPIO) {
1033         create_gpio_keys(ms->fdt, pl061_dev, phandle);
1034     } else {
1035         create_secure_gpio_pwr(ms->fdt, pl061_dev, phandle);
1036     }
1037 }
1038 
1039 static void create_virtio_devices(const VirtMachineState *vms)
1040 {
1041     int i;
1042     hwaddr size = vms->memmap[VIRT_MMIO].size;
1043     MachineState *ms = MACHINE(vms);
1044 
1045     /* We create the transports in forwards order. Since qbus_realize()
1046      * prepends (not appends) new child buses, the incrementing loop below will
1047      * create a list of virtio-mmio buses with decreasing base addresses.
1048      *
1049      * When a -device option is processed from the command line,
1050      * qbus_find_recursive() picks the next free virtio-mmio bus in forwards
1051      * order. The upshot is that -device options in increasing command line
1052      * order are mapped to virtio-mmio buses with decreasing base addresses.
1053      *
1054      * When this code was originally written, that arrangement ensured that the
1055      * guest Linux kernel would give the lowest "name" (/dev/vda, eth0, etc) to
1056      * the first -device on the command line. (The end-to-end order is a
1057      * function of this loop, qbus_realize(), qbus_find_recursive(), and the
1058      * guest kernel's name-to-address assignment strategy.)
1059      *
1060      * Meanwhile, the kernel's traversal seems to have been reversed; see eg.
1061      * the message, if not necessarily the code, of commit 70161ff336.
1062      * Therefore the loop now establishes the inverse of the original intent.
1063      *
1064      * Unfortunately, we can't counteract the kernel change by reversing the
1065      * loop; it would break existing command lines.
1066      *
1067      * In any case, the kernel makes no guarantee about the stability of
1068      * enumeration order of virtio devices (as demonstrated by it changing
1069      * between kernel versions). For reliable and stable identification
1070      * of disks users must use UUIDs or similar mechanisms.
1071      */
1072     for (i = 0; i < NUM_VIRTIO_TRANSPORTS; i++) {
1073         int irq = vms->irqmap[VIRT_MMIO] + i;
1074         hwaddr base = vms->memmap[VIRT_MMIO].base + i * size;
1075 
1076         sysbus_create_simple("virtio-mmio", base,
1077                              qdev_get_gpio_in(vms->gic, irq));
1078     }
1079 
1080     /* We add dtb nodes in reverse order so that they appear in the finished
1081      * device tree lowest address first.
1082      *
1083      * Note that this mapping is independent of the loop above. The previous
1084      * loop influences virtio device to virtio transport assignment, whereas
1085      * this loop controls how virtio transports are laid out in the dtb.
1086      */
1087     for (i = NUM_VIRTIO_TRANSPORTS - 1; i >= 0; i--) {
1088         char *nodename;
1089         int irq = vms->irqmap[VIRT_MMIO] + i;
1090         hwaddr base = vms->memmap[VIRT_MMIO].base + i * size;
1091 
1092         nodename = g_strdup_printf("/virtio_mmio@%" PRIx64, base);
1093         qemu_fdt_add_subnode(ms->fdt, nodename);
1094         qemu_fdt_setprop_string(ms->fdt, nodename,
1095                                 "compatible", "virtio,mmio");
1096         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1097                                      2, base, 2, size);
1098         qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
1099                                GIC_FDT_IRQ_TYPE_SPI, irq,
1100                                GIC_FDT_IRQ_FLAGS_EDGE_LO_HI);
1101         qemu_fdt_setprop(ms->fdt, nodename, "dma-coherent", NULL, 0);
1102         g_free(nodename);
1103     }
1104 }
1105 
1106 #define VIRT_FLASH_SECTOR_SIZE (256 * KiB)
1107 
1108 static PFlashCFI01 *virt_flash_create1(VirtMachineState *vms,
1109                                         const char *name,
1110                                         const char *alias_prop_name)
1111 {
1112     /*
1113      * Create a single flash device.  We use the same parameters as
1114      * the flash devices on the Versatile Express board.
1115      */
1116     DeviceState *dev = qdev_new(TYPE_PFLASH_CFI01);
1117 
1118     qdev_prop_set_uint64(dev, "sector-length", VIRT_FLASH_SECTOR_SIZE);
1119     qdev_prop_set_uint8(dev, "width", 4);
1120     qdev_prop_set_uint8(dev, "device-width", 2);
1121     qdev_prop_set_bit(dev, "big-endian", false);
1122     qdev_prop_set_uint16(dev, "id0", 0x89);
1123     qdev_prop_set_uint16(dev, "id1", 0x18);
1124     qdev_prop_set_uint16(dev, "id2", 0x00);
1125     qdev_prop_set_uint16(dev, "id3", 0x00);
1126     qdev_prop_set_string(dev, "name", name);
1127     object_property_add_child(OBJECT(vms), name, OBJECT(dev));
1128     object_property_add_alias(OBJECT(vms), alias_prop_name,
1129                               OBJECT(dev), "drive");
1130     return PFLASH_CFI01(dev);
1131 }
1132 
1133 static void virt_flash_create(VirtMachineState *vms)
1134 {
1135     vms->flash[0] = virt_flash_create1(vms, "virt.flash0", "pflash0");
1136     vms->flash[1] = virt_flash_create1(vms, "virt.flash1", "pflash1");
1137 }
1138 
1139 static void virt_flash_map1(PFlashCFI01 *flash,
1140                             hwaddr base, hwaddr size,
1141                             MemoryRegion *sysmem)
1142 {
1143     DeviceState *dev = DEVICE(flash);
1144 
1145     assert(QEMU_IS_ALIGNED(size, VIRT_FLASH_SECTOR_SIZE));
1146     assert(size / VIRT_FLASH_SECTOR_SIZE <= UINT32_MAX);
1147     qdev_prop_set_uint32(dev, "num-blocks", size / VIRT_FLASH_SECTOR_SIZE);
1148     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1149 
1150     memory_region_add_subregion(sysmem, base,
1151                                 sysbus_mmio_get_region(SYS_BUS_DEVICE(dev),
1152                                                        0));
1153 }
1154 
1155 static void virt_flash_map(VirtMachineState *vms,
1156                            MemoryRegion *sysmem,
1157                            MemoryRegion *secure_sysmem)
1158 {
1159     /*
1160      * Map two flash devices to fill the VIRT_FLASH space in the memmap.
1161      * sysmem is the system memory space. secure_sysmem is the secure view
1162      * of the system, and the first flash device should be made visible only
1163      * there. The second flash device is visible to both secure and nonsecure.
1164      * If sysmem == secure_sysmem this means there is no separate Secure
1165      * address space and both flash devices are generally visible.
1166      */
1167     hwaddr flashsize = vms->memmap[VIRT_FLASH].size / 2;
1168     hwaddr flashbase = vms->memmap[VIRT_FLASH].base;
1169 
1170     virt_flash_map1(vms->flash[0], flashbase, flashsize,
1171                     secure_sysmem);
1172     virt_flash_map1(vms->flash[1], flashbase + flashsize, flashsize,
1173                     sysmem);
1174 }
1175 
1176 static void virt_flash_fdt(VirtMachineState *vms,
1177                            MemoryRegion *sysmem,
1178                            MemoryRegion *secure_sysmem)
1179 {
1180     hwaddr flashsize = vms->memmap[VIRT_FLASH].size / 2;
1181     hwaddr flashbase = vms->memmap[VIRT_FLASH].base;
1182     MachineState *ms = MACHINE(vms);
1183     char *nodename;
1184 
1185     if (sysmem == secure_sysmem) {
1186         /* Report both flash devices as a single node in the DT */
1187         nodename = g_strdup_printf("/flash@%" PRIx64, flashbase);
1188         qemu_fdt_add_subnode(ms->fdt, nodename);
1189         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible", "cfi-flash");
1190         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1191                                      2, flashbase, 2, flashsize,
1192                                      2, flashbase + flashsize, 2, flashsize);
1193         qemu_fdt_setprop_cell(ms->fdt, nodename, "bank-width", 4);
1194         g_free(nodename);
1195     } else {
1196         /*
1197          * Report the devices as separate nodes so we can mark one as
1198          * only visible to the secure world.
1199          */
1200         nodename = g_strdup_printf("/secflash@%" 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         qemu_fdt_setprop_cell(ms->fdt, nodename, "bank-width", 4);
1206         qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
1207         qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
1208         g_free(nodename);
1209 
1210         nodename = g_strdup_printf("/flash@%" PRIx64, flashbase + flashsize);
1211         qemu_fdt_add_subnode(ms->fdt, nodename);
1212         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible", "cfi-flash");
1213         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1214                                      2, flashbase + flashsize, 2, flashsize);
1215         qemu_fdt_setprop_cell(ms->fdt, nodename, "bank-width", 4);
1216         g_free(nodename);
1217     }
1218 }
1219 
1220 static bool virt_firmware_init(VirtMachineState *vms,
1221                                MemoryRegion *sysmem,
1222                                MemoryRegion *secure_sysmem)
1223 {
1224     int i;
1225     const char *bios_name;
1226     BlockBackend *pflash_blk0;
1227 
1228     /* Map legacy -drive if=pflash to machine properties */
1229     for (i = 0; i < ARRAY_SIZE(vms->flash); i++) {
1230         pflash_cfi01_legacy_drive(vms->flash[i],
1231                                   drive_get(IF_PFLASH, 0, i));
1232     }
1233 
1234     virt_flash_map(vms, sysmem, secure_sysmem);
1235 
1236     pflash_blk0 = pflash_cfi01_get_blk(vms->flash[0]);
1237 
1238     bios_name = MACHINE(vms)->firmware;
1239     if (bios_name) {
1240         char *fname;
1241         MemoryRegion *mr;
1242         int image_size;
1243 
1244         if (pflash_blk0) {
1245             error_report("The contents of the first flash device may be "
1246                          "specified with -bios or with -drive if=pflash... "
1247                          "but you cannot use both options at once");
1248             exit(1);
1249         }
1250 
1251         /* Fall back to -bios */
1252 
1253         fname = qemu_find_file(QEMU_FILE_TYPE_BIOS, bios_name);
1254         if (!fname) {
1255             error_report("Could not find ROM image '%s'", bios_name);
1256             exit(1);
1257         }
1258         mr = sysbus_mmio_get_region(SYS_BUS_DEVICE(vms->flash[0]), 0);
1259         image_size = load_image_mr(fname, mr);
1260         g_free(fname);
1261         if (image_size < 0) {
1262             error_report("Could not load ROM image '%s'", bios_name);
1263             exit(1);
1264         }
1265     }
1266 
1267     return pflash_blk0 || bios_name;
1268 }
1269 
1270 static FWCfgState *create_fw_cfg(const VirtMachineState *vms, AddressSpace *as)
1271 {
1272     MachineState *ms = MACHINE(vms);
1273     hwaddr base = vms->memmap[VIRT_FW_CFG].base;
1274     hwaddr size = vms->memmap[VIRT_FW_CFG].size;
1275     FWCfgState *fw_cfg;
1276     char *nodename;
1277 
1278     fw_cfg = fw_cfg_init_mem_wide(base + 8, base, 8, base + 16, as);
1279     fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, (uint16_t)ms->smp.cpus);
1280 
1281     nodename = g_strdup_printf("/fw-cfg@%" PRIx64, base);
1282     qemu_fdt_add_subnode(ms->fdt, nodename);
1283     qemu_fdt_setprop_string(ms->fdt, nodename,
1284                             "compatible", "qemu,fw-cfg-mmio");
1285     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1286                                  2, base, 2, size);
1287     qemu_fdt_setprop(ms->fdt, nodename, "dma-coherent", NULL, 0);
1288     g_free(nodename);
1289     return fw_cfg;
1290 }
1291 
1292 static void create_pcie_irq_map(const MachineState *ms,
1293                                 uint32_t gic_phandle,
1294                                 int first_irq, const char *nodename)
1295 {
1296     int devfn, pin;
1297     uint32_t full_irq_map[4 * 4 * 10] = { 0 };
1298     uint32_t *irq_map = full_irq_map;
1299 
1300     for (devfn = 0; devfn <= 0x18; devfn += 0x8) {
1301         for (pin = 0; pin < 4; pin++) {
1302             int irq_type = GIC_FDT_IRQ_TYPE_SPI;
1303             int irq_nr = first_irq + ((pin + PCI_SLOT(devfn)) % PCI_NUM_PINS);
1304             int irq_level = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
1305             int i;
1306 
1307             uint32_t map[] = {
1308                 devfn << 8, 0, 0,                           /* devfn */
1309                 pin + 1,                                    /* PCI pin */
1310                 gic_phandle, 0, 0, irq_type, irq_nr, irq_level }; /* GIC irq */
1311 
1312             /* Convert map to big endian */
1313             for (i = 0; i < 10; i++) {
1314                 irq_map[i] = cpu_to_be32(map[i]);
1315             }
1316             irq_map += 10;
1317         }
1318     }
1319 
1320     qemu_fdt_setprop(ms->fdt, nodename, "interrupt-map",
1321                      full_irq_map, sizeof(full_irq_map));
1322 
1323     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupt-map-mask",
1324                            cpu_to_be16(PCI_DEVFN(3, 0)), /* Slot 3 */
1325                            0, 0,
1326                            0x7           /* PCI irq */);
1327 }
1328 
1329 static void create_smmu(const VirtMachineState *vms,
1330                         PCIBus *bus)
1331 {
1332     char *node;
1333     const char compat[] = "arm,smmu-v3";
1334     int irq =  vms->irqmap[VIRT_SMMU];
1335     int i;
1336     hwaddr base = vms->memmap[VIRT_SMMU].base;
1337     hwaddr size = vms->memmap[VIRT_SMMU].size;
1338     const char irq_names[] = "eventq\0priq\0cmdq-sync\0gerror";
1339     DeviceState *dev;
1340     MachineState *ms = MACHINE(vms);
1341 
1342     if (vms->iommu != VIRT_IOMMU_SMMUV3 || !vms->iommu_phandle) {
1343         return;
1344     }
1345 
1346     dev = qdev_new("arm-smmuv3");
1347 
1348     object_property_set_link(OBJECT(dev), "primary-bus", OBJECT(bus),
1349                              &error_abort);
1350     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1351     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, base);
1352     for (i = 0; i < NUM_SMMU_IRQS; i++) {
1353         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
1354                            qdev_get_gpio_in(vms->gic, irq + i));
1355     }
1356 
1357     node = g_strdup_printf("/smmuv3@%" PRIx64, base);
1358     qemu_fdt_add_subnode(ms->fdt, node);
1359     qemu_fdt_setprop(ms->fdt, node, "compatible", compat, sizeof(compat));
1360     qemu_fdt_setprop_sized_cells(ms->fdt, node, "reg", 2, base, 2, size);
1361 
1362     qemu_fdt_setprop_cells(ms->fdt, node, "interrupts",
1363             GIC_FDT_IRQ_TYPE_SPI, irq    , GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1364             GIC_FDT_IRQ_TYPE_SPI, irq + 1, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1365             GIC_FDT_IRQ_TYPE_SPI, irq + 2, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1366             GIC_FDT_IRQ_TYPE_SPI, irq + 3, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI);
1367 
1368     qemu_fdt_setprop(ms->fdt, node, "interrupt-names", irq_names,
1369                      sizeof(irq_names));
1370 
1371     qemu_fdt_setprop(ms->fdt, node, "dma-coherent", NULL, 0);
1372 
1373     qemu_fdt_setprop_cell(ms->fdt, node, "#iommu-cells", 1);
1374 
1375     qemu_fdt_setprop_cell(ms->fdt, node, "phandle", vms->iommu_phandle);
1376     g_free(node);
1377 }
1378 
1379 static void create_virtio_iommu_dt_bindings(VirtMachineState *vms)
1380 {
1381     const char compat[] = "virtio,pci-iommu\0pci1af4,1057";
1382     uint16_t bdf = vms->virtio_iommu_bdf;
1383     MachineState *ms = MACHINE(vms);
1384     char *node;
1385 
1386     vms->iommu_phandle = qemu_fdt_alloc_phandle(ms->fdt);
1387 
1388     node = g_strdup_printf("%s/virtio_iommu@%x,%x", vms->pciehb_nodename,
1389                            PCI_SLOT(bdf), PCI_FUNC(bdf));
1390     qemu_fdt_add_subnode(ms->fdt, node);
1391     qemu_fdt_setprop(ms->fdt, node, "compatible", compat, sizeof(compat));
1392     qemu_fdt_setprop_sized_cells(ms->fdt, node, "reg",
1393                                  1, bdf << 8, 1, 0, 1, 0,
1394                                  1, 0, 1, 0);
1395 
1396     qemu_fdt_setprop_cell(ms->fdt, node, "#iommu-cells", 1);
1397     qemu_fdt_setprop_cell(ms->fdt, node, "phandle", vms->iommu_phandle);
1398     g_free(node);
1399 
1400     qemu_fdt_setprop_cells(ms->fdt, vms->pciehb_nodename, "iommu-map",
1401                            0x0, vms->iommu_phandle, 0x0, bdf,
1402                            bdf + 1, vms->iommu_phandle, bdf + 1, 0xffff - bdf);
1403 }
1404 
1405 static void create_pcie(VirtMachineState *vms)
1406 {
1407     hwaddr base_mmio = vms->memmap[VIRT_PCIE_MMIO].base;
1408     hwaddr size_mmio = vms->memmap[VIRT_PCIE_MMIO].size;
1409     hwaddr base_mmio_high = vms->memmap[VIRT_HIGH_PCIE_MMIO].base;
1410     hwaddr size_mmio_high = vms->memmap[VIRT_HIGH_PCIE_MMIO].size;
1411     hwaddr base_pio = vms->memmap[VIRT_PCIE_PIO].base;
1412     hwaddr size_pio = vms->memmap[VIRT_PCIE_PIO].size;
1413     hwaddr base_ecam, size_ecam;
1414     hwaddr base = base_mmio;
1415     int nr_pcie_buses;
1416     int irq = vms->irqmap[VIRT_PCIE];
1417     MemoryRegion *mmio_alias;
1418     MemoryRegion *mmio_reg;
1419     MemoryRegion *ecam_alias;
1420     MemoryRegion *ecam_reg;
1421     DeviceState *dev;
1422     char *nodename;
1423     int i, ecam_id;
1424     PCIHostState *pci;
1425     MachineState *ms = MACHINE(vms);
1426 
1427     dev = qdev_new(TYPE_GPEX_HOST);
1428     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1429 
1430     ecam_id = VIRT_ECAM_ID(vms->highmem_ecam);
1431     base_ecam = vms->memmap[ecam_id].base;
1432     size_ecam = vms->memmap[ecam_id].size;
1433     nr_pcie_buses = size_ecam / PCIE_MMCFG_SIZE_MIN;
1434     /* Map only the first size_ecam bytes of ECAM space */
1435     ecam_alias = g_new0(MemoryRegion, 1);
1436     ecam_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 0);
1437     memory_region_init_alias(ecam_alias, OBJECT(dev), "pcie-ecam",
1438                              ecam_reg, 0, size_ecam);
1439     memory_region_add_subregion(get_system_memory(), base_ecam, ecam_alias);
1440 
1441     /* Map the MMIO window into system address space so as to expose
1442      * the section of PCI MMIO space which starts at the same base address
1443      * (ie 1:1 mapping for that part of PCI MMIO space visible through
1444      * the window).
1445      */
1446     mmio_alias = g_new0(MemoryRegion, 1);
1447     mmio_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 1);
1448     memory_region_init_alias(mmio_alias, OBJECT(dev), "pcie-mmio",
1449                              mmio_reg, base_mmio, size_mmio);
1450     memory_region_add_subregion(get_system_memory(), base_mmio, mmio_alias);
1451 
1452     if (vms->highmem_mmio) {
1453         /* Map high MMIO space */
1454         MemoryRegion *high_mmio_alias = g_new0(MemoryRegion, 1);
1455 
1456         memory_region_init_alias(high_mmio_alias, OBJECT(dev), "pcie-mmio-high",
1457                                  mmio_reg, base_mmio_high, size_mmio_high);
1458         memory_region_add_subregion(get_system_memory(), base_mmio_high,
1459                                     high_mmio_alias);
1460     }
1461 
1462     /* Map IO port space */
1463     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 2, base_pio);
1464 
1465     for (i = 0; i < GPEX_NUM_IRQS; i++) {
1466         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
1467                            qdev_get_gpio_in(vms->gic, irq + i));
1468         gpex_set_irq_num(GPEX_HOST(dev), i, irq + i);
1469     }
1470 
1471     pci = PCI_HOST_BRIDGE(dev);
1472     pci->bypass_iommu = vms->default_bus_bypass_iommu;
1473     vms->bus = pci->bus;
1474     if (vms->bus) {
1475         for (i = 0; i < nb_nics; i++) {
1476             NICInfo *nd = &nd_table[i];
1477 
1478             if (!nd->model) {
1479                 nd->model = g_strdup("virtio");
1480             }
1481 
1482             pci_nic_init_nofail(nd, pci->bus, nd->model, NULL);
1483         }
1484     }
1485 
1486     nodename = vms->pciehb_nodename = g_strdup_printf("/pcie@%" PRIx64, base);
1487     qemu_fdt_add_subnode(ms->fdt, nodename);
1488     qemu_fdt_setprop_string(ms->fdt, nodename,
1489                             "compatible", "pci-host-ecam-generic");
1490     qemu_fdt_setprop_string(ms->fdt, nodename, "device_type", "pci");
1491     qemu_fdt_setprop_cell(ms->fdt, nodename, "#address-cells", 3);
1492     qemu_fdt_setprop_cell(ms->fdt, nodename, "#size-cells", 2);
1493     qemu_fdt_setprop_cell(ms->fdt, nodename, "linux,pci-domain", 0);
1494     qemu_fdt_setprop_cells(ms->fdt, nodename, "bus-range", 0,
1495                            nr_pcie_buses - 1);
1496     qemu_fdt_setprop(ms->fdt, nodename, "dma-coherent", NULL, 0);
1497 
1498     if (vms->msi_phandle) {
1499         qemu_fdt_setprop_cells(ms->fdt, nodename, "msi-map",
1500                                0, vms->msi_phandle, 0, 0x10000);
1501     }
1502 
1503     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1504                                  2, base_ecam, 2, size_ecam);
1505 
1506     if (vms->highmem_mmio) {
1507         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "ranges",
1508                                      1, FDT_PCI_RANGE_IOPORT, 2, 0,
1509                                      2, base_pio, 2, size_pio,
1510                                      1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
1511                                      2, base_mmio, 2, size_mmio,
1512                                      1, FDT_PCI_RANGE_MMIO_64BIT,
1513                                      2, base_mmio_high,
1514                                      2, base_mmio_high, 2, size_mmio_high);
1515     } else {
1516         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "ranges",
1517                                      1, FDT_PCI_RANGE_IOPORT, 2, 0,
1518                                      2, base_pio, 2, size_pio,
1519                                      1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
1520                                      2, base_mmio, 2, size_mmio);
1521     }
1522 
1523     qemu_fdt_setprop_cell(ms->fdt, nodename, "#interrupt-cells", 1);
1524     create_pcie_irq_map(ms, vms->gic_phandle, irq, nodename);
1525 
1526     if (vms->iommu) {
1527         vms->iommu_phandle = qemu_fdt_alloc_phandle(ms->fdt);
1528 
1529         switch (vms->iommu) {
1530         case VIRT_IOMMU_SMMUV3:
1531             create_smmu(vms, vms->bus);
1532             qemu_fdt_setprop_cells(ms->fdt, nodename, "iommu-map",
1533                                    0x0, vms->iommu_phandle, 0x0, 0x10000);
1534             break;
1535         default:
1536             g_assert_not_reached();
1537         }
1538     }
1539 }
1540 
1541 static void create_platform_bus(VirtMachineState *vms)
1542 {
1543     DeviceState *dev;
1544     SysBusDevice *s;
1545     int i;
1546     MemoryRegion *sysmem = get_system_memory();
1547 
1548     dev = qdev_new(TYPE_PLATFORM_BUS_DEVICE);
1549     dev->id = g_strdup(TYPE_PLATFORM_BUS_DEVICE);
1550     qdev_prop_set_uint32(dev, "num_irqs", PLATFORM_BUS_NUM_IRQS);
1551     qdev_prop_set_uint32(dev, "mmio_size", vms->memmap[VIRT_PLATFORM_BUS].size);
1552     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1553     vms->platform_bus_dev = dev;
1554 
1555     s = SYS_BUS_DEVICE(dev);
1556     for (i = 0; i < PLATFORM_BUS_NUM_IRQS; i++) {
1557         int irq = vms->irqmap[VIRT_PLATFORM_BUS] + i;
1558         sysbus_connect_irq(s, i, qdev_get_gpio_in(vms->gic, irq));
1559     }
1560 
1561     memory_region_add_subregion(sysmem,
1562                                 vms->memmap[VIRT_PLATFORM_BUS].base,
1563                                 sysbus_mmio_get_region(s, 0));
1564 }
1565 
1566 static void create_tag_ram(MemoryRegion *tag_sysmem,
1567                            hwaddr base, hwaddr size,
1568                            const char *name)
1569 {
1570     MemoryRegion *tagram = g_new(MemoryRegion, 1);
1571 
1572     memory_region_init_ram(tagram, NULL, name, size / 32, &error_fatal);
1573     memory_region_add_subregion(tag_sysmem, base / 32, tagram);
1574 }
1575 
1576 static void create_secure_ram(VirtMachineState *vms,
1577                               MemoryRegion *secure_sysmem,
1578                               MemoryRegion *secure_tag_sysmem)
1579 {
1580     MemoryRegion *secram = g_new(MemoryRegion, 1);
1581     char *nodename;
1582     hwaddr base = vms->memmap[VIRT_SECURE_MEM].base;
1583     hwaddr size = vms->memmap[VIRT_SECURE_MEM].size;
1584     MachineState *ms = MACHINE(vms);
1585 
1586     memory_region_init_ram(secram, NULL, "virt.secure-ram", size,
1587                            &error_fatal);
1588     memory_region_add_subregion(secure_sysmem, base, secram);
1589 
1590     nodename = g_strdup_printf("/secram@%" PRIx64, base);
1591     qemu_fdt_add_subnode(ms->fdt, nodename);
1592     qemu_fdt_setprop_string(ms->fdt, nodename, "device_type", "memory");
1593     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg", 2, base, 2, size);
1594     qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
1595     qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
1596 
1597     if (secure_tag_sysmem) {
1598         create_tag_ram(secure_tag_sysmem, base, size, "mach-virt.secure-tag");
1599     }
1600 
1601     g_free(nodename);
1602 }
1603 
1604 static void *machvirt_dtb(const struct arm_boot_info *binfo, int *fdt_size)
1605 {
1606     const VirtMachineState *board = container_of(binfo, VirtMachineState,
1607                                                  bootinfo);
1608     MachineState *ms = MACHINE(board);
1609 
1610 
1611     *fdt_size = board->fdt_size;
1612     return ms->fdt;
1613 }
1614 
1615 static void virt_build_smbios(VirtMachineState *vms)
1616 {
1617     MachineClass *mc = MACHINE_GET_CLASS(vms);
1618     MachineState *ms = MACHINE(vms);
1619     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
1620     uint8_t *smbios_tables, *smbios_anchor;
1621     size_t smbios_tables_len, smbios_anchor_len;
1622     struct smbios_phys_mem_area mem_array;
1623     const char *product = "QEMU Virtual Machine";
1624 
1625     if (kvm_enabled()) {
1626         product = "KVM Virtual Machine";
1627     }
1628 
1629     smbios_set_defaults("QEMU", product,
1630                         vmc->smbios_old_sys_ver ? "1.0" : mc->name, false,
1631                         true, SMBIOS_ENTRY_POINT_TYPE_64);
1632 
1633     /* build the array of physical mem area from base_memmap */
1634     mem_array.address = vms->memmap[VIRT_MEM].base;
1635     mem_array.length = ms->ram_size;
1636 
1637     smbios_get_tables(ms, &mem_array, 1,
1638                       &smbios_tables, &smbios_tables_len,
1639                       &smbios_anchor, &smbios_anchor_len,
1640                       &error_fatal);
1641 
1642     if (smbios_anchor) {
1643         fw_cfg_add_file(vms->fw_cfg, "etc/smbios/smbios-tables",
1644                         smbios_tables, smbios_tables_len);
1645         fw_cfg_add_file(vms->fw_cfg, "etc/smbios/smbios-anchor",
1646                         smbios_anchor, smbios_anchor_len);
1647     }
1648 }
1649 
1650 static
1651 void virt_machine_done(Notifier *notifier, void *data)
1652 {
1653     VirtMachineState *vms = container_of(notifier, VirtMachineState,
1654                                          machine_done);
1655     MachineState *ms = MACHINE(vms);
1656     ARMCPU *cpu = ARM_CPU(first_cpu);
1657     struct arm_boot_info *info = &vms->bootinfo;
1658     AddressSpace *as = arm_boot_address_space(cpu, info);
1659 
1660     /*
1661      * If the user provided a dtb, we assume the dynamic sysbus nodes
1662      * already are integrated there. This corresponds to a use case where
1663      * the dynamic sysbus nodes are complex and their generation is not yet
1664      * supported. In that case the user can take charge of the guest dt
1665      * while qemu takes charge of the qom stuff.
1666      */
1667     if (info->dtb_filename == NULL) {
1668         platform_bus_add_all_fdt_nodes(ms->fdt, "/intc",
1669                                        vms->memmap[VIRT_PLATFORM_BUS].base,
1670                                        vms->memmap[VIRT_PLATFORM_BUS].size,
1671                                        vms->irqmap[VIRT_PLATFORM_BUS]);
1672     }
1673     if (arm_load_dtb(info->dtb_start, info, info->dtb_limit, as, ms) < 0) {
1674         exit(1);
1675     }
1676 
1677     fw_cfg_add_extra_pci_roots(vms->bus, vms->fw_cfg);
1678 
1679     virt_acpi_setup(vms);
1680     virt_build_smbios(vms);
1681 }
1682 
1683 static uint64_t virt_cpu_mp_affinity(VirtMachineState *vms, int idx)
1684 {
1685     uint8_t clustersz = ARM_DEFAULT_CPUS_PER_CLUSTER;
1686     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
1687 
1688     if (!vmc->disallow_affinity_adjustment) {
1689         /* Adjust MPIDR like 64-bit KVM hosts, which incorporate the
1690          * GIC's target-list limitations. 32-bit KVM hosts currently
1691          * always create clusters of 4 CPUs, but that is expected to
1692          * change when they gain support for gicv3. When KVM is enabled
1693          * it will override the changes we make here, therefore our
1694          * purposes are to make TCG consistent (with 64-bit KVM hosts)
1695          * and to improve SGI efficiency.
1696          */
1697         if (vms->gic_version == VIRT_GIC_VERSION_2) {
1698             clustersz = GIC_TARGETLIST_BITS;
1699         } else {
1700             clustersz = GICV3_TARGETLIST_BITS;
1701         }
1702     }
1703     return arm_cpu_mp_affinity(idx, clustersz);
1704 }
1705 
1706 static inline bool *virt_get_high_memmap_enabled(VirtMachineState *vms,
1707                                                  int index)
1708 {
1709     bool *enabled_array[] = {
1710         &vms->highmem_redists,
1711         &vms->highmem_ecam,
1712         &vms->highmem_mmio,
1713     };
1714 
1715     assert(ARRAY_SIZE(extended_memmap) - VIRT_LOWMEMMAP_LAST ==
1716            ARRAY_SIZE(enabled_array));
1717     assert(index - VIRT_LOWMEMMAP_LAST < ARRAY_SIZE(enabled_array));
1718 
1719     return enabled_array[index - VIRT_LOWMEMMAP_LAST];
1720 }
1721 
1722 static void virt_set_high_memmap(VirtMachineState *vms,
1723                                  hwaddr base, int pa_bits)
1724 {
1725     hwaddr region_base, region_size;
1726     bool *region_enabled, fits;
1727     int i;
1728 
1729     for (i = VIRT_LOWMEMMAP_LAST; i < ARRAY_SIZE(extended_memmap); i++) {
1730         region_enabled = virt_get_high_memmap_enabled(vms, i);
1731         region_base = ROUND_UP(base, extended_memmap[i].size);
1732         region_size = extended_memmap[i].size;
1733 
1734         vms->memmap[i].base = region_base;
1735         vms->memmap[i].size = region_size;
1736 
1737         /*
1738          * Check each device to see if it fits in the PA space,
1739          * moving highest_gpa as we go. For compatibility, move
1740          * highest_gpa for disabled fitting devices as well, if
1741          * the compact layout has been disabled.
1742          *
1743          * For each device that doesn't fit, disable it.
1744          */
1745         fits = (region_base + region_size) <= BIT_ULL(pa_bits);
1746         *region_enabled &= fits;
1747         if (vms->highmem_compact && !*region_enabled) {
1748             continue;
1749         }
1750 
1751         base = region_base + region_size;
1752         if (fits) {
1753             vms->highest_gpa = base - 1;
1754         }
1755     }
1756 }
1757 
1758 static void virt_set_memmap(VirtMachineState *vms, int pa_bits)
1759 {
1760     MachineState *ms = MACHINE(vms);
1761     hwaddr base, device_memory_base, device_memory_size, memtop;
1762     int i;
1763 
1764     vms->memmap = extended_memmap;
1765 
1766     for (i = 0; i < ARRAY_SIZE(base_memmap); i++) {
1767         vms->memmap[i] = base_memmap[i];
1768     }
1769 
1770     if (ms->ram_slots > ACPI_MAX_RAM_SLOTS) {
1771         error_report("unsupported number of memory slots: %"PRIu64,
1772                      ms->ram_slots);
1773         exit(EXIT_FAILURE);
1774     }
1775 
1776     /*
1777      * !highmem is exactly the same as limiting the PA space to 32bit,
1778      * irrespective of the underlying capabilities of the HW.
1779      */
1780     if (!vms->highmem) {
1781         pa_bits = 32;
1782     }
1783 
1784     /*
1785      * We compute the base of the high IO region depending on the
1786      * amount of initial and device memory. The device memory start/size
1787      * is aligned on 1GiB. We never put the high IO region below 256GiB
1788      * so that if maxram_size is < 255GiB we keep the legacy memory map.
1789      * The device region size assumes 1GiB page max alignment per slot.
1790      */
1791     device_memory_base =
1792         ROUND_UP(vms->memmap[VIRT_MEM].base + ms->ram_size, GiB);
1793     device_memory_size = ms->maxram_size - ms->ram_size + ms->ram_slots * GiB;
1794 
1795     /* Base address of the high IO region */
1796     memtop = base = device_memory_base + ROUND_UP(device_memory_size, GiB);
1797     if (memtop > BIT_ULL(pa_bits)) {
1798 	    error_report("Addressing limited to %d bits, but memory exceeds it by %llu bytes\n",
1799 			 pa_bits, memtop - BIT_ULL(pa_bits));
1800         exit(EXIT_FAILURE);
1801     }
1802     if (base < device_memory_base) {
1803         error_report("maxmem/slots too huge");
1804         exit(EXIT_FAILURE);
1805     }
1806     if (base < vms->memmap[VIRT_MEM].base + LEGACY_RAMLIMIT_BYTES) {
1807         base = vms->memmap[VIRT_MEM].base + LEGACY_RAMLIMIT_BYTES;
1808     }
1809 
1810     /* We know for sure that at least the memory fits in the PA space */
1811     vms->highest_gpa = memtop - 1;
1812 
1813     virt_set_high_memmap(vms, base, pa_bits);
1814 
1815     if (device_memory_size > 0) {
1816         ms->device_memory = g_malloc0(sizeof(*ms->device_memory));
1817         ms->device_memory->base = device_memory_base;
1818         memory_region_init(&ms->device_memory->mr, OBJECT(vms),
1819                            "device-memory", device_memory_size);
1820     }
1821 }
1822 
1823 /*
1824  * finalize_gic_version - Determines the final gic_version
1825  * according to the gic-version property
1826  *
1827  * Default GIC type is v2
1828  */
1829 static void finalize_gic_version(VirtMachineState *vms)
1830 {
1831     unsigned int max_cpus = MACHINE(vms)->smp.max_cpus;
1832 
1833     if (kvm_enabled()) {
1834         int probe_bitmap;
1835 
1836         if (!kvm_irqchip_in_kernel()) {
1837             switch (vms->gic_version) {
1838             case VIRT_GIC_VERSION_HOST:
1839                 warn_report(
1840                     "gic-version=host not relevant with kernel-irqchip=off "
1841                      "as only userspace GICv2 is supported. Using v2 ...");
1842                 return;
1843             case VIRT_GIC_VERSION_MAX:
1844             case VIRT_GIC_VERSION_NOSEL:
1845                 vms->gic_version = VIRT_GIC_VERSION_2;
1846                 return;
1847             case VIRT_GIC_VERSION_2:
1848                 return;
1849             case VIRT_GIC_VERSION_3:
1850                 error_report(
1851                     "gic-version=3 is not supported with kernel-irqchip=off");
1852                 exit(1);
1853             case VIRT_GIC_VERSION_4:
1854                 error_report(
1855                     "gic-version=4 is not supported with kernel-irqchip=off");
1856                 exit(1);
1857             }
1858         }
1859 
1860         probe_bitmap = kvm_arm_vgic_probe();
1861         if (!probe_bitmap) {
1862             error_report("Unable to determine GIC version supported by host");
1863             exit(1);
1864         }
1865 
1866         switch (vms->gic_version) {
1867         case VIRT_GIC_VERSION_HOST:
1868         case VIRT_GIC_VERSION_MAX:
1869             if (probe_bitmap & KVM_ARM_VGIC_V3) {
1870                 vms->gic_version = VIRT_GIC_VERSION_3;
1871             } else {
1872                 vms->gic_version = VIRT_GIC_VERSION_2;
1873             }
1874             return;
1875         case VIRT_GIC_VERSION_NOSEL:
1876             if ((probe_bitmap & KVM_ARM_VGIC_V2) && max_cpus <= GIC_NCPU) {
1877                 vms->gic_version = VIRT_GIC_VERSION_2;
1878             } else if (probe_bitmap & KVM_ARM_VGIC_V3) {
1879                 /*
1880                  * in case the host does not support v2 in-kernel emulation or
1881                  * the end-user requested more than 8 VCPUs we now default
1882                  * to v3. In any case defaulting to v2 would be broken.
1883                  */
1884                 vms->gic_version = VIRT_GIC_VERSION_3;
1885             } else if (max_cpus > GIC_NCPU) {
1886                 error_report("host only supports in-kernel GICv2 emulation "
1887                              "but more than 8 vcpus are requested");
1888                 exit(1);
1889             }
1890             break;
1891         case VIRT_GIC_VERSION_2:
1892         case VIRT_GIC_VERSION_3:
1893             break;
1894         case VIRT_GIC_VERSION_4:
1895             error_report("gic-version=4 is not supported with KVM");
1896             exit(1);
1897         }
1898 
1899         /* Check chosen version is effectively supported by the host */
1900         if (vms->gic_version == VIRT_GIC_VERSION_2 &&
1901             !(probe_bitmap & KVM_ARM_VGIC_V2)) {
1902             error_report("host does not support in-kernel GICv2 emulation");
1903             exit(1);
1904         } else if (vms->gic_version == VIRT_GIC_VERSION_3 &&
1905                    !(probe_bitmap & KVM_ARM_VGIC_V3)) {
1906             error_report("host does not support in-kernel GICv3 emulation");
1907             exit(1);
1908         }
1909         return;
1910     }
1911 
1912     /* TCG mode */
1913     switch (vms->gic_version) {
1914     case VIRT_GIC_VERSION_NOSEL:
1915         vms->gic_version = VIRT_GIC_VERSION_2;
1916         break;
1917     case VIRT_GIC_VERSION_MAX:
1918         if (module_object_class_by_name("arm-gicv3")) {
1919             /* CONFIG_ARM_GICV3_TCG was set */
1920             if (vms->virt) {
1921                 /* GICv4 only makes sense if CPU has EL2 */
1922                 vms->gic_version = VIRT_GIC_VERSION_4;
1923             } else {
1924                 vms->gic_version = VIRT_GIC_VERSION_3;
1925             }
1926         } else {
1927             vms->gic_version = VIRT_GIC_VERSION_2;
1928         }
1929         break;
1930     case VIRT_GIC_VERSION_HOST:
1931         error_report("gic-version=host requires KVM");
1932         exit(1);
1933     case VIRT_GIC_VERSION_4:
1934         if (!vms->virt) {
1935             error_report("gic-version=4 requires virtualization enabled");
1936             exit(1);
1937         }
1938         break;
1939     case VIRT_GIC_VERSION_2:
1940     case VIRT_GIC_VERSION_3:
1941         break;
1942     }
1943 }
1944 
1945 /*
1946  * virt_cpu_post_init() must be called after the CPUs have
1947  * been realized and the GIC has been created.
1948  */
1949 static void virt_cpu_post_init(VirtMachineState *vms, MemoryRegion *sysmem)
1950 {
1951     int max_cpus = MACHINE(vms)->smp.max_cpus;
1952     bool aarch64, pmu, steal_time;
1953     CPUState *cpu;
1954 
1955     aarch64 = object_property_get_bool(OBJECT(first_cpu), "aarch64", NULL);
1956     pmu = object_property_get_bool(OBJECT(first_cpu), "pmu", NULL);
1957     steal_time = object_property_get_bool(OBJECT(first_cpu),
1958                                           "kvm-steal-time", NULL);
1959 
1960     if (kvm_enabled()) {
1961         hwaddr pvtime_reg_base = vms->memmap[VIRT_PVTIME].base;
1962         hwaddr pvtime_reg_size = vms->memmap[VIRT_PVTIME].size;
1963 
1964         if (steal_time) {
1965             MemoryRegion *pvtime = g_new(MemoryRegion, 1);
1966             hwaddr pvtime_size = max_cpus * PVTIME_SIZE_PER_CPU;
1967 
1968             /* The memory region size must be a multiple of host page size. */
1969             pvtime_size = REAL_HOST_PAGE_ALIGN(pvtime_size);
1970 
1971             if (pvtime_size > pvtime_reg_size) {
1972                 error_report("pvtime requires a %" HWADDR_PRId
1973                              " byte memory region for %d CPUs,"
1974                              " but only %" HWADDR_PRId " has been reserved",
1975                              pvtime_size, max_cpus, pvtime_reg_size);
1976                 exit(1);
1977             }
1978 
1979             memory_region_init_ram(pvtime, NULL, "pvtime", pvtime_size, NULL);
1980             memory_region_add_subregion(sysmem, pvtime_reg_base, pvtime);
1981         }
1982 
1983         CPU_FOREACH(cpu) {
1984             if (pmu) {
1985                 assert(arm_feature(&ARM_CPU(cpu)->env, ARM_FEATURE_PMU));
1986                 if (kvm_irqchip_in_kernel()) {
1987                     kvm_arm_pmu_set_irq(cpu, PPI(VIRTUAL_PMU_IRQ));
1988                 }
1989                 kvm_arm_pmu_init(cpu);
1990             }
1991             if (steal_time) {
1992                 kvm_arm_pvtime_init(cpu, pvtime_reg_base +
1993                                          cpu->cpu_index * PVTIME_SIZE_PER_CPU);
1994             }
1995         }
1996     } else {
1997         if (aarch64 && vms->highmem) {
1998             int requested_pa_size = 64 - clz64(vms->highest_gpa);
1999             int pamax = arm_pamax(ARM_CPU(first_cpu));
2000 
2001             if (pamax < requested_pa_size) {
2002                 error_report("VCPU supports less PA bits (%d) than "
2003                              "requested by the memory map (%d)",
2004                              pamax, requested_pa_size);
2005                 exit(1);
2006             }
2007         }
2008     }
2009 }
2010 
2011 static void machvirt_init(MachineState *machine)
2012 {
2013     VirtMachineState *vms = VIRT_MACHINE(machine);
2014     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(machine);
2015     MachineClass *mc = MACHINE_GET_CLASS(machine);
2016     const CPUArchIdList *possible_cpus;
2017     MemoryRegion *sysmem = get_system_memory();
2018     MemoryRegion *secure_sysmem = NULL;
2019     MemoryRegion *tag_sysmem = NULL;
2020     MemoryRegion *secure_tag_sysmem = NULL;
2021     int n, virt_max_cpus;
2022     bool firmware_loaded;
2023     bool aarch64 = true;
2024     bool has_ged = !vmc->no_ged;
2025     unsigned int smp_cpus = machine->smp.cpus;
2026     unsigned int max_cpus = machine->smp.max_cpus;
2027 
2028     if (!cpu_type_valid(machine->cpu_type)) {
2029         error_report("mach-virt: CPU type %s not supported", machine->cpu_type);
2030         exit(1);
2031     }
2032 
2033     possible_cpus = mc->possible_cpu_arch_ids(machine);
2034 
2035     /*
2036      * In accelerated mode, the memory map is computed earlier in kvm_type()
2037      * to create a VM with the right number of IPA bits.
2038      */
2039     if (!vms->memmap) {
2040         Object *cpuobj;
2041         ARMCPU *armcpu;
2042         int pa_bits;
2043 
2044         /*
2045          * Instanciate a temporary CPU object to find out about what
2046          * we are about to deal with. Once this is done, get rid of
2047          * the object.
2048          */
2049         cpuobj = object_new(possible_cpus->cpus[0].type);
2050         armcpu = ARM_CPU(cpuobj);
2051 
2052         pa_bits = arm_pamax(armcpu);
2053 
2054         object_unref(cpuobj);
2055 
2056         virt_set_memmap(vms, pa_bits);
2057     }
2058 
2059     /* We can probe only here because during property set
2060      * KVM is not available yet
2061      */
2062     finalize_gic_version(vms);
2063 
2064     if (vms->secure) {
2065         /*
2066          * The Secure view of the world is the same as the NonSecure,
2067          * but with a few extra devices. Create it as a container region
2068          * containing the system memory at low priority; any secure-only
2069          * devices go in at higher priority and take precedence.
2070          */
2071         secure_sysmem = g_new(MemoryRegion, 1);
2072         memory_region_init(secure_sysmem, OBJECT(machine), "secure-memory",
2073                            UINT64_MAX);
2074         memory_region_add_subregion_overlap(secure_sysmem, 0, sysmem, -1);
2075     }
2076 
2077     firmware_loaded = virt_firmware_init(vms, sysmem,
2078                                          secure_sysmem ?: sysmem);
2079 
2080     /* If we have an EL3 boot ROM then the assumption is that it will
2081      * implement PSCI itself, so disable QEMU's internal implementation
2082      * so it doesn't get in the way. Instead of starting secondary
2083      * CPUs in PSCI powerdown state we will start them all running and
2084      * let the boot ROM sort them out.
2085      * The usual case is that we do use QEMU's PSCI implementation;
2086      * if the guest has EL2 then we will use SMC as the conduit,
2087      * and otherwise we will use HVC (for backwards compatibility and
2088      * because if we're using KVM then we must use HVC).
2089      */
2090     if (vms->secure && firmware_loaded) {
2091         vms->psci_conduit = QEMU_PSCI_CONDUIT_DISABLED;
2092     } else if (vms->virt) {
2093         vms->psci_conduit = QEMU_PSCI_CONDUIT_SMC;
2094     } else {
2095         vms->psci_conduit = QEMU_PSCI_CONDUIT_HVC;
2096     }
2097 
2098     /*
2099      * The maximum number of CPUs depends on the GIC version, or on how
2100      * many redistributors we can fit into the memory map (which in turn
2101      * depends on whether this is a GICv3 or v4).
2102      */
2103     if (vms->gic_version == VIRT_GIC_VERSION_2) {
2104         virt_max_cpus = GIC_NCPU;
2105     } else {
2106         virt_max_cpus = virt_redist_capacity(vms, VIRT_GIC_REDIST);
2107         if (vms->highmem_redists) {
2108             virt_max_cpus += virt_redist_capacity(vms, VIRT_HIGH_GIC_REDIST2);
2109         }
2110     }
2111 
2112     if (max_cpus > virt_max_cpus) {
2113         error_report("Number of SMP CPUs requested (%d) exceeds max CPUs "
2114                      "supported by machine 'mach-virt' (%d)",
2115                      max_cpus, virt_max_cpus);
2116         if (vms->gic_version != VIRT_GIC_VERSION_2 && !vms->highmem_redists) {
2117             error_printf("Try 'highmem-redists=on' for more CPUs\n");
2118         }
2119 
2120         exit(1);
2121     }
2122 
2123     if (vms->secure && (kvm_enabled() || hvf_enabled())) {
2124         error_report("mach-virt: %s does not support providing "
2125                      "Security extensions (TrustZone) to the guest CPU",
2126                      kvm_enabled() ? "KVM" : "HVF");
2127         exit(1);
2128     }
2129 
2130     if (vms->virt && (kvm_enabled() || hvf_enabled())) {
2131         error_report("mach-virt: %s does not support providing "
2132                      "Virtualization extensions to the guest CPU",
2133                      kvm_enabled() ? "KVM" : "HVF");
2134         exit(1);
2135     }
2136 
2137     if (vms->mte && (kvm_enabled() || hvf_enabled())) {
2138         error_report("mach-virt: %s does not support providing "
2139                      "MTE to the guest CPU",
2140                      kvm_enabled() ? "KVM" : "HVF");
2141         exit(1);
2142     }
2143 
2144     create_fdt(vms);
2145 
2146     assert(possible_cpus->len == max_cpus);
2147     for (n = 0; n < possible_cpus->len; n++) {
2148         Object *cpuobj;
2149         CPUState *cs;
2150 
2151         if (n >= smp_cpus) {
2152             break;
2153         }
2154 
2155         cpuobj = object_new(possible_cpus->cpus[n].type);
2156         object_property_set_int(cpuobj, "mp-affinity",
2157                                 possible_cpus->cpus[n].arch_id, NULL);
2158 
2159         cs = CPU(cpuobj);
2160         cs->cpu_index = n;
2161 
2162         numa_cpu_pre_plug(&possible_cpus->cpus[cs->cpu_index], DEVICE(cpuobj),
2163                           &error_fatal);
2164 
2165         aarch64 &= object_property_get_bool(cpuobj, "aarch64", NULL);
2166 
2167         if (!vms->secure) {
2168             object_property_set_bool(cpuobj, "has_el3", false, NULL);
2169         }
2170 
2171         if (!vms->virt && object_property_find(cpuobj, "has_el2")) {
2172             object_property_set_bool(cpuobj, "has_el2", false, NULL);
2173         }
2174 
2175         if (vmc->kvm_no_adjvtime &&
2176             object_property_find(cpuobj, "kvm-no-adjvtime")) {
2177             object_property_set_bool(cpuobj, "kvm-no-adjvtime", true, NULL);
2178         }
2179 
2180         if (vmc->no_kvm_steal_time &&
2181             object_property_find(cpuobj, "kvm-steal-time")) {
2182             object_property_set_bool(cpuobj, "kvm-steal-time", false, NULL);
2183         }
2184 
2185         if (vmc->no_pmu && object_property_find(cpuobj, "pmu")) {
2186             object_property_set_bool(cpuobj, "pmu", false, NULL);
2187         }
2188 
2189         if (vmc->no_tcg_lpa2 && object_property_find(cpuobj, "lpa2")) {
2190             object_property_set_bool(cpuobj, "lpa2", false, NULL);
2191         }
2192 
2193         if (object_property_find(cpuobj, "reset-cbar")) {
2194             object_property_set_int(cpuobj, "reset-cbar",
2195                                     vms->memmap[VIRT_CPUPERIPHS].base,
2196                                     &error_abort);
2197         }
2198 
2199         object_property_set_link(cpuobj, "memory", OBJECT(sysmem),
2200                                  &error_abort);
2201         if (vms->secure) {
2202             object_property_set_link(cpuobj, "secure-memory",
2203                                      OBJECT(secure_sysmem), &error_abort);
2204         }
2205 
2206         if (vms->mte) {
2207             /* Create the memory region only once, but link to all cpus. */
2208             if (!tag_sysmem) {
2209                 /*
2210                  * The property exists only if MemTag is supported.
2211                  * If it is, we must allocate the ram to back that up.
2212                  */
2213                 if (!object_property_find(cpuobj, "tag-memory")) {
2214                     error_report("MTE requested, but not supported "
2215                                  "by the guest CPU");
2216                     exit(1);
2217                 }
2218 
2219                 tag_sysmem = g_new(MemoryRegion, 1);
2220                 memory_region_init(tag_sysmem, OBJECT(machine),
2221                                    "tag-memory", UINT64_MAX / 32);
2222 
2223                 if (vms->secure) {
2224                     secure_tag_sysmem = g_new(MemoryRegion, 1);
2225                     memory_region_init(secure_tag_sysmem, OBJECT(machine),
2226                                        "secure-tag-memory", UINT64_MAX / 32);
2227 
2228                     /* As with ram, secure-tag takes precedence over tag.  */
2229                     memory_region_add_subregion_overlap(secure_tag_sysmem, 0,
2230                                                         tag_sysmem, -1);
2231                 }
2232             }
2233 
2234             object_property_set_link(cpuobj, "tag-memory", OBJECT(tag_sysmem),
2235                                      &error_abort);
2236             if (vms->secure) {
2237                 object_property_set_link(cpuobj, "secure-tag-memory",
2238                                          OBJECT(secure_tag_sysmem),
2239                                          &error_abort);
2240             }
2241         }
2242 
2243         qdev_realize(DEVICE(cpuobj), NULL, &error_fatal);
2244         object_unref(cpuobj);
2245     }
2246     fdt_add_timer_nodes(vms);
2247     fdt_add_cpu_nodes(vms);
2248 
2249     memory_region_add_subregion(sysmem, vms->memmap[VIRT_MEM].base,
2250                                 machine->ram);
2251     if (machine->device_memory) {
2252         memory_region_add_subregion(sysmem, machine->device_memory->base,
2253                                     &machine->device_memory->mr);
2254     }
2255 
2256     virt_flash_fdt(vms, sysmem, secure_sysmem ?: sysmem);
2257 
2258     create_gic(vms, sysmem);
2259 
2260     virt_cpu_post_init(vms, sysmem);
2261 
2262     fdt_add_pmu_nodes(vms);
2263 
2264     create_uart(vms, VIRT_UART, sysmem, serial_hd(0));
2265 
2266     if (vms->secure) {
2267         create_secure_ram(vms, secure_sysmem, secure_tag_sysmem);
2268         create_uart(vms, VIRT_SECURE_UART, secure_sysmem, serial_hd(1));
2269     }
2270 
2271     if (tag_sysmem) {
2272         create_tag_ram(tag_sysmem, vms->memmap[VIRT_MEM].base,
2273                        machine->ram_size, "mach-virt.tag");
2274     }
2275 
2276     vms->highmem_ecam &= (!firmware_loaded || aarch64);
2277 
2278     create_rtc(vms);
2279 
2280     create_pcie(vms);
2281 
2282     if (has_ged && aarch64 && firmware_loaded && virt_is_acpi_enabled(vms)) {
2283         vms->acpi_dev = create_acpi_ged(vms);
2284     } else {
2285         create_gpio_devices(vms, VIRT_GPIO, sysmem);
2286     }
2287 
2288     if (vms->secure && !vmc->no_secure_gpio) {
2289         create_gpio_devices(vms, VIRT_SECURE_GPIO, secure_sysmem);
2290     }
2291 
2292      /* connect powerdown request */
2293      vms->powerdown_notifier.notify = virt_powerdown_req;
2294      qemu_register_powerdown_notifier(&vms->powerdown_notifier);
2295 
2296     /* Create mmio transports, so the user can create virtio backends
2297      * (which will be automatically plugged in to the transports). If
2298      * no backend is created the transport will just sit harmlessly idle.
2299      */
2300     create_virtio_devices(vms);
2301 
2302     vms->fw_cfg = create_fw_cfg(vms, &address_space_memory);
2303     rom_set_fw(vms->fw_cfg);
2304 
2305     create_platform_bus(vms);
2306 
2307     if (machine->nvdimms_state->is_enabled) {
2308         const struct AcpiGenericAddress arm_virt_nvdimm_acpi_dsmio = {
2309             .space_id = AML_AS_SYSTEM_MEMORY,
2310             .address = vms->memmap[VIRT_NVDIMM_ACPI].base,
2311             .bit_width = NVDIMM_ACPI_IO_LEN << 3
2312         };
2313 
2314         nvdimm_init_acpi_state(machine->nvdimms_state, sysmem,
2315                                arm_virt_nvdimm_acpi_dsmio,
2316                                vms->fw_cfg, OBJECT(vms));
2317     }
2318 
2319     vms->bootinfo.ram_size = machine->ram_size;
2320     vms->bootinfo.board_id = -1;
2321     vms->bootinfo.loader_start = vms->memmap[VIRT_MEM].base;
2322     vms->bootinfo.get_dtb = machvirt_dtb;
2323     vms->bootinfo.skip_dtb_autoload = true;
2324     vms->bootinfo.firmware_loaded = firmware_loaded;
2325     vms->bootinfo.psci_conduit = vms->psci_conduit;
2326     arm_load_kernel(ARM_CPU(first_cpu), machine, &vms->bootinfo);
2327 
2328     vms->machine_done.notify = virt_machine_done;
2329     qemu_add_machine_init_done_notifier(&vms->machine_done);
2330 }
2331 
2332 static bool virt_get_secure(Object *obj, Error **errp)
2333 {
2334     VirtMachineState *vms = VIRT_MACHINE(obj);
2335 
2336     return vms->secure;
2337 }
2338 
2339 static void virt_set_secure(Object *obj, bool value, Error **errp)
2340 {
2341     VirtMachineState *vms = VIRT_MACHINE(obj);
2342 
2343     vms->secure = value;
2344 }
2345 
2346 static bool virt_get_virt(Object *obj, Error **errp)
2347 {
2348     VirtMachineState *vms = VIRT_MACHINE(obj);
2349 
2350     return vms->virt;
2351 }
2352 
2353 static void virt_set_virt(Object *obj, bool value, Error **errp)
2354 {
2355     VirtMachineState *vms = VIRT_MACHINE(obj);
2356 
2357     vms->virt = value;
2358 }
2359 
2360 static bool virt_get_highmem(Object *obj, Error **errp)
2361 {
2362     VirtMachineState *vms = VIRT_MACHINE(obj);
2363 
2364     return vms->highmem;
2365 }
2366 
2367 static void virt_set_highmem(Object *obj, bool value, Error **errp)
2368 {
2369     VirtMachineState *vms = VIRT_MACHINE(obj);
2370 
2371     vms->highmem = value;
2372 }
2373 
2374 static bool virt_get_compact_highmem(Object *obj, Error **errp)
2375 {
2376     VirtMachineState *vms = VIRT_MACHINE(obj);
2377 
2378     return vms->highmem_compact;
2379 }
2380 
2381 static void virt_set_compact_highmem(Object *obj, bool value, Error **errp)
2382 {
2383     VirtMachineState *vms = VIRT_MACHINE(obj);
2384 
2385     vms->highmem_compact = value;
2386 }
2387 
2388 static bool virt_get_highmem_redists(Object *obj, Error **errp)
2389 {
2390     VirtMachineState *vms = VIRT_MACHINE(obj);
2391 
2392     return vms->highmem_redists;
2393 }
2394 
2395 static void virt_set_highmem_redists(Object *obj, bool value, Error **errp)
2396 {
2397     VirtMachineState *vms = VIRT_MACHINE(obj);
2398 
2399     vms->highmem_redists = value;
2400 }
2401 
2402 static bool virt_get_highmem_ecam(Object *obj, Error **errp)
2403 {
2404     VirtMachineState *vms = VIRT_MACHINE(obj);
2405 
2406     return vms->highmem_ecam;
2407 }
2408 
2409 static void virt_set_highmem_ecam(Object *obj, bool value, Error **errp)
2410 {
2411     VirtMachineState *vms = VIRT_MACHINE(obj);
2412 
2413     vms->highmem_ecam = value;
2414 }
2415 
2416 static bool virt_get_highmem_mmio(Object *obj, Error **errp)
2417 {
2418     VirtMachineState *vms = VIRT_MACHINE(obj);
2419 
2420     return vms->highmem_mmio;
2421 }
2422 
2423 static void virt_set_highmem_mmio(Object *obj, bool value, Error **errp)
2424 {
2425     VirtMachineState *vms = VIRT_MACHINE(obj);
2426 
2427     vms->highmem_mmio = value;
2428 }
2429 
2430 
2431 static bool virt_get_its(Object *obj, Error **errp)
2432 {
2433     VirtMachineState *vms = VIRT_MACHINE(obj);
2434 
2435     return vms->its;
2436 }
2437 
2438 static void virt_set_its(Object *obj, bool value, Error **errp)
2439 {
2440     VirtMachineState *vms = VIRT_MACHINE(obj);
2441 
2442     vms->its = value;
2443 }
2444 
2445 static bool virt_get_dtb_randomness(Object *obj, Error **errp)
2446 {
2447     VirtMachineState *vms = VIRT_MACHINE(obj);
2448 
2449     return vms->dtb_randomness;
2450 }
2451 
2452 static void virt_set_dtb_randomness(Object *obj, bool value, Error **errp)
2453 {
2454     VirtMachineState *vms = VIRT_MACHINE(obj);
2455 
2456     vms->dtb_randomness = value;
2457 }
2458 
2459 static char *virt_get_oem_id(Object *obj, Error **errp)
2460 {
2461     VirtMachineState *vms = VIRT_MACHINE(obj);
2462 
2463     return g_strdup(vms->oem_id);
2464 }
2465 
2466 static void virt_set_oem_id(Object *obj, const char *value, Error **errp)
2467 {
2468     VirtMachineState *vms = VIRT_MACHINE(obj);
2469     size_t len = strlen(value);
2470 
2471     if (len > 6) {
2472         error_setg(errp,
2473                    "User specified oem-id value is bigger than 6 bytes in size");
2474         return;
2475     }
2476 
2477     strncpy(vms->oem_id, value, 6);
2478 }
2479 
2480 static char *virt_get_oem_table_id(Object *obj, Error **errp)
2481 {
2482     VirtMachineState *vms = VIRT_MACHINE(obj);
2483 
2484     return g_strdup(vms->oem_table_id);
2485 }
2486 
2487 static void virt_set_oem_table_id(Object *obj, const char *value,
2488                                   Error **errp)
2489 {
2490     VirtMachineState *vms = VIRT_MACHINE(obj);
2491     size_t len = strlen(value);
2492 
2493     if (len > 8) {
2494         error_setg(errp,
2495                    "User specified oem-table-id value is bigger than 8 bytes in size");
2496         return;
2497     }
2498     strncpy(vms->oem_table_id, value, 8);
2499 }
2500 
2501 
2502 bool virt_is_acpi_enabled(VirtMachineState *vms)
2503 {
2504     if (vms->acpi == ON_OFF_AUTO_OFF) {
2505         return false;
2506     }
2507     return true;
2508 }
2509 
2510 static void virt_get_acpi(Object *obj, Visitor *v, const char *name,
2511                           void *opaque, Error **errp)
2512 {
2513     VirtMachineState *vms = VIRT_MACHINE(obj);
2514     OnOffAuto acpi = vms->acpi;
2515 
2516     visit_type_OnOffAuto(v, name, &acpi, errp);
2517 }
2518 
2519 static void virt_set_acpi(Object *obj, Visitor *v, const char *name,
2520                           void *opaque, Error **errp)
2521 {
2522     VirtMachineState *vms = VIRT_MACHINE(obj);
2523 
2524     visit_type_OnOffAuto(v, name, &vms->acpi, errp);
2525 }
2526 
2527 static bool virt_get_ras(Object *obj, Error **errp)
2528 {
2529     VirtMachineState *vms = VIRT_MACHINE(obj);
2530 
2531     return vms->ras;
2532 }
2533 
2534 static void virt_set_ras(Object *obj, bool value, Error **errp)
2535 {
2536     VirtMachineState *vms = VIRT_MACHINE(obj);
2537 
2538     vms->ras = value;
2539 }
2540 
2541 static bool virt_get_mte(Object *obj, Error **errp)
2542 {
2543     VirtMachineState *vms = VIRT_MACHINE(obj);
2544 
2545     return vms->mte;
2546 }
2547 
2548 static void virt_set_mte(Object *obj, bool value, Error **errp)
2549 {
2550     VirtMachineState *vms = VIRT_MACHINE(obj);
2551 
2552     vms->mte = value;
2553 }
2554 
2555 static char *virt_get_gic_version(Object *obj, Error **errp)
2556 {
2557     VirtMachineState *vms = VIRT_MACHINE(obj);
2558     const char *val;
2559 
2560     switch (vms->gic_version) {
2561     case VIRT_GIC_VERSION_4:
2562         val = "4";
2563         break;
2564     case VIRT_GIC_VERSION_3:
2565         val = "3";
2566         break;
2567     default:
2568         val = "2";
2569         break;
2570     }
2571     return g_strdup(val);
2572 }
2573 
2574 static void virt_set_gic_version(Object *obj, const char *value, Error **errp)
2575 {
2576     VirtMachineState *vms = VIRT_MACHINE(obj);
2577 
2578     if (!strcmp(value, "4")) {
2579         vms->gic_version = VIRT_GIC_VERSION_4;
2580     } else if (!strcmp(value, "3")) {
2581         vms->gic_version = VIRT_GIC_VERSION_3;
2582     } else if (!strcmp(value, "2")) {
2583         vms->gic_version = VIRT_GIC_VERSION_2;
2584     } else if (!strcmp(value, "host")) {
2585         vms->gic_version = VIRT_GIC_VERSION_HOST; /* Will probe later */
2586     } else if (!strcmp(value, "max")) {
2587         vms->gic_version = VIRT_GIC_VERSION_MAX; /* Will probe later */
2588     } else {
2589         error_setg(errp, "Invalid gic-version value");
2590         error_append_hint(errp, "Valid values are 3, 2, host, max.\n");
2591     }
2592 }
2593 
2594 static char *virt_get_iommu(Object *obj, Error **errp)
2595 {
2596     VirtMachineState *vms = VIRT_MACHINE(obj);
2597 
2598     switch (vms->iommu) {
2599     case VIRT_IOMMU_NONE:
2600         return g_strdup("none");
2601     case VIRT_IOMMU_SMMUV3:
2602         return g_strdup("smmuv3");
2603     default:
2604         g_assert_not_reached();
2605     }
2606 }
2607 
2608 static void virt_set_iommu(Object *obj, const char *value, Error **errp)
2609 {
2610     VirtMachineState *vms = VIRT_MACHINE(obj);
2611 
2612     if (!strcmp(value, "smmuv3")) {
2613         vms->iommu = VIRT_IOMMU_SMMUV3;
2614     } else if (!strcmp(value, "none")) {
2615         vms->iommu = VIRT_IOMMU_NONE;
2616     } else {
2617         error_setg(errp, "Invalid iommu value");
2618         error_append_hint(errp, "Valid values are none, smmuv3.\n");
2619     }
2620 }
2621 
2622 static bool virt_get_default_bus_bypass_iommu(Object *obj, Error **errp)
2623 {
2624     VirtMachineState *vms = VIRT_MACHINE(obj);
2625 
2626     return vms->default_bus_bypass_iommu;
2627 }
2628 
2629 static void virt_set_default_bus_bypass_iommu(Object *obj, bool value,
2630                                               Error **errp)
2631 {
2632     VirtMachineState *vms = VIRT_MACHINE(obj);
2633 
2634     vms->default_bus_bypass_iommu = value;
2635 }
2636 
2637 static CpuInstanceProperties
2638 virt_cpu_index_to_props(MachineState *ms, unsigned cpu_index)
2639 {
2640     MachineClass *mc = MACHINE_GET_CLASS(ms);
2641     const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(ms);
2642 
2643     assert(cpu_index < possible_cpus->len);
2644     return possible_cpus->cpus[cpu_index].props;
2645 }
2646 
2647 static int64_t virt_get_default_cpu_node_id(const MachineState *ms, int idx)
2648 {
2649     int64_t socket_id = ms->possible_cpus->cpus[idx].props.socket_id;
2650 
2651     return socket_id % ms->numa_state->num_nodes;
2652 }
2653 
2654 static const CPUArchIdList *virt_possible_cpu_arch_ids(MachineState *ms)
2655 {
2656     int n;
2657     unsigned int max_cpus = ms->smp.max_cpus;
2658     VirtMachineState *vms = VIRT_MACHINE(ms);
2659     MachineClass *mc = MACHINE_GET_CLASS(vms);
2660 
2661     if (ms->possible_cpus) {
2662         assert(ms->possible_cpus->len == max_cpus);
2663         return ms->possible_cpus;
2664     }
2665 
2666     ms->possible_cpus = g_malloc0(sizeof(CPUArchIdList) +
2667                                   sizeof(CPUArchId) * max_cpus);
2668     ms->possible_cpus->len = max_cpus;
2669     for (n = 0; n < ms->possible_cpus->len; n++) {
2670         ms->possible_cpus->cpus[n].type = ms->cpu_type;
2671         ms->possible_cpus->cpus[n].arch_id =
2672             virt_cpu_mp_affinity(vms, n);
2673 
2674         assert(!mc->smp_props.dies_supported);
2675         ms->possible_cpus->cpus[n].props.has_socket_id = true;
2676         ms->possible_cpus->cpus[n].props.socket_id =
2677             n / (ms->smp.clusters * ms->smp.cores * ms->smp.threads);
2678         ms->possible_cpus->cpus[n].props.has_cluster_id = true;
2679         ms->possible_cpus->cpus[n].props.cluster_id =
2680             (n / (ms->smp.cores * ms->smp.threads)) % ms->smp.clusters;
2681         ms->possible_cpus->cpus[n].props.has_core_id = true;
2682         ms->possible_cpus->cpus[n].props.core_id =
2683             (n / ms->smp.threads) % ms->smp.cores;
2684         ms->possible_cpus->cpus[n].props.has_thread_id = true;
2685         ms->possible_cpus->cpus[n].props.thread_id =
2686             n % ms->smp.threads;
2687     }
2688     return ms->possible_cpus;
2689 }
2690 
2691 static void virt_memory_pre_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
2692                                  Error **errp)
2693 {
2694     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2695     const MachineState *ms = MACHINE(hotplug_dev);
2696     const bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2697 
2698     if (!vms->acpi_dev) {
2699         error_setg(errp,
2700                    "memory hotplug is not enabled: missing acpi-ged device");
2701         return;
2702     }
2703 
2704     if (vms->mte) {
2705         error_setg(errp, "memory hotplug is not enabled: MTE is enabled");
2706         return;
2707     }
2708 
2709     if (is_nvdimm && !ms->nvdimms_state->is_enabled) {
2710         error_setg(errp, "nvdimm is not enabled: add 'nvdimm=on' to '-M'");
2711         return;
2712     }
2713 
2714     pc_dimm_pre_plug(PC_DIMM(dev), MACHINE(hotplug_dev), NULL, errp);
2715 }
2716 
2717 static void virt_memory_plug(HotplugHandler *hotplug_dev,
2718                              DeviceState *dev, Error **errp)
2719 {
2720     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2721     MachineState *ms = MACHINE(hotplug_dev);
2722     bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2723 
2724     pc_dimm_plug(PC_DIMM(dev), MACHINE(vms));
2725 
2726     if (is_nvdimm) {
2727         nvdimm_plug(ms->nvdimms_state);
2728     }
2729 
2730     hotplug_handler_plug(HOTPLUG_HANDLER(vms->acpi_dev),
2731                          dev, &error_abort);
2732 }
2733 
2734 static void virt_virtio_md_pci_pre_plug(HotplugHandler *hotplug_dev,
2735                                         DeviceState *dev, Error **errp)
2736 {
2737     HotplugHandler *hotplug_dev2 = qdev_get_bus_hotplug_handler(dev);
2738     Error *local_err = NULL;
2739 
2740     if (!hotplug_dev2 && dev->hotplugged) {
2741         /*
2742          * Without a bus hotplug handler, we cannot control the plug/unplug
2743          * order. We should never reach this point when hotplugging on ARM.
2744          * However, it's nice to add a safety net, similar to what we have
2745          * on x86.
2746          */
2747         error_setg(errp, "hotplug of virtio based memory devices not supported"
2748                    " on this bus.");
2749         return;
2750     }
2751     /*
2752      * First, see if we can plug this memory device at all. If that
2753      * succeeds, branch of to the actual hotplug handler.
2754      */
2755     memory_device_pre_plug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev), NULL,
2756                            &local_err);
2757     if (!local_err && hotplug_dev2) {
2758         hotplug_handler_pre_plug(hotplug_dev2, dev, &local_err);
2759     }
2760     error_propagate(errp, local_err);
2761 }
2762 
2763 static void virt_virtio_md_pci_plug(HotplugHandler *hotplug_dev,
2764                                     DeviceState *dev, Error **errp)
2765 {
2766     HotplugHandler *hotplug_dev2 = qdev_get_bus_hotplug_handler(dev);
2767     Error *local_err = NULL;
2768 
2769     /*
2770      * Plug the memory device first and then branch off to the actual
2771      * hotplug handler. If that one fails, we can easily undo the memory
2772      * device bits.
2773      */
2774     memory_device_plug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev));
2775     if (hotplug_dev2) {
2776         hotplug_handler_plug(hotplug_dev2, dev, &local_err);
2777         if (local_err) {
2778             memory_device_unplug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev));
2779         }
2780     }
2781     error_propagate(errp, local_err);
2782 }
2783 
2784 static void virt_virtio_md_pci_unplug_request(HotplugHandler *hotplug_dev,
2785                                               DeviceState *dev, Error **errp)
2786 {
2787     /* We don't support hot unplug of virtio based memory devices */
2788     error_setg(errp, "virtio based memory devices cannot be unplugged.");
2789 }
2790 
2791 
2792 static void virt_machine_device_pre_plug_cb(HotplugHandler *hotplug_dev,
2793                                             DeviceState *dev, Error **errp)
2794 {
2795     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2796 
2797     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2798         virt_memory_pre_plug(hotplug_dev, dev, errp);
2799     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI)) {
2800         virt_virtio_md_pci_pre_plug(hotplug_dev, dev, errp);
2801     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2802         hwaddr db_start = 0, db_end = 0;
2803         char *resv_prop_str;
2804 
2805         if (vms->iommu != VIRT_IOMMU_NONE) {
2806             error_setg(errp, "virt machine does not support multiple IOMMUs");
2807             return;
2808         }
2809 
2810         switch (vms->msi_controller) {
2811         case VIRT_MSI_CTRL_NONE:
2812             return;
2813         case VIRT_MSI_CTRL_ITS:
2814             /* GITS_TRANSLATER page */
2815             db_start = base_memmap[VIRT_GIC_ITS].base + 0x10000;
2816             db_end = base_memmap[VIRT_GIC_ITS].base +
2817                      base_memmap[VIRT_GIC_ITS].size - 1;
2818             break;
2819         case VIRT_MSI_CTRL_GICV2M:
2820             /* MSI_SETSPI_NS page */
2821             db_start = base_memmap[VIRT_GIC_V2M].base;
2822             db_end = db_start + base_memmap[VIRT_GIC_V2M].size - 1;
2823             break;
2824         }
2825         resv_prop_str = g_strdup_printf("0x%"PRIx64":0x%"PRIx64":%u",
2826                                         db_start, db_end,
2827                                         VIRTIO_IOMMU_RESV_MEM_T_MSI);
2828 
2829         object_property_set_uint(OBJECT(dev), "len-reserved-regions", 1, errp);
2830         object_property_set_str(OBJECT(dev), "reserved-regions[0]",
2831                                 resv_prop_str, errp);
2832         g_free(resv_prop_str);
2833     }
2834 }
2835 
2836 static void virt_machine_device_plug_cb(HotplugHandler *hotplug_dev,
2837                                         DeviceState *dev, Error **errp)
2838 {
2839     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2840 
2841     if (vms->platform_bus_dev) {
2842         MachineClass *mc = MACHINE_GET_CLASS(vms);
2843 
2844         if (device_is_dynamic_sysbus(mc, dev)) {
2845             platform_bus_link_device(PLATFORM_BUS_DEVICE(vms->platform_bus_dev),
2846                                      SYS_BUS_DEVICE(dev));
2847         }
2848     }
2849     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2850         virt_memory_plug(hotplug_dev, dev, errp);
2851     }
2852 
2853     if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI)) {
2854         virt_virtio_md_pci_plug(hotplug_dev, dev, errp);
2855     }
2856 
2857     if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2858         PCIDevice *pdev = PCI_DEVICE(dev);
2859 
2860         vms->iommu = VIRT_IOMMU_VIRTIO;
2861         vms->virtio_iommu_bdf = pci_get_bdf(pdev);
2862         create_virtio_iommu_dt_bindings(vms);
2863     }
2864 }
2865 
2866 static void virt_dimm_unplug_request(HotplugHandler *hotplug_dev,
2867                                      DeviceState *dev, Error **errp)
2868 {
2869     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2870 
2871     if (!vms->acpi_dev) {
2872         error_setg(errp,
2873                    "memory hotplug is not enabled: missing acpi-ged device");
2874         return;
2875     }
2876 
2877     if (object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM)) {
2878         error_setg(errp, "nvdimm device hot unplug is not supported yet.");
2879         return;
2880     }
2881 
2882     hotplug_handler_unplug_request(HOTPLUG_HANDLER(vms->acpi_dev), dev,
2883                                    errp);
2884 }
2885 
2886 static void virt_dimm_unplug(HotplugHandler *hotplug_dev,
2887                              DeviceState *dev, Error **errp)
2888 {
2889     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2890     Error *local_err = NULL;
2891 
2892     hotplug_handler_unplug(HOTPLUG_HANDLER(vms->acpi_dev), dev, &local_err);
2893     if (local_err) {
2894         goto out;
2895     }
2896 
2897     pc_dimm_unplug(PC_DIMM(dev), MACHINE(vms));
2898     qdev_unrealize(dev);
2899 
2900 out:
2901     error_propagate(errp, local_err);
2902 }
2903 
2904 static void virt_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
2905                                           DeviceState *dev, Error **errp)
2906 {
2907     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2908         virt_dimm_unplug_request(hotplug_dev, dev, errp);
2909     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI)) {
2910         virt_virtio_md_pci_unplug_request(hotplug_dev, dev, errp);
2911     } else {
2912         error_setg(errp, "device unplug request for unsupported device"
2913                    " type: %s", object_get_typename(OBJECT(dev)));
2914     }
2915 }
2916 
2917 static void virt_machine_device_unplug_cb(HotplugHandler *hotplug_dev,
2918                                           DeviceState *dev, Error **errp)
2919 {
2920     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2921         virt_dimm_unplug(hotplug_dev, dev, errp);
2922     } else {
2923         error_setg(errp, "virt: device unplug for unsupported device"
2924                    " type: %s", object_get_typename(OBJECT(dev)));
2925     }
2926 }
2927 
2928 static HotplugHandler *virt_machine_get_hotplug_handler(MachineState *machine,
2929                                                         DeviceState *dev)
2930 {
2931     MachineClass *mc = MACHINE_GET_CLASS(machine);
2932 
2933     if (device_is_dynamic_sysbus(mc, dev) ||
2934         object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) ||
2935         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI) ||
2936         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2937         return HOTPLUG_HANDLER(machine);
2938     }
2939     return NULL;
2940 }
2941 
2942 /*
2943  * for arm64 kvm_type [7-0] encodes the requested number of bits
2944  * in the IPA address space
2945  */
2946 static int virt_kvm_type(MachineState *ms, const char *type_str)
2947 {
2948     VirtMachineState *vms = VIRT_MACHINE(ms);
2949     int max_vm_pa_size, requested_pa_size;
2950     bool fixed_ipa;
2951 
2952     max_vm_pa_size = kvm_arm_get_max_vm_ipa_size(ms, &fixed_ipa);
2953 
2954     /* we freeze the memory map to compute the highest gpa */
2955     virt_set_memmap(vms, max_vm_pa_size);
2956 
2957     requested_pa_size = 64 - clz64(vms->highest_gpa);
2958 
2959     /*
2960      * KVM requires the IPA size to be at least 32 bits.
2961      */
2962     if (requested_pa_size < 32) {
2963         requested_pa_size = 32;
2964     }
2965 
2966     if (requested_pa_size > max_vm_pa_size) {
2967         error_report("-m and ,maxmem option values "
2968                      "require an IPA range (%d bits) larger than "
2969                      "the one supported by the host (%d bits)",
2970                      requested_pa_size, max_vm_pa_size);
2971         exit(1);
2972     }
2973     /*
2974      * We return the requested PA log size, unless KVM only supports
2975      * the implicit legacy 40b IPA setting, in which case the kvm_type
2976      * must be 0.
2977      */
2978     return fixed_ipa ? 0 : requested_pa_size;
2979 }
2980 
2981 static void virt_machine_class_init(ObjectClass *oc, void *data)
2982 {
2983     MachineClass *mc = MACHINE_CLASS(oc);
2984     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
2985 
2986     mc->init = machvirt_init;
2987     /* Start with max_cpus set to 512, which is the maximum supported by KVM.
2988      * The value may be reduced later when we have more information about the
2989      * configuration of the particular instance.
2990      */
2991     mc->max_cpus = 512;
2992     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_CALXEDA_XGMAC);
2993     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_AMD_XGBE);
2994     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_RAMFB_DEVICE);
2995     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_PLATFORM);
2996 #ifdef CONFIG_TPM
2997     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_TPM_TIS_SYSBUS);
2998 #endif
2999     mc->block_default_type = IF_VIRTIO;
3000     mc->no_cdrom = 1;
3001     mc->pci_allow_0_address = true;
3002     /* We know we will never create a pre-ARMv7 CPU which needs 1K pages */
3003     mc->minimum_page_bits = 12;
3004     mc->possible_cpu_arch_ids = virt_possible_cpu_arch_ids;
3005     mc->cpu_index_to_instance_props = virt_cpu_index_to_props;
3006     mc->default_cpu_type = ARM_CPU_TYPE_NAME("cortex-a15");
3007     mc->get_default_cpu_node_id = virt_get_default_cpu_node_id;
3008     mc->kvm_type = virt_kvm_type;
3009     assert(!mc->get_hotplug_handler);
3010     mc->get_hotplug_handler = virt_machine_get_hotplug_handler;
3011     hc->pre_plug = virt_machine_device_pre_plug_cb;
3012     hc->plug = virt_machine_device_plug_cb;
3013     hc->unplug_request = virt_machine_device_unplug_request_cb;
3014     hc->unplug = virt_machine_device_unplug_cb;
3015     mc->nvdimm_supported = true;
3016     mc->smp_props.clusters_supported = true;
3017     mc->auto_enable_numa_with_memhp = true;
3018     mc->auto_enable_numa_with_memdev = true;
3019     mc->default_ram_id = "mach-virt.ram";
3020 
3021     object_class_property_add(oc, "acpi", "OnOffAuto",
3022         virt_get_acpi, virt_set_acpi,
3023         NULL, NULL);
3024     object_class_property_set_description(oc, "acpi",
3025         "Enable ACPI");
3026     object_class_property_add_bool(oc, "secure", virt_get_secure,
3027                                    virt_set_secure);
3028     object_class_property_set_description(oc, "secure",
3029                                                 "Set on/off to enable/disable the ARM "
3030                                                 "Security Extensions (TrustZone)");
3031 
3032     object_class_property_add_bool(oc, "virtualization", virt_get_virt,
3033                                    virt_set_virt);
3034     object_class_property_set_description(oc, "virtualization",
3035                                           "Set on/off to enable/disable emulating a "
3036                                           "guest CPU which implements the ARM "
3037                                           "Virtualization Extensions");
3038 
3039     object_class_property_add_bool(oc, "highmem", virt_get_highmem,
3040                                    virt_set_highmem);
3041     object_class_property_set_description(oc, "highmem",
3042                                           "Set on/off to enable/disable using "
3043                                           "physical address space above 32 bits");
3044 
3045     object_class_property_add_bool(oc, "compact-highmem",
3046                                    virt_get_compact_highmem,
3047                                    virt_set_compact_highmem);
3048     object_class_property_set_description(oc, "compact-highmem",
3049                                           "Set on/off to enable/disable compact "
3050                                           "layout for high memory regions");
3051 
3052     object_class_property_add_bool(oc, "highmem-redists",
3053                                    virt_get_highmem_redists,
3054                                    virt_set_highmem_redists);
3055     object_class_property_set_description(oc, "highmem-redists",
3056                                           "Set on/off to enable/disable high "
3057                                           "memory region for GICv3 or GICv4 "
3058                                           "redistributor");
3059 
3060     object_class_property_add_bool(oc, "highmem-ecam",
3061                                    virt_get_highmem_ecam,
3062                                    virt_set_highmem_ecam);
3063     object_class_property_set_description(oc, "highmem-ecam",
3064                                           "Set on/off to enable/disable high "
3065                                           "memory region for PCI ECAM");
3066 
3067     object_class_property_add_bool(oc, "highmem-mmio",
3068                                    virt_get_highmem_mmio,
3069                                    virt_set_highmem_mmio);
3070     object_class_property_set_description(oc, "highmem-mmio",
3071                                           "Set on/off to enable/disable high "
3072                                           "memory region for PCI MMIO");
3073 
3074     object_class_property_add_str(oc, "gic-version", virt_get_gic_version,
3075                                   virt_set_gic_version);
3076     object_class_property_set_description(oc, "gic-version",
3077                                           "Set GIC version. "
3078                                           "Valid values are 2, 3, 4, host and max");
3079 
3080     object_class_property_add_str(oc, "iommu", virt_get_iommu, virt_set_iommu);
3081     object_class_property_set_description(oc, "iommu",
3082                                           "Set the IOMMU type. "
3083                                           "Valid values are none and smmuv3");
3084 
3085     object_class_property_add_bool(oc, "default-bus-bypass-iommu",
3086                                    virt_get_default_bus_bypass_iommu,
3087                                    virt_set_default_bus_bypass_iommu);
3088     object_class_property_set_description(oc, "default-bus-bypass-iommu",
3089                                           "Set on/off to enable/disable "
3090                                           "bypass_iommu for default root bus");
3091 
3092     object_class_property_add_bool(oc, "ras", virt_get_ras,
3093                                    virt_set_ras);
3094     object_class_property_set_description(oc, "ras",
3095                                           "Set on/off to enable/disable reporting host memory errors "
3096                                           "to a KVM guest using ACPI and guest external abort exceptions");
3097 
3098     object_class_property_add_bool(oc, "mte", virt_get_mte, virt_set_mte);
3099     object_class_property_set_description(oc, "mte",
3100                                           "Set on/off to enable/disable emulating a "
3101                                           "guest CPU which implements the ARM "
3102                                           "Memory Tagging Extension");
3103 
3104     object_class_property_add_bool(oc, "its", virt_get_its,
3105                                    virt_set_its);
3106     object_class_property_set_description(oc, "its",
3107                                           "Set on/off to enable/disable "
3108                                           "ITS instantiation");
3109 
3110     object_class_property_add_bool(oc, "dtb-randomness",
3111                                    virt_get_dtb_randomness,
3112                                    virt_set_dtb_randomness);
3113     object_class_property_set_description(oc, "dtb-randomness",
3114                                           "Set off to disable passing random or "
3115                                           "non-deterministic dtb nodes to guest");
3116 
3117     object_class_property_add_bool(oc, "dtb-kaslr-seed",
3118                                    virt_get_dtb_randomness,
3119                                    virt_set_dtb_randomness);
3120     object_class_property_set_description(oc, "dtb-kaslr-seed",
3121                                           "Deprecated synonym of dtb-randomness");
3122 
3123     object_class_property_add_str(oc, "x-oem-id",
3124                                   virt_get_oem_id,
3125                                   virt_set_oem_id);
3126     object_class_property_set_description(oc, "x-oem-id",
3127                                           "Override the default value of field OEMID "
3128                                           "in ACPI table header."
3129                                           "The string may be up to 6 bytes in size");
3130 
3131 
3132     object_class_property_add_str(oc, "x-oem-table-id",
3133                                   virt_get_oem_table_id,
3134                                   virt_set_oem_table_id);
3135     object_class_property_set_description(oc, "x-oem-table-id",
3136                                           "Override the default value of field OEM Table ID "
3137                                           "in ACPI table header."
3138                                           "The string may be up to 8 bytes in size");
3139 
3140 }
3141 
3142 static void virt_instance_init(Object *obj)
3143 {
3144     VirtMachineState *vms = VIRT_MACHINE(obj);
3145     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
3146 
3147     /* EL3 is disabled by default on virt: this makes us consistent
3148      * between KVM and TCG for this board, and it also allows us to
3149      * boot UEFI blobs which assume no TrustZone support.
3150      */
3151     vms->secure = false;
3152 
3153     /* EL2 is also disabled by default, for similar reasons */
3154     vms->virt = false;
3155 
3156     /* High memory is enabled by default */
3157     vms->highmem = true;
3158     vms->highmem_compact = !vmc->no_highmem_compact;
3159     vms->gic_version = VIRT_GIC_VERSION_NOSEL;
3160 
3161     vms->highmem_ecam = !vmc->no_highmem_ecam;
3162     vms->highmem_mmio = true;
3163     vms->highmem_redists = true;
3164 
3165     if (vmc->no_its) {
3166         vms->its = false;
3167     } else {
3168         /* Default allows ITS instantiation */
3169         vms->its = true;
3170 
3171         if (vmc->no_tcg_its) {
3172             vms->tcg_its = false;
3173         } else {
3174             vms->tcg_its = true;
3175         }
3176     }
3177 
3178     /* Default disallows iommu instantiation */
3179     vms->iommu = VIRT_IOMMU_NONE;
3180 
3181     /* The default root bus is attached to iommu by default */
3182     vms->default_bus_bypass_iommu = false;
3183 
3184     /* Default disallows RAS instantiation */
3185     vms->ras = false;
3186 
3187     /* MTE is disabled by default.  */
3188     vms->mte = false;
3189 
3190     /* Supply kaslr-seed and rng-seed by default */
3191     vms->dtb_randomness = true;
3192 
3193     vms->irqmap = a15irqmap;
3194 
3195     virt_flash_create(vms);
3196 
3197     vms->oem_id = g_strndup(ACPI_BUILD_APPNAME6, 6);
3198     vms->oem_table_id = g_strndup(ACPI_BUILD_APPNAME8, 8);
3199 }
3200 
3201 static const TypeInfo virt_machine_info = {
3202     .name          = TYPE_VIRT_MACHINE,
3203     .parent        = TYPE_MACHINE,
3204     .abstract      = true,
3205     .instance_size = sizeof(VirtMachineState),
3206     .class_size    = sizeof(VirtMachineClass),
3207     .class_init    = virt_machine_class_init,
3208     .instance_init = virt_instance_init,
3209     .interfaces = (InterfaceInfo[]) {
3210          { TYPE_HOTPLUG_HANDLER },
3211          { }
3212     },
3213 };
3214 
3215 static void machvirt_machine_init(void)
3216 {
3217     type_register_static(&virt_machine_info);
3218 }
3219 type_init(machvirt_machine_init);
3220 
3221 static void virt_machine_8_0_options(MachineClass *mc)
3222 {
3223 }
3224 DEFINE_VIRT_MACHINE_AS_LATEST(8, 0)
3225 
3226 static void virt_machine_7_2_options(MachineClass *mc)
3227 {
3228     virt_machine_8_0_options(mc);
3229     compat_props_add(mc->compat_props, hw_compat_7_2, hw_compat_7_2_len);
3230 }
3231 DEFINE_VIRT_MACHINE(7, 2)
3232 
3233 static void virt_machine_7_1_options(MachineClass *mc)
3234 {
3235     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3236 
3237     virt_machine_7_2_options(mc);
3238     compat_props_add(mc->compat_props, hw_compat_7_1, hw_compat_7_1_len);
3239     /* Compact layout for high memory regions was introduced with 7.2 */
3240     vmc->no_highmem_compact = true;
3241 }
3242 DEFINE_VIRT_MACHINE(7, 1)
3243 
3244 static void virt_machine_7_0_options(MachineClass *mc)
3245 {
3246     virt_machine_7_1_options(mc);
3247     compat_props_add(mc->compat_props, hw_compat_7_0, hw_compat_7_0_len);
3248 }
3249 DEFINE_VIRT_MACHINE(7, 0)
3250 
3251 static void virt_machine_6_2_options(MachineClass *mc)
3252 {
3253     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3254 
3255     virt_machine_7_0_options(mc);
3256     compat_props_add(mc->compat_props, hw_compat_6_2, hw_compat_6_2_len);
3257     vmc->no_tcg_lpa2 = true;
3258 }
3259 DEFINE_VIRT_MACHINE(6, 2)
3260 
3261 static void virt_machine_6_1_options(MachineClass *mc)
3262 {
3263     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3264 
3265     virt_machine_6_2_options(mc);
3266     compat_props_add(mc->compat_props, hw_compat_6_1, hw_compat_6_1_len);
3267     mc->smp_props.prefer_sockets = true;
3268     vmc->no_cpu_topology = true;
3269 
3270     /* qemu ITS was introduced with 6.2 */
3271     vmc->no_tcg_its = true;
3272 }
3273 DEFINE_VIRT_MACHINE(6, 1)
3274 
3275 static void virt_machine_6_0_options(MachineClass *mc)
3276 {
3277     virt_machine_6_1_options(mc);
3278     compat_props_add(mc->compat_props, hw_compat_6_0, hw_compat_6_0_len);
3279 }
3280 DEFINE_VIRT_MACHINE(6, 0)
3281 
3282 static void virt_machine_5_2_options(MachineClass *mc)
3283 {
3284     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3285 
3286     virt_machine_6_0_options(mc);
3287     compat_props_add(mc->compat_props, hw_compat_5_2, hw_compat_5_2_len);
3288     vmc->no_secure_gpio = true;
3289 }
3290 DEFINE_VIRT_MACHINE(5, 2)
3291 
3292 static void virt_machine_5_1_options(MachineClass *mc)
3293 {
3294     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3295 
3296     virt_machine_5_2_options(mc);
3297     compat_props_add(mc->compat_props, hw_compat_5_1, hw_compat_5_1_len);
3298     vmc->no_kvm_steal_time = true;
3299 }
3300 DEFINE_VIRT_MACHINE(5, 1)
3301 
3302 static void virt_machine_5_0_options(MachineClass *mc)
3303 {
3304     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3305 
3306     virt_machine_5_1_options(mc);
3307     compat_props_add(mc->compat_props, hw_compat_5_0, hw_compat_5_0_len);
3308     mc->numa_mem_supported = true;
3309     vmc->acpi_expose_flash = true;
3310     mc->auto_enable_numa_with_memdev = false;
3311 }
3312 DEFINE_VIRT_MACHINE(5, 0)
3313 
3314 static void virt_machine_4_2_options(MachineClass *mc)
3315 {
3316     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3317 
3318     virt_machine_5_0_options(mc);
3319     compat_props_add(mc->compat_props, hw_compat_4_2, hw_compat_4_2_len);
3320     vmc->kvm_no_adjvtime = true;
3321 }
3322 DEFINE_VIRT_MACHINE(4, 2)
3323 
3324 static void virt_machine_4_1_options(MachineClass *mc)
3325 {
3326     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3327 
3328     virt_machine_4_2_options(mc);
3329     compat_props_add(mc->compat_props, hw_compat_4_1, hw_compat_4_1_len);
3330     vmc->no_ged = true;
3331     mc->auto_enable_numa_with_memhp = false;
3332 }
3333 DEFINE_VIRT_MACHINE(4, 1)
3334 
3335 static void virt_machine_4_0_options(MachineClass *mc)
3336 {
3337     virt_machine_4_1_options(mc);
3338     compat_props_add(mc->compat_props, hw_compat_4_0, hw_compat_4_0_len);
3339 }
3340 DEFINE_VIRT_MACHINE(4, 0)
3341 
3342 static void virt_machine_3_1_options(MachineClass *mc)
3343 {
3344     virt_machine_4_0_options(mc);
3345     compat_props_add(mc->compat_props, hw_compat_3_1, hw_compat_3_1_len);
3346 }
3347 DEFINE_VIRT_MACHINE(3, 1)
3348 
3349 static void virt_machine_3_0_options(MachineClass *mc)
3350 {
3351     virt_machine_3_1_options(mc);
3352     compat_props_add(mc->compat_props, hw_compat_3_0, hw_compat_3_0_len);
3353 }
3354 DEFINE_VIRT_MACHINE(3, 0)
3355 
3356 static void virt_machine_2_12_options(MachineClass *mc)
3357 {
3358     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3359 
3360     virt_machine_3_0_options(mc);
3361     compat_props_add(mc->compat_props, hw_compat_2_12, hw_compat_2_12_len);
3362     vmc->no_highmem_ecam = true;
3363     mc->max_cpus = 255;
3364 }
3365 DEFINE_VIRT_MACHINE(2, 12)
3366 
3367 static void virt_machine_2_11_options(MachineClass *mc)
3368 {
3369     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3370 
3371     virt_machine_2_12_options(mc);
3372     compat_props_add(mc->compat_props, hw_compat_2_11, hw_compat_2_11_len);
3373     vmc->smbios_old_sys_ver = true;
3374 }
3375 DEFINE_VIRT_MACHINE(2, 11)
3376 
3377 static void virt_machine_2_10_options(MachineClass *mc)
3378 {
3379     virt_machine_2_11_options(mc);
3380     compat_props_add(mc->compat_props, hw_compat_2_10, hw_compat_2_10_len);
3381     /* before 2.11 we never faulted accesses to bad addresses */
3382     mc->ignore_memory_transaction_failures = true;
3383 }
3384 DEFINE_VIRT_MACHINE(2, 10)
3385 
3386 static void virt_machine_2_9_options(MachineClass *mc)
3387 {
3388     virt_machine_2_10_options(mc);
3389     compat_props_add(mc->compat_props, hw_compat_2_9, hw_compat_2_9_len);
3390 }
3391 DEFINE_VIRT_MACHINE(2, 9)
3392 
3393 static void virt_machine_2_8_options(MachineClass *mc)
3394 {
3395     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3396 
3397     virt_machine_2_9_options(mc);
3398     compat_props_add(mc->compat_props, hw_compat_2_8, hw_compat_2_8_len);
3399     /* For 2.8 and earlier we falsely claimed in the DT that
3400      * our timers were edge-triggered, not level-triggered.
3401      */
3402     vmc->claim_edge_triggered_timers = true;
3403 }
3404 DEFINE_VIRT_MACHINE(2, 8)
3405 
3406 static void virt_machine_2_7_options(MachineClass *mc)
3407 {
3408     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3409 
3410     virt_machine_2_8_options(mc);
3411     compat_props_add(mc->compat_props, hw_compat_2_7, hw_compat_2_7_len);
3412     /* ITS was introduced with 2.8 */
3413     vmc->no_its = true;
3414     /* Stick with 1K pages for migration compatibility */
3415     mc->minimum_page_bits = 0;
3416 }
3417 DEFINE_VIRT_MACHINE(2, 7)
3418 
3419 static void virt_machine_2_6_options(MachineClass *mc)
3420 {
3421     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3422 
3423     virt_machine_2_7_options(mc);
3424     compat_props_add(mc->compat_props, hw_compat_2_6, hw_compat_2_6_len);
3425     vmc->disallow_affinity_adjustment = true;
3426     /* Disable PMU for 2.6 as PMU support was first introduced in 2.7 */
3427     vmc->no_pmu = true;
3428 }
3429 DEFINE_VIRT_MACHINE(2, 6)
3430