xref: /qemu/hw/arm/virt.c (revision 52581c71)
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 static MemMapEntry extended_memmap[] = {
179     /* Additional 64 MB redist region (can contain up to 512 redistributors) */
180     [VIRT_HIGH_GIC_REDIST2] =   { 0x0, 64 * MiB },
181     [VIRT_HIGH_PCIE_ECAM] =     { 0x0, 256 * MiB },
182     /* Second PCIe window */
183     [VIRT_HIGH_PCIE_MMIO] =     { 0x0, 512 * GiB },
184 };
185 
186 static const int a15irqmap[] = {
187     [VIRT_UART] = 1,
188     [VIRT_RTC] = 2,
189     [VIRT_PCIE] = 3, /* ... to 6 */
190     [VIRT_GPIO] = 7,
191     [VIRT_SECURE_UART] = 8,
192     [VIRT_ACPI_GED] = 9,
193     [VIRT_MMIO] = 16, /* ...to 16 + NUM_VIRTIO_TRANSPORTS - 1 */
194     [VIRT_GIC_V2M] = 48, /* ...to 48 + NUM_GICV2M_SPIS - 1 */
195     [VIRT_SMMU] = 74,    /* ...to 74 + NUM_SMMU_IRQS - 1 */
196     [VIRT_PLATFORM_BUS] = 112, /* ...to 112 + PLATFORM_BUS_NUM_IRQS -1 */
197 };
198 
199 static const char *valid_cpus[] = {
200     ARM_CPU_TYPE_NAME("cortex-a7"),
201     ARM_CPU_TYPE_NAME("cortex-a15"),
202     ARM_CPU_TYPE_NAME("cortex-a53"),
203     ARM_CPU_TYPE_NAME("cortex-a57"),
204     ARM_CPU_TYPE_NAME("cortex-a72"),
205     ARM_CPU_TYPE_NAME("cortex-a76"),
206     ARM_CPU_TYPE_NAME("a64fx"),
207     ARM_CPU_TYPE_NAME("neoverse-n1"),
208     ARM_CPU_TYPE_NAME("host"),
209     ARM_CPU_TYPE_NAME("max"),
210 };
211 
212 static bool cpu_type_valid(const char *cpu)
213 {
214     int i;
215 
216     for (i = 0; i < ARRAY_SIZE(valid_cpus); i++) {
217         if (strcmp(cpu, valid_cpus[i]) == 0) {
218             return true;
219         }
220     }
221     return false;
222 }
223 
224 static void create_kaslr_seed(MachineState *ms, const char *node)
225 {
226     uint64_t seed;
227 
228     if (qemu_guest_getrandom(&seed, sizeof(seed), NULL)) {
229         return;
230     }
231     qemu_fdt_setprop_u64(ms->fdt, node, "kaslr-seed", seed);
232 }
233 
234 static void create_fdt(VirtMachineState *vms)
235 {
236     MachineState *ms = MACHINE(vms);
237     int nb_numa_nodes = ms->numa_state->num_nodes;
238     void *fdt = create_device_tree(&vms->fdt_size);
239 
240     if (!fdt) {
241         error_report("create_device_tree() failed");
242         exit(1);
243     }
244 
245     ms->fdt = fdt;
246 
247     /* Header */
248     qemu_fdt_setprop_string(fdt, "/", "compatible", "linux,dummy-virt");
249     qemu_fdt_setprop_cell(fdt, "/", "#address-cells", 0x2);
250     qemu_fdt_setprop_cell(fdt, "/", "#size-cells", 0x2);
251 
252     /* /chosen must exist for load_dtb to fill in necessary properties later */
253     qemu_fdt_add_subnode(fdt, "/chosen");
254     if (vms->dtb_kaslr_seed) {
255         create_kaslr_seed(ms, "/chosen");
256     }
257 
258     if (vms->secure) {
259         qemu_fdt_add_subnode(fdt, "/secure-chosen");
260         if (vms->dtb_kaslr_seed) {
261             create_kaslr_seed(ms, "/secure-chosen");
262         }
263     }
264 
265     /* Clock node, for the benefit of the UART. The kernel device tree
266      * binding documentation claims the PL011 node clock properties are
267      * optional but in practice if you omit them the kernel refuses to
268      * probe for the device.
269      */
270     vms->clock_phandle = qemu_fdt_alloc_phandle(fdt);
271     qemu_fdt_add_subnode(fdt, "/apb-pclk");
272     qemu_fdt_setprop_string(fdt, "/apb-pclk", "compatible", "fixed-clock");
273     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "#clock-cells", 0x0);
274     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "clock-frequency", 24000000);
275     qemu_fdt_setprop_string(fdt, "/apb-pclk", "clock-output-names",
276                                 "clk24mhz");
277     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "phandle", vms->clock_phandle);
278 
279     if (nb_numa_nodes > 0 && ms->numa_state->have_numa_distance) {
280         int size = nb_numa_nodes * nb_numa_nodes * 3 * sizeof(uint32_t);
281         uint32_t *matrix = g_malloc0(size);
282         int idx, i, j;
283 
284         for (i = 0; i < nb_numa_nodes; i++) {
285             for (j = 0; j < nb_numa_nodes; j++) {
286                 idx = (i * nb_numa_nodes + j) * 3;
287                 matrix[idx + 0] = cpu_to_be32(i);
288                 matrix[idx + 1] = cpu_to_be32(j);
289                 matrix[idx + 2] =
290                     cpu_to_be32(ms->numa_state->nodes[i].distance[j]);
291             }
292         }
293 
294         qemu_fdt_add_subnode(fdt, "/distance-map");
295         qemu_fdt_setprop_string(fdt, "/distance-map", "compatible",
296                                 "numa-distance-map-v1");
297         qemu_fdt_setprop(fdt, "/distance-map", "distance-matrix",
298                          matrix, size);
299         g_free(matrix);
300     }
301 }
302 
303 static void fdt_add_timer_nodes(const VirtMachineState *vms)
304 {
305     /* On real hardware these interrupts are level-triggered.
306      * On KVM they were edge-triggered before host kernel version 4.4,
307      * and level-triggered afterwards.
308      * On emulated QEMU they are level-triggered.
309      *
310      * Getting the DTB info about them wrong is awkward for some
311      * guest kernels:
312      *  pre-4.8 ignore the DT and leave the interrupt configured
313      *   with whatever the GIC reset value (or the bootloader) left it at
314      *  4.8 before rc6 honour the incorrect data by programming it back
315      *   into the GIC, causing problems
316      *  4.8rc6 and later ignore the DT and always write "level triggered"
317      *   into the GIC
318      *
319      * For backwards-compatibility, virt-2.8 and earlier will continue
320      * to say these are edge-triggered, but later machines will report
321      * the correct information.
322      */
323     ARMCPU *armcpu;
324     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
325     uint32_t irqflags = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
326     MachineState *ms = MACHINE(vms);
327 
328     if (vmc->claim_edge_triggered_timers) {
329         irqflags = GIC_FDT_IRQ_FLAGS_EDGE_LO_HI;
330     }
331 
332     if (vms->gic_version == VIRT_GIC_VERSION_2) {
333         irqflags = deposit32(irqflags, GIC_FDT_IRQ_PPI_CPU_START,
334                              GIC_FDT_IRQ_PPI_CPU_WIDTH,
335                              (1 << MACHINE(vms)->smp.cpus) - 1);
336     }
337 
338     qemu_fdt_add_subnode(ms->fdt, "/timer");
339 
340     armcpu = ARM_CPU(qemu_get_cpu(0));
341     if (arm_feature(&armcpu->env, ARM_FEATURE_V8)) {
342         const char compat[] = "arm,armv8-timer\0arm,armv7-timer";
343         qemu_fdt_setprop(ms->fdt, "/timer", "compatible",
344                          compat, sizeof(compat));
345     } else {
346         qemu_fdt_setprop_string(ms->fdt, "/timer", "compatible",
347                                 "arm,armv7-timer");
348     }
349     qemu_fdt_setprop(ms->fdt, "/timer", "always-on", NULL, 0);
350     qemu_fdt_setprop_cells(ms->fdt, "/timer", "interrupts",
351                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_S_EL1_IRQ, irqflags,
352                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_NS_EL1_IRQ, irqflags,
353                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_VIRT_IRQ, irqflags,
354                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_NS_EL2_IRQ, irqflags);
355 }
356 
357 static void fdt_add_cpu_nodes(const VirtMachineState *vms)
358 {
359     int cpu;
360     int addr_cells = 1;
361     const MachineState *ms = MACHINE(vms);
362     const VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
363     int smp_cpus = ms->smp.cpus;
364 
365     /*
366      * See Linux Documentation/devicetree/bindings/arm/cpus.yaml
367      * On ARM v8 64-bit systems value should be set to 2,
368      * that corresponds to the MPIDR_EL1 register size.
369      * If MPIDR_EL1[63:32] value is equal to 0 on all CPUs
370      * in the system, #address-cells can be set to 1, since
371      * MPIDR_EL1[63:32] bits are not used for CPUs
372      * identification.
373      *
374      * Here we actually don't know whether our system is 32- or 64-bit one.
375      * The simplest way to go is to examine affinity IDs of all our CPUs. If
376      * at least one of them has Aff3 populated, we set #address-cells to 2.
377      */
378     for (cpu = 0; cpu < smp_cpus; cpu++) {
379         ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(cpu));
380 
381         if (armcpu->mp_affinity & ARM_AFF3_MASK) {
382             addr_cells = 2;
383             break;
384         }
385     }
386 
387     qemu_fdt_add_subnode(ms->fdt, "/cpus");
388     qemu_fdt_setprop_cell(ms->fdt, "/cpus", "#address-cells", addr_cells);
389     qemu_fdt_setprop_cell(ms->fdt, "/cpus", "#size-cells", 0x0);
390 
391     for (cpu = smp_cpus - 1; cpu >= 0; cpu--) {
392         char *nodename = g_strdup_printf("/cpus/cpu@%d", cpu);
393         ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(cpu));
394         CPUState *cs = CPU(armcpu);
395 
396         qemu_fdt_add_subnode(ms->fdt, nodename);
397         qemu_fdt_setprop_string(ms->fdt, nodename, "device_type", "cpu");
398         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
399                                     armcpu->dtb_compatible);
400 
401         if (vms->psci_conduit != QEMU_PSCI_CONDUIT_DISABLED && smp_cpus > 1) {
402             qemu_fdt_setprop_string(ms->fdt, nodename,
403                                         "enable-method", "psci");
404         }
405 
406         if (addr_cells == 2) {
407             qemu_fdt_setprop_u64(ms->fdt, nodename, "reg",
408                                  armcpu->mp_affinity);
409         } else {
410             qemu_fdt_setprop_cell(ms->fdt, nodename, "reg",
411                                   armcpu->mp_affinity);
412         }
413 
414         if (ms->possible_cpus->cpus[cs->cpu_index].props.has_node_id) {
415             qemu_fdt_setprop_cell(ms->fdt, nodename, "numa-node-id",
416                 ms->possible_cpus->cpus[cs->cpu_index].props.node_id);
417         }
418 
419         if (!vmc->no_cpu_topology) {
420             qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle",
421                                   qemu_fdt_alloc_phandle(ms->fdt));
422         }
423 
424         g_free(nodename);
425     }
426 
427     if (!vmc->no_cpu_topology) {
428         /*
429          * Add vCPU topology description through fdt node cpu-map.
430          *
431          * See Linux Documentation/devicetree/bindings/cpu/cpu-topology.txt
432          * In a SMP system, the hierarchy of CPUs can be defined through
433          * four entities that are used to describe the layout of CPUs in
434          * the system: socket/cluster/core/thread.
435          *
436          * A socket node represents the boundary of system physical package
437          * and its child nodes must be one or more cluster nodes. A system
438          * can contain several layers of clustering within a single physical
439          * package and cluster nodes can be contained in parent cluster nodes.
440          *
441          * Note: currently we only support one layer of clustering within
442          * each physical package.
443          */
444         qemu_fdt_add_subnode(ms->fdt, "/cpus/cpu-map");
445 
446         for (cpu = smp_cpus - 1; cpu >= 0; cpu--) {
447             char *cpu_path = g_strdup_printf("/cpus/cpu@%d", cpu);
448             char *map_path;
449 
450             if (ms->smp.threads > 1) {
451                 map_path = g_strdup_printf(
452                     "/cpus/cpu-map/socket%d/cluster%d/core%d/thread%d",
453                     cpu / (ms->smp.clusters * ms->smp.cores * ms->smp.threads),
454                     (cpu / (ms->smp.cores * ms->smp.threads)) % ms->smp.clusters,
455                     (cpu / ms->smp.threads) % ms->smp.cores,
456                     cpu % ms->smp.threads);
457             } else {
458                 map_path = g_strdup_printf(
459                     "/cpus/cpu-map/socket%d/cluster%d/core%d",
460                     cpu / (ms->smp.clusters * ms->smp.cores),
461                     (cpu / ms->smp.cores) % ms->smp.clusters,
462                     cpu % ms->smp.cores);
463             }
464             qemu_fdt_add_path(ms->fdt, map_path);
465             qemu_fdt_setprop_phandle(ms->fdt, map_path, "cpu", cpu_path);
466 
467             g_free(map_path);
468             g_free(cpu_path);
469         }
470     }
471 }
472 
473 static void fdt_add_its_gic_node(VirtMachineState *vms)
474 {
475     char *nodename;
476     MachineState *ms = MACHINE(vms);
477 
478     vms->msi_phandle = qemu_fdt_alloc_phandle(ms->fdt);
479     nodename = g_strdup_printf("/intc/its@%" PRIx64,
480                                vms->memmap[VIRT_GIC_ITS].base);
481     qemu_fdt_add_subnode(ms->fdt, nodename);
482     qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
483                             "arm,gic-v3-its");
484     qemu_fdt_setprop(ms->fdt, nodename, "msi-controller", NULL, 0);
485     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
486                                  2, vms->memmap[VIRT_GIC_ITS].base,
487                                  2, vms->memmap[VIRT_GIC_ITS].size);
488     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", vms->msi_phandle);
489     g_free(nodename);
490 }
491 
492 static void fdt_add_v2m_gic_node(VirtMachineState *vms)
493 {
494     MachineState *ms = MACHINE(vms);
495     char *nodename;
496 
497     nodename = g_strdup_printf("/intc/v2m@%" PRIx64,
498                                vms->memmap[VIRT_GIC_V2M].base);
499     vms->msi_phandle = qemu_fdt_alloc_phandle(ms->fdt);
500     qemu_fdt_add_subnode(ms->fdt, nodename);
501     qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
502                             "arm,gic-v2m-frame");
503     qemu_fdt_setprop(ms->fdt, nodename, "msi-controller", NULL, 0);
504     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
505                                  2, vms->memmap[VIRT_GIC_V2M].base,
506                                  2, vms->memmap[VIRT_GIC_V2M].size);
507     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", vms->msi_phandle);
508     g_free(nodename);
509 }
510 
511 static void fdt_add_gic_node(VirtMachineState *vms)
512 {
513     MachineState *ms = MACHINE(vms);
514     char *nodename;
515 
516     vms->gic_phandle = qemu_fdt_alloc_phandle(ms->fdt);
517     qemu_fdt_setprop_cell(ms->fdt, "/", "interrupt-parent", vms->gic_phandle);
518 
519     nodename = g_strdup_printf("/intc@%" PRIx64,
520                                vms->memmap[VIRT_GIC_DIST].base);
521     qemu_fdt_add_subnode(ms->fdt, nodename);
522     qemu_fdt_setprop_cell(ms->fdt, nodename, "#interrupt-cells", 3);
523     qemu_fdt_setprop(ms->fdt, nodename, "interrupt-controller", NULL, 0);
524     qemu_fdt_setprop_cell(ms->fdt, nodename, "#address-cells", 0x2);
525     qemu_fdt_setprop_cell(ms->fdt, nodename, "#size-cells", 0x2);
526     qemu_fdt_setprop(ms->fdt, nodename, "ranges", NULL, 0);
527     if (vms->gic_version != VIRT_GIC_VERSION_2) {
528         int nb_redist_regions = virt_gicv3_redist_region_count(vms);
529 
530         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
531                                 "arm,gic-v3");
532 
533         qemu_fdt_setprop_cell(ms->fdt, nodename,
534                               "#redistributor-regions", nb_redist_regions);
535 
536         if (nb_redist_regions == 1) {
537             qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
538                                          2, vms->memmap[VIRT_GIC_DIST].base,
539                                          2, vms->memmap[VIRT_GIC_DIST].size,
540                                          2, vms->memmap[VIRT_GIC_REDIST].base,
541                                          2, vms->memmap[VIRT_GIC_REDIST].size);
542         } else {
543             qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
544                                  2, vms->memmap[VIRT_GIC_DIST].base,
545                                  2, vms->memmap[VIRT_GIC_DIST].size,
546                                  2, vms->memmap[VIRT_GIC_REDIST].base,
547                                  2, vms->memmap[VIRT_GIC_REDIST].size,
548                                  2, vms->memmap[VIRT_HIGH_GIC_REDIST2].base,
549                                  2, vms->memmap[VIRT_HIGH_GIC_REDIST2].size);
550         }
551 
552         if (vms->virt) {
553             qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
554                                    GIC_FDT_IRQ_TYPE_PPI, ARCH_GIC_MAINT_IRQ,
555                                    GIC_FDT_IRQ_FLAGS_LEVEL_HI);
556         }
557     } else {
558         /* 'cortex-a15-gic' means 'GIC v2' */
559         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
560                                 "arm,cortex-a15-gic");
561         if (!vms->virt) {
562             qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
563                                          2, vms->memmap[VIRT_GIC_DIST].base,
564                                          2, vms->memmap[VIRT_GIC_DIST].size,
565                                          2, vms->memmap[VIRT_GIC_CPU].base,
566                                          2, vms->memmap[VIRT_GIC_CPU].size);
567         } else {
568             qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
569                                          2, vms->memmap[VIRT_GIC_DIST].base,
570                                          2, vms->memmap[VIRT_GIC_DIST].size,
571                                          2, vms->memmap[VIRT_GIC_CPU].base,
572                                          2, vms->memmap[VIRT_GIC_CPU].size,
573                                          2, vms->memmap[VIRT_GIC_HYP].base,
574                                          2, vms->memmap[VIRT_GIC_HYP].size,
575                                          2, vms->memmap[VIRT_GIC_VCPU].base,
576                                          2, vms->memmap[VIRT_GIC_VCPU].size);
577             qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
578                                    GIC_FDT_IRQ_TYPE_PPI, ARCH_GIC_MAINT_IRQ,
579                                    GIC_FDT_IRQ_FLAGS_LEVEL_HI);
580         }
581     }
582 
583     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", vms->gic_phandle);
584     g_free(nodename);
585 }
586 
587 static void fdt_add_pmu_nodes(const VirtMachineState *vms)
588 {
589     ARMCPU *armcpu = ARM_CPU(first_cpu);
590     uint32_t irqflags = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
591     MachineState *ms = MACHINE(vms);
592 
593     if (!arm_feature(&armcpu->env, ARM_FEATURE_PMU)) {
594         assert(!object_property_get_bool(OBJECT(armcpu), "pmu", NULL));
595         return;
596     }
597 
598     if (vms->gic_version == VIRT_GIC_VERSION_2) {
599         irqflags = deposit32(irqflags, GIC_FDT_IRQ_PPI_CPU_START,
600                              GIC_FDT_IRQ_PPI_CPU_WIDTH,
601                              (1 << MACHINE(vms)->smp.cpus) - 1);
602     }
603 
604     qemu_fdt_add_subnode(ms->fdt, "/pmu");
605     if (arm_feature(&armcpu->env, ARM_FEATURE_V8)) {
606         const char compat[] = "arm,armv8-pmuv3";
607         qemu_fdt_setprop(ms->fdt, "/pmu", "compatible",
608                          compat, sizeof(compat));
609         qemu_fdt_setprop_cells(ms->fdt, "/pmu", "interrupts",
610                                GIC_FDT_IRQ_TYPE_PPI, VIRTUAL_PMU_IRQ, irqflags);
611     }
612 }
613 
614 static inline DeviceState *create_acpi_ged(VirtMachineState *vms)
615 {
616     DeviceState *dev;
617     MachineState *ms = MACHINE(vms);
618     int irq = vms->irqmap[VIRT_ACPI_GED];
619     uint32_t event = ACPI_GED_PWR_DOWN_EVT;
620 
621     if (ms->ram_slots) {
622         event |= ACPI_GED_MEM_HOTPLUG_EVT;
623     }
624 
625     if (ms->nvdimms_state->is_enabled) {
626         event |= ACPI_GED_NVDIMM_HOTPLUG_EVT;
627     }
628 
629     dev = qdev_new(TYPE_ACPI_GED);
630     qdev_prop_set_uint32(dev, "ged-event", event);
631 
632     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_ACPI_GED].base);
633     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 1, vms->memmap[VIRT_PCDIMM_ACPI].base);
634     sysbus_connect_irq(SYS_BUS_DEVICE(dev), 0, qdev_get_gpio_in(vms->gic, irq));
635 
636     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
637 
638     return dev;
639 }
640 
641 static void create_its(VirtMachineState *vms)
642 {
643     const char *itsclass = its_class_name();
644     DeviceState *dev;
645 
646     if (!strcmp(itsclass, "arm-gicv3-its")) {
647         if (!vms->tcg_its) {
648             itsclass = NULL;
649         }
650     }
651 
652     if (!itsclass) {
653         /* Do nothing if not supported */
654         return;
655     }
656 
657     dev = qdev_new(itsclass);
658 
659     object_property_set_link(OBJECT(dev), "parent-gicv3", OBJECT(vms->gic),
660                              &error_abort);
661     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
662     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_GIC_ITS].base);
663 
664     fdt_add_its_gic_node(vms);
665     vms->msi_controller = VIRT_MSI_CTRL_ITS;
666 }
667 
668 static void create_v2m(VirtMachineState *vms)
669 {
670     int i;
671     int irq = vms->irqmap[VIRT_GIC_V2M];
672     DeviceState *dev;
673 
674     dev = qdev_new("arm-gicv2m");
675     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_GIC_V2M].base);
676     qdev_prop_set_uint32(dev, "base-spi", irq);
677     qdev_prop_set_uint32(dev, "num-spi", NUM_GICV2M_SPIS);
678     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
679 
680     for (i = 0; i < NUM_GICV2M_SPIS; i++) {
681         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
682                            qdev_get_gpio_in(vms->gic, irq + i));
683     }
684 
685     fdt_add_v2m_gic_node(vms);
686     vms->msi_controller = VIRT_MSI_CTRL_GICV2M;
687 }
688 
689 static void create_gic(VirtMachineState *vms, MemoryRegion *mem)
690 {
691     MachineState *ms = MACHINE(vms);
692     /* We create a standalone GIC */
693     SysBusDevice *gicbusdev;
694     const char *gictype;
695     int i;
696     unsigned int smp_cpus = ms->smp.cpus;
697     uint32_t nb_redist_regions = 0;
698     int revision;
699 
700     if (vms->gic_version == VIRT_GIC_VERSION_2) {
701         gictype = gic_class_name();
702     } else {
703         gictype = gicv3_class_name();
704     }
705 
706     switch (vms->gic_version) {
707     case VIRT_GIC_VERSION_2:
708         revision = 2;
709         break;
710     case VIRT_GIC_VERSION_3:
711         revision = 3;
712         break;
713     case VIRT_GIC_VERSION_4:
714         revision = 4;
715         break;
716     default:
717         g_assert_not_reached();
718     }
719     vms->gic = qdev_new(gictype);
720     qdev_prop_set_uint32(vms->gic, "revision", revision);
721     qdev_prop_set_uint32(vms->gic, "num-cpu", smp_cpus);
722     /* Note that the num-irq property counts both internal and external
723      * interrupts; there are always 32 of the former (mandated by GIC spec).
724      */
725     qdev_prop_set_uint32(vms->gic, "num-irq", NUM_IRQS + 32);
726     if (!kvm_irqchip_in_kernel()) {
727         qdev_prop_set_bit(vms->gic, "has-security-extensions", vms->secure);
728     }
729 
730     if (vms->gic_version != VIRT_GIC_VERSION_2) {
731         uint32_t redist0_capacity = virt_redist_capacity(vms, VIRT_GIC_REDIST);
732         uint32_t redist0_count = MIN(smp_cpus, redist0_capacity);
733 
734         nb_redist_regions = virt_gicv3_redist_region_count(vms);
735 
736         qdev_prop_set_uint32(vms->gic, "len-redist-region-count",
737                              nb_redist_regions);
738         qdev_prop_set_uint32(vms->gic, "redist-region-count[0]", redist0_count);
739 
740         if (!kvm_irqchip_in_kernel()) {
741             if (vms->tcg_its) {
742                 object_property_set_link(OBJECT(vms->gic), "sysmem",
743                                          OBJECT(mem), &error_fatal);
744                 qdev_prop_set_bit(vms->gic, "has-lpi", true);
745             }
746         }
747 
748         if (nb_redist_regions == 2) {
749             uint32_t redist1_capacity =
750                 virt_redist_capacity(vms, VIRT_HIGH_GIC_REDIST2);
751 
752             qdev_prop_set_uint32(vms->gic, "redist-region-count[1]",
753                 MIN(smp_cpus - redist0_count, redist1_capacity));
754         }
755     } else {
756         if (!kvm_irqchip_in_kernel()) {
757             qdev_prop_set_bit(vms->gic, "has-virtualization-extensions",
758                               vms->virt);
759         }
760     }
761     gicbusdev = SYS_BUS_DEVICE(vms->gic);
762     sysbus_realize_and_unref(gicbusdev, &error_fatal);
763     sysbus_mmio_map(gicbusdev, 0, vms->memmap[VIRT_GIC_DIST].base);
764     if (vms->gic_version != VIRT_GIC_VERSION_2) {
765         sysbus_mmio_map(gicbusdev, 1, vms->memmap[VIRT_GIC_REDIST].base);
766         if (nb_redist_regions == 2) {
767             sysbus_mmio_map(gicbusdev, 2,
768                             vms->memmap[VIRT_HIGH_GIC_REDIST2].base);
769         }
770     } else {
771         sysbus_mmio_map(gicbusdev, 1, vms->memmap[VIRT_GIC_CPU].base);
772         if (vms->virt) {
773             sysbus_mmio_map(gicbusdev, 2, vms->memmap[VIRT_GIC_HYP].base);
774             sysbus_mmio_map(gicbusdev, 3, vms->memmap[VIRT_GIC_VCPU].base);
775         }
776     }
777 
778     /* Wire the outputs from each CPU's generic timer and the GICv3
779      * maintenance interrupt signal to the appropriate GIC PPI inputs,
780      * and the GIC's IRQ/FIQ/VIRQ/VFIQ interrupt outputs to the CPU's inputs.
781      */
782     for (i = 0; i < smp_cpus; i++) {
783         DeviceState *cpudev = DEVICE(qemu_get_cpu(i));
784         int ppibase = NUM_IRQS + i * GIC_INTERNAL + GIC_NR_SGIS;
785         int irq;
786         /* Mapping from the output timer irq lines from the CPU to the
787          * GIC PPI inputs we use for the virt board.
788          */
789         const int timer_irq[] = {
790             [GTIMER_PHYS] = ARCH_TIMER_NS_EL1_IRQ,
791             [GTIMER_VIRT] = ARCH_TIMER_VIRT_IRQ,
792             [GTIMER_HYP]  = ARCH_TIMER_NS_EL2_IRQ,
793             [GTIMER_SEC]  = ARCH_TIMER_S_EL1_IRQ,
794         };
795 
796         for (irq = 0; irq < ARRAY_SIZE(timer_irq); irq++) {
797             qdev_connect_gpio_out(cpudev, irq,
798                                   qdev_get_gpio_in(vms->gic,
799                                                    ppibase + timer_irq[irq]));
800         }
801 
802         if (vms->gic_version != VIRT_GIC_VERSION_2) {
803             qemu_irq irq = qdev_get_gpio_in(vms->gic,
804                                             ppibase + ARCH_GIC_MAINT_IRQ);
805             qdev_connect_gpio_out_named(cpudev, "gicv3-maintenance-interrupt",
806                                         0, irq);
807         } else if (vms->virt) {
808             qemu_irq irq = qdev_get_gpio_in(vms->gic,
809                                             ppibase + ARCH_GIC_MAINT_IRQ);
810             sysbus_connect_irq(gicbusdev, i + 4 * smp_cpus, irq);
811         }
812 
813         qdev_connect_gpio_out_named(cpudev, "pmu-interrupt", 0,
814                                     qdev_get_gpio_in(vms->gic, ppibase
815                                                      + VIRTUAL_PMU_IRQ));
816 
817         sysbus_connect_irq(gicbusdev, i, qdev_get_gpio_in(cpudev, ARM_CPU_IRQ));
818         sysbus_connect_irq(gicbusdev, i + smp_cpus,
819                            qdev_get_gpio_in(cpudev, ARM_CPU_FIQ));
820         sysbus_connect_irq(gicbusdev, i + 2 * smp_cpus,
821                            qdev_get_gpio_in(cpudev, ARM_CPU_VIRQ));
822         sysbus_connect_irq(gicbusdev, i + 3 * smp_cpus,
823                            qdev_get_gpio_in(cpudev, ARM_CPU_VFIQ));
824     }
825 
826     fdt_add_gic_node(vms);
827 
828     if (vms->gic_version != VIRT_GIC_VERSION_2 && vms->its) {
829         create_its(vms);
830     } else if (vms->gic_version == VIRT_GIC_VERSION_2) {
831         create_v2m(vms);
832     }
833 }
834 
835 static void create_uart(const VirtMachineState *vms, int uart,
836                         MemoryRegion *mem, Chardev *chr)
837 {
838     char *nodename;
839     hwaddr base = vms->memmap[uart].base;
840     hwaddr size = vms->memmap[uart].size;
841     int irq = vms->irqmap[uart];
842     const char compat[] = "arm,pl011\0arm,primecell";
843     const char clocknames[] = "uartclk\0apb_pclk";
844     DeviceState *dev = qdev_new(TYPE_PL011);
845     SysBusDevice *s = SYS_BUS_DEVICE(dev);
846     MachineState *ms = MACHINE(vms);
847 
848     qdev_prop_set_chr(dev, "chardev", chr);
849     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
850     memory_region_add_subregion(mem, base,
851                                 sysbus_mmio_get_region(s, 0));
852     sysbus_connect_irq(s, 0, qdev_get_gpio_in(vms->gic, irq));
853 
854     nodename = g_strdup_printf("/pl011@%" PRIx64, base);
855     qemu_fdt_add_subnode(ms->fdt, nodename);
856     /* Note that we can't use setprop_string because of the embedded NUL */
857     qemu_fdt_setprop(ms->fdt, nodename, "compatible",
858                          compat, sizeof(compat));
859     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
860                                      2, base, 2, size);
861     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
862                                GIC_FDT_IRQ_TYPE_SPI, irq,
863                                GIC_FDT_IRQ_FLAGS_LEVEL_HI);
864     qemu_fdt_setprop_cells(ms->fdt, nodename, "clocks",
865                                vms->clock_phandle, vms->clock_phandle);
866     qemu_fdt_setprop(ms->fdt, nodename, "clock-names",
867                          clocknames, sizeof(clocknames));
868 
869     if (uart == VIRT_UART) {
870         qemu_fdt_setprop_string(ms->fdt, "/chosen", "stdout-path", nodename);
871     } else {
872         /* Mark as not usable by the normal world */
873         qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
874         qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
875 
876         qemu_fdt_setprop_string(ms->fdt, "/secure-chosen", "stdout-path",
877                                 nodename);
878     }
879 
880     g_free(nodename);
881 }
882 
883 static void create_rtc(const VirtMachineState *vms)
884 {
885     char *nodename;
886     hwaddr base = vms->memmap[VIRT_RTC].base;
887     hwaddr size = vms->memmap[VIRT_RTC].size;
888     int irq = vms->irqmap[VIRT_RTC];
889     const char compat[] = "arm,pl031\0arm,primecell";
890     MachineState *ms = MACHINE(vms);
891 
892     sysbus_create_simple("pl031", base, qdev_get_gpio_in(vms->gic, irq));
893 
894     nodename = g_strdup_printf("/pl031@%" PRIx64, base);
895     qemu_fdt_add_subnode(ms->fdt, nodename);
896     qemu_fdt_setprop(ms->fdt, nodename, "compatible", compat, sizeof(compat));
897     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
898                                  2, base, 2, size);
899     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
900                            GIC_FDT_IRQ_TYPE_SPI, irq,
901                            GIC_FDT_IRQ_FLAGS_LEVEL_HI);
902     qemu_fdt_setprop_cell(ms->fdt, nodename, "clocks", vms->clock_phandle);
903     qemu_fdt_setprop_string(ms->fdt, nodename, "clock-names", "apb_pclk");
904     g_free(nodename);
905 }
906 
907 static DeviceState *gpio_key_dev;
908 static void virt_powerdown_req(Notifier *n, void *opaque)
909 {
910     VirtMachineState *s = container_of(n, VirtMachineState, powerdown_notifier);
911 
912     if (s->acpi_dev) {
913         acpi_send_event(s->acpi_dev, ACPI_POWER_DOWN_STATUS);
914     } else {
915         /* use gpio Pin 3 for power button event */
916         qemu_set_irq(qdev_get_gpio_in(gpio_key_dev, 0), 1);
917     }
918 }
919 
920 static void create_gpio_keys(char *fdt, DeviceState *pl061_dev,
921                              uint32_t phandle)
922 {
923     gpio_key_dev = sysbus_create_simple("gpio-key", -1,
924                                         qdev_get_gpio_in(pl061_dev, 3));
925 
926     qemu_fdt_add_subnode(fdt, "/gpio-keys");
927     qemu_fdt_setprop_string(fdt, "/gpio-keys", "compatible", "gpio-keys");
928     qemu_fdt_setprop_cell(fdt, "/gpio-keys", "#size-cells", 0);
929     qemu_fdt_setprop_cell(fdt, "/gpio-keys", "#address-cells", 1);
930 
931     qemu_fdt_add_subnode(fdt, "/gpio-keys/poweroff");
932     qemu_fdt_setprop_string(fdt, "/gpio-keys/poweroff",
933                             "label", "GPIO Key Poweroff");
934     qemu_fdt_setprop_cell(fdt, "/gpio-keys/poweroff", "linux,code",
935                           KEY_POWER);
936     qemu_fdt_setprop_cells(fdt, "/gpio-keys/poweroff",
937                            "gpios", phandle, 3, 0);
938 }
939 
940 #define SECURE_GPIO_POWEROFF 0
941 #define SECURE_GPIO_RESET    1
942 
943 static void create_secure_gpio_pwr(char *fdt, DeviceState *pl061_dev,
944                                    uint32_t phandle)
945 {
946     DeviceState *gpio_pwr_dev;
947 
948     /* gpio-pwr */
949     gpio_pwr_dev = sysbus_create_simple("gpio-pwr", -1, NULL);
950 
951     /* connect secure pl061 to gpio-pwr */
952     qdev_connect_gpio_out(pl061_dev, SECURE_GPIO_RESET,
953                           qdev_get_gpio_in_named(gpio_pwr_dev, "reset", 0));
954     qdev_connect_gpio_out(pl061_dev, SECURE_GPIO_POWEROFF,
955                           qdev_get_gpio_in_named(gpio_pwr_dev, "shutdown", 0));
956 
957     qemu_fdt_add_subnode(fdt, "/gpio-poweroff");
958     qemu_fdt_setprop_string(fdt, "/gpio-poweroff", "compatible",
959                             "gpio-poweroff");
960     qemu_fdt_setprop_cells(fdt, "/gpio-poweroff",
961                            "gpios", phandle, SECURE_GPIO_POWEROFF, 0);
962     qemu_fdt_setprop_string(fdt, "/gpio-poweroff", "status", "disabled");
963     qemu_fdt_setprop_string(fdt, "/gpio-poweroff", "secure-status",
964                             "okay");
965 
966     qemu_fdt_add_subnode(fdt, "/gpio-restart");
967     qemu_fdt_setprop_string(fdt, "/gpio-restart", "compatible",
968                             "gpio-restart");
969     qemu_fdt_setprop_cells(fdt, "/gpio-restart",
970                            "gpios", phandle, SECURE_GPIO_RESET, 0);
971     qemu_fdt_setprop_string(fdt, "/gpio-restart", "status", "disabled");
972     qemu_fdt_setprop_string(fdt, "/gpio-restart", "secure-status",
973                             "okay");
974 }
975 
976 static void create_gpio_devices(const VirtMachineState *vms, int gpio,
977                                 MemoryRegion *mem)
978 {
979     char *nodename;
980     DeviceState *pl061_dev;
981     hwaddr base = vms->memmap[gpio].base;
982     hwaddr size = vms->memmap[gpio].size;
983     int irq = vms->irqmap[gpio];
984     const char compat[] = "arm,pl061\0arm,primecell";
985     SysBusDevice *s;
986     MachineState *ms = MACHINE(vms);
987 
988     pl061_dev = qdev_new("pl061");
989     /* Pull lines down to 0 if not driven by the PL061 */
990     qdev_prop_set_uint32(pl061_dev, "pullups", 0);
991     qdev_prop_set_uint32(pl061_dev, "pulldowns", 0xff);
992     s = SYS_BUS_DEVICE(pl061_dev);
993     sysbus_realize_and_unref(s, &error_fatal);
994     memory_region_add_subregion(mem, base, sysbus_mmio_get_region(s, 0));
995     sysbus_connect_irq(s, 0, qdev_get_gpio_in(vms->gic, irq));
996 
997     uint32_t phandle = qemu_fdt_alloc_phandle(ms->fdt);
998     nodename = g_strdup_printf("/pl061@%" PRIx64, base);
999     qemu_fdt_add_subnode(ms->fdt, nodename);
1000     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1001                                  2, base, 2, size);
1002     qemu_fdt_setprop(ms->fdt, nodename, "compatible", compat, sizeof(compat));
1003     qemu_fdt_setprop_cell(ms->fdt, nodename, "#gpio-cells", 2);
1004     qemu_fdt_setprop(ms->fdt, nodename, "gpio-controller", NULL, 0);
1005     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
1006                            GIC_FDT_IRQ_TYPE_SPI, irq,
1007                            GIC_FDT_IRQ_FLAGS_LEVEL_HI);
1008     qemu_fdt_setprop_cell(ms->fdt, nodename, "clocks", vms->clock_phandle);
1009     qemu_fdt_setprop_string(ms->fdt, nodename, "clock-names", "apb_pclk");
1010     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", phandle);
1011 
1012     if (gpio != VIRT_GPIO) {
1013         /* Mark as not usable by the normal world */
1014         qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
1015         qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
1016     }
1017     g_free(nodename);
1018 
1019     /* Child gpio devices */
1020     if (gpio == VIRT_GPIO) {
1021         create_gpio_keys(ms->fdt, pl061_dev, phandle);
1022     } else {
1023         create_secure_gpio_pwr(ms->fdt, pl061_dev, phandle);
1024     }
1025 }
1026 
1027 static void create_virtio_devices(const VirtMachineState *vms)
1028 {
1029     int i;
1030     hwaddr size = vms->memmap[VIRT_MMIO].size;
1031     MachineState *ms = MACHINE(vms);
1032 
1033     /* We create the transports in forwards order. Since qbus_realize()
1034      * prepends (not appends) new child buses, the incrementing loop below will
1035      * create a list of virtio-mmio buses with decreasing base addresses.
1036      *
1037      * When a -device option is processed from the command line,
1038      * qbus_find_recursive() picks the next free virtio-mmio bus in forwards
1039      * order. The upshot is that -device options in increasing command line
1040      * order are mapped to virtio-mmio buses with decreasing base addresses.
1041      *
1042      * When this code was originally written, that arrangement ensured that the
1043      * guest Linux kernel would give the lowest "name" (/dev/vda, eth0, etc) to
1044      * the first -device on the command line. (The end-to-end order is a
1045      * function of this loop, qbus_realize(), qbus_find_recursive(), and the
1046      * guest kernel's name-to-address assignment strategy.)
1047      *
1048      * Meanwhile, the kernel's traversal seems to have been reversed; see eg.
1049      * the message, if not necessarily the code, of commit 70161ff336.
1050      * Therefore the loop now establishes the inverse of the original intent.
1051      *
1052      * Unfortunately, we can't counteract the kernel change by reversing the
1053      * loop; it would break existing command lines.
1054      *
1055      * In any case, the kernel makes no guarantee about the stability of
1056      * enumeration order of virtio devices (as demonstrated by it changing
1057      * between kernel versions). For reliable and stable identification
1058      * of disks users must use UUIDs or similar mechanisms.
1059      */
1060     for (i = 0; i < NUM_VIRTIO_TRANSPORTS; i++) {
1061         int irq = vms->irqmap[VIRT_MMIO] + i;
1062         hwaddr base = vms->memmap[VIRT_MMIO].base + i * size;
1063 
1064         sysbus_create_simple("virtio-mmio", base,
1065                              qdev_get_gpio_in(vms->gic, irq));
1066     }
1067 
1068     /* We add dtb nodes in reverse order so that they appear in the finished
1069      * device tree lowest address first.
1070      *
1071      * Note that this mapping is independent of the loop above. The previous
1072      * loop influences virtio device to virtio transport assignment, whereas
1073      * this loop controls how virtio transports are laid out in the dtb.
1074      */
1075     for (i = NUM_VIRTIO_TRANSPORTS - 1; i >= 0; i--) {
1076         char *nodename;
1077         int irq = vms->irqmap[VIRT_MMIO] + i;
1078         hwaddr base = vms->memmap[VIRT_MMIO].base + i * size;
1079 
1080         nodename = g_strdup_printf("/virtio_mmio@%" PRIx64, base);
1081         qemu_fdt_add_subnode(ms->fdt, nodename);
1082         qemu_fdt_setprop_string(ms->fdt, nodename,
1083                                 "compatible", "virtio,mmio");
1084         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1085                                      2, base, 2, size);
1086         qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
1087                                GIC_FDT_IRQ_TYPE_SPI, irq,
1088                                GIC_FDT_IRQ_FLAGS_EDGE_LO_HI);
1089         qemu_fdt_setprop(ms->fdt, nodename, "dma-coherent", NULL, 0);
1090         g_free(nodename);
1091     }
1092 }
1093 
1094 #define VIRT_FLASH_SECTOR_SIZE (256 * KiB)
1095 
1096 static PFlashCFI01 *virt_flash_create1(VirtMachineState *vms,
1097                                         const char *name,
1098                                         const char *alias_prop_name)
1099 {
1100     /*
1101      * Create a single flash device.  We use the same parameters as
1102      * the flash devices on the Versatile Express board.
1103      */
1104     DeviceState *dev = qdev_new(TYPE_PFLASH_CFI01);
1105 
1106     qdev_prop_set_uint64(dev, "sector-length", VIRT_FLASH_SECTOR_SIZE);
1107     qdev_prop_set_uint8(dev, "width", 4);
1108     qdev_prop_set_uint8(dev, "device-width", 2);
1109     qdev_prop_set_bit(dev, "big-endian", false);
1110     qdev_prop_set_uint16(dev, "id0", 0x89);
1111     qdev_prop_set_uint16(dev, "id1", 0x18);
1112     qdev_prop_set_uint16(dev, "id2", 0x00);
1113     qdev_prop_set_uint16(dev, "id3", 0x00);
1114     qdev_prop_set_string(dev, "name", name);
1115     object_property_add_child(OBJECT(vms), name, OBJECT(dev));
1116     object_property_add_alias(OBJECT(vms), alias_prop_name,
1117                               OBJECT(dev), "drive");
1118     return PFLASH_CFI01(dev);
1119 }
1120 
1121 static void virt_flash_create(VirtMachineState *vms)
1122 {
1123     vms->flash[0] = virt_flash_create1(vms, "virt.flash0", "pflash0");
1124     vms->flash[1] = virt_flash_create1(vms, "virt.flash1", "pflash1");
1125 }
1126 
1127 static void virt_flash_map1(PFlashCFI01 *flash,
1128                             hwaddr base, hwaddr size,
1129                             MemoryRegion *sysmem)
1130 {
1131     DeviceState *dev = DEVICE(flash);
1132 
1133     assert(QEMU_IS_ALIGNED(size, VIRT_FLASH_SECTOR_SIZE));
1134     assert(size / VIRT_FLASH_SECTOR_SIZE <= UINT32_MAX);
1135     qdev_prop_set_uint32(dev, "num-blocks", size / VIRT_FLASH_SECTOR_SIZE);
1136     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1137 
1138     memory_region_add_subregion(sysmem, base,
1139                                 sysbus_mmio_get_region(SYS_BUS_DEVICE(dev),
1140                                                        0));
1141 }
1142 
1143 static void virt_flash_map(VirtMachineState *vms,
1144                            MemoryRegion *sysmem,
1145                            MemoryRegion *secure_sysmem)
1146 {
1147     /*
1148      * Map two flash devices to fill the VIRT_FLASH space in the memmap.
1149      * sysmem is the system memory space. secure_sysmem is the secure view
1150      * of the system, and the first flash device should be made visible only
1151      * there. The second flash device is visible to both secure and nonsecure.
1152      * If sysmem == secure_sysmem this means there is no separate Secure
1153      * address space and both flash devices are generally visible.
1154      */
1155     hwaddr flashsize = vms->memmap[VIRT_FLASH].size / 2;
1156     hwaddr flashbase = vms->memmap[VIRT_FLASH].base;
1157 
1158     virt_flash_map1(vms->flash[0], flashbase, flashsize,
1159                     secure_sysmem);
1160     virt_flash_map1(vms->flash[1], flashbase + flashsize, flashsize,
1161                     sysmem);
1162 }
1163 
1164 static void virt_flash_fdt(VirtMachineState *vms,
1165                            MemoryRegion *sysmem,
1166                            MemoryRegion *secure_sysmem)
1167 {
1168     hwaddr flashsize = vms->memmap[VIRT_FLASH].size / 2;
1169     hwaddr flashbase = vms->memmap[VIRT_FLASH].base;
1170     MachineState *ms = MACHINE(vms);
1171     char *nodename;
1172 
1173     if (sysmem == secure_sysmem) {
1174         /* Report both flash devices as a single node in the DT */
1175         nodename = g_strdup_printf("/flash@%" PRIx64, flashbase);
1176         qemu_fdt_add_subnode(ms->fdt, nodename);
1177         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible", "cfi-flash");
1178         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1179                                      2, flashbase, 2, flashsize,
1180                                      2, flashbase + flashsize, 2, flashsize);
1181         qemu_fdt_setprop_cell(ms->fdt, nodename, "bank-width", 4);
1182         g_free(nodename);
1183     } else {
1184         /*
1185          * Report the devices as separate nodes so we can mark one as
1186          * only visible to the secure world.
1187          */
1188         nodename = g_strdup_printf("/secflash@%" PRIx64, flashbase);
1189         qemu_fdt_add_subnode(ms->fdt, nodename);
1190         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible", "cfi-flash");
1191         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1192                                      2, flashbase, 2, flashsize);
1193         qemu_fdt_setprop_cell(ms->fdt, nodename, "bank-width", 4);
1194         qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
1195         qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
1196         g_free(nodename);
1197 
1198         nodename = g_strdup_printf("/flash@%" PRIx64, flashbase);
1199         qemu_fdt_add_subnode(ms->fdt, nodename);
1200         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible", "cfi-flash");
1201         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1202                                      2, flashbase + flashsize, 2, flashsize);
1203         qemu_fdt_setprop_cell(ms->fdt, nodename, "bank-width", 4);
1204         g_free(nodename);
1205     }
1206 }
1207 
1208 static bool virt_firmware_init(VirtMachineState *vms,
1209                                MemoryRegion *sysmem,
1210                                MemoryRegion *secure_sysmem)
1211 {
1212     int i;
1213     const char *bios_name;
1214     BlockBackend *pflash_blk0;
1215 
1216     /* Map legacy -drive if=pflash to machine properties */
1217     for (i = 0; i < ARRAY_SIZE(vms->flash); i++) {
1218         pflash_cfi01_legacy_drive(vms->flash[i],
1219                                   drive_get(IF_PFLASH, 0, i));
1220     }
1221 
1222     virt_flash_map(vms, sysmem, secure_sysmem);
1223 
1224     pflash_blk0 = pflash_cfi01_get_blk(vms->flash[0]);
1225 
1226     bios_name = MACHINE(vms)->firmware;
1227     if (bios_name) {
1228         char *fname;
1229         MemoryRegion *mr;
1230         int image_size;
1231 
1232         if (pflash_blk0) {
1233             error_report("The contents of the first flash device may be "
1234                          "specified with -bios or with -drive if=pflash... "
1235                          "but you cannot use both options at once");
1236             exit(1);
1237         }
1238 
1239         /* Fall back to -bios */
1240 
1241         fname = qemu_find_file(QEMU_FILE_TYPE_BIOS, bios_name);
1242         if (!fname) {
1243             error_report("Could not find ROM image '%s'", bios_name);
1244             exit(1);
1245         }
1246         mr = sysbus_mmio_get_region(SYS_BUS_DEVICE(vms->flash[0]), 0);
1247         image_size = load_image_mr(fname, mr);
1248         g_free(fname);
1249         if (image_size < 0) {
1250             error_report("Could not load ROM image '%s'", bios_name);
1251             exit(1);
1252         }
1253     }
1254 
1255     return pflash_blk0 || bios_name;
1256 }
1257 
1258 static FWCfgState *create_fw_cfg(const VirtMachineState *vms, AddressSpace *as)
1259 {
1260     MachineState *ms = MACHINE(vms);
1261     hwaddr base = vms->memmap[VIRT_FW_CFG].base;
1262     hwaddr size = vms->memmap[VIRT_FW_CFG].size;
1263     FWCfgState *fw_cfg;
1264     char *nodename;
1265 
1266     fw_cfg = fw_cfg_init_mem_wide(base + 8, base, 8, base + 16, as);
1267     fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, (uint16_t)ms->smp.cpus);
1268 
1269     nodename = g_strdup_printf("/fw-cfg@%" PRIx64, base);
1270     qemu_fdt_add_subnode(ms->fdt, nodename);
1271     qemu_fdt_setprop_string(ms->fdt, nodename,
1272                             "compatible", "qemu,fw-cfg-mmio");
1273     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1274                                  2, base, 2, size);
1275     qemu_fdt_setprop(ms->fdt, nodename, "dma-coherent", NULL, 0);
1276     g_free(nodename);
1277     return fw_cfg;
1278 }
1279 
1280 static void create_pcie_irq_map(const MachineState *ms,
1281                                 uint32_t gic_phandle,
1282                                 int first_irq, const char *nodename)
1283 {
1284     int devfn, pin;
1285     uint32_t full_irq_map[4 * 4 * 10] = { 0 };
1286     uint32_t *irq_map = full_irq_map;
1287 
1288     for (devfn = 0; devfn <= 0x18; devfn += 0x8) {
1289         for (pin = 0; pin < 4; pin++) {
1290             int irq_type = GIC_FDT_IRQ_TYPE_SPI;
1291             int irq_nr = first_irq + ((pin + PCI_SLOT(devfn)) % PCI_NUM_PINS);
1292             int irq_level = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
1293             int i;
1294 
1295             uint32_t map[] = {
1296                 devfn << 8, 0, 0,                           /* devfn */
1297                 pin + 1,                                    /* PCI pin */
1298                 gic_phandle, 0, 0, irq_type, irq_nr, irq_level }; /* GIC irq */
1299 
1300             /* Convert map to big endian */
1301             for (i = 0; i < 10; i++) {
1302                 irq_map[i] = cpu_to_be32(map[i]);
1303             }
1304             irq_map += 10;
1305         }
1306     }
1307 
1308     qemu_fdt_setprop(ms->fdt, nodename, "interrupt-map",
1309                      full_irq_map, sizeof(full_irq_map));
1310 
1311     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupt-map-mask",
1312                            cpu_to_be16(PCI_DEVFN(3, 0)), /* Slot 3 */
1313                            0, 0,
1314                            0x7           /* PCI irq */);
1315 }
1316 
1317 static void create_smmu(const VirtMachineState *vms,
1318                         PCIBus *bus)
1319 {
1320     char *node;
1321     const char compat[] = "arm,smmu-v3";
1322     int irq =  vms->irqmap[VIRT_SMMU];
1323     int i;
1324     hwaddr base = vms->memmap[VIRT_SMMU].base;
1325     hwaddr size = vms->memmap[VIRT_SMMU].size;
1326     const char irq_names[] = "eventq\0priq\0cmdq-sync\0gerror";
1327     DeviceState *dev;
1328     MachineState *ms = MACHINE(vms);
1329 
1330     if (vms->iommu != VIRT_IOMMU_SMMUV3 || !vms->iommu_phandle) {
1331         return;
1332     }
1333 
1334     dev = qdev_new("arm-smmuv3");
1335 
1336     object_property_set_link(OBJECT(dev), "primary-bus", OBJECT(bus),
1337                              &error_abort);
1338     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1339     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, base);
1340     for (i = 0; i < NUM_SMMU_IRQS; i++) {
1341         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
1342                            qdev_get_gpio_in(vms->gic, irq + i));
1343     }
1344 
1345     node = g_strdup_printf("/smmuv3@%" PRIx64, base);
1346     qemu_fdt_add_subnode(ms->fdt, node);
1347     qemu_fdt_setprop(ms->fdt, node, "compatible", compat, sizeof(compat));
1348     qemu_fdt_setprop_sized_cells(ms->fdt, node, "reg", 2, base, 2, size);
1349 
1350     qemu_fdt_setprop_cells(ms->fdt, node, "interrupts",
1351             GIC_FDT_IRQ_TYPE_SPI, irq    , GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1352             GIC_FDT_IRQ_TYPE_SPI, irq + 1, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1353             GIC_FDT_IRQ_TYPE_SPI, irq + 2, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1354             GIC_FDT_IRQ_TYPE_SPI, irq + 3, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI);
1355 
1356     qemu_fdt_setprop(ms->fdt, node, "interrupt-names", irq_names,
1357                      sizeof(irq_names));
1358 
1359     qemu_fdt_setprop_cell(ms->fdt, node, "clocks", vms->clock_phandle);
1360     qemu_fdt_setprop_string(ms->fdt, node, "clock-names", "apb_pclk");
1361     qemu_fdt_setprop(ms->fdt, node, "dma-coherent", NULL, 0);
1362 
1363     qemu_fdt_setprop_cell(ms->fdt, node, "#iommu-cells", 1);
1364 
1365     qemu_fdt_setprop_cell(ms->fdt, node, "phandle", vms->iommu_phandle);
1366     g_free(node);
1367 }
1368 
1369 static void create_virtio_iommu_dt_bindings(VirtMachineState *vms)
1370 {
1371     const char compat[] = "virtio,pci-iommu";
1372     uint16_t bdf = vms->virtio_iommu_bdf;
1373     MachineState *ms = MACHINE(vms);
1374     char *node;
1375 
1376     vms->iommu_phandle = qemu_fdt_alloc_phandle(ms->fdt);
1377 
1378     node = g_strdup_printf("%s/virtio_iommu@%d", vms->pciehb_nodename, bdf);
1379     qemu_fdt_add_subnode(ms->fdt, node);
1380     qemu_fdt_setprop(ms->fdt, node, "compatible", compat, sizeof(compat));
1381     qemu_fdt_setprop_sized_cells(ms->fdt, node, "reg",
1382                                  1, bdf << 8, 1, 0, 1, 0,
1383                                  1, 0, 1, 0);
1384 
1385     qemu_fdt_setprop_cell(ms->fdt, node, "#iommu-cells", 1);
1386     qemu_fdt_setprop_cell(ms->fdt, node, "phandle", vms->iommu_phandle);
1387     g_free(node);
1388 
1389     qemu_fdt_setprop_cells(ms->fdt, vms->pciehb_nodename, "iommu-map",
1390                            0x0, vms->iommu_phandle, 0x0, bdf,
1391                            bdf + 1, vms->iommu_phandle, bdf + 1, 0xffff - bdf);
1392 }
1393 
1394 static void create_pcie(VirtMachineState *vms)
1395 {
1396     hwaddr base_mmio = vms->memmap[VIRT_PCIE_MMIO].base;
1397     hwaddr size_mmio = vms->memmap[VIRT_PCIE_MMIO].size;
1398     hwaddr base_mmio_high = vms->memmap[VIRT_HIGH_PCIE_MMIO].base;
1399     hwaddr size_mmio_high = vms->memmap[VIRT_HIGH_PCIE_MMIO].size;
1400     hwaddr base_pio = vms->memmap[VIRT_PCIE_PIO].base;
1401     hwaddr size_pio = vms->memmap[VIRT_PCIE_PIO].size;
1402     hwaddr base_ecam, size_ecam;
1403     hwaddr base = base_mmio;
1404     int nr_pcie_buses;
1405     int irq = vms->irqmap[VIRT_PCIE];
1406     MemoryRegion *mmio_alias;
1407     MemoryRegion *mmio_reg;
1408     MemoryRegion *ecam_alias;
1409     MemoryRegion *ecam_reg;
1410     DeviceState *dev;
1411     char *nodename;
1412     int i, ecam_id;
1413     PCIHostState *pci;
1414     MachineState *ms = MACHINE(vms);
1415 
1416     dev = qdev_new(TYPE_GPEX_HOST);
1417     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1418 
1419     ecam_id = VIRT_ECAM_ID(vms->highmem_ecam);
1420     base_ecam = vms->memmap[ecam_id].base;
1421     size_ecam = vms->memmap[ecam_id].size;
1422     nr_pcie_buses = size_ecam / PCIE_MMCFG_SIZE_MIN;
1423     /* Map only the first size_ecam bytes of ECAM space */
1424     ecam_alias = g_new0(MemoryRegion, 1);
1425     ecam_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 0);
1426     memory_region_init_alias(ecam_alias, OBJECT(dev), "pcie-ecam",
1427                              ecam_reg, 0, size_ecam);
1428     memory_region_add_subregion(get_system_memory(), base_ecam, ecam_alias);
1429 
1430     /* Map the MMIO window into system address space so as to expose
1431      * the section of PCI MMIO space which starts at the same base address
1432      * (ie 1:1 mapping for that part of PCI MMIO space visible through
1433      * the window).
1434      */
1435     mmio_alias = g_new0(MemoryRegion, 1);
1436     mmio_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 1);
1437     memory_region_init_alias(mmio_alias, OBJECT(dev), "pcie-mmio",
1438                              mmio_reg, base_mmio, size_mmio);
1439     memory_region_add_subregion(get_system_memory(), base_mmio, mmio_alias);
1440 
1441     if (vms->highmem_mmio) {
1442         /* Map high MMIO space */
1443         MemoryRegion *high_mmio_alias = g_new0(MemoryRegion, 1);
1444 
1445         memory_region_init_alias(high_mmio_alias, OBJECT(dev), "pcie-mmio-high",
1446                                  mmio_reg, base_mmio_high, size_mmio_high);
1447         memory_region_add_subregion(get_system_memory(), base_mmio_high,
1448                                     high_mmio_alias);
1449     }
1450 
1451     /* Map IO port space */
1452     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 2, base_pio);
1453 
1454     for (i = 0; i < GPEX_NUM_IRQS; i++) {
1455         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
1456                            qdev_get_gpio_in(vms->gic, irq + i));
1457         gpex_set_irq_num(GPEX_HOST(dev), i, irq + i);
1458     }
1459 
1460     pci = PCI_HOST_BRIDGE(dev);
1461     pci->bypass_iommu = vms->default_bus_bypass_iommu;
1462     vms->bus = pci->bus;
1463     if (vms->bus) {
1464         for (i = 0; i < nb_nics; i++) {
1465             NICInfo *nd = &nd_table[i];
1466 
1467             if (!nd->model) {
1468                 nd->model = g_strdup("virtio");
1469             }
1470 
1471             pci_nic_init_nofail(nd, pci->bus, nd->model, NULL);
1472         }
1473     }
1474 
1475     nodename = vms->pciehb_nodename = g_strdup_printf("/pcie@%" PRIx64, base);
1476     qemu_fdt_add_subnode(ms->fdt, nodename);
1477     qemu_fdt_setprop_string(ms->fdt, nodename,
1478                             "compatible", "pci-host-ecam-generic");
1479     qemu_fdt_setprop_string(ms->fdt, nodename, "device_type", "pci");
1480     qemu_fdt_setprop_cell(ms->fdt, nodename, "#address-cells", 3);
1481     qemu_fdt_setprop_cell(ms->fdt, nodename, "#size-cells", 2);
1482     qemu_fdt_setprop_cell(ms->fdt, nodename, "linux,pci-domain", 0);
1483     qemu_fdt_setprop_cells(ms->fdt, nodename, "bus-range", 0,
1484                            nr_pcie_buses - 1);
1485     qemu_fdt_setprop(ms->fdt, nodename, "dma-coherent", NULL, 0);
1486 
1487     if (vms->msi_phandle) {
1488         qemu_fdt_setprop_cells(ms->fdt, nodename, "msi-parent",
1489                                vms->msi_phandle);
1490     }
1491 
1492     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1493                                  2, base_ecam, 2, size_ecam);
1494 
1495     if (vms->highmem_mmio) {
1496         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "ranges",
1497                                      1, FDT_PCI_RANGE_IOPORT, 2, 0,
1498                                      2, base_pio, 2, size_pio,
1499                                      1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
1500                                      2, base_mmio, 2, size_mmio,
1501                                      1, FDT_PCI_RANGE_MMIO_64BIT,
1502                                      2, base_mmio_high,
1503                                      2, base_mmio_high, 2, size_mmio_high);
1504     } else {
1505         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "ranges",
1506                                      1, FDT_PCI_RANGE_IOPORT, 2, 0,
1507                                      2, base_pio, 2, size_pio,
1508                                      1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
1509                                      2, base_mmio, 2, size_mmio);
1510     }
1511 
1512     qemu_fdt_setprop_cell(ms->fdt, nodename, "#interrupt-cells", 1);
1513     create_pcie_irq_map(ms, vms->gic_phandle, irq, nodename);
1514 
1515     if (vms->iommu) {
1516         vms->iommu_phandle = qemu_fdt_alloc_phandle(ms->fdt);
1517 
1518         switch (vms->iommu) {
1519         case VIRT_IOMMU_SMMUV3:
1520             create_smmu(vms, vms->bus);
1521             qemu_fdt_setprop_cells(ms->fdt, nodename, "iommu-map",
1522                                    0x0, vms->iommu_phandle, 0x0, 0x10000);
1523             break;
1524         default:
1525             g_assert_not_reached();
1526         }
1527     }
1528 }
1529 
1530 static void create_platform_bus(VirtMachineState *vms)
1531 {
1532     DeviceState *dev;
1533     SysBusDevice *s;
1534     int i;
1535     MemoryRegion *sysmem = get_system_memory();
1536 
1537     dev = qdev_new(TYPE_PLATFORM_BUS_DEVICE);
1538     dev->id = g_strdup(TYPE_PLATFORM_BUS_DEVICE);
1539     qdev_prop_set_uint32(dev, "num_irqs", PLATFORM_BUS_NUM_IRQS);
1540     qdev_prop_set_uint32(dev, "mmio_size", vms->memmap[VIRT_PLATFORM_BUS].size);
1541     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1542     vms->platform_bus_dev = dev;
1543 
1544     s = SYS_BUS_DEVICE(dev);
1545     for (i = 0; i < PLATFORM_BUS_NUM_IRQS; i++) {
1546         int irq = vms->irqmap[VIRT_PLATFORM_BUS] + i;
1547         sysbus_connect_irq(s, i, qdev_get_gpio_in(vms->gic, irq));
1548     }
1549 
1550     memory_region_add_subregion(sysmem,
1551                                 vms->memmap[VIRT_PLATFORM_BUS].base,
1552                                 sysbus_mmio_get_region(s, 0));
1553 }
1554 
1555 static void create_tag_ram(MemoryRegion *tag_sysmem,
1556                            hwaddr base, hwaddr size,
1557                            const char *name)
1558 {
1559     MemoryRegion *tagram = g_new(MemoryRegion, 1);
1560 
1561     memory_region_init_ram(tagram, NULL, name, size / 32, &error_fatal);
1562     memory_region_add_subregion(tag_sysmem, base / 32, tagram);
1563 }
1564 
1565 static void create_secure_ram(VirtMachineState *vms,
1566                               MemoryRegion *secure_sysmem,
1567                               MemoryRegion *secure_tag_sysmem)
1568 {
1569     MemoryRegion *secram = g_new(MemoryRegion, 1);
1570     char *nodename;
1571     hwaddr base = vms->memmap[VIRT_SECURE_MEM].base;
1572     hwaddr size = vms->memmap[VIRT_SECURE_MEM].size;
1573     MachineState *ms = MACHINE(vms);
1574 
1575     memory_region_init_ram(secram, NULL, "virt.secure-ram", size,
1576                            &error_fatal);
1577     memory_region_add_subregion(secure_sysmem, base, secram);
1578 
1579     nodename = g_strdup_printf("/secram@%" PRIx64, base);
1580     qemu_fdt_add_subnode(ms->fdt, nodename);
1581     qemu_fdt_setprop_string(ms->fdt, nodename, "device_type", "memory");
1582     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg", 2, base, 2, size);
1583     qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
1584     qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
1585 
1586     if (secure_tag_sysmem) {
1587         create_tag_ram(secure_tag_sysmem, base, size, "mach-virt.secure-tag");
1588     }
1589 
1590     g_free(nodename);
1591 }
1592 
1593 static void *machvirt_dtb(const struct arm_boot_info *binfo, int *fdt_size)
1594 {
1595     const VirtMachineState *board = container_of(binfo, VirtMachineState,
1596                                                  bootinfo);
1597     MachineState *ms = MACHINE(board);
1598 
1599 
1600     *fdt_size = board->fdt_size;
1601     return ms->fdt;
1602 }
1603 
1604 static void virt_build_smbios(VirtMachineState *vms)
1605 {
1606     MachineClass *mc = MACHINE_GET_CLASS(vms);
1607     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
1608     uint8_t *smbios_tables, *smbios_anchor;
1609     size_t smbios_tables_len, smbios_anchor_len;
1610     const char *product = "QEMU Virtual Machine";
1611 
1612     if (kvm_enabled()) {
1613         product = "KVM Virtual Machine";
1614     }
1615 
1616     smbios_set_defaults("QEMU", product,
1617                         vmc->smbios_old_sys_ver ? "1.0" : mc->name, false,
1618                         true, SMBIOS_ENTRY_POINT_TYPE_64);
1619 
1620     smbios_get_tables(MACHINE(vms), NULL, 0,
1621                       &smbios_tables, &smbios_tables_len,
1622                       &smbios_anchor, &smbios_anchor_len,
1623                       &error_fatal);
1624 
1625     if (smbios_anchor) {
1626         fw_cfg_add_file(vms->fw_cfg, "etc/smbios/smbios-tables",
1627                         smbios_tables, smbios_tables_len);
1628         fw_cfg_add_file(vms->fw_cfg, "etc/smbios/smbios-anchor",
1629                         smbios_anchor, smbios_anchor_len);
1630     }
1631 }
1632 
1633 static
1634 void virt_machine_done(Notifier *notifier, void *data)
1635 {
1636     VirtMachineState *vms = container_of(notifier, VirtMachineState,
1637                                          machine_done);
1638     MachineState *ms = MACHINE(vms);
1639     ARMCPU *cpu = ARM_CPU(first_cpu);
1640     struct arm_boot_info *info = &vms->bootinfo;
1641     AddressSpace *as = arm_boot_address_space(cpu, info);
1642 
1643     /*
1644      * If the user provided a dtb, we assume the dynamic sysbus nodes
1645      * already are integrated there. This corresponds to a use case where
1646      * the dynamic sysbus nodes are complex and their generation is not yet
1647      * supported. In that case the user can take charge of the guest dt
1648      * while qemu takes charge of the qom stuff.
1649      */
1650     if (info->dtb_filename == NULL) {
1651         platform_bus_add_all_fdt_nodes(ms->fdt, "/intc",
1652                                        vms->memmap[VIRT_PLATFORM_BUS].base,
1653                                        vms->memmap[VIRT_PLATFORM_BUS].size,
1654                                        vms->irqmap[VIRT_PLATFORM_BUS]);
1655     }
1656     if (arm_load_dtb(info->dtb_start, info, info->dtb_limit, as, ms) < 0) {
1657         exit(1);
1658     }
1659 
1660     fw_cfg_add_extra_pci_roots(vms->bus, vms->fw_cfg);
1661 
1662     virt_acpi_setup(vms);
1663     virt_build_smbios(vms);
1664 }
1665 
1666 static uint64_t virt_cpu_mp_affinity(VirtMachineState *vms, int idx)
1667 {
1668     uint8_t clustersz = ARM_DEFAULT_CPUS_PER_CLUSTER;
1669     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
1670 
1671     if (!vmc->disallow_affinity_adjustment) {
1672         /* Adjust MPIDR like 64-bit KVM hosts, which incorporate the
1673          * GIC's target-list limitations. 32-bit KVM hosts currently
1674          * always create clusters of 4 CPUs, but that is expected to
1675          * change when they gain support for gicv3. When KVM is enabled
1676          * it will override the changes we make here, therefore our
1677          * purposes are to make TCG consistent (with 64-bit KVM hosts)
1678          * and to improve SGI efficiency.
1679          */
1680         if (vms->gic_version == VIRT_GIC_VERSION_2) {
1681             clustersz = GIC_TARGETLIST_BITS;
1682         } else {
1683             clustersz = GICV3_TARGETLIST_BITS;
1684         }
1685     }
1686     return arm_cpu_mp_affinity(idx, clustersz);
1687 }
1688 
1689 static void virt_set_memmap(VirtMachineState *vms, int pa_bits)
1690 {
1691     MachineState *ms = MACHINE(vms);
1692     hwaddr base, device_memory_base, device_memory_size, memtop;
1693     int i;
1694 
1695     vms->memmap = extended_memmap;
1696 
1697     for (i = 0; i < ARRAY_SIZE(base_memmap); i++) {
1698         vms->memmap[i] = base_memmap[i];
1699     }
1700 
1701     if (ms->ram_slots > ACPI_MAX_RAM_SLOTS) {
1702         error_report("unsupported number of memory slots: %"PRIu64,
1703                      ms->ram_slots);
1704         exit(EXIT_FAILURE);
1705     }
1706 
1707     /*
1708      * !highmem is exactly the same as limiting the PA space to 32bit,
1709      * irrespective of the underlying capabilities of the HW.
1710      */
1711     if (!vms->highmem) {
1712         pa_bits = 32;
1713     }
1714 
1715     /*
1716      * We compute the base of the high IO region depending on the
1717      * amount of initial and device memory. The device memory start/size
1718      * is aligned on 1GiB. We never put the high IO region below 256GiB
1719      * so that if maxram_size is < 255GiB we keep the legacy memory map.
1720      * The device region size assumes 1GiB page max alignment per slot.
1721      */
1722     device_memory_base =
1723         ROUND_UP(vms->memmap[VIRT_MEM].base + ms->ram_size, GiB);
1724     device_memory_size = ms->maxram_size - ms->ram_size + ms->ram_slots * GiB;
1725 
1726     /* Base address of the high IO region */
1727     memtop = base = device_memory_base + ROUND_UP(device_memory_size, GiB);
1728     if (memtop > BIT_ULL(pa_bits)) {
1729 	    error_report("Addressing limited to %d bits, but memory exceeds it by %llu bytes\n",
1730 			 pa_bits, memtop - BIT_ULL(pa_bits));
1731         exit(EXIT_FAILURE);
1732     }
1733     if (base < device_memory_base) {
1734         error_report("maxmem/slots too huge");
1735         exit(EXIT_FAILURE);
1736     }
1737     if (base < vms->memmap[VIRT_MEM].base + LEGACY_RAMLIMIT_BYTES) {
1738         base = vms->memmap[VIRT_MEM].base + LEGACY_RAMLIMIT_BYTES;
1739     }
1740 
1741     /* We know for sure that at least the memory fits in the PA space */
1742     vms->highest_gpa = memtop - 1;
1743 
1744     for (i = VIRT_LOWMEMMAP_LAST; i < ARRAY_SIZE(extended_memmap); i++) {
1745         hwaddr size = extended_memmap[i].size;
1746         bool fits;
1747 
1748         base = ROUND_UP(base, size);
1749         vms->memmap[i].base = base;
1750         vms->memmap[i].size = size;
1751 
1752         /*
1753          * Check each device to see if they fit in the PA space,
1754          * moving highest_gpa as we go.
1755          *
1756          * For each device that doesn't fit, disable it.
1757          */
1758         fits = (base + size) <= BIT_ULL(pa_bits);
1759         if (fits) {
1760             vms->highest_gpa = base + size - 1;
1761         }
1762 
1763         switch (i) {
1764         case VIRT_HIGH_GIC_REDIST2:
1765             vms->highmem_redists &= fits;
1766             break;
1767         case VIRT_HIGH_PCIE_ECAM:
1768             vms->highmem_ecam &= fits;
1769             break;
1770         case VIRT_HIGH_PCIE_MMIO:
1771             vms->highmem_mmio &= fits;
1772             break;
1773         }
1774 
1775         base += size;
1776     }
1777 
1778     if (device_memory_size > 0) {
1779         ms->device_memory = g_malloc0(sizeof(*ms->device_memory));
1780         ms->device_memory->base = device_memory_base;
1781         memory_region_init(&ms->device_memory->mr, OBJECT(vms),
1782                            "device-memory", device_memory_size);
1783     }
1784 }
1785 
1786 /*
1787  * finalize_gic_version - Determines the final gic_version
1788  * according to the gic-version property
1789  *
1790  * Default GIC type is v2
1791  */
1792 static void finalize_gic_version(VirtMachineState *vms)
1793 {
1794     unsigned int max_cpus = MACHINE(vms)->smp.max_cpus;
1795 
1796     if (kvm_enabled()) {
1797         int probe_bitmap;
1798 
1799         if (!kvm_irqchip_in_kernel()) {
1800             switch (vms->gic_version) {
1801             case VIRT_GIC_VERSION_HOST:
1802                 warn_report(
1803                     "gic-version=host not relevant with kernel-irqchip=off "
1804                      "as only userspace GICv2 is supported. Using v2 ...");
1805                 return;
1806             case VIRT_GIC_VERSION_MAX:
1807             case VIRT_GIC_VERSION_NOSEL:
1808                 vms->gic_version = VIRT_GIC_VERSION_2;
1809                 return;
1810             case VIRT_GIC_VERSION_2:
1811                 return;
1812             case VIRT_GIC_VERSION_3:
1813                 error_report(
1814                     "gic-version=3 is not supported with kernel-irqchip=off");
1815                 exit(1);
1816             case VIRT_GIC_VERSION_4:
1817                 error_report(
1818                     "gic-version=4 is not supported with kernel-irqchip=off");
1819                 exit(1);
1820             }
1821         }
1822 
1823         probe_bitmap = kvm_arm_vgic_probe();
1824         if (!probe_bitmap) {
1825             error_report("Unable to determine GIC version supported by host");
1826             exit(1);
1827         }
1828 
1829         switch (vms->gic_version) {
1830         case VIRT_GIC_VERSION_HOST:
1831         case VIRT_GIC_VERSION_MAX:
1832             if (probe_bitmap & KVM_ARM_VGIC_V3) {
1833                 vms->gic_version = VIRT_GIC_VERSION_3;
1834             } else {
1835                 vms->gic_version = VIRT_GIC_VERSION_2;
1836             }
1837             return;
1838         case VIRT_GIC_VERSION_NOSEL:
1839             if ((probe_bitmap & KVM_ARM_VGIC_V2) && max_cpus <= GIC_NCPU) {
1840                 vms->gic_version = VIRT_GIC_VERSION_2;
1841             } else if (probe_bitmap & KVM_ARM_VGIC_V3) {
1842                 /*
1843                  * in case the host does not support v2 in-kernel emulation or
1844                  * the end-user requested more than 8 VCPUs we now default
1845                  * to v3. In any case defaulting to v2 would be broken.
1846                  */
1847                 vms->gic_version = VIRT_GIC_VERSION_3;
1848             } else if (max_cpus > GIC_NCPU) {
1849                 error_report("host only supports in-kernel GICv2 emulation "
1850                              "but more than 8 vcpus are requested");
1851                 exit(1);
1852             }
1853             break;
1854         case VIRT_GIC_VERSION_2:
1855         case VIRT_GIC_VERSION_3:
1856             break;
1857         case VIRT_GIC_VERSION_4:
1858             error_report("gic-version=4 is not supported with KVM");
1859             exit(1);
1860         }
1861 
1862         /* Check chosen version is effectively supported by the host */
1863         if (vms->gic_version == VIRT_GIC_VERSION_2 &&
1864             !(probe_bitmap & KVM_ARM_VGIC_V2)) {
1865             error_report("host does not support in-kernel GICv2 emulation");
1866             exit(1);
1867         } else if (vms->gic_version == VIRT_GIC_VERSION_3 &&
1868                    !(probe_bitmap & KVM_ARM_VGIC_V3)) {
1869             error_report("host does not support in-kernel GICv3 emulation");
1870             exit(1);
1871         }
1872         return;
1873     }
1874 
1875     /* TCG mode */
1876     switch (vms->gic_version) {
1877     case VIRT_GIC_VERSION_NOSEL:
1878         vms->gic_version = VIRT_GIC_VERSION_2;
1879         break;
1880     case VIRT_GIC_VERSION_MAX:
1881         if (module_object_class_by_name("arm-gicv3")) {
1882             /* CONFIG_ARM_GICV3_TCG was set */
1883             if (vms->virt) {
1884                 /* GICv4 only makes sense if CPU has EL2 */
1885                 vms->gic_version = VIRT_GIC_VERSION_4;
1886             } else {
1887                 vms->gic_version = VIRT_GIC_VERSION_3;
1888             }
1889         } else {
1890             vms->gic_version = VIRT_GIC_VERSION_2;
1891         }
1892         break;
1893     case VIRT_GIC_VERSION_HOST:
1894         error_report("gic-version=host requires KVM");
1895         exit(1);
1896     case VIRT_GIC_VERSION_4:
1897         if (!vms->virt) {
1898             error_report("gic-version=4 requires virtualization enabled");
1899             exit(1);
1900         }
1901         break;
1902     case VIRT_GIC_VERSION_2:
1903     case VIRT_GIC_VERSION_3:
1904         break;
1905     }
1906 }
1907 
1908 /*
1909  * virt_cpu_post_init() must be called after the CPUs have
1910  * been realized and the GIC has been created.
1911  */
1912 static void virt_cpu_post_init(VirtMachineState *vms, MemoryRegion *sysmem)
1913 {
1914     int max_cpus = MACHINE(vms)->smp.max_cpus;
1915     bool aarch64, pmu, steal_time;
1916     CPUState *cpu;
1917 
1918     aarch64 = object_property_get_bool(OBJECT(first_cpu), "aarch64", NULL);
1919     pmu = object_property_get_bool(OBJECT(first_cpu), "pmu", NULL);
1920     steal_time = object_property_get_bool(OBJECT(first_cpu),
1921                                           "kvm-steal-time", NULL);
1922 
1923     if (kvm_enabled()) {
1924         hwaddr pvtime_reg_base = vms->memmap[VIRT_PVTIME].base;
1925         hwaddr pvtime_reg_size = vms->memmap[VIRT_PVTIME].size;
1926 
1927         if (steal_time) {
1928             MemoryRegion *pvtime = g_new(MemoryRegion, 1);
1929             hwaddr pvtime_size = max_cpus * PVTIME_SIZE_PER_CPU;
1930 
1931             /* The memory region size must be a multiple of host page size. */
1932             pvtime_size = REAL_HOST_PAGE_ALIGN(pvtime_size);
1933 
1934             if (pvtime_size > pvtime_reg_size) {
1935                 error_report("pvtime requires a %" HWADDR_PRId
1936                              " byte memory region for %d CPUs,"
1937                              " but only %" HWADDR_PRId " has been reserved",
1938                              pvtime_size, max_cpus, pvtime_reg_size);
1939                 exit(1);
1940             }
1941 
1942             memory_region_init_ram(pvtime, NULL, "pvtime", pvtime_size, NULL);
1943             memory_region_add_subregion(sysmem, pvtime_reg_base, pvtime);
1944         }
1945 
1946         CPU_FOREACH(cpu) {
1947             if (pmu) {
1948                 assert(arm_feature(&ARM_CPU(cpu)->env, ARM_FEATURE_PMU));
1949                 if (kvm_irqchip_in_kernel()) {
1950                     kvm_arm_pmu_set_irq(cpu, PPI(VIRTUAL_PMU_IRQ));
1951                 }
1952                 kvm_arm_pmu_init(cpu);
1953             }
1954             if (steal_time) {
1955                 kvm_arm_pvtime_init(cpu, pvtime_reg_base +
1956                                          cpu->cpu_index * PVTIME_SIZE_PER_CPU);
1957             }
1958         }
1959     } else {
1960         if (aarch64 && vms->highmem) {
1961             int requested_pa_size = 64 - clz64(vms->highest_gpa);
1962             int pamax = arm_pamax(ARM_CPU(first_cpu));
1963 
1964             if (pamax < requested_pa_size) {
1965                 error_report("VCPU supports less PA bits (%d) than "
1966                              "requested by the memory map (%d)",
1967                              pamax, requested_pa_size);
1968                 exit(1);
1969             }
1970         }
1971     }
1972 }
1973 
1974 static void machvirt_init(MachineState *machine)
1975 {
1976     VirtMachineState *vms = VIRT_MACHINE(machine);
1977     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(machine);
1978     MachineClass *mc = MACHINE_GET_CLASS(machine);
1979     const CPUArchIdList *possible_cpus;
1980     MemoryRegion *sysmem = get_system_memory();
1981     MemoryRegion *secure_sysmem = NULL;
1982     MemoryRegion *tag_sysmem = NULL;
1983     MemoryRegion *secure_tag_sysmem = NULL;
1984     int n, virt_max_cpus;
1985     bool firmware_loaded;
1986     bool aarch64 = true;
1987     bool has_ged = !vmc->no_ged;
1988     unsigned int smp_cpus = machine->smp.cpus;
1989     unsigned int max_cpus = machine->smp.max_cpus;
1990 
1991     if (!cpu_type_valid(machine->cpu_type)) {
1992         error_report("mach-virt: CPU type %s not supported", machine->cpu_type);
1993         exit(1);
1994     }
1995 
1996     possible_cpus = mc->possible_cpu_arch_ids(machine);
1997 
1998     /*
1999      * In accelerated mode, the memory map is computed earlier in kvm_type()
2000      * to create a VM with the right number of IPA bits.
2001      */
2002     if (!vms->memmap) {
2003         Object *cpuobj;
2004         ARMCPU *armcpu;
2005         int pa_bits;
2006 
2007         /*
2008          * Instanciate a temporary CPU object to find out about what
2009          * we are about to deal with. Once this is done, get rid of
2010          * the object.
2011          */
2012         cpuobj = object_new(possible_cpus->cpus[0].type);
2013         armcpu = ARM_CPU(cpuobj);
2014 
2015         if (object_property_get_bool(cpuobj, "aarch64", NULL)) {
2016             pa_bits = arm_pamax(armcpu);
2017         } else if (arm_feature(&armcpu->env, ARM_FEATURE_LPAE)) {
2018             /* v7 with LPAE */
2019             pa_bits = 40;
2020         } else {
2021             /* Anything else */
2022             pa_bits = 32;
2023         }
2024 
2025         object_unref(cpuobj);
2026 
2027         virt_set_memmap(vms, pa_bits);
2028     }
2029 
2030     /* We can probe only here because during property set
2031      * KVM is not available yet
2032      */
2033     finalize_gic_version(vms);
2034 
2035     if (vms->secure) {
2036         /*
2037          * The Secure view of the world is the same as the NonSecure,
2038          * but with a few extra devices. Create it as a container region
2039          * containing the system memory at low priority; any secure-only
2040          * devices go in at higher priority and take precedence.
2041          */
2042         secure_sysmem = g_new(MemoryRegion, 1);
2043         memory_region_init(secure_sysmem, OBJECT(machine), "secure-memory",
2044                            UINT64_MAX);
2045         memory_region_add_subregion_overlap(secure_sysmem, 0, sysmem, -1);
2046     }
2047 
2048     firmware_loaded = virt_firmware_init(vms, sysmem,
2049                                          secure_sysmem ?: sysmem);
2050 
2051     /* If we have an EL3 boot ROM then the assumption is that it will
2052      * implement PSCI itself, so disable QEMU's internal implementation
2053      * so it doesn't get in the way. Instead of starting secondary
2054      * CPUs in PSCI powerdown state we will start them all running and
2055      * let the boot ROM sort them out.
2056      * The usual case is that we do use QEMU's PSCI implementation;
2057      * if the guest has EL2 then we will use SMC as the conduit,
2058      * and otherwise we will use HVC (for backwards compatibility and
2059      * because if we're using KVM then we must use HVC).
2060      */
2061     if (vms->secure && firmware_loaded) {
2062         vms->psci_conduit = QEMU_PSCI_CONDUIT_DISABLED;
2063     } else if (vms->virt) {
2064         vms->psci_conduit = QEMU_PSCI_CONDUIT_SMC;
2065     } else {
2066         vms->psci_conduit = QEMU_PSCI_CONDUIT_HVC;
2067     }
2068 
2069     /*
2070      * The maximum number of CPUs depends on the GIC version, or on how
2071      * many redistributors we can fit into the memory map (which in turn
2072      * depends on whether this is a GICv3 or v4).
2073      */
2074     if (vms->gic_version == VIRT_GIC_VERSION_2) {
2075         virt_max_cpus = GIC_NCPU;
2076     } else {
2077         virt_max_cpus = virt_redist_capacity(vms, VIRT_GIC_REDIST) +
2078             virt_redist_capacity(vms, VIRT_HIGH_GIC_REDIST2);
2079     }
2080 
2081     if (max_cpus > virt_max_cpus) {
2082         error_report("Number of SMP CPUs requested (%d) exceeds max CPUs "
2083                      "supported by machine 'mach-virt' (%d)",
2084                      max_cpus, virt_max_cpus);
2085         exit(1);
2086     }
2087 
2088     if (vms->secure && (kvm_enabled() || hvf_enabled())) {
2089         error_report("mach-virt: %s does not support providing "
2090                      "Security extensions (TrustZone) to the guest CPU",
2091                      kvm_enabled() ? "KVM" : "HVF");
2092         exit(1);
2093     }
2094 
2095     if (vms->virt && (kvm_enabled() || hvf_enabled())) {
2096         error_report("mach-virt: %s does not support providing "
2097                      "Virtualization extensions to the guest CPU",
2098                      kvm_enabled() ? "KVM" : "HVF");
2099         exit(1);
2100     }
2101 
2102     if (vms->mte && (kvm_enabled() || hvf_enabled())) {
2103         error_report("mach-virt: %s does not support providing "
2104                      "MTE to the guest CPU",
2105                      kvm_enabled() ? "KVM" : "HVF");
2106         exit(1);
2107     }
2108 
2109     create_fdt(vms);
2110 
2111     assert(possible_cpus->len == max_cpus);
2112     for (n = 0; n < possible_cpus->len; n++) {
2113         Object *cpuobj;
2114         CPUState *cs;
2115 
2116         if (n >= smp_cpus) {
2117             break;
2118         }
2119 
2120         cpuobj = object_new(possible_cpus->cpus[n].type);
2121         object_property_set_int(cpuobj, "mp-affinity",
2122                                 possible_cpus->cpus[n].arch_id, NULL);
2123 
2124         cs = CPU(cpuobj);
2125         cs->cpu_index = n;
2126 
2127         numa_cpu_pre_plug(&possible_cpus->cpus[cs->cpu_index], DEVICE(cpuobj),
2128                           &error_fatal);
2129 
2130         aarch64 &= object_property_get_bool(cpuobj, "aarch64", NULL);
2131 
2132         if (!vms->secure) {
2133             object_property_set_bool(cpuobj, "has_el3", false, NULL);
2134         }
2135 
2136         if (!vms->virt && object_property_find(cpuobj, "has_el2")) {
2137             object_property_set_bool(cpuobj, "has_el2", false, NULL);
2138         }
2139 
2140         if (vmc->kvm_no_adjvtime &&
2141             object_property_find(cpuobj, "kvm-no-adjvtime")) {
2142             object_property_set_bool(cpuobj, "kvm-no-adjvtime", true, NULL);
2143         }
2144 
2145         if (vmc->no_kvm_steal_time &&
2146             object_property_find(cpuobj, "kvm-steal-time")) {
2147             object_property_set_bool(cpuobj, "kvm-steal-time", false, NULL);
2148         }
2149 
2150         if (vmc->no_pmu && object_property_find(cpuobj, "pmu")) {
2151             object_property_set_bool(cpuobj, "pmu", false, NULL);
2152         }
2153 
2154         if (vmc->no_tcg_lpa2 && object_property_find(cpuobj, "lpa2")) {
2155             object_property_set_bool(cpuobj, "lpa2", false, NULL);
2156         }
2157 
2158         if (object_property_find(cpuobj, "reset-cbar")) {
2159             object_property_set_int(cpuobj, "reset-cbar",
2160                                     vms->memmap[VIRT_CPUPERIPHS].base,
2161                                     &error_abort);
2162         }
2163 
2164         object_property_set_link(cpuobj, "memory", OBJECT(sysmem),
2165                                  &error_abort);
2166         if (vms->secure) {
2167             object_property_set_link(cpuobj, "secure-memory",
2168                                      OBJECT(secure_sysmem), &error_abort);
2169         }
2170 
2171         if (vms->mte) {
2172             /* Create the memory region only once, but link to all cpus. */
2173             if (!tag_sysmem) {
2174                 /*
2175                  * The property exists only if MemTag is supported.
2176                  * If it is, we must allocate the ram to back that up.
2177                  */
2178                 if (!object_property_find(cpuobj, "tag-memory")) {
2179                     error_report("MTE requested, but not supported "
2180                                  "by the guest CPU");
2181                     exit(1);
2182                 }
2183 
2184                 tag_sysmem = g_new(MemoryRegion, 1);
2185                 memory_region_init(tag_sysmem, OBJECT(machine),
2186                                    "tag-memory", UINT64_MAX / 32);
2187 
2188                 if (vms->secure) {
2189                     secure_tag_sysmem = g_new(MemoryRegion, 1);
2190                     memory_region_init(secure_tag_sysmem, OBJECT(machine),
2191                                        "secure-tag-memory", UINT64_MAX / 32);
2192 
2193                     /* As with ram, secure-tag takes precedence over tag.  */
2194                     memory_region_add_subregion_overlap(secure_tag_sysmem, 0,
2195                                                         tag_sysmem, -1);
2196                 }
2197             }
2198 
2199             object_property_set_link(cpuobj, "tag-memory", OBJECT(tag_sysmem),
2200                                      &error_abort);
2201             if (vms->secure) {
2202                 object_property_set_link(cpuobj, "secure-tag-memory",
2203                                          OBJECT(secure_tag_sysmem),
2204                                          &error_abort);
2205             }
2206         }
2207 
2208         qdev_realize(DEVICE(cpuobj), NULL, &error_fatal);
2209         object_unref(cpuobj);
2210     }
2211     fdt_add_timer_nodes(vms);
2212     fdt_add_cpu_nodes(vms);
2213 
2214     memory_region_add_subregion(sysmem, vms->memmap[VIRT_MEM].base,
2215                                 machine->ram);
2216     if (machine->device_memory) {
2217         memory_region_add_subregion(sysmem, machine->device_memory->base,
2218                                     &machine->device_memory->mr);
2219     }
2220 
2221     virt_flash_fdt(vms, sysmem, secure_sysmem ?: sysmem);
2222 
2223     create_gic(vms, sysmem);
2224 
2225     virt_cpu_post_init(vms, sysmem);
2226 
2227     fdt_add_pmu_nodes(vms);
2228 
2229     create_uart(vms, VIRT_UART, sysmem, serial_hd(0));
2230 
2231     if (vms->secure) {
2232         create_secure_ram(vms, secure_sysmem, secure_tag_sysmem);
2233         create_uart(vms, VIRT_SECURE_UART, secure_sysmem, serial_hd(1));
2234     }
2235 
2236     if (tag_sysmem) {
2237         create_tag_ram(tag_sysmem, vms->memmap[VIRT_MEM].base,
2238                        machine->ram_size, "mach-virt.tag");
2239     }
2240 
2241     vms->highmem_ecam &= (!firmware_loaded || aarch64);
2242 
2243     create_rtc(vms);
2244 
2245     create_pcie(vms);
2246 
2247     if (has_ged && aarch64 && firmware_loaded && virt_is_acpi_enabled(vms)) {
2248         vms->acpi_dev = create_acpi_ged(vms);
2249     } else {
2250         create_gpio_devices(vms, VIRT_GPIO, sysmem);
2251     }
2252 
2253     if (vms->secure && !vmc->no_secure_gpio) {
2254         create_gpio_devices(vms, VIRT_SECURE_GPIO, secure_sysmem);
2255     }
2256 
2257      /* connect powerdown request */
2258      vms->powerdown_notifier.notify = virt_powerdown_req;
2259      qemu_register_powerdown_notifier(&vms->powerdown_notifier);
2260 
2261     /* Create mmio transports, so the user can create virtio backends
2262      * (which will be automatically plugged in to the transports). If
2263      * no backend is created the transport will just sit harmlessly idle.
2264      */
2265     create_virtio_devices(vms);
2266 
2267     vms->fw_cfg = create_fw_cfg(vms, &address_space_memory);
2268     rom_set_fw(vms->fw_cfg);
2269 
2270     create_platform_bus(vms);
2271 
2272     if (machine->nvdimms_state->is_enabled) {
2273         const struct AcpiGenericAddress arm_virt_nvdimm_acpi_dsmio = {
2274             .space_id = AML_AS_SYSTEM_MEMORY,
2275             .address = vms->memmap[VIRT_NVDIMM_ACPI].base,
2276             .bit_width = NVDIMM_ACPI_IO_LEN << 3
2277         };
2278 
2279         nvdimm_init_acpi_state(machine->nvdimms_state, sysmem,
2280                                arm_virt_nvdimm_acpi_dsmio,
2281                                vms->fw_cfg, OBJECT(vms));
2282     }
2283 
2284     vms->bootinfo.ram_size = machine->ram_size;
2285     vms->bootinfo.board_id = -1;
2286     vms->bootinfo.loader_start = vms->memmap[VIRT_MEM].base;
2287     vms->bootinfo.get_dtb = machvirt_dtb;
2288     vms->bootinfo.skip_dtb_autoload = true;
2289     vms->bootinfo.firmware_loaded = firmware_loaded;
2290     vms->bootinfo.psci_conduit = vms->psci_conduit;
2291     arm_load_kernel(ARM_CPU(first_cpu), machine, &vms->bootinfo);
2292 
2293     vms->machine_done.notify = virt_machine_done;
2294     qemu_add_machine_init_done_notifier(&vms->machine_done);
2295 }
2296 
2297 static bool virt_get_secure(Object *obj, Error **errp)
2298 {
2299     VirtMachineState *vms = VIRT_MACHINE(obj);
2300 
2301     return vms->secure;
2302 }
2303 
2304 static void virt_set_secure(Object *obj, bool value, Error **errp)
2305 {
2306     VirtMachineState *vms = VIRT_MACHINE(obj);
2307 
2308     vms->secure = value;
2309 }
2310 
2311 static bool virt_get_virt(Object *obj, Error **errp)
2312 {
2313     VirtMachineState *vms = VIRT_MACHINE(obj);
2314 
2315     return vms->virt;
2316 }
2317 
2318 static void virt_set_virt(Object *obj, bool value, Error **errp)
2319 {
2320     VirtMachineState *vms = VIRT_MACHINE(obj);
2321 
2322     vms->virt = value;
2323 }
2324 
2325 static bool virt_get_highmem(Object *obj, Error **errp)
2326 {
2327     VirtMachineState *vms = VIRT_MACHINE(obj);
2328 
2329     return vms->highmem;
2330 }
2331 
2332 static void virt_set_highmem(Object *obj, bool value, Error **errp)
2333 {
2334     VirtMachineState *vms = VIRT_MACHINE(obj);
2335 
2336     vms->highmem = value;
2337 }
2338 
2339 static bool virt_get_its(Object *obj, Error **errp)
2340 {
2341     VirtMachineState *vms = VIRT_MACHINE(obj);
2342 
2343     return vms->its;
2344 }
2345 
2346 static void virt_set_its(Object *obj, bool value, Error **errp)
2347 {
2348     VirtMachineState *vms = VIRT_MACHINE(obj);
2349 
2350     vms->its = value;
2351 }
2352 
2353 static bool virt_get_dtb_kaslr_seed(Object *obj, Error **errp)
2354 {
2355     VirtMachineState *vms = VIRT_MACHINE(obj);
2356 
2357     return vms->dtb_kaslr_seed;
2358 }
2359 
2360 static void virt_set_dtb_kaslr_seed(Object *obj, bool value, Error **errp)
2361 {
2362     VirtMachineState *vms = VIRT_MACHINE(obj);
2363 
2364     vms->dtb_kaslr_seed = value;
2365 }
2366 
2367 static char *virt_get_oem_id(Object *obj, Error **errp)
2368 {
2369     VirtMachineState *vms = VIRT_MACHINE(obj);
2370 
2371     return g_strdup(vms->oem_id);
2372 }
2373 
2374 static void virt_set_oem_id(Object *obj, const char *value, Error **errp)
2375 {
2376     VirtMachineState *vms = VIRT_MACHINE(obj);
2377     size_t len = strlen(value);
2378 
2379     if (len > 6) {
2380         error_setg(errp,
2381                    "User specified oem-id value is bigger than 6 bytes in size");
2382         return;
2383     }
2384 
2385     strncpy(vms->oem_id, value, 6);
2386 }
2387 
2388 static char *virt_get_oem_table_id(Object *obj, Error **errp)
2389 {
2390     VirtMachineState *vms = VIRT_MACHINE(obj);
2391 
2392     return g_strdup(vms->oem_table_id);
2393 }
2394 
2395 static void virt_set_oem_table_id(Object *obj, const char *value,
2396                                   Error **errp)
2397 {
2398     VirtMachineState *vms = VIRT_MACHINE(obj);
2399     size_t len = strlen(value);
2400 
2401     if (len > 8) {
2402         error_setg(errp,
2403                    "User specified oem-table-id value is bigger than 8 bytes in size");
2404         return;
2405     }
2406     strncpy(vms->oem_table_id, value, 8);
2407 }
2408 
2409 
2410 bool virt_is_acpi_enabled(VirtMachineState *vms)
2411 {
2412     if (vms->acpi == ON_OFF_AUTO_OFF) {
2413         return false;
2414     }
2415     return true;
2416 }
2417 
2418 static void virt_get_acpi(Object *obj, Visitor *v, const char *name,
2419                           void *opaque, Error **errp)
2420 {
2421     VirtMachineState *vms = VIRT_MACHINE(obj);
2422     OnOffAuto acpi = vms->acpi;
2423 
2424     visit_type_OnOffAuto(v, name, &acpi, errp);
2425 }
2426 
2427 static void virt_set_acpi(Object *obj, Visitor *v, const char *name,
2428                           void *opaque, Error **errp)
2429 {
2430     VirtMachineState *vms = VIRT_MACHINE(obj);
2431 
2432     visit_type_OnOffAuto(v, name, &vms->acpi, errp);
2433 }
2434 
2435 static bool virt_get_ras(Object *obj, Error **errp)
2436 {
2437     VirtMachineState *vms = VIRT_MACHINE(obj);
2438 
2439     return vms->ras;
2440 }
2441 
2442 static void virt_set_ras(Object *obj, bool value, Error **errp)
2443 {
2444     VirtMachineState *vms = VIRT_MACHINE(obj);
2445 
2446     vms->ras = value;
2447 }
2448 
2449 static bool virt_get_mte(Object *obj, Error **errp)
2450 {
2451     VirtMachineState *vms = VIRT_MACHINE(obj);
2452 
2453     return vms->mte;
2454 }
2455 
2456 static void virt_set_mte(Object *obj, bool value, Error **errp)
2457 {
2458     VirtMachineState *vms = VIRT_MACHINE(obj);
2459 
2460     vms->mte = value;
2461 }
2462 
2463 static char *virt_get_gic_version(Object *obj, Error **errp)
2464 {
2465     VirtMachineState *vms = VIRT_MACHINE(obj);
2466     const char *val;
2467 
2468     switch (vms->gic_version) {
2469     case VIRT_GIC_VERSION_4:
2470         val = "4";
2471         break;
2472     case VIRT_GIC_VERSION_3:
2473         val = "3";
2474         break;
2475     default:
2476         val = "2";
2477         break;
2478     }
2479     return g_strdup(val);
2480 }
2481 
2482 static void virt_set_gic_version(Object *obj, const char *value, Error **errp)
2483 {
2484     VirtMachineState *vms = VIRT_MACHINE(obj);
2485 
2486     if (!strcmp(value, "4")) {
2487         vms->gic_version = VIRT_GIC_VERSION_4;
2488     } else if (!strcmp(value, "3")) {
2489         vms->gic_version = VIRT_GIC_VERSION_3;
2490     } else if (!strcmp(value, "2")) {
2491         vms->gic_version = VIRT_GIC_VERSION_2;
2492     } else if (!strcmp(value, "host")) {
2493         vms->gic_version = VIRT_GIC_VERSION_HOST; /* Will probe later */
2494     } else if (!strcmp(value, "max")) {
2495         vms->gic_version = VIRT_GIC_VERSION_MAX; /* Will probe later */
2496     } else {
2497         error_setg(errp, "Invalid gic-version value");
2498         error_append_hint(errp, "Valid values are 3, 2, host, max.\n");
2499     }
2500 }
2501 
2502 static char *virt_get_iommu(Object *obj, Error **errp)
2503 {
2504     VirtMachineState *vms = VIRT_MACHINE(obj);
2505 
2506     switch (vms->iommu) {
2507     case VIRT_IOMMU_NONE:
2508         return g_strdup("none");
2509     case VIRT_IOMMU_SMMUV3:
2510         return g_strdup("smmuv3");
2511     default:
2512         g_assert_not_reached();
2513     }
2514 }
2515 
2516 static void virt_set_iommu(Object *obj, const char *value, Error **errp)
2517 {
2518     VirtMachineState *vms = VIRT_MACHINE(obj);
2519 
2520     if (!strcmp(value, "smmuv3")) {
2521         vms->iommu = VIRT_IOMMU_SMMUV3;
2522     } else if (!strcmp(value, "none")) {
2523         vms->iommu = VIRT_IOMMU_NONE;
2524     } else {
2525         error_setg(errp, "Invalid iommu value");
2526         error_append_hint(errp, "Valid values are none, smmuv3.\n");
2527     }
2528 }
2529 
2530 static bool virt_get_default_bus_bypass_iommu(Object *obj, Error **errp)
2531 {
2532     VirtMachineState *vms = VIRT_MACHINE(obj);
2533 
2534     return vms->default_bus_bypass_iommu;
2535 }
2536 
2537 static void virt_set_default_bus_bypass_iommu(Object *obj, bool value,
2538                                               Error **errp)
2539 {
2540     VirtMachineState *vms = VIRT_MACHINE(obj);
2541 
2542     vms->default_bus_bypass_iommu = value;
2543 }
2544 
2545 static CpuInstanceProperties
2546 virt_cpu_index_to_props(MachineState *ms, unsigned cpu_index)
2547 {
2548     MachineClass *mc = MACHINE_GET_CLASS(ms);
2549     const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(ms);
2550 
2551     assert(cpu_index < possible_cpus->len);
2552     return possible_cpus->cpus[cpu_index].props;
2553 }
2554 
2555 static int64_t virt_get_default_cpu_node_id(const MachineState *ms, int idx)
2556 {
2557     int64_t socket_id = ms->possible_cpus->cpus[idx].props.socket_id;
2558 
2559     return socket_id % ms->numa_state->num_nodes;
2560 }
2561 
2562 static const CPUArchIdList *virt_possible_cpu_arch_ids(MachineState *ms)
2563 {
2564     int n;
2565     unsigned int max_cpus = ms->smp.max_cpus;
2566     VirtMachineState *vms = VIRT_MACHINE(ms);
2567     MachineClass *mc = MACHINE_GET_CLASS(vms);
2568 
2569     if (ms->possible_cpus) {
2570         assert(ms->possible_cpus->len == max_cpus);
2571         return ms->possible_cpus;
2572     }
2573 
2574     ms->possible_cpus = g_malloc0(sizeof(CPUArchIdList) +
2575                                   sizeof(CPUArchId) * max_cpus);
2576     ms->possible_cpus->len = max_cpus;
2577     for (n = 0; n < ms->possible_cpus->len; n++) {
2578         ms->possible_cpus->cpus[n].type = ms->cpu_type;
2579         ms->possible_cpus->cpus[n].arch_id =
2580             virt_cpu_mp_affinity(vms, n);
2581 
2582         assert(!mc->smp_props.dies_supported);
2583         ms->possible_cpus->cpus[n].props.has_socket_id = true;
2584         ms->possible_cpus->cpus[n].props.socket_id =
2585             n / (ms->smp.clusters * ms->smp.cores * ms->smp.threads);
2586         ms->possible_cpus->cpus[n].props.has_cluster_id = true;
2587         ms->possible_cpus->cpus[n].props.cluster_id =
2588             (n / (ms->smp.cores * ms->smp.threads)) % ms->smp.clusters;
2589         ms->possible_cpus->cpus[n].props.has_core_id = true;
2590         ms->possible_cpus->cpus[n].props.core_id =
2591             (n / ms->smp.threads) % ms->smp.cores;
2592         ms->possible_cpus->cpus[n].props.has_thread_id = true;
2593         ms->possible_cpus->cpus[n].props.thread_id =
2594             n % ms->smp.threads;
2595     }
2596     return ms->possible_cpus;
2597 }
2598 
2599 static void virt_memory_pre_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
2600                                  Error **errp)
2601 {
2602     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2603     const MachineState *ms = MACHINE(hotplug_dev);
2604     const bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2605 
2606     if (!vms->acpi_dev) {
2607         error_setg(errp,
2608                    "memory hotplug is not enabled: missing acpi-ged device");
2609         return;
2610     }
2611 
2612     if (vms->mte) {
2613         error_setg(errp, "memory hotplug is not enabled: MTE is enabled");
2614         return;
2615     }
2616 
2617     if (is_nvdimm && !ms->nvdimms_state->is_enabled) {
2618         error_setg(errp, "nvdimm is not enabled: add 'nvdimm=on' to '-M'");
2619         return;
2620     }
2621 
2622     pc_dimm_pre_plug(PC_DIMM(dev), MACHINE(hotplug_dev), NULL, errp);
2623 }
2624 
2625 static void virt_memory_plug(HotplugHandler *hotplug_dev,
2626                              DeviceState *dev, Error **errp)
2627 {
2628     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2629     MachineState *ms = MACHINE(hotplug_dev);
2630     bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2631 
2632     pc_dimm_plug(PC_DIMM(dev), MACHINE(vms));
2633 
2634     if (is_nvdimm) {
2635         nvdimm_plug(ms->nvdimms_state);
2636     }
2637 
2638     hotplug_handler_plug(HOTPLUG_HANDLER(vms->acpi_dev),
2639                          dev, &error_abort);
2640 }
2641 
2642 static void virt_virtio_md_pci_pre_plug(HotplugHandler *hotplug_dev,
2643                                         DeviceState *dev, Error **errp)
2644 {
2645     HotplugHandler *hotplug_dev2 = qdev_get_bus_hotplug_handler(dev);
2646     Error *local_err = NULL;
2647 
2648     if (!hotplug_dev2 && dev->hotplugged) {
2649         /*
2650          * Without a bus hotplug handler, we cannot control the plug/unplug
2651          * order. We should never reach this point when hotplugging on ARM.
2652          * However, it's nice to add a safety net, similar to what we have
2653          * on x86.
2654          */
2655         error_setg(errp, "hotplug of virtio based memory devices not supported"
2656                    " on this bus.");
2657         return;
2658     }
2659     /*
2660      * First, see if we can plug this memory device at all. If that
2661      * succeeds, branch of to the actual hotplug handler.
2662      */
2663     memory_device_pre_plug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev), NULL,
2664                            &local_err);
2665     if (!local_err && hotplug_dev2) {
2666         hotplug_handler_pre_plug(hotplug_dev2, dev, &local_err);
2667     }
2668     error_propagate(errp, local_err);
2669 }
2670 
2671 static void virt_virtio_md_pci_plug(HotplugHandler *hotplug_dev,
2672                                     DeviceState *dev, Error **errp)
2673 {
2674     HotplugHandler *hotplug_dev2 = qdev_get_bus_hotplug_handler(dev);
2675     Error *local_err = NULL;
2676 
2677     /*
2678      * Plug the memory device first and then branch off to the actual
2679      * hotplug handler. If that one fails, we can easily undo the memory
2680      * device bits.
2681      */
2682     memory_device_plug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev));
2683     if (hotplug_dev2) {
2684         hotplug_handler_plug(hotplug_dev2, dev, &local_err);
2685         if (local_err) {
2686             memory_device_unplug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev));
2687         }
2688     }
2689     error_propagate(errp, local_err);
2690 }
2691 
2692 static void virt_virtio_md_pci_unplug_request(HotplugHandler *hotplug_dev,
2693                                               DeviceState *dev, Error **errp)
2694 {
2695     /* We don't support hot unplug of virtio based memory devices */
2696     error_setg(errp, "virtio based memory devices cannot be unplugged.");
2697 }
2698 
2699 
2700 static void virt_machine_device_pre_plug_cb(HotplugHandler *hotplug_dev,
2701                                             DeviceState *dev, Error **errp)
2702 {
2703     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2704 
2705     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2706         virt_memory_pre_plug(hotplug_dev, dev, errp);
2707     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI)) {
2708         virt_virtio_md_pci_pre_plug(hotplug_dev, dev, errp);
2709     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2710         hwaddr db_start = 0, db_end = 0;
2711         char *resv_prop_str;
2712 
2713         if (vms->iommu != VIRT_IOMMU_NONE) {
2714             error_setg(errp, "virt machine does not support multiple IOMMUs");
2715             return;
2716         }
2717 
2718         switch (vms->msi_controller) {
2719         case VIRT_MSI_CTRL_NONE:
2720             return;
2721         case VIRT_MSI_CTRL_ITS:
2722             /* GITS_TRANSLATER page */
2723             db_start = base_memmap[VIRT_GIC_ITS].base + 0x10000;
2724             db_end = base_memmap[VIRT_GIC_ITS].base +
2725                      base_memmap[VIRT_GIC_ITS].size - 1;
2726             break;
2727         case VIRT_MSI_CTRL_GICV2M:
2728             /* MSI_SETSPI_NS page */
2729             db_start = base_memmap[VIRT_GIC_V2M].base;
2730             db_end = db_start + base_memmap[VIRT_GIC_V2M].size - 1;
2731             break;
2732         }
2733         resv_prop_str = g_strdup_printf("0x%"PRIx64":0x%"PRIx64":%u",
2734                                         db_start, db_end,
2735                                         VIRTIO_IOMMU_RESV_MEM_T_MSI);
2736 
2737         object_property_set_uint(OBJECT(dev), "len-reserved-regions", 1, errp);
2738         object_property_set_str(OBJECT(dev), "reserved-regions[0]",
2739                                 resv_prop_str, errp);
2740         g_free(resv_prop_str);
2741     }
2742 }
2743 
2744 static void virt_machine_device_plug_cb(HotplugHandler *hotplug_dev,
2745                                         DeviceState *dev, Error **errp)
2746 {
2747     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2748 
2749     if (vms->platform_bus_dev) {
2750         MachineClass *mc = MACHINE_GET_CLASS(vms);
2751 
2752         if (device_is_dynamic_sysbus(mc, dev)) {
2753             platform_bus_link_device(PLATFORM_BUS_DEVICE(vms->platform_bus_dev),
2754                                      SYS_BUS_DEVICE(dev));
2755         }
2756     }
2757     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2758         virt_memory_plug(hotplug_dev, dev, errp);
2759     }
2760 
2761     if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI)) {
2762         virt_virtio_md_pci_plug(hotplug_dev, dev, errp);
2763     }
2764 
2765     if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2766         PCIDevice *pdev = PCI_DEVICE(dev);
2767 
2768         vms->iommu = VIRT_IOMMU_VIRTIO;
2769         vms->virtio_iommu_bdf = pci_get_bdf(pdev);
2770         create_virtio_iommu_dt_bindings(vms);
2771     }
2772 }
2773 
2774 static void virt_dimm_unplug_request(HotplugHandler *hotplug_dev,
2775                                      DeviceState *dev, Error **errp)
2776 {
2777     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2778     Error *local_err = NULL;
2779 
2780     if (!vms->acpi_dev) {
2781         error_setg(&local_err,
2782                    "memory hotplug is not enabled: missing acpi-ged device");
2783         goto out;
2784     }
2785 
2786     if (object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM)) {
2787         error_setg(&local_err,
2788                    "nvdimm device hot unplug is not supported yet.");
2789         goto out;
2790     }
2791 
2792     hotplug_handler_unplug_request(HOTPLUG_HANDLER(vms->acpi_dev), dev,
2793                                    &local_err);
2794 out:
2795     error_propagate(errp, local_err);
2796 }
2797 
2798 static void virt_dimm_unplug(HotplugHandler *hotplug_dev,
2799                              DeviceState *dev, Error **errp)
2800 {
2801     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2802     Error *local_err = NULL;
2803 
2804     hotplug_handler_unplug(HOTPLUG_HANDLER(vms->acpi_dev), dev, &local_err);
2805     if (local_err) {
2806         goto out;
2807     }
2808 
2809     pc_dimm_unplug(PC_DIMM(dev), MACHINE(vms));
2810     qdev_unrealize(dev);
2811 
2812 out:
2813     error_propagate(errp, local_err);
2814 }
2815 
2816 static void virt_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
2817                                           DeviceState *dev, Error **errp)
2818 {
2819     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2820         virt_dimm_unplug_request(hotplug_dev, dev, errp);
2821     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI)) {
2822         virt_virtio_md_pci_unplug_request(hotplug_dev, dev, errp);
2823     } else {
2824         error_setg(errp, "device unplug request for unsupported device"
2825                    " type: %s", object_get_typename(OBJECT(dev)));
2826     }
2827 }
2828 
2829 static void virt_machine_device_unplug_cb(HotplugHandler *hotplug_dev,
2830                                           DeviceState *dev, Error **errp)
2831 {
2832     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2833         virt_dimm_unplug(hotplug_dev, dev, errp);
2834     } else {
2835         error_setg(errp, "virt: device unplug for unsupported device"
2836                    " type: %s", object_get_typename(OBJECT(dev)));
2837     }
2838 }
2839 
2840 static HotplugHandler *virt_machine_get_hotplug_handler(MachineState *machine,
2841                                                         DeviceState *dev)
2842 {
2843     MachineClass *mc = MACHINE_GET_CLASS(machine);
2844 
2845     if (device_is_dynamic_sysbus(mc, dev) ||
2846         object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) ||
2847         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI) ||
2848         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2849         return HOTPLUG_HANDLER(machine);
2850     }
2851     return NULL;
2852 }
2853 
2854 /*
2855  * for arm64 kvm_type [7-0] encodes the requested number of bits
2856  * in the IPA address space
2857  */
2858 static int virt_kvm_type(MachineState *ms, const char *type_str)
2859 {
2860     VirtMachineState *vms = VIRT_MACHINE(ms);
2861     int max_vm_pa_size, requested_pa_size;
2862     bool fixed_ipa;
2863 
2864     max_vm_pa_size = kvm_arm_get_max_vm_ipa_size(ms, &fixed_ipa);
2865 
2866     /* we freeze the memory map to compute the highest gpa */
2867     virt_set_memmap(vms, max_vm_pa_size);
2868 
2869     requested_pa_size = 64 - clz64(vms->highest_gpa);
2870 
2871     /*
2872      * KVM requires the IPA size to be at least 32 bits.
2873      */
2874     if (requested_pa_size < 32) {
2875         requested_pa_size = 32;
2876     }
2877 
2878     if (requested_pa_size > max_vm_pa_size) {
2879         error_report("-m and ,maxmem option values "
2880                      "require an IPA range (%d bits) larger than "
2881                      "the one supported by the host (%d bits)",
2882                      requested_pa_size, max_vm_pa_size);
2883         exit(1);
2884     }
2885     /*
2886      * We return the requested PA log size, unless KVM only supports
2887      * the implicit legacy 40b IPA setting, in which case the kvm_type
2888      * must be 0.
2889      */
2890     return fixed_ipa ? 0 : requested_pa_size;
2891 }
2892 
2893 static void virt_machine_class_init(ObjectClass *oc, void *data)
2894 {
2895     MachineClass *mc = MACHINE_CLASS(oc);
2896     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
2897 
2898     mc->init = machvirt_init;
2899     /* Start with max_cpus set to 512, which is the maximum supported by KVM.
2900      * The value may be reduced later when we have more information about the
2901      * configuration of the particular instance.
2902      */
2903     mc->max_cpus = 512;
2904     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_CALXEDA_XGMAC);
2905     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_AMD_XGBE);
2906     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_RAMFB_DEVICE);
2907     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_PLATFORM);
2908 #ifdef CONFIG_TPM
2909     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_TPM_TIS_SYSBUS);
2910 #endif
2911     mc->block_default_type = IF_VIRTIO;
2912     mc->no_cdrom = 1;
2913     mc->pci_allow_0_address = true;
2914     /* We know we will never create a pre-ARMv7 CPU which needs 1K pages */
2915     mc->minimum_page_bits = 12;
2916     mc->possible_cpu_arch_ids = virt_possible_cpu_arch_ids;
2917     mc->cpu_index_to_instance_props = virt_cpu_index_to_props;
2918     mc->default_cpu_type = ARM_CPU_TYPE_NAME("cortex-a15");
2919     mc->get_default_cpu_node_id = virt_get_default_cpu_node_id;
2920     mc->kvm_type = virt_kvm_type;
2921     assert(!mc->get_hotplug_handler);
2922     mc->get_hotplug_handler = virt_machine_get_hotplug_handler;
2923     hc->pre_plug = virt_machine_device_pre_plug_cb;
2924     hc->plug = virt_machine_device_plug_cb;
2925     hc->unplug_request = virt_machine_device_unplug_request_cb;
2926     hc->unplug = virt_machine_device_unplug_cb;
2927     mc->nvdimm_supported = true;
2928     mc->smp_props.clusters_supported = true;
2929     mc->auto_enable_numa_with_memhp = true;
2930     mc->auto_enable_numa_with_memdev = true;
2931     mc->default_ram_id = "mach-virt.ram";
2932 
2933     object_class_property_add(oc, "acpi", "OnOffAuto",
2934         virt_get_acpi, virt_set_acpi,
2935         NULL, NULL);
2936     object_class_property_set_description(oc, "acpi",
2937         "Enable ACPI");
2938     object_class_property_add_bool(oc, "secure", virt_get_secure,
2939                                    virt_set_secure);
2940     object_class_property_set_description(oc, "secure",
2941                                                 "Set on/off to enable/disable the ARM "
2942                                                 "Security Extensions (TrustZone)");
2943 
2944     object_class_property_add_bool(oc, "virtualization", virt_get_virt,
2945                                    virt_set_virt);
2946     object_class_property_set_description(oc, "virtualization",
2947                                           "Set on/off to enable/disable emulating a "
2948                                           "guest CPU which implements the ARM "
2949                                           "Virtualization Extensions");
2950 
2951     object_class_property_add_bool(oc, "highmem", virt_get_highmem,
2952                                    virt_set_highmem);
2953     object_class_property_set_description(oc, "highmem",
2954                                           "Set on/off to enable/disable using "
2955                                           "physical address space above 32 bits");
2956 
2957     object_class_property_add_str(oc, "gic-version", virt_get_gic_version,
2958                                   virt_set_gic_version);
2959     object_class_property_set_description(oc, "gic-version",
2960                                           "Set GIC version. "
2961                                           "Valid values are 2, 3, 4, host and max");
2962 
2963     object_class_property_add_str(oc, "iommu", virt_get_iommu, virt_set_iommu);
2964     object_class_property_set_description(oc, "iommu",
2965                                           "Set the IOMMU type. "
2966                                           "Valid values are none and smmuv3");
2967 
2968     object_class_property_add_bool(oc, "default-bus-bypass-iommu",
2969                                    virt_get_default_bus_bypass_iommu,
2970                                    virt_set_default_bus_bypass_iommu);
2971     object_class_property_set_description(oc, "default-bus-bypass-iommu",
2972                                           "Set on/off to enable/disable "
2973                                           "bypass_iommu for default root bus");
2974 
2975     object_class_property_add_bool(oc, "ras", virt_get_ras,
2976                                    virt_set_ras);
2977     object_class_property_set_description(oc, "ras",
2978                                           "Set on/off to enable/disable reporting host memory errors "
2979                                           "to a KVM guest using ACPI and guest external abort exceptions");
2980 
2981     object_class_property_add_bool(oc, "mte", virt_get_mte, virt_set_mte);
2982     object_class_property_set_description(oc, "mte",
2983                                           "Set on/off to enable/disable emulating a "
2984                                           "guest CPU which implements the ARM "
2985                                           "Memory Tagging Extension");
2986 
2987     object_class_property_add_bool(oc, "its", virt_get_its,
2988                                    virt_set_its);
2989     object_class_property_set_description(oc, "its",
2990                                           "Set on/off to enable/disable "
2991                                           "ITS instantiation");
2992 
2993     object_class_property_add_bool(oc, "dtb-kaslr-seed",
2994                                    virt_get_dtb_kaslr_seed,
2995                                    virt_set_dtb_kaslr_seed);
2996     object_class_property_set_description(oc, "dtb-kaslr-seed",
2997                                           "Set off to disable passing of kaslr-seed "
2998                                           "dtb node to guest");
2999 
3000     object_class_property_add_str(oc, "x-oem-id",
3001                                   virt_get_oem_id,
3002                                   virt_set_oem_id);
3003     object_class_property_set_description(oc, "x-oem-id",
3004                                           "Override the default value of field OEMID "
3005                                           "in ACPI table header."
3006                                           "The string may be up to 6 bytes in size");
3007 
3008 
3009     object_class_property_add_str(oc, "x-oem-table-id",
3010                                   virt_get_oem_table_id,
3011                                   virt_set_oem_table_id);
3012     object_class_property_set_description(oc, "x-oem-table-id",
3013                                           "Override the default value of field OEM Table ID "
3014                                           "in ACPI table header."
3015                                           "The string may be up to 8 bytes in size");
3016 
3017 }
3018 
3019 static void virt_instance_init(Object *obj)
3020 {
3021     VirtMachineState *vms = VIRT_MACHINE(obj);
3022     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
3023 
3024     /* EL3 is disabled by default on virt: this makes us consistent
3025      * between KVM and TCG for this board, and it also allows us to
3026      * boot UEFI blobs which assume no TrustZone support.
3027      */
3028     vms->secure = false;
3029 
3030     /* EL2 is also disabled by default, for similar reasons */
3031     vms->virt = false;
3032 
3033     /* High memory is enabled by default */
3034     vms->highmem = true;
3035     vms->gic_version = VIRT_GIC_VERSION_NOSEL;
3036 
3037     vms->highmem_ecam = !vmc->no_highmem_ecam;
3038     vms->highmem_mmio = true;
3039     vms->highmem_redists = true;
3040 
3041     if (vmc->no_its) {
3042         vms->its = false;
3043     } else {
3044         /* Default allows ITS instantiation */
3045         vms->its = true;
3046 
3047         if (vmc->no_tcg_its) {
3048             vms->tcg_its = false;
3049         } else {
3050             vms->tcg_its = true;
3051         }
3052     }
3053 
3054     /* Default disallows iommu instantiation */
3055     vms->iommu = VIRT_IOMMU_NONE;
3056 
3057     /* The default root bus is attached to iommu by default */
3058     vms->default_bus_bypass_iommu = false;
3059 
3060     /* Default disallows RAS instantiation */
3061     vms->ras = false;
3062 
3063     /* MTE is disabled by default.  */
3064     vms->mte = false;
3065 
3066     /* Supply a kaslr-seed by default */
3067     vms->dtb_kaslr_seed = true;
3068 
3069     vms->irqmap = a15irqmap;
3070 
3071     virt_flash_create(vms);
3072 
3073     vms->oem_id = g_strndup(ACPI_BUILD_APPNAME6, 6);
3074     vms->oem_table_id = g_strndup(ACPI_BUILD_APPNAME8, 8);
3075 }
3076 
3077 static const TypeInfo virt_machine_info = {
3078     .name          = TYPE_VIRT_MACHINE,
3079     .parent        = TYPE_MACHINE,
3080     .abstract      = true,
3081     .instance_size = sizeof(VirtMachineState),
3082     .class_size    = sizeof(VirtMachineClass),
3083     .class_init    = virt_machine_class_init,
3084     .instance_init = virt_instance_init,
3085     .interfaces = (InterfaceInfo[]) {
3086          { TYPE_HOTPLUG_HANDLER },
3087          { }
3088     },
3089 };
3090 
3091 static void machvirt_machine_init(void)
3092 {
3093     type_register_static(&virt_machine_info);
3094 }
3095 type_init(machvirt_machine_init);
3096 
3097 static void virt_machine_7_1_options(MachineClass *mc)
3098 {
3099 }
3100 DEFINE_VIRT_MACHINE_AS_LATEST(7, 1)
3101 
3102 static void virt_machine_7_0_options(MachineClass *mc)
3103 {
3104     virt_machine_7_1_options(mc);
3105     compat_props_add(mc->compat_props, hw_compat_7_0, hw_compat_7_0_len);
3106 }
3107 DEFINE_VIRT_MACHINE(7, 0)
3108 
3109 static void virt_machine_6_2_options(MachineClass *mc)
3110 {
3111     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3112 
3113     virt_machine_7_0_options(mc);
3114     compat_props_add(mc->compat_props, hw_compat_6_2, hw_compat_6_2_len);
3115     vmc->no_tcg_lpa2 = true;
3116 }
3117 DEFINE_VIRT_MACHINE(6, 2)
3118 
3119 static void virt_machine_6_1_options(MachineClass *mc)
3120 {
3121     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3122 
3123     virt_machine_6_2_options(mc);
3124     compat_props_add(mc->compat_props, hw_compat_6_1, hw_compat_6_1_len);
3125     mc->smp_props.prefer_sockets = true;
3126     vmc->no_cpu_topology = true;
3127 
3128     /* qemu ITS was introduced with 6.2 */
3129     vmc->no_tcg_its = true;
3130 }
3131 DEFINE_VIRT_MACHINE(6, 1)
3132 
3133 static void virt_machine_6_0_options(MachineClass *mc)
3134 {
3135     virt_machine_6_1_options(mc);
3136     compat_props_add(mc->compat_props, hw_compat_6_0, hw_compat_6_0_len);
3137 }
3138 DEFINE_VIRT_MACHINE(6, 0)
3139 
3140 static void virt_machine_5_2_options(MachineClass *mc)
3141 {
3142     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3143 
3144     virt_machine_6_0_options(mc);
3145     compat_props_add(mc->compat_props, hw_compat_5_2, hw_compat_5_2_len);
3146     vmc->no_secure_gpio = true;
3147 }
3148 DEFINE_VIRT_MACHINE(5, 2)
3149 
3150 static void virt_machine_5_1_options(MachineClass *mc)
3151 {
3152     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3153 
3154     virt_machine_5_2_options(mc);
3155     compat_props_add(mc->compat_props, hw_compat_5_1, hw_compat_5_1_len);
3156     vmc->no_kvm_steal_time = true;
3157 }
3158 DEFINE_VIRT_MACHINE(5, 1)
3159 
3160 static void virt_machine_5_0_options(MachineClass *mc)
3161 {
3162     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3163 
3164     virt_machine_5_1_options(mc);
3165     compat_props_add(mc->compat_props, hw_compat_5_0, hw_compat_5_0_len);
3166     mc->numa_mem_supported = true;
3167     vmc->acpi_expose_flash = true;
3168     mc->auto_enable_numa_with_memdev = false;
3169 }
3170 DEFINE_VIRT_MACHINE(5, 0)
3171 
3172 static void virt_machine_4_2_options(MachineClass *mc)
3173 {
3174     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3175 
3176     virt_machine_5_0_options(mc);
3177     compat_props_add(mc->compat_props, hw_compat_4_2, hw_compat_4_2_len);
3178     vmc->kvm_no_adjvtime = true;
3179 }
3180 DEFINE_VIRT_MACHINE(4, 2)
3181 
3182 static void virt_machine_4_1_options(MachineClass *mc)
3183 {
3184     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3185 
3186     virt_machine_4_2_options(mc);
3187     compat_props_add(mc->compat_props, hw_compat_4_1, hw_compat_4_1_len);
3188     vmc->no_ged = true;
3189     mc->auto_enable_numa_with_memhp = false;
3190 }
3191 DEFINE_VIRT_MACHINE(4, 1)
3192 
3193 static void virt_machine_4_0_options(MachineClass *mc)
3194 {
3195     virt_machine_4_1_options(mc);
3196     compat_props_add(mc->compat_props, hw_compat_4_0, hw_compat_4_0_len);
3197 }
3198 DEFINE_VIRT_MACHINE(4, 0)
3199 
3200 static void virt_machine_3_1_options(MachineClass *mc)
3201 {
3202     virt_machine_4_0_options(mc);
3203     compat_props_add(mc->compat_props, hw_compat_3_1, hw_compat_3_1_len);
3204 }
3205 DEFINE_VIRT_MACHINE(3, 1)
3206 
3207 static void virt_machine_3_0_options(MachineClass *mc)
3208 {
3209     virt_machine_3_1_options(mc);
3210     compat_props_add(mc->compat_props, hw_compat_3_0, hw_compat_3_0_len);
3211 }
3212 DEFINE_VIRT_MACHINE(3, 0)
3213 
3214 static void virt_machine_2_12_options(MachineClass *mc)
3215 {
3216     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3217 
3218     virt_machine_3_0_options(mc);
3219     compat_props_add(mc->compat_props, hw_compat_2_12, hw_compat_2_12_len);
3220     vmc->no_highmem_ecam = true;
3221     mc->max_cpus = 255;
3222 }
3223 DEFINE_VIRT_MACHINE(2, 12)
3224 
3225 static void virt_machine_2_11_options(MachineClass *mc)
3226 {
3227     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3228 
3229     virt_machine_2_12_options(mc);
3230     compat_props_add(mc->compat_props, hw_compat_2_11, hw_compat_2_11_len);
3231     vmc->smbios_old_sys_ver = true;
3232 }
3233 DEFINE_VIRT_MACHINE(2, 11)
3234 
3235 static void virt_machine_2_10_options(MachineClass *mc)
3236 {
3237     virt_machine_2_11_options(mc);
3238     compat_props_add(mc->compat_props, hw_compat_2_10, hw_compat_2_10_len);
3239     /* before 2.11 we never faulted accesses to bad addresses */
3240     mc->ignore_memory_transaction_failures = true;
3241 }
3242 DEFINE_VIRT_MACHINE(2, 10)
3243 
3244 static void virt_machine_2_9_options(MachineClass *mc)
3245 {
3246     virt_machine_2_10_options(mc);
3247     compat_props_add(mc->compat_props, hw_compat_2_9, hw_compat_2_9_len);
3248 }
3249 DEFINE_VIRT_MACHINE(2, 9)
3250 
3251 static void virt_machine_2_8_options(MachineClass *mc)
3252 {
3253     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3254 
3255     virt_machine_2_9_options(mc);
3256     compat_props_add(mc->compat_props, hw_compat_2_8, hw_compat_2_8_len);
3257     /* For 2.8 and earlier we falsely claimed in the DT that
3258      * our timers were edge-triggered, not level-triggered.
3259      */
3260     vmc->claim_edge_triggered_timers = true;
3261 }
3262 DEFINE_VIRT_MACHINE(2, 8)
3263 
3264 static void virt_machine_2_7_options(MachineClass *mc)
3265 {
3266     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3267 
3268     virt_machine_2_8_options(mc);
3269     compat_props_add(mc->compat_props, hw_compat_2_7, hw_compat_2_7_len);
3270     /* ITS was introduced with 2.8 */
3271     vmc->no_its = true;
3272     /* Stick with 1K pages for migration compatibility */
3273     mc->minimum_page_bits = 0;
3274 }
3275 DEFINE_VIRT_MACHINE(2, 7)
3276 
3277 static void virt_machine_2_6_options(MachineClass *mc)
3278 {
3279     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3280 
3281     virt_machine_2_7_options(mc);
3282     compat_props_add(mc->compat_props, hw_compat_2_6, hw_compat_2_6_len);
3283     vmc->disallow_affinity_adjustment = true;
3284     /* Disable PMU for 2.6 as PMU support was first introduced in 2.7 */
3285     vmc->no_pmu = true;
3286 }
3287 DEFINE_VIRT_MACHINE(2, 6)
3288