xref: /openbmc/qemu/hw/arm/virt.c (revision 954a6c4f)
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-a35"),
203     ARM_CPU_TYPE_NAME("cortex-a53"),
204     ARM_CPU_TYPE_NAME("cortex-a57"),
205     ARM_CPU_TYPE_NAME("cortex-a72"),
206     ARM_CPU_TYPE_NAME("cortex-a76"),
207     ARM_CPU_TYPE_NAME("a64fx"),
208     ARM_CPU_TYPE_NAME("neoverse-n1"),
209     ARM_CPU_TYPE_NAME("host"),
210     ARM_CPU_TYPE_NAME("max"),
211 };
212 
213 static bool cpu_type_valid(const char *cpu)
214 {
215     int i;
216 
217     for (i = 0; i < ARRAY_SIZE(valid_cpus); i++) {
218         if (strcmp(cpu, valid_cpus[i]) == 0) {
219             return true;
220         }
221     }
222     return false;
223 }
224 
225 static void create_randomness(MachineState *ms, const char *node)
226 {
227     struct {
228         uint64_t kaslr;
229         uint8_t rng[32];
230     } seed;
231 
232     if (qemu_guest_getrandom(&seed, sizeof(seed), NULL)) {
233         return;
234     }
235     qemu_fdt_setprop_u64(ms->fdt, node, "kaslr-seed", seed.kaslr);
236     qemu_fdt_setprop(ms->fdt, node, "rng-seed", seed.rng, sizeof(seed.rng));
237 }
238 
239 static void create_fdt(VirtMachineState *vms)
240 {
241     MachineState *ms = MACHINE(vms);
242     int nb_numa_nodes = ms->numa_state->num_nodes;
243     void *fdt = create_device_tree(&vms->fdt_size);
244 
245     if (!fdt) {
246         error_report("create_device_tree() failed");
247         exit(1);
248     }
249 
250     ms->fdt = fdt;
251 
252     /* Header */
253     qemu_fdt_setprop_string(fdt, "/", "compatible", "linux,dummy-virt");
254     qemu_fdt_setprop_cell(fdt, "/", "#address-cells", 0x2);
255     qemu_fdt_setprop_cell(fdt, "/", "#size-cells", 0x2);
256     qemu_fdt_setprop_string(fdt, "/", "model", "linux,dummy-virt");
257 
258     /* /chosen must exist for load_dtb to fill in necessary properties later */
259     qemu_fdt_add_subnode(fdt, "/chosen");
260     if (vms->dtb_randomness) {
261         create_randomness(ms, "/chosen");
262     }
263 
264     if (vms->secure) {
265         qemu_fdt_add_subnode(fdt, "/secure-chosen");
266         if (vms->dtb_randomness) {
267             create_randomness(ms, "/secure-chosen");
268         }
269     }
270 
271     /* Clock node, for the benefit of the UART. The kernel device tree
272      * binding documentation claims the PL011 node clock properties are
273      * optional but in practice if you omit them the kernel refuses to
274      * probe for the device.
275      */
276     vms->clock_phandle = qemu_fdt_alloc_phandle(fdt);
277     qemu_fdt_add_subnode(fdt, "/apb-pclk");
278     qemu_fdt_setprop_string(fdt, "/apb-pclk", "compatible", "fixed-clock");
279     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "#clock-cells", 0x0);
280     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "clock-frequency", 24000000);
281     qemu_fdt_setprop_string(fdt, "/apb-pclk", "clock-output-names",
282                                 "clk24mhz");
283     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "phandle", vms->clock_phandle);
284 
285     if (nb_numa_nodes > 0 && ms->numa_state->have_numa_distance) {
286         int size = nb_numa_nodes * nb_numa_nodes * 3 * sizeof(uint32_t);
287         uint32_t *matrix = g_malloc0(size);
288         int idx, i, j;
289 
290         for (i = 0; i < nb_numa_nodes; i++) {
291             for (j = 0; j < nb_numa_nodes; j++) {
292                 idx = (i * nb_numa_nodes + j) * 3;
293                 matrix[idx + 0] = cpu_to_be32(i);
294                 matrix[idx + 1] = cpu_to_be32(j);
295                 matrix[idx + 2] =
296                     cpu_to_be32(ms->numa_state->nodes[i].distance[j]);
297             }
298         }
299 
300         qemu_fdt_add_subnode(fdt, "/distance-map");
301         qemu_fdt_setprop_string(fdt, "/distance-map", "compatible",
302                                 "numa-distance-map-v1");
303         qemu_fdt_setprop(fdt, "/distance-map", "distance-matrix",
304                          matrix, size);
305         g_free(matrix);
306     }
307 }
308 
309 static void fdt_add_timer_nodes(const VirtMachineState *vms)
310 {
311     /* On real hardware these interrupts are level-triggered.
312      * On KVM they were edge-triggered before host kernel version 4.4,
313      * and level-triggered afterwards.
314      * On emulated QEMU they are level-triggered.
315      *
316      * Getting the DTB info about them wrong is awkward for some
317      * guest kernels:
318      *  pre-4.8 ignore the DT and leave the interrupt configured
319      *   with whatever the GIC reset value (or the bootloader) left it at
320      *  4.8 before rc6 honour the incorrect data by programming it back
321      *   into the GIC, causing problems
322      *  4.8rc6 and later ignore the DT and always write "level triggered"
323      *   into the GIC
324      *
325      * For backwards-compatibility, virt-2.8 and earlier will continue
326      * to say these are edge-triggered, but later machines will report
327      * the correct information.
328      */
329     ARMCPU *armcpu;
330     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
331     uint32_t irqflags = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
332     MachineState *ms = MACHINE(vms);
333 
334     if (vmc->claim_edge_triggered_timers) {
335         irqflags = GIC_FDT_IRQ_FLAGS_EDGE_LO_HI;
336     }
337 
338     if (vms->gic_version == VIRT_GIC_VERSION_2) {
339         irqflags = deposit32(irqflags, GIC_FDT_IRQ_PPI_CPU_START,
340                              GIC_FDT_IRQ_PPI_CPU_WIDTH,
341                              (1 << MACHINE(vms)->smp.cpus) - 1);
342     }
343 
344     qemu_fdt_add_subnode(ms->fdt, "/timer");
345 
346     armcpu = ARM_CPU(qemu_get_cpu(0));
347     if (arm_feature(&armcpu->env, ARM_FEATURE_V8)) {
348         const char compat[] = "arm,armv8-timer\0arm,armv7-timer";
349         qemu_fdt_setprop(ms->fdt, "/timer", "compatible",
350                          compat, sizeof(compat));
351     } else {
352         qemu_fdt_setprop_string(ms->fdt, "/timer", "compatible",
353                                 "arm,armv7-timer");
354     }
355     qemu_fdt_setprop(ms->fdt, "/timer", "always-on", NULL, 0);
356     qemu_fdt_setprop_cells(ms->fdt, "/timer", "interrupts",
357                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_S_EL1_IRQ, irqflags,
358                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_NS_EL1_IRQ, irqflags,
359                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_VIRT_IRQ, irqflags,
360                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_NS_EL2_IRQ, irqflags);
361 }
362 
363 static void fdt_add_cpu_nodes(const VirtMachineState *vms)
364 {
365     int cpu;
366     int addr_cells = 1;
367     const MachineState *ms = MACHINE(vms);
368     const VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
369     int smp_cpus = ms->smp.cpus;
370 
371     /*
372      * See Linux Documentation/devicetree/bindings/arm/cpus.yaml
373      * On ARM v8 64-bit systems value should be set to 2,
374      * that corresponds to the MPIDR_EL1 register size.
375      * If MPIDR_EL1[63:32] value is equal to 0 on all CPUs
376      * in the system, #address-cells can be set to 1, since
377      * MPIDR_EL1[63:32] bits are not used for CPUs
378      * identification.
379      *
380      * Here we actually don't know whether our system is 32- or 64-bit one.
381      * The simplest way to go is to examine affinity IDs of all our CPUs. If
382      * at least one of them has Aff3 populated, we set #address-cells to 2.
383      */
384     for (cpu = 0; cpu < smp_cpus; cpu++) {
385         ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(cpu));
386 
387         if (armcpu->mp_affinity & ARM_AFF3_MASK) {
388             addr_cells = 2;
389             break;
390         }
391     }
392 
393     qemu_fdt_add_subnode(ms->fdt, "/cpus");
394     qemu_fdt_setprop_cell(ms->fdt, "/cpus", "#address-cells", addr_cells);
395     qemu_fdt_setprop_cell(ms->fdt, "/cpus", "#size-cells", 0x0);
396 
397     for (cpu = smp_cpus - 1; cpu >= 0; cpu--) {
398         char *nodename = g_strdup_printf("/cpus/cpu@%d", cpu);
399         ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(cpu));
400         CPUState *cs = CPU(armcpu);
401 
402         qemu_fdt_add_subnode(ms->fdt, nodename);
403         qemu_fdt_setprop_string(ms->fdt, nodename, "device_type", "cpu");
404         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
405                                     armcpu->dtb_compatible);
406 
407         if (vms->psci_conduit != QEMU_PSCI_CONDUIT_DISABLED && smp_cpus > 1) {
408             qemu_fdt_setprop_string(ms->fdt, nodename,
409                                         "enable-method", "psci");
410         }
411 
412         if (addr_cells == 2) {
413             qemu_fdt_setprop_u64(ms->fdt, nodename, "reg",
414                                  armcpu->mp_affinity);
415         } else {
416             qemu_fdt_setprop_cell(ms->fdt, nodename, "reg",
417                                   armcpu->mp_affinity);
418         }
419 
420         if (ms->possible_cpus->cpus[cs->cpu_index].props.has_node_id) {
421             qemu_fdt_setprop_cell(ms->fdt, nodename, "numa-node-id",
422                 ms->possible_cpus->cpus[cs->cpu_index].props.node_id);
423         }
424 
425         if (!vmc->no_cpu_topology) {
426             qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle",
427                                   qemu_fdt_alloc_phandle(ms->fdt));
428         }
429 
430         g_free(nodename);
431     }
432 
433     if (!vmc->no_cpu_topology) {
434         /*
435          * Add vCPU topology description through fdt node cpu-map.
436          *
437          * See Linux Documentation/devicetree/bindings/cpu/cpu-topology.txt
438          * In a SMP system, the hierarchy of CPUs can be defined through
439          * four entities that are used to describe the layout of CPUs in
440          * the system: socket/cluster/core/thread.
441          *
442          * A socket node represents the boundary of system physical package
443          * and its child nodes must be one or more cluster nodes. A system
444          * can contain several layers of clustering within a single physical
445          * package and cluster nodes can be contained in parent cluster nodes.
446          *
447          * Note: currently we only support one layer of clustering within
448          * each physical package.
449          */
450         qemu_fdt_add_subnode(ms->fdt, "/cpus/cpu-map");
451 
452         for (cpu = smp_cpus - 1; cpu >= 0; cpu--) {
453             char *cpu_path = g_strdup_printf("/cpus/cpu@%d", cpu);
454             char *map_path;
455 
456             if (ms->smp.threads > 1) {
457                 map_path = g_strdup_printf(
458                     "/cpus/cpu-map/socket%d/cluster%d/core%d/thread%d",
459                     cpu / (ms->smp.clusters * ms->smp.cores * ms->smp.threads),
460                     (cpu / (ms->smp.cores * ms->smp.threads)) % ms->smp.clusters,
461                     (cpu / ms->smp.threads) % ms->smp.cores,
462                     cpu % ms->smp.threads);
463             } else {
464                 map_path = g_strdup_printf(
465                     "/cpus/cpu-map/socket%d/cluster%d/core%d",
466                     cpu / (ms->smp.clusters * ms->smp.cores),
467                     (cpu / ms->smp.cores) % ms->smp.clusters,
468                     cpu % ms->smp.cores);
469             }
470             qemu_fdt_add_path(ms->fdt, map_path);
471             qemu_fdt_setprop_phandle(ms->fdt, map_path, "cpu", cpu_path);
472 
473             g_free(map_path);
474             g_free(cpu_path);
475         }
476     }
477 }
478 
479 static void fdt_add_its_gic_node(VirtMachineState *vms)
480 {
481     char *nodename;
482     MachineState *ms = MACHINE(vms);
483 
484     vms->msi_phandle = qemu_fdt_alloc_phandle(ms->fdt);
485     nodename = g_strdup_printf("/intc/its@%" PRIx64,
486                                vms->memmap[VIRT_GIC_ITS].base);
487     qemu_fdt_add_subnode(ms->fdt, nodename);
488     qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
489                             "arm,gic-v3-its");
490     qemu_fdt_setprop(ms->fdt, nodename, "msi-controller", NULL, 0);
491     qemu_fdt_setprop_cell(ms->fdt, nodename, "#msi-cells", 1);
492     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
493                                  2, vms->memmap[VIRT_GIC_ITS].base,
494                                  2, vms->memmap[VIRT_GIC_ITS].size);
495     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", vms->msi_phandle);
496     g_free(nodename);
497 }
498 
499 static void fdt_add_v2m_gic_node(VirtMachineState *vms)
500 {
501     MachineState *ms = MACHINE(vms);
502     char *nodename;
503 
504     nodename = g_strdup_printf("/intc/v2m@%" PRIx64,
505                                vms->memmap[VIRT_GIC_V2M].base);
506     vms->msi_phandle = qemu_fdt_alloc_phandle(ms->fdt);
507     qemu_fdt_add_subnode(ms->fdt, nodename);
508     qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
509                             "arm,gic-v2m-frame");
510     qemu_fdt_setprop(ms->fdt, nodename, "msi-controller", NULL, 0);
511     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
512                                  2, vms->memmap[VIRT_GIC_V2M].base,
513                                  2, vms->memmap[VIRT_GIC_V2M].size);
514     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", vms->msi_phandle);
515     g_free(nodename);
516 }
517 
518 static void fdt_add_gic_node(VirtMachineState *vms)
519 {
520     MachineState *ms = MACHINE(vms);
521     char *nodename;
522 
523     vms->gic_phandle = qemu_fdt_alloc_phandle(ms->fdt);
524     qemu_fdt_setprop_cell(ms->fdt, "/", "interrupt-parent", vms->gic_phandle);
525 
526     nodename = g_strdup_printf("/intc@%" PRIx64,
527                                vms->memmap[VIRT_GIC_DIST].base);
528     qemu_fdt_add_subnode(ms->fdt, nodename);
529     qemu_fdt_setprop_cell(ms->fdt, nodename, "#interrupt-cells", 3);
530     qemu_fdt_setprop(ms->fdt, nodename, "interrupt-controller", NULL, 0);
531     qemu_fdt_setprop_cell(ms->fdt, nodename, "#address-cells", 0x2);
532     qemu_fdt_setprop_cell(ms->fdt, nodename, "#size-cells", 0x2);
533     qemu_fdt_setprop(ms->fdt, nodename, "ranges", NULL, 0);
534     if (vms->gic_version != VIRT_GIC_VERSION_2) {
535         int nb_redist_regions = virt_gicv3_redist_region_count(vms);
536 
537         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
538                                 "arm,gic-v3");
539 
540         qemu_fdt_setprop_cell(ms->fdt, nodename,
541                               "#redistributor-regions", nb_redist_regions);
542 
543         if (nb_redist_regions == 1) {
544             qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
545                                          2, vms->memmap[VIRT_GIC_DIST].base,
546                                          2, vms->memmap[VIRT_GIC_DIST].size,
547                                          2, vms->memmap[VIRT_GIC_REDIST].base,
548                                          2, vms->memmap[VIRT_GIC_REDIST].size);
549         } else {
550             qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
551                                  2, vms->memmap[VIRT_GIC_DIST].base,
552                                  2, vms->memmap[VIRT_GIC_DIST].size,
553                                  2, vms->memmap[VIRT_GIC_REDIST].base,
554                                  2, vms->memmap[VIRT_GIC_REDIST].size,
555                                  2, vms->memmap[VIRT_HIGH_GIC_REDIST2].base,
556                                  2, vms->memmap[VIRT_HIGH_GIC_REDIST2].size);
557         }
558 
559         if (vms->virt) {
560             qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
561                                    GIC_FDT_IRQ_TYPE_PPI, ARCH_GIC_MAINT_IRQ,
562                                    GIC_FDT_IRQ_FLAGS_LEVEL_HI);
563         }
564     } else {
565         /* 'cortex-a15-gic' means 'GIC v2' */
566         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
567                                 "arm,cortex-a15-gic");
568         if (!vms->virt) {
569             qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
570                                          2, vms->memmap[VIRT_GIC_DIST].base,
571                                          2, vms->memmap[VIRT_GIC_DIST].size,
572                                          2, vms->memmap[VIRT_GIC_CPU].base,
573                                          2, vms->memmap[VIRT_GIC_CPU].size);
574         } else {
575             qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
576                                          2, vms->memmap[VIRT_GIC_DIST].base,
577                                          2, vms->memmap[VIRT_GIC_DIST].size,
578                                          2, vms->memmap[VIRT_GIC_CPU].base,
579                                          2, vms->memmap[VIRT_GIC_CPU].size,
580                                          2, vms->memmap[VIRT_GIC_HYP].base,
581                                          2, vms->memmap[VIRT_GIC_HYP].size,
582                                          2, vms->memmap[VIRT_GIC_VCPU].base,
583                                          2, vms->memmap[VIRT_GIC_VCPU].size);
584             qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
585                                    GIC_FDT_IRQ_TYPE_PPI, ARCH_GIC_MAINT_IRQ,
586                                    GIC_FDT_IRQ_FLAGS_LEVEL_HI);
587         }
588     }
589 
590     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", vms->gic_phandle);
591     g_free(nodename);
592 }
593 
594 static void fdt_add_pmu_nodes(const VirtMachineState *vms)
595 {
596     ARMCPU *armcpu = ARM_CPU(first_cpu);
597     uint32_t irqflags = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
598     MachineState *ms = MACHINE(vms);
599 
600     if (!arm_feature(&armcpu->env, ARM_FEATURE_PMU)) {
601         assert(!object_property_get_bool(OBJECT(armcpu), "pmu", NULL));
602         return;
603     }
604 
605     if (vms->gic_version == VIRT_GIC_VERSION_2) {
606         irqflags = deposit32(irqflags, GIC_FDT_IRQ_PPI_CPU_START,
607                              GIC_FDT_IRQ_PPI_CPU_WIDTH,
608                              (1 << MACHINE(vms)->smp.cpus) - 1);
609     }
610 
611     qemu_fdt_add_subnode(ms->fdt, "/pmu");
612     if (arm_feature(&armcpu->env, ARM_FEATURE_V8)) {
613         const char compat[] = "arm,armv8-pmuv3";
614         qemu_fdt_setprop(ms->fdt, "/pmu", "compatible",
615                          compat, sizeof(compat));
616         qemu_fdt_setprop_cells(ms->fdt, "/pmu", "interrupts",
617                                GIC_FDT_IRQ_TYPE_PPI, VIRTUAL_PMU_IRQ, irqflags);
618     }
619 }
620 
621 static inline DeviceState *create_acpi_ged(VirtMachineState *vms)
622 {
623     DeviceState *dev;
624     MachineState *ms = MACHINE(vms);
625     int irq = vms->irqmap[VIRT_ACPI_GED];
626     uint32_t event = ACPI_GED_PWR_DOWN_EVT;
627 
628     if (ms->ram_slots) {
629         event |= ACPI_GED_MEM_HOTPLUG_EVT;
630     }
631 
632     if (ms->nvdimms_state->is_enabled) {
633         event |= ACPI_GED_NVDIMM_HOTPLUG_EVT;
634     }
635 
636     dev = qdev_new(TYPE_ACPI_GED);
637     qdev_prop_set_uint32(dev, "ged-event", event);
638 
639     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_ACPI_GED].base);
640     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 1, vms->memmap[VIRT_PCDIMM_ACPI].base);
641     sysbus_connect_irq(SYS_BUS_DEVICE(dev), 0, qdev_get_gpio_in(vms->gic, irq));
642 
643     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
644 
645     return dev;
646 }
647 
648 static void create_its(VirtMachineState *vms)
649 {
650     const char *itsclass = its_class_name();
651     DeviceState *dev;
652 
653     if (!strcmp(itsclass, "arm-gicv3-its")) {
654         if (!vms->tcg_its) {
655             itsclass = NULL;
656         }
657     }
658 
659     if (!itsclass) {
660         /* Do nothing if not supported */
661         return;
662     }
663 
664     dev = qdev_new(itsclass);
665 
666     object_property_set_link(OBJECT(dev), "parent-gicv3", OBJECT(vms->gic),
667                              &error_abort);
668     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
669     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_GIC_ITS].base);
670 
671     fdt_add_its_gic_node(vms);
672     vms->msi_controller = VIRT_MSI_CTRL_ITS;
673 }
674 
675 static void create_v2m(VirtMachineState *vms)
676 {
677     int i;
678     int irq = vms->irqmap[VIRT_GIC_V2M];
679     DeviceState *dev;
680 
681     dev = qdev_new("arm-gicv2m");
682     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_GIC_V2M].base);
683     qdev_prop_set_uint32(dev, "base-spi", irq);
684     qdev_prop_set_uint32(dev, "num-spi", NUM_GICV2M_SPIS);
685     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
686 
687     for (i = 0; i < NUM_GICV2M_SPIS; i++) {
688         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
689                            qdev_get_gpio_in(vms->gic, irq + i));
690     }
691 
692     fdt_add_v2m_gic_node(vms);
693     vms->msi_controller = VIRT_MSI_CTRL_GICV2M;
694 }
695 
696 static void create_gic(VirtMachineState *vms, MemoryRegion *mem)
697 {
698     MachineState *ms = MACHINE(vms);
699     /* We create a standalone GIC */
700     SysBusDevice *gicbusdev;
701     const char *gictype;
702     int i;
703     unsigned int smp_cpus = ms->smp.cpus;
704     uint32_t nb_redist_regions = 0;
705     int revision;
706 
707     if (vms->gic_version == VIRT_GIC_VERSION_2) {
708         gictype = gic_class_name();
709     } else {
710         gictype = gicv3_class_name();
711     }
712 
713     switch (vms->gic_version) {
714     case VIRT_GIC_VERSION_2:
715         revision = 2;
716         break;
717     case VIRT_GIC_VERSION_3:
718         revision = 3;
719         break;
720     case VIRT_GIC_VERSION_4:
721         revision = 4;
722         break;
723     default:
724         g_assert_not_reached();
725     }
726     vms->gic = qdev_new(gictype);
727     qdev_prop_set_uint32(vms->gic, "revision", revision);
728     qdev_prop_set_uint32(vms->gic, "num-cpu", smp_cpus);
729     /* Note that the num-irq property counts both internal and external
730      * interrupts; there are always 32 of the former (mandated by GIC spec).
731      */
732     qdev_prop_set_uint32(vms->gic, "num-irq", NUM_IRQS + 32);
733     if (!kvm_irqchip_in_kernel()) {
734         qdev_prop_set_bit(vms->gic, "has-security-extensions", vms->secure);
735     }
736 
737     if (vms->gic_version != VIRT_GIC_VERSION_2) {
738         uint32_t redist0_capacity = virt_redist_capacity(vms, VIRT_GIC_REDIST);
739         uint32_t redist0_count = MIN(smp_cpus, redist0_capacity);
740 
741         nb_redist_regions = virt_gicv3_redist_region_count(vms);
742 
743         qdev_prop_set_uint32(vms->gic, "len-redist-region-count",
744                              nb_redist_regions);
745         qdev_prop_set_uint32(vms->gic, "redist-region-count[0]", redist0_count);
746 
747         if (!kvm_irqchip_in_kernel()) {
748             if (vms->tcg_its) {
749                 object_property_set_link(OBJECT(vms->gic), "sysmem",
750                                          OBJECT(mem), &error_fatal);
751                 qdev_prop_set_bit(vms->gic, "has-lpi", true);
752             }
753         }
754 
755         if (nb_redist_regions == 2) {
756             uint32_t redist1_capacity =
757                 virt_redist_capacity(vms, VIRT_HIGH_GIC_REDIST2);
758 
759             qdev_prop_set_uint32(vms->gic, "redist-region-count[1]",
760                 MIN(smp_cpus - redist0_count, redist1_capacity));
761         }
762     } else {
763         if (!kvm_irqchip_in_kernel()) {
764             qdev_prop_set_bit(vms->gic, "has-virtualization-extensions",
765                               vms->virt);
766         }
767     }
768     gicbusdev = SYS_BUS_DEVICE(vms->gic);
769     sysbus_realize_and_unref(gicbusdev, &error_fatal);
770     sysbus_mmio_map(gicbusdev, 0, vms->memmap[VIRT_GIC_DIST].base);
771     if (vms->gic_version != VIRT_GIC_VERSION_2) {
772         sysbus_mmio_map(gicbusdev, 1, vms->memmap[VIRT_GIC_REDIST].base);
773         if (nb_redist_regions == 2) {
774             sysbus_mmio_map(gicbusdev, 2,
775                             vms->memmap[VIRT_HIGH_GIC_REDIST2].base);
776         }
777     } else {
778         sysbus_mmio_map(gicbusdev, 1, vms->memmap[VIRT_GIC_CPU].base);
779         if (vms->virt) {
780             sysbus_mmio_map(gicbusdev, 2, vms->memmap[VIRT_GIC_HYP].base);
781             sysbus_mmio_map(gicbusdev, 3, vms->memmap[VIRT_GIC_VCPU].base);
782         }
783     }
784 
785     /* Wire the outputs from each CPU's generic timer and the GICv3
786      * maintenance interrupt signal to the appropriate GIC PPI inputs,
787      * and the GIC's IRQ/FIQ/VIRQ/VFIQ interrupt outputs to the CPU's inputs.
788      */
789     for (i = 0; i < smp_cpus; i++) {
790         DeviceState *cpudev = DEVICE(qemu_get_cpu(i));
791         int ppibase = NUM_IRQS + i * GIC_INTERNAL + GIC_NR_SGIS;
792         int irq;
793         /* Mapping from the output timer irq lines from the CPU to the
794          * GIC PPI inputs we use for the virt board.
795          */
796         const int timer_irq[] = {
797             [GTIMER_PHYS] = ARCH_TIMER_NS_EL1_IRQ,
798             [GTIMER_VIRT] = ARCH_TIMER_VIRT_IRQ,
799             [GTIMER_HYP]  = ARCH_TIMER_NS_EL2_IRQ,
800             [GTIMER_SEC]  = ARCH_TIMER_S_EL1_IRQ,
801         };
802 
803         for (irq = 0; irq < ARRAY_SIZE(timer_irq); irq++) {
804             qdev_connect_gpio_out(cpudev, irq,
805                                   qdev_get_gpio_in(vms->gic,
806                                                    ppibase + timer_irq[irq]));
807         }
808 
809         if (vms->gic_version != VIRT_GIC_VERSION_2) {
810             qemu_irq irq = qdev_get_gpio_in(vms->gic,
811                                             ppibase + ARCH_GIC_MAINT_IRQ);
812             qdev_connect_gpio_out_named(cpudev, "gicv3-maintenance-interrupt",
813                                         0, irq);
814         } else if (vms->virt) {
815             qemu_irq irq = qdev_get_gpio_in(vms->gic,
816                                             ppibase + ARCH_GIC_MAINT_IRQ);
817             sysbus_connect_irq(gicbusdev, i + 4 * smp_cpus, irq);
818         }
819 
820         qdev_connect_gpio_out_named(cpudev, "pmu-interrupt", 0,
821                                     qdev_get_gpio_in(vms->gic, ppibase
822                                                      + VIRTUAL_PMU_IRQ));
823 
824         sysbus_connect_irq(gicbusdev, i, qdev_get_gpio_in(cpudev, ARM_CPU_IRQ));
825         sysbus_connect_irq(gicbusdev, i + smp_cpus,
826                            qdev_get_gpio_in(cpudev, ARM_CPU_FIQ));
827         sysbus_connect_irq(gicbusdev, i + 2 * smp_cpus,
828                            qdev_get_gpio_in(cpudev, ARM_CPU_VIRQ));
829         sysbus_connect_irq(gicbusdev, i + 3 * smp_cpus,
830                            qdev_get_gpio_in(cpudev, ARM_CPU_VFIQ));
831     }
832 
833     fdt_add_gic_node(vms);
834 
835     if (vms->gic_version != VIRT_GIC_VERSION_2 && vms->its) {
836         create_its(vms);
837     } else if (vms->gic_version == VIRT_GIC_VERSION_2) {
838         create_v2m(vms);
839     }
840 }
841 
842 static void create_uart(const VirtMachineState *vms, int uart,
843                         MemoryRegion *mem, Chardev *chr)
844 {
845     char *nodename;
846     hwaddr base = vms->memmap[uart].base;
847     hwaddr size = vms->memmap[uart].size;
848     int irq = vms->irqmap[uart];
849     const char compat[] = "arm,pl011\0arm,primecell";
850     const char clocknames[] = "uartclk\0apb_pclk";
851     DeviceState *dev = qdev_new(TYPE_PL011);
852     SysBusDevice *s = SYS_BUS_DEVICE(dev);
853     MachineState *ms = MACHINE(vms);
854 
855     qdev_prop_set_chr(dev, "chardev", chr);
856     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
857     memory_region_add_subregion(mem, base,
858                                 sysbus_mmio_get_region(s, 0));
859     sysbus_connect_irq(s, 0, qdev_get_gpio_in(vms->gic, irq));
860 
861     nodename = g_strdup_printf("/pl011@%" PRIx64, base);
862     qemu_fdt_add_subnode(ms->fdt, nodename);
863     /* Note that we can't use setprop_string because of the embedded NUL */
864     qemu_fdt_setprop(ms->fdt, nodename, "compatible",
865                          compat, sizeof(compat));
866     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
867                                      2, base, 2, size);
868     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
869                                GIC_FDT_IRQ_TYPE_SPI, irq,
870                                GIC_FDT_IRQ_FLAGS_LEVEL_HI);
871     qemu_fdt_setprop_cells(ms->fdt, nodename, "clocks",
872                                vms->clock_phandle, vms->clock_phandle);
873     qemu_fdt_setprop(ms->fdt, nodename, "clock-names",
874                          clocknames, sizeof(clocknames));
875 
876     if (uart == VIRT_UART) {
877         qemu_fdt_setprop_string(ms->fdt, "/chosen", "stdout-path", nodename);
878     } else {
879         /* Mark as not usable by the normal world */
880         qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
881         qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
882 
883         qemu_fdt_setprop_string(ms->fdt, "/secure-chosen", "stdout-path",
884                                 nodename);
885     }
886 
887     g_free(nodename);
888 }
889 
890 static void create_rtc(const VirtMachineState *vms)
891 {
892     char *nodename;
893     hwaddr base = vms->memmap[VIRT_RTC].base;
894     hwaddr size = vms->memmap[VIRT_RTC].size;
895     int irq = vms->irqmap[VIRT_RTC];
896     const char compat[] = "arm,pl031\0arm,primecell";
897     MachineState *ms = MACHINE(vms);
898 
899     sysbus_create_simple("pl031", base, qdev_get_gpio_in(vms->gic, irq));
900 
901     nodename = g_strdup_printf("/pl031@%" PRIx64, base);
902     qemu_fdt_add_subnode(ms->fdt, nodename);
903     qemu_fdt_setprop(ms->fdt, nodename, "compatible", compat, sizeof(compat));
904     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
905                                  2, base, 2, size);
906     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
907                            GIC_FDT_IRQ_TYPE_SPI, irq,
908                            GIC_FDT_IRQ_FLAGS_LEVEL_HI);
909     qemu_fdt_setprop_cell(ms->fdt, nodename, "clocks", vms->clock_phandle);
910     qemu_fdt_setprop_string(ms->fdt, nodename, "clock-names", "apb_pclk");
911     g_free(nodename);
912 }
913 
914 static DeviceState *gpio_key_dev;
915 static void virt_powerdown_req(Notifier *n, void *opaque)
916 {
917     VirtMachineState *s = container_of(n, VirtMachineState, powerdown_notifier);
918 
919     if (s->acpi_dev) {
920         acpi_send_event(s->acpi_dev, ACPI_POWER_DOWN_STATUS);
921     } else {
922         /* use gpio Pin 3 for power button event */
923         qemu_set_irq(qdev_get_gpio_in(gpio_key_dev, 0), 1);
924     }
925 }
926 
927 static void create_gpio_keys(char *fdt, DeviceState *pl061_dev,
928                              uint32_t phandle)
929 {
930     gpio_key_dev = sysbus_create_simple("gpio-key", -1,
931                                         qdev_get_gpio_in(pl061_dev, 3));
932 
933     qemu_fdt_add_subnode(fdt, "/gpio-keys");
934     qemu_fdt_setprop_string(fdt, "/gpio-keys", "compatible", "gpio-keys");
935 
936     qemu_fdt_add_subnode(fdt, "/gpio-keys/poweroff");
937     qemu_fdt_setprop_string(fdt, "/gpio-keys/poweroff",
938                             "label", "GPIO Key Poweroff");
939     qemu_fdt_setprop_cell(fdt, "/gpio-keys/poweroff", "linux,code",
940                           KEY_POWER);
941     qemu_fdt_setprop_cells(fdt, "/gpio-keys/poweroff",
942                            "gpios", phandle, 3, 0);
943 }
944 
945 #define SECURE_GPIO_POWEROFF 0
946 #define SECURE_GPIO_RESET    1
947 
948 static void create_secure_gpio_pwr(char *fdt, DeviceState *pl061_dev,
949                                    uint32_t phandle)
950 {
951     DeviceState *gpio_pwr_dev;
952 
953     /* gpio-pwr */
954     gpio_pwr_dev = sysbus_create_simple("gpio-pwr", -1, NULL);
955 
956     /* connect secure pl061 to gpio-pwr */
957     qdev_connect_gpio_out(pl061_dev, SECURE_GPIO_RESET,
958                           qdev_get_gpio_in_named(gpio_pwr_dev, "reset", 0));
959     qdev_connect_gpio_out(pl061_dev, SECURE_GPIO_POWEROFF,
960                           qdev_get_gpio_in_named(gpio_pwr_dev, "shutdown", 0));
961 
962     qemu_fdt_add_subnode(fdt, "/gpio-poweroff");
963     qemu_fdt_setprop_string(fdt, "/gpio-poweroff", "compatible",
964                             "gpio-poweroff");
965     qemu_fdt_setprop_cells(fdt, "/gpio-poweroff",
966                            "gpios", phandle, SECURE_GPIO_POWEROFF, 0);
967     qemu_fdt_setprop_string(fdt, "/gpio-poweroff", "status", "disabled");
968     qemu_fdt_setprop_string(fdt, "/gpio-poweroff", "secure-status",
969                             "okay");
970 
971     qemu_fdt_add_subnode(fdt, "/gpio-restart");
972     qemu_fdt_setprop_string(fdt, "/gpio-restart", "compatible",
973                             "gpio-restart");
974     qemu_fdt_setprop_cells(fdt, "/gpio-restart",
975                            "gpios", phandle, SECURE_GPIO_RESET, 0);
976     qemu_fdt_setprop_string(fdt, "/gpio-restart", "status", "disabled");
977     qemu_fdt_setprop_string(fdt, "/gpio-restart", "secure-status",
978                             "okay");
979 }
980 
981 static void create_gpio_devices(const VirtMachineState *vms, int gpio,
982                                 MemoryRegion *mem)
983 {
984     char *nodename;
985     DeviceState *pl061_dev;
986     hwaddr base = vms->memmap[gpio].base;
987     hwaddr size = vms->memmap[gpio].size;
988     int irq = vms->irqmap[gpio];
989     const char compat[] = "arm,pl061\0arm,primecell";
990     SysBusDevice *s;
991     MachineState *ms = MACHINE(vms);
992 
993     pl061_dev = qdev_new("pl061");
994     /* Pull lines down to 0 if not driven by the PL061 */
995     qdev_prop_set_uint32(pl061_dev, "pullups", 0);
996     qdev_prop_set_uint32(pl061_dev, "pulldowns", 0xff);
997     s = SYS_BUS_DEVICE(pl061_dev);
998     sysbus_realize_and_unref(s, &error_fatal);
999     memory_region_add_subregion(mem, base, sysbus_mmio_get_region(s, 0));
1000     sysbus_connect_irq(s, 0, qdev_get_gpio_in(vms->gic, irq));
1001 
1002     uint32_t phandle = qemu_fdt_alloc_phandle(ms->fdt);
1003     nodename = g_strdup_printf("/pl061@%" PRIx64, base);
1004     qemu_fdt_add_subnode(ms->fdt, nodename);
1005     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1006                                  2, base, 2, size);
1007     qemu_fdt_setprop(ms->fdt, nodename, "compatible", compat, sizeof(compat));
1008     qemu_fdt_setprop_cell(ms->fdt, nodename, "#gpio-cells", 2);
1009     qemu_fdt_setprop(ms->fdt, nodename, "gpio-controller", NULL, 0);
1010     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
1011                            GIC_FDT_IRQ_TYPE_SPI, irq,
1012                            GIC_FDT_IRQ_FLAGS_LEVEL_HI);
1013     qemu_fdt_setprop_cell(ms->fdt, nodename, "clocks", vms->clock_phandle);
1014     qemu_fdt_setprop_string(ms->fdt, nodename, "clock-names", "apb_pclk");
1015     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", phandle);
1016 
1017     if (gpio != VIRT_GPIO) {
1018         /* Mark as not usable by the normal world */
1019         qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
1020         qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
1021     }
1022     g_free(nodename);
1023 
1024     /* Child gpio devices */
1025     if (gpio == VIRT_GPIO) {
1026         create_gpio_keys(ms->fdt, pl061_dev, phandle);
1027     } else {
1028         create_secure_gpio_pwr(ms->fdt, pl061_dev, phandle);
1029     }
1030 }
1031 
1032 static void create_virtio_devices(const VirtMachineState *vms)
1033 {
1034     int i;
1035     hwaddr size = vms->memmap[VIRT_MMIO].size;
1036     MachineState *ms = MACHINE(vms);
1037 
1038     /* We create the transports in forwards order. Since qbus_realize()
1039      * prepends (not appends) new child buses, the incrementing loop below will
1040      * create a list of virtio-mmio buses with decreasing base addresses.
1041      *
1042      * When a -device option is processed from the command line,
1043      * qbus_find_recursive() picks the next free virtio-mmio bus in forwards
1044      * order. The upshot is that -device options in increasing command line
1045      * order are mapped to virtio-mmio buses with decreasing base addresses.
1046      *
1047      * When this code was originally written, that arrangement ensured that the
1048      * guest Linux kernel would give the lowest "name" (/dev/vda, eth0, etc) to
1049      * the first -device on the command line. (The end-to-end order is a
1050      * function of this loop, qbus_realize(), qbus_find_recursive(), and the
1051      * guest kernel's name-to-address assignment strategy.)
1052      *
1053      * Meanwhile, the kernel's traversal seems to have been reversed; see eg.
1054      * the message, if not necessarily the code, of commit 70161ff336.
1055      * Therefore the loop now establishes the inverse of the original intent.
1056      *
1057      * Unfortunately, we can't counteract the kernel change by reversing the
1058      * loop; it would break existing command lines.
1059      *
1060      * In any case, the kernel makes no guarantee about the stability of
1061      * enumeration order of virtio devices (as demonstrated by it changing
1062      * between kernel versions). For reliable and stable identification
1063      * of disks users must use UUIDs or similar mechanisms.
1064      */
1065     for (i = 0; i < NUM_VIRTIO_TRANSPORTS; i++) {
1066         int irq = vms->irqmap[VIRT_MMIO] + i;
1067         hwaddr base = vms->memmap[VIRT_MMIO].base + i * size;
1068 
1069         sysbus_create_simple("virtio-mmio", base,
1070                              qdev_get_gpio_in(vms->gic, irq));
1071     }
1072 
1073     /* We add dtb nodes in reverse order so that they appear in the finished
1074      * device tree lowest address first.
1075      *
1076      * Note that this mapping is independent of the loop above. The previous
1077      * loop influences virtio device to virtio transport assignment, whereas
1078      * this loop controls how virtio transports are laid out in the dtb.
1079      */
1080     for (i = NUM_VIRTIO_TRANSPORTS - 1; i >= 0; i--) {
1081         char *nodename;
1082         int irq = vms->irqmap[VIRT_MMIO] + i;
1083         hwaddr base = vms->memmap[VIRT_MMIO].base + i * size;
1084 
1085         nodename = g_strdup_printf("/virtio_mmio@%" PRIx64, base);
1086         qemu_fdt_add_subnode(ms->fdt, nodename);
1087         qemu_fdt_setprop_string(ms->fdt, nodename,
1088                                 "compatible", "virtio,mmio");
1089         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1090                                      2, base, 2, size);
1091         qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
1092                                GIC_FDT_IRQ_TYPE_SPI, irq,
1093                                GIC_FDT_IRQ_FLAGS_EDGE_LO_HI);
1094         qemu_fdt_setprop(ms->fdt, nodename, "dma-coherent", NULL, 0);
1095         g_free(nodename);
1096     }
1097 }
1098 
1099 #define VIRT_FLASH_SECTOR_SIZE (256 * KiB)
1100 
1101 static PFlashCFI01 *virt_flash_create1(VirtMachineState *vms,
1102                                         const char *name,
1103                                         const char *alias_prop_name)
1104 {
1105     /*
1106      * Create a single flash device.  We use the same parameters as
1107      * the flash devices on the Versatile Express board.
1108      */
1109     DeviceState *dev = qdev_new(TYPE_PFLASH_CFI01);
1110 
1111     qdev_prop_set_uint64(dev, "sector-length", VIRT_FLASH_SECTOR_SIZE);
1112     qdev_prop_set_uint8(dev, "width", 4);
1113     qdev_prop_set_uint8(dev, "device-width", 2);
1114     qdev_prop_set_bit(dev, "big-endian", false);
1115     qdev_prop_set_uint16(dev, "id0", 0x89);
1116     qdev_prop_set_uint16(dev, "id1", 0x18);
1117     qdev_prop_set_uint16(dev, "id2", 0x00);
1118     qdev_prop_set_uint16(dev, "id3", 0x00);
1119     qdev_prop_set_string(dev, "name", name);
1120     object_property_add_child(OBJECT(vms), name, OBJECT(dev));
1121     object_property_add_alias(OBJECT(vms), alias_prop_name,
1122                               OBJECT(dev), "drive");
1123     return PFLASH_CFI01(dev);
1124 }
1125 
1126 static void virt_flash_create(VirtMachineState *vms)
1127 {
1128     vms->flash[0] = virt_flash_create1(vms, "virt.flash0", "pflash0");
1129     vms->flash[1] = virt_flash_create1(vms, "virt.flash1", "pflash1");
1130 }
1131 
1132 static void virt_flash_map1(PFlashCFI01 *flash,
1133                             hwaddr base, hwaddr size,
1134                             MemoryRegion *sysmem)
1135 {
1136     DeviceState *dev = DEVICE(flash);
1137 
1138     assert(QEMU_IS_ALIGNED(size, VIRT_FLASH_SECTOR_SIZE));
1139     assert(size / VIRT_FLASH_SECTOR_SIZE <= UINT32_MAX);
1140     qdev_prop_set_uint32(dev, "num-blocks", size / VIRT_FLASH_SECTOR_SIZE);
1141     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1142 
1143     memory_region_add_subregion(sysmem, base,
1144                                 sysbus_mmio_get_region(SYS_BUS_DEVICE(dev),
1145                                                        0));
1146 }
1147 
1148 static void virt_flash_map(VirtMachineState *vms,
1149                            MemoryRegion *sysmem,
1150                            MemoryRegion *secure_sysmem)
1151 {
1152     /*
1153      * Map two flash devices to fill the VIRT_FLASH space in the memmap.
1154      * sysmem is the system memory space. secure_sysmem is the secure view
1155      * of the system, and the first flash device should be made visible only
1156      * there. The second flash device is visible to both secure and nonsecure.
1157      * If sysmem == secure_sysmem this means there is no separate Secure
1158      * address space and both flash devices are generally visible.
1159      */
1160     hwaddr flashsize = vms->memmap[VIRT_FLASH].size / 2;
1161     hwaddr flashbase = vms->memmap[VIRT_FLASH].base;
1162 
1163     virt_flash_map1(vms->flash[0], flashbase, flashsize,
1164                     secure_sysmem);
1165     virt_flash_map1(vms->flash[1], flashbase + flashsize, flashsize,
1166                     sysmem);
1167 }
1168 
1169 static void virt_flash_fdt(VirtMachineState *vms,
1170                            MemoryRegion *sysmem,
1171                            MemoryRegion *secure_sysmem)
1172 {
1173     hwaddr flashsize = vms->memmap[VIRT_FLASH].size / 2;
1174     hwaddr flashbase = vms->memmap[VIRT_FLASH].base;
1175     MachineState *ms = MACHINE(vms);
1176     char *nodename;
1177 
1178     if (sysmem == secure_sysmem) {
1179         /* Report both flash devices as a single node in the DT */
1180         nodename = g_strdup_printf("/flash@%" PRIx64, flashbase);
1181         qemu_fdt_add_subnode(ms->fdt, nodename);
1182         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible", "cfi-flash");
1183         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1184                                      2, flashbase, 2, flashsize,
1185                                      2, flashbase + flashsize, 2, flashsize);
1186         qemu_fdt_setprop_cell(ms->fdt, nodename, "bank-width", 4);
1187         g_free(nodename);
1188     } else {
1189         /*
1190          * Report the devices as separate nodes so we can mark one as
1191          * only visible to the secure world.
1192          */
1193         nodename = g_strdup_printf("/secflash@%" PRIx64, flashbase);
1194         qemu_fdt_add_subnode(ms->fdt, nodename);
1195         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible", "cfi-flash");
1196         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1197                                      2, flashbase, 2, flashsize);
1198         qemu_fdt_setprop_cell(ms->fdt, nodename, "bank-width", 4);
1199         qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
1200         qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
1201         g_free(nodename);
1202 
1203         nodename = g_strdup_printf("/flash@%" PRIx64, flashbase + flashsize);
1204         qemu_fdt_add_subnode(ms->fdt, nodename);
1205         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible", "cfi-flash");
1206         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1207                                      2, flashbase + flashsize, 2, flashsize);
1208         qemu_fdt_setprop_cell(ms->fdt, nodename, "bank-width", 4);
1209         g_free(nodename);
1210     }
1211 }
1212 
1213 static bool virt_firmware_init(VirtMachineState *vms,
1214                                MemoryRegion *sysmem,
1215                                MemoryRegion *secure_sysmem)
1216 {
1217     int i;
1218     const char *bios_name;
1219     BlockBackend *pflash_blk0;
1220 
1221     /* Map legacy -drive if=pflash to machine properties */
1222     for (i = 0; i < ARRAY_SIZE(vms->flash); i++) {
1223         pflash_cfi01_legacy_drive(vms->flash[i],
1224                                   drive_get(IF_PFLASH, 0, i));
1225     }
1226 
1227     virt_flash_map(vms, sysmem, secure_sysmem);
1228 
1229     pflash_blk0 = pflash_cfi01_get_blk(vms->flash[0]);
1230 
1231     bios_name = MACHINE(vms)->firmware;
1232     if (bios_name) {
1233         char *fname;
1234         MemoryRegion *mr;
1235         int image_size;
1236 
1237         if (pflash_blk0) {
1238             error_report("The contents of the first flash device may be "
1239                          "specified with -bios or with -drive if=pflash... "
1240                          "but you cannot use both options at once");
1241             exit(1);
1242         }
1243 
1244         /* Fall back to -bios */
1245 
1246         fname = qemu_find_file(QEMU_FILE_TYPE_BIOS, bios_name);
1247         if (!fname) {
1248             error_report("Could not find ROM image '%s'", bios_name);
1249             exit(1);
1250         }
1251         mr = sysbus_mmio_get_region(SYS_BUS_DEVICE(vms->flash[0]), 0);
1252         image_size = load_image_mr(fname, mr);
1253         g_free(fname);
1254         if (image_size < 0) {
1255             error_report("Could not load ROM image '%s'", bios_name);
1256             exit(1);
1257         }
1258     }
1259 
1260     return pflash_blk0 || bios_name;
1261 }
1262 
1263 static FWCfgState *create_fw_cfg(const VirtMachineState *vms, AddressSpace *as)
1264 {
1265     MachineState *ms = MACHINE(vms);
1266     hwaddr base = vms->memmap[VIRT_FW_CFG].base;
1267     hwaddr size = vms->memmap[VIRT_FW_CFG].size;
1268     FWCfgState *fw_cfg;
1269     char *nodename;
1270 
1271     fw_cfg = fw_cfg_init_mem_wide(base + 8, base, 8, base + 16, as);
1272     fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, (uint16_t)ms->smp.cpus);
1273 
1274     nodename = g_strdup_printf("/fw-cfg@%" PRIx64, base);
1275     qemu_fdt_add_subnode(ms->fdt, nodename);
1276     qemu_fdt_setprop_string(ms->fdt, nodename,
1277                             "compatible", "qemu,fw-cfg-mmio");
1278     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1279                                  2, base, 2, size);
1280     qemu_fdt_setprop(ms->fdt, nodename, "dma-coherent", NULL, 0);
1281     g_free(nodename);
1282     return fw_cfg;
1283 }
1284 
1285 static void create_pcie_irq_map(const MachineState *ms,
1286                                 uint32_t gic_phandle,
1287                                 int first_irq, const char *nodename)
1288 {
1289     int devfn, pin;
1290     uint32_t full_irq_map[4 * 4 * 10] = { 0 };
1291     uint32_t *irq_map = full_irq_map;
1292 
1293     for (devfn = 0; devfn <= 0x18; devfn += 0x8) {
1294         for (pin = 0; pin < 4; pin++) {
1295             int irq_type = GIC_FDT_IRQ_TYPE_SPI;
1296             int irq_nr = first_irq + ((pin + PCI_SLOT(devfn)) % PCI_NUM_PINS);
1297             int irq_level = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
1298             int i;
1299 
1300             uint32_t map[] = {
1301                 devfn << 8, 0, 0,                           /* devfn */
1302                 pin + 1,                                    /* PCI pin */
1303                 gic_phandle, 0, 0, irq_type, irq_nr, irq_level }; /* GIC irq */
1304 
1305             /* Convert map to big endian */
1306             for (i = 0; i < 10; i++) {
1307                 irq_map[i] = cpu_to_be32(map[i]);
1308             }
1309             irq_map += 10;
1310         }
1311     }
1312 
1313     qemu_fdt_setprop(ms->fdt, nodename, "interrupt-map",
1314                      full_irq_map, sizeof(full_irq_map));
1315 
1316     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupt-map-mask",
1317                            cpu_to_be16(PCI_DEVFN(3, 0)), /* Slot 3 */
1318                            0, 0,
1319                            0x7           /* PCI irq */);
1320 }
1321 
1322 static void create_smmu(const VirtMachineState *vms,
1323                         PCIBus *bus)
1324 {
1325     char *node;
1326     const char compat[] = "arm,smmu-v3";
1327     int irq =  vms->irqmap[VIRT_SMMU];
1328     int i;
1329     hwaddr base = vms->memmap[VIRT_SMMU].base;
1330     hwaddr size = vms->memmap[VIRT_SMMU].size;
1331     const char irq_names[] = "eventq\0priq\0cmdq-sync\0gerror";
1332     DeviceState *dev;
1333     MachineState *ms = MACHINE(vms);
1334 
1335     if (vms->iommu != VIRT_IOMMU_SMMUV3 || !vms->iommu_phandle) {
1336         return;
1337     }
1338 
1339     dev = qdev_new("arm-smmuv3");
1340 
1341     object_property_set_link(OBJECT(dev), "primary-bus", OBJECT(bus),
1342                              &error_abort);
1343     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1344     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, base);
1345     for (i = 0; i < NUM_SMMU_IRQS; i++) {
1346         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
1347                            qdev_get_gpio_in(vms->gic, irq + i));
1348     }
1349 
1350     node = g_strdup_printf("/smmuv3@%" PRIx64, base);
1351     qemu_fdt_add_subnode(ms->fdt, node);
1352     qemu_fdt_setprop(ms->fdt, node, "compatible", compat, sizeof(compat));
1353     qemu_fdt_setprop_sized_cells(ms->fdt, node, "reg", 2, base, 2, size);
1354 
1355     qemu_fdt_setprop_cells(ms->fdt, node, "interrupts",
1356             GIC_FDT_IRQ_TYPE_SPI, irq    , GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1357             GIC_FDT_IRQ_TYPE_SPI, irq + 1, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1358             GIC_FDT_IRQ_TYPE_SPI, irq + 2, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1359             GIC_FDT_IRQ_TYPE_SPI, irq + 3, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI);
1360 
1361     qemu_fdt_setprop(ms->fdt, node, "interrupt-names", irq_names,
1362                      sizeof(irq_names));
1363 
1364     qemu_fdt_setprop(ms->fdt, node, "dma-coherent", NULL, 0);
1365 
1366     qemu_fdt_setprop_cell(ms->fdt, node, "#iommu-cells", 1);
1367 
1368     qemu_fdt_setprop_cell(ms->fdt, node, "phandle", vms->iommu_phandle);
1369     g_free(node);
1370 }
1371 
1372 static void create_virtio_iommu_dt_bindings(VirtMachineState *vms)
1373 {
1374     const char compat[] = "virtio,pci-iommu\0pci1af4,1057";
1375     uint16_t bdf = vms->virtio_iommu_bdf;
1376     MachineState *ms = MACHINE(vms);
1377     char *node;
1378 
1379     vms->iommu_phandle = qemu_fdt_alloc_phandle(ms->fdt);
1380 
1381     node = g_strdup_printf("%s/virtio_iommu@%x,%x", vms->pciehb_nodename,
1382                            PCI_SLOT(bdf), PCI_FUNC(bdf));
1383     qemu_fdt_add_subnode(ms->fdt, node);
1384     qemu_fdt_setprop(ms->fdt, node, "compatible", compat, sizeof(compat));
1385     qemu_fdt_setprop_sized_cells(ms->fdt, node, "reg",
1386                                  1, bdf << 8, 1, 0, 1, 0,
1387                                  1, 0, 1, 0);
1388 
1389     qemu_fdt_setprop_cell(ms->fdt, node, "#iommu-cells", 1);
1390     qemu_fdt_setprop_cell(ms->fdt, node, "phandle", vms->iommu_phandle);
1391     g_free(node);
1392 
1393     qemu_fdt_setprop_cells(ms->fdt, vms->pciehb_nodename, "iommu-map",
1394                            0x0, vms->iommu_phandle, 0x0, bdf,
1395                            bdf + 1, vms->iommu_phandle, bdf + 1, 0xffff - bdf);
1396 }
1397 
1398 static void create_pcie(VirtMachineState *vms)
1399 {
1400     hwaddr base_mmio = vms->memmap[VIRT_PCIE_MMIO].base;
1401     hwaddr size_mmio = vms->memmap[VIRT_PCIE_MMIO].size;
1402     hwaddr base_mmio_high = vms->memmap[VIRT_HIGH_PCIE_MMIO].base;
1403     hwaddr size_mmio_high = vms->memmap[VIRT_HIGH_PCIE_MMIO].size;
1404     hwaddr base_pio = vms->memmap[VIRT_PCIE_PIO].base;
1405     hwaddr size_pio = vms->memmap[VIRT_PCIE_PIO].size;
1406     hwaddr base_ecam, size_ecam;
1407     hwaddr base = base_mmio;
1408     int nr_pcie_buses;
1409     int irq = vms->irqmap[VIRT_PCIE];
1410     MemoryRegion *mmio_alias;
1411     MemoryRegion *mmio_reg;
1412     MemoryRegion *ecam_alias;
1413     MemoryRegion *ecam_reg;
1414     DeviceState *dev;
1415     char *nodename;
1416     int i, ecam_id;
1417     PCIHostState *pci;
1418     MachineState *ms = MACHINE(vms);
1419 
1420     dev = qdev_new(TYPE_GPEX_HOST);
1421     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1422 
1423     ecam_id = VIRT_ECAM_ID(vms->highmem_ecam);
1424     base_ecam = vms->memmap[ecam_id].base;
1425     size_ecam = vms->memmap[ecam_id].size;
1426     nr_pcie_buses = size_ecam / PCIE_MMCFG_SIZE_MIN;
1427     /* Map only the first size_ecam bytes of ECAM space */
1428     ecam_alias = g_new0(MemoryRegion, 1);
1429     ecam_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 0);
1430     memory_region_init_alias(ecam_alias, OBJECT(dev), "pcie-ecam",
1431                              ecam_reg, 0, size_ecam);
1432     memory_region_add_subregion(get_system_memory(), base_ecam, ecam_alias);
1433 
1434     /* Map the MMIO window into system address space so as to expose
1435      * the section of PCI MMIO space which starts at the same base address
1436      * (ie 1:1 mapping for that part of PCI MMIO space visible through
1437      * the window).
1438      */
1439     mmio_alias = g_new0(MemoryRegion, 1);
1440     mmio_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 1);
1441     memory_region_init_alias(mmio_alias, OBJECT(dev), "pcie-mmio",
1442                              mmio_reg, base_mmio, size_mmio);
1443     memory_region_add_subregion(get_system_memory(), base_mmio, mmio_alias);
1444 
1445     if (vms->highmem_mmio) {
1446         /* Map high MMIO space */
1447         MemoryRegion *high_mmio_alias = g_new0(MemoryRegion, 1);
1448 
1449         memory_region_init_alias(high_mmio_alias, OBJECT(dev), "pcie-mmio-high",
1450                                  mmio_reg, base_mmio_high, size_mmio_high);
1451         memory_region_add_subregion(get_system_memory(), base_mmio_high,
1452                                     high_mmio_alias);
1453     }
1454 
1455     /* Map IO port space */
1456     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 2, base_pio);
1457 
1458     for (i = 0; i < GPEX_NUM_IRQS; i++) {
1459         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
1460                            qdev_get_gpio_in(vms->gic, irq + i));
1461         gpex_set_irq_num(GPEX_HOST(dev), i, irq + i);
1462     }
1463 
1464     pci = PCI_HOST_BRIDGE(dev);
1465     pci->bypass_iommu = vms->default_bus_bypass_iommu;
1466     vms->bus = pci->bus;
1467     if (vms->bus) {
1468         for (i = 0; i < nb_nics; i++) {
1469             NICInfo *nd = &nd_table[i];
1470 
1471             if (!nd->model) {
1472                 nd->model = g_strdup("virtio");
1473             }
1474 
1475             pci_nic_init_nofail(nd, pci->bus, nd->model, NULL);
1476         }
1477     }
1478 
1479     nodename = vms->pciehb_nodename = g_strdup_printf("/pcie@%" PRIx64, base);
1480     qemu_fdt_add_subnode(ms->fdt, nodename);
1481     qemu_fdt_setprop_string(ms->fdt, nodename,
1482                             "compatible", "pci-host-ecam-generic");
1483     qemu_fdt_setprop_string(ms->fdt, nodename, "device_type", "pci");
1484     qemu_fdt_setprop_cell(ms->fdt, nodename, "#address-cells", 3);
1485     qemu_fdt_setprop_cell(ms->fdt, nodename, "#size-cells", 2);
1486     qemu_fdt_setprop_cell(ms->fdt, nodename, "linux,pci-domain", 0);
1487     qemu_fdt_setprop_cells(ms->fdt, nodename, "bus-range", 0,
1488                            nr_pcie_buses - 1);
1489     qemu_fdt_setprop(ms->fdt, nodename, "dma-coherent", NULL, 0);
1490 
1491     if (vms->msi_phandle) {
1492         qemu_fdt_setprop_cells(ms->fdt, nodename, "msi-map",
1493                                0, vms->msi_phandle, 0, 0x10000);
1494     }
1495 
1496     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1497                                  2, base_ecam, 2, size_ecam);
1498 
1499     if (vms->highmem_mmio) {
1500         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "ranges",
1501                                      1, FDT_PCI_RANGE_IOPORT, 2, 0,
1502                                      2, base_pio, 2, size_pio,
1503                                      1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
1504                                      2, base_mmio, 2, size_mmio,
1505                                      1, FDT_PCI_RANGE_MMIO_64BIT,
1506                                      2, base_mmio_high,
1507                                      2, base_mmio_high, 2, size_mmio_high);
1508     } else {
1509         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "ranges",
1510                                      1, FDT_PCI_RANGE_IOPORT, 2, 0,
1511                                      2, base_pio, 2, size_pio,
1512                                      1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
1513                                      2, base_mmio, 2, size_mmio);
1514     }
1515 
1516     qemu_fdt_setprop_cell(ms->fdt, nodename, "#interrupt-cells", 1);
1517     create_pcie_irq_map(ms, vms->gic_phandle, irq, nodename);
1518 
1519     if (vms->iommu) {
1520         vms->iommu_phandle = qemu_fdt_alloc_phandle(ms->fdt);
1521 
1522         switch (vms->iommu) {
1523         case VIRT_IOMMU_SMMUV3:
1524             create_smmu(vms, vms->bus);
1525             qemu_fdt_setprop_cells(ms->fdt, nodename, "iommu-map",
1526                                    0x0, vms->iommu_phandle, 0x0, 0x10000);
1527             break;
1528         default:
1529             g_assert_not_reached();
1530         }
1531     }
1532 }
1533 
1534 static void create_platform_bus(VirtMachineState *vms)
1535 {
1536     DeviceState *dev;
1537     SysBusDevice *s;
1538     int i;
1539     MemoryRegion *sysmem = get_system_memory();
1540 
1541     dev = qdev_new(TYPE_PLATFORM_BUS_DEVICE);
1542     dev->id = g_strdup(TYPE_PLATFORM_BUS_DEVICE);
1543     qdev_prop_set_uint32(dev, "num_irqs", PLATFORM_BUS_NUM_IRQS);
1544     qdev_prop_set_uint32(dev, "mmio_size", vms->memmap[VIRT_PLATFORM_BUS].size);
1545     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1546     vms->platform_bus_dev = dev;
1547 
1548     s = SYS_BUS_DEVICE(dev);
1549     for (i = 0; i < PLATFORM_BUS_NUM_IRQS; i++) {
1550         int irq = vms->irqmap[VIRT_PLATFORM_BUS] + i;
1551         sysbus_connect_irq(s, i, qdev_get_gpio_in(vms->gic, irq));
1552     }
1553 
1554     memory_region_add_subregion(sysmem,
1555                                 vms->memmap[VIRT_PLATFORM_BUS].base,
1556                                 sysbus_mmio_get_region(s, 0));
1557 }
1558 
1559 static void create_tag_ram(MemoryRegion *tag_sysmem,
1560                            hwaddr base, hwaddr size,
1561                            const char *name)
1562 {
1563     MemoryRegion *tagram = g_new(MemoryRegion, 1);
1564 
1565     memory_region_init_ram(tagram, NULL, name, size / 32, &error_fatal);
1566     memory_region_add_subregion(tag_sysmem, base / 32, tagram);
1567 }
1568 
1569 static void create_secure_ram(VirtMachineState *vms,
1570                               MemoryRegion *secure_sysmem,
1571                               MemoryRegion *secure_tag_sysmem)
1572 {
1573     MemoryRegion *secram = g_new(MemoryRegion, 1);
1574     char *nodename;
1575     hwaddr base = vms->memmap[VIRT_SECURE_MEM].base;
1576     hwaddr size = vms->memmap[VIRT_SECURE_MEM].size;
1577     MachineState *ms = MACHINE(vms);
1578 
1579     memory_region_init_ram(secram, NULL, "virt.secure-ram", size,
1580                            &error_fatal);
1581     memory_region_add_subregion(secure_sysmem, base, secram);
1582 
1583     nodename = g_strdup_printf("/secram@%" PRIx64, base);
1584     qemu_fdt_add_subnode(ms->fdt, nodename);
1585     qemu_fdt_setprop_string(ms->fdt, nodename, "device_type", "memory");
1586     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg", 2, base, 2, size);
1587     qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
1588     qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
1589 
1590     if (secure_tag_sysmem) {
1591         create_tag_ram(secure_tag_sysmem, base, size, "mach-virt.secure-tag");
1592     }
1593 
1594     g_free(nodename);
1595 }
1596 
1597 static void *machvirt_dtb(const struct arm_boot_info *binfo, int *fdt_size)
1598 {
1599     const VirtMachineState *board = container_of(binfo, VirtMachineState,
1600                                                  bootinfo);
1601     MachineState *ms = MACHINE(board);
1602 
1603 
1604     *fdt_size = board->fdt_size;
1605     return ms->fdt;
1606 }
1607 
1608 static void virt_build_smbios(VirtMachineState *vms)
1609 {
1610     MachineClass *mc = MACHINE_GET_CLASS(vms);
1611     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
1612     uint8_t *smbios_tables, *smbios_anchor;
1613     size_t smbios_tables_len, smbios_anchor_len;
1614     const char *product = "QEMU Virtual Machine";
1615 
1616     if (kvm_enabled()) {
1617         product = "KVM Virtual Machine";
1618     }
1619 
1620     smbios_set_defaults("QEMU", product,
1621                         vmc->smbios_old_sys_ver ? "1.0" : mc->name, false,
1622                         true, SMBIOS_ENTRY_POINT_TYPE_64);
1623 
1624     smbios_get_tables(MACHINE(vms), NULL, 0,
1625                       &smbios_tables, &smbios_tables_len,
1626                       &smbios_anchor, &smbios_anchor_len,
1627                       &error_fatal);
1628 
1629     if (smbios_anchor) {
1630         fw_cfg_add_file(vms->fw_cfg, "etc/smbios/smbios-tables",
1631                         smbios_tables, smbios_tables_len);
1632         fw_cfg_add_file(vms->fw_cfg, "etc/smbios/smbios-anchor",
1633                         smbios_anchor, smbios_anchor_len);
1634     }
1635 }
1636 
1637 static
1638 void virt_machine_done(Notifier *notifier, void *data)
1639 {
1640     VirtMachineState *vms = container_of(notifier, VirtMachineState,
1641                                          machine_done);
1642     MachineState *ms = MACHINE(vms);
1643     ARMCPU *cpu = ARM_CPU(first_cpu);
1644     struct arm_boot_info *info = &vms->bootinfo;
1645     AddressSpace *as = arm_boot_address_space(cpu, info);
1646 
1647     /*
1648      * If the user provided a dtb, we assume the dynamic sysbus nodes
1649      * already are integrated there. This corresponds to a use case where
1650      * the dynamic sysbus nodes are complex and their generation is not yet
1651      * supported. In that case the user can take charge of the guest dt
1652      * while qemu takes charge of the qom stuff.
1653      */
1654     if (info->dtb_filename == NULL) {
1655         platform_bus_add_all_fdt_nodes(ms->fdt, "/intc",
1656                                        vms->memmap[VIRT_PLATFORM_BUS].base,
1657                                        vms->memmap[VIRT_PLATFORM_BUS].size,
1658                                        vms->irqmap[VIRT_PLATFORM_BUS]);
1659     }
1660     if (arm_load_dtb(info->dtb_start, info, info->dtb_limit, as, ms) < 0) {
1661         exit(1);
1662     }
1663 
1664     fw_cfg_add_extra_pci_roots(vms->bus, vms->fw_cfg);
1665 
1666     virt_acpi_setup(vms);
1667     virt_build_smbios(vms);
1668 }
1669 
1670 static uint64_t virt_cpu_mp_affinity(VirtMachineState *vms, int idx)
1671 {
1672     uint8_t clustersz = ARM_DEFAULT_CPUS_PER_CLUSTER;
1673     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
1674 
1675     if (!vmc->disallow_affinity_adjustment) {
1676         /* Adjust MPIDR like 64-bit KVM hosts, which incorporate the
1677          * GIC's target-list limitations. 32-bit KVM hosts currently
1678          * always create clusters of 4 CPUs, but that is expected to
1679          * change when they gain support for gicv3. When KVM is enabled
1680          * it will override the changes we make here, therefore our
1681          * purposes are to make TCG consistent (with 64-bit KVM hosts)
1682          * and to improve SGI efficiency.
1683          */
1684         if (vms->gic_version == VIRT_GIC_VERSION_2) {
1685             clustersz = GIC_TARGETLIST_BITS;
1686         } else {
1687             clustersz = GICV3_TARGETLIST_BITS;
1688         }
1689     }
1690     return arm_cpu_mp_affinity(idx, clustersz);
1691 }
1692 
1693 static void virt_set_memmap(VirtMachineState *vms, int pa_bits)
1694 {
1695     MachineState *ms = MACHINE(vms);
1696     hwaddr base, device_memory_base, device_memory_size, memtop;
1697     int i;
1698 
1699     vms->memmap = extended_memmap;
1700 
1701     for (i = 0; i < ARRAY_SIZE(base_memmap); i++) {
1702         vms->memmap[i] = base_memmap[i];
1703     }
1704 
1705     if (ms->ram_slots > ACPI_MAX_RAM_SLOTS) {
1706         error_report("unsupported number of memory slots: %"PRIu64,
1707                      ms->ram_slots);
1708         exit(EXIT_FAILURE);
1709     }
1710 
1711     /*
1712      * !highmem is exactly the same as limiting the PA space to 32bit,
1713      * irrespective of the underlying capabilities of the HW.
1714      */
1715     if (!vms->highmem) {
1716         pa_bits = 32;
1717     }
1718 
1719     /*
1720      * We compute the base of the high IO region depending on the
1721      * amount of initial and device memory. The device memory start/size
1722      * is aligned on 1GiB. We never put the high IO region below 256GiB
1723      * so that if maxram_size is < 255GiB we keep the legacy memory map.
1724      * The device region size assumes 1GiB page max alignment per slot.
1725      */
1726     device_memory_base =
1727         ROUND_UP(vms->memmap[VIRT_MEM].base + ms->ram_size, GiB);
1728     device_memory_size = ms->maxram_size - ms->ram_size + ms->ram_slots * GiB;
1729 
1730     /* Base address of the high IO region */
1731     memtop = base = device_memory_base + ROUND_UP(device_memory_size, GiB);
1732     if (memtop > BIT_ULL(pa_bits)) {
1733 	    error_report("Addressing limited to %d bits, but memory exceeds it by %llu bytes\n",
1734 			 pa_bits, memtop - BIT_ULL(pa_bits));
1735         exit(EXIT_FAILURE);
1736     }
1737     if (base < device_memory_base) {
1738         error_report("maxmem/slots too huge");
1739         exit(EXIT_FAILURE);
1740     }
1741     if (base < vms->memmap[VIRT_MEM].base + LEGACY_RAMLIMIT_BYTES) {
1742         base = vms->memmap[VIRT_MEM].base + LEGACY_RAMLIMIT_BYTES;
1743     }
1744 
1745     /* We know for sure that at least the memory fits in the PA space */
1746     vms->highest_gpa = memtop - 1;
1747 
1748     for (i = VIRT_LOWMEMMAP_LAST; i < ARRAY_SIZE(extended_memmap); i++) {
1749         hwaddr size = extended_memmap[i].size;
1750         bool fits;
1751 
1752         base = ROUND_UP(base, size);
1753         vms->memmap[i].base = base;
1754         vms->memmap[i].size = size;
1755 
1756         /*
1757          * Check each device to see if they fit in the PA space,
1758          * moving highest_gpa as we go.
1759          *
1760          * For each device that doesn't fit, disable it.
1761          */
1762         fits = (base + size) <= BIT_ULL(pa_bits);
1763         if (fits) {
1764             vms->highest_gpa = base + size - 1;
1765         }
1766 
1767         switch (i) {
1768         case VIRT_HIGH_GIC_REDIST2:
1769             vms->highmem_redists &= fits;
1770             break;
1771         case VIRT_HIGH_PCIE_ECAM:
1772             vms->highmem_ecam &= fits;
1773             break;
1774         case VIRT_HIGH_PCIE_MMIO:
1775             vms->highmem_mmio &= fits;
1776             break;
1777         }
1778 
1779         base += size;
1780     }
1781 
1782     if (device_memory_size > 0) {
1783         ms->device_memory = g_malloc0(sizeof(*ms->device_memory));
1784         ms->device_memory->base = device_memory_base;
1785         memory_region_init(&ms->device_memory->mr, OBJECT(vms),
1786                            "device-memory", device_memory_size);
1787     }
1788 }
1789 
1790 /*
1791  * finalize_gic_version - Determines the final gic_version
1792  * according to the gic-version property
1793  *
1794  * Default GIC type is v2
1795  */
1796 static void finalize_gic_version(VirtMachineState *vms)
1797 {
1798     unsigned int max_cpus = MACHINE(vms)->smp.max_cpus;
1799 
1800     if (kvm_enabled()) {
1801         int probe_bitmap;
1802 
1803         if (!kvm_irqchip_in_kernel()) {
1804             switch (vms->gic_version) {
1805             case VIRT_GIC_VERSION_HOST:
1806                 warn_report(
1807                     "gic-version=host not relevant with kernel-irqchip=off "
1808                      "as only userspace GICv2 is supported. Using v2 ...");
1809                 return;
1810             case VIRT_GIC_VERSION_MAX:
1811             case VIRT_GIC_VERSION_NOSEL:
1812                 vms->gic_version = VIRT_GIC_VERSION_2;
1813                 return;
1814             case VIRT_GIC_VERSION_2:
1815                 return;
1816             case VIRT_GIC_VERSION_3:
1817                 error_report(
1818                     "gic-version=3 is not supported with kernel-irqchip=off");
1819                 exit(1);
1820             case VIRT_GIC_VERSION_4:
1821                 error_report(
1822                     "gic-version=4 is not supported with kernel-irqchip=off");
1823                 exit(1);
1824             }
1825         }
1826 
1827         probe_bitmap = kvm_arm_vgic_probe();
1828         if (!probe_bitmap) {
1829             error_report("Unable to determine GIC version supported by host");
1830             exit(1);
1831         }
1832 
1833         switch (vms->gic_version) {
1834         case VIRT_GIC_VERSION_HOST:
1835         case VIRT_GIC_VERSION_MAX:
1836             if (probe_bitmap & KVM_ARM_VGIC_V3) {
1837                 vms->gic_version = VIRT_GIC_VERSION_3;
1838             } else {
1839                 vms->gic_version = VIRT_GIC_VERSION_2;
1840             }
1841             return;
1842         case VIRT_GIC_VERSION_NOSEL:
1843             if ((probe_bitmap & KVM_ARM_VGIC_V2) && max_cpus <= GIC_NCPU) {
1844                 vms->gic_version = VIRT_GIC_VERSION_2;
1845             } else if (probe_bitmap & KVM_ARM_VGIC_V3) {
1846                 /*
1847                  * in case the host does not support v2 in-kernel emulation or
1848                  * the end-user requested more than 8 VCPUs we now default
1849                  * to v3. In any case defaulting to v2 would be broken.
1850                  */
1851                 vms->gic_version = VIRT_GIC_VERSION_3;
1852             } else if (max_cpus > GIC_NCPU) {
1853                 error_report("host only supports in-kernel GICv2 emulation "
1854                              "but more than 8 vcpus are requested");
1855                 exit(1);
1856             }
1857             break;
1858         case VIRT_GIC_VERSION_2:
1859         case VIRT_GIC_VERSION_3:
1860             break;
1861         case VIRT_GIC_VERSION_4:
1862             error_report("gic-version=4 is not supported with KVM");
1863             exit(1);
1864         }
1865 
1866         /* Check chosen version is effectively supported by the host */
1867         if (vms->gic_version == VIRT_GIC_VERSION_2 &&
1868             !(probe_bitmap & KVM_ARM_VGIC_V2)) {
1869             error_report("host does not support in-kernel GICv2 emulation");
1870             exit(1);
1871         } else if (vms->gic_version == VIRT_GIC_VERSION_3 &&
1872                    !(probe_bitmap & KVM_ARM_VGIC_V3)) {
1873             error_report("host does not support in-kernel GICv3 emulation");
1874             exit(1);
1875         }
1876         return;
1877     }
1878 
1879     /* TCG mode */
1880     switch (vms->gic_version) {
1881     case VIRT_GIC_VERSION_NOSEL:
1882         vms->gic_version = VIRT_GIC_VERSION_2;
1883         break;
1884     case VIRT_GIC_VERSION_MAX:
1885         if (module_object_class_by_name("arm-gicv3")) {
1886             /* CONFIG_ARM_GICV3_TCG was set */
1887             if (vms->virt) {
1888                 /* GICv4 only makes sense if CPU has EL2 */
1889                 vms->gic_version = VIRT_GIC_VERSION_4;
1890             } else {
1891                 vms->gic_version = VIRT_GIC_VERSION_3;
1892             }
1893         } else {
1894             vms->gic_version = VIRT_GIC_VERSION_2;
1895         }
1896         break;
1897     case VIRT_GIC_VERSION_HOST:
1898         error_report("gic-version=host requires KVM");
1899         exit(1);
1900     case VIRT_GIC_VERSION_4:
1901         if (!vms->virt) {
1902             error_report("gic-version=4 requires virtualization enabled");
1903             exit(1);
1904         }
1905         break;
1906     case VIRT_GIC_VERSION_2:
1907     case VIRT_GIC_VERSION_3:
1908         break;
1909     }
1910 }
1911 
1912 /*
1913  * virt_cpu_post_init() must be called after the CPUs have
1914  * been realized and the GIC has been created.
1915  */
1916 static void virt_cpu_post_init(VirtMachineState *vms, MemoryRegion *sysmem)
1917 {
1918     int max_cpus = MACHINE(vms)->smp.max_cpus;
1919     bool aarch64, pmu, steal_time;
1920     CPUState *cpu;
1921 
1922     aarch64 = object_property_get_bool(OBJECT(first_cpu), "aarch64", NULL);
1923     pmu = object_property_get_bool(OBJECT(first_cpu), "pmu", NULL);
1924     steal_time = object_property_get_bool(OBJECT(first_cpu),
1925                                           "kvm-steal-time", NULL);
1926 
1927     if (kvm_enabled()) {
1928         hwaddr pvtime_reg_base = vms->memmap[VIRT_PVTIME].base;
1929         hwaddr pvtime_reg_size = vms->memmap[VIRT_PVTIME].size;
1930 
1931         if (steal_time) {
1932             MemoryRegion *pvtime = g_new(MemoryRegion, 1);
1933             hwaddr pvtime_size = max_cpus * PVTIME_SIZE_PER_CPU;
1934 
1935             /* The memory region size must be a multiple of host page size. */
1936             pvtime_size = REAL_HOST_PAGE_ALIGN(pvtime_size);
1937 
1938             if (pvtime_size > pvtime_reg_size) {
1939                 error_report("pvtime requires a %" HWADDR_PRId
1940                              " byte memory region for %d CPUs,"
1941                              " but only %" HWADDR_PRId " has been reserved",
1942                              pvtime_size, max_cpus, pvtime_reg_size);
1943                 exit(1);
1944             }
1945 
1946             memory_region_init_ram(pvtime, NULL, "pvtime", pvtime_size, NULL);
1947             memory_region_add_subregion(sysmem, pvtime_reg_base, pvtime);
1948         }
1949 
1950         CPU_FOREACH(cpu) {
1951             if (pmu) {
1952                 assert(arm_feature(&ARM_CPU(cpu)->env, ARM_FEATURE_PMU));
1953                 if (kvm_irqchip_in_kernel()) {
1954                     kvm_arm_pmu_set_irq(cpu, PPI(VIRTUAL_PMU_IRQ));
1955                 }
1956                 kvm_arm_pmu_init(cpu);
1957             }
1958             if (steal_time) {
1959                 kvm_arm_pvtime_init(cpu, pvtime_reg_base +
1960                                          cpu->cpu_index * PVTIME_SIZE_PER_CPU);
1961             }
1962         }
1963     } else {
1964         if (aarch64 && vms->highmem) {
1965             int requested_pa_size = 64 - clz64(vms->highest_gpa);
1966             int pamax = arm_pamax(ARM_CPU(first_cpu));
1967 
1968             if (pamax < requested_pa_size) {
1969                 error_report("VCPU supports less PA bits (%d) than "
1970                              "requested by the memory map (%d)",
1971                              pamax, requested_pa_size);
1972                 exit(1);
1973             }
1974         }
1975     }
1976 }
1977 
1978 static void machvirt_init(MachineState *machine)
1979 {
1980     VirtMachineState *vms = VIRT_MACHINE(machine);
1981     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(machine);
1982     MachineClass *mc = MACHINE_GET_CLASS(machine);
1983     const CPUArchIdList *possible_cpus;
1984     MemoryRegion *sysmem = get_system_memory();
1985     MemoryRegion *secure_sysmem = NULL;
1986     MemoryRegion *tag_sysmem = NULL;
1987     MemoryRegion *secure_tag_sysmem = NULL;
1988     int n, virt_max_cpus;
1989     bool firmware_loaded;
1990     bool aarch64 = true;
1991     bool has_ged = !vmc->no_ged;
1992     unsigned int smp_cpus = machine->smp.cpus;
1993     unsigned int max_cpus = machine->smp.max_cpus;
1994 
1995     if (!cpu_type_valid(machine->cpu_type)) {
1996         error_report("mach-virt: CPU type %s not supported", machine->cpu_type);
1997         exit(1);
1998     }
1999 
2000     possible_cpus = mc->possible_cpu_arch_ids(machine);
2001 
2002     /*
2003      * In accelerated mode, the memory map is computed earlier in kvm_type()
2004      * to create a VM with the right number of IPA bits.
2005      */
2006     if (!vms->memmap) {
2007         Object *cpuobj;
2008         ARMCPU *armcpu;
2009         int pa_bits;
2010 
2011         /*
2012          * Instanciate a temporary CPU object to find out about what
2013          * we are about to deal with. Once this is done, get rid of
2014          * the object.
2015          */
2016         cpuobj = object_new(possible_cpus->cpus[0].type);
2017         armcpu = ARM_CPU(cpuobj);
2018 
2019         pa_bits = arm_pamax(armcpu);
2020 
2021         object_unref(cpuobj);
2022 
2023         virt_set_memmap(vms, pa_bits);
2024     }
2025 
2026     /* We can probe only here because during property set
2027      * KVM is not available yet
2028      */
2029     finalize_gic_version(vms);
2030 
2031     if (vms->secure) {
2032         /*
2033          * The Secure view of the world is the same as the NonSecure,
2034          * but with a few extra devices. Create it as a container region
2035          * containing the system memory at low priority; any secure-only
2036          * devices go in at higher priority and take precedence.
2037          */
2038         secure_sysmem = g_new(MemoryRegion, 1);
2039         memory_region_init(secure_sysmem, OBJECT(machine), "secure-memory",
2040                            UINT64_MAX);
2041         memory_region_add_subregion_overlap(secure_sysmem, 0, sysmem, -1);
2042     }
2043 
2044     firmware_loaded = virt_firmware_init(vms, sysmem,
2045                                          secure_sysmem ?: sysmem);
2046 
2047     /* If we have an EL3 boot ROM then the assumption is that it will
2048      * implement PSCI itself, so disable QEMU's internal implementation
2049      * so it doesn't get in the way. Instead of starting secondary
2050      * CPUs in PSCI powerdown state we will start them all running and
2051      * let the boot ROM sort them out.
2052      * The usual case is that we do use QEMU's PSCI implementation;
2053      * if the guest has EL2 then we will use SMC as the conduit,
2054      * and otherwise we will use HVC (for backwards compatibility and
2055      * because if we're using KVM then we must use HVC).
2056      */
2057     if (vms->secure && firmware_loaded) {
2058         vms->psci_conduit = QEMU_PSCI_CONDUIT_DISABLED;
2059     } else if (vms->virt) {
2060         vms->psci_conduit = QEMU_PSCI_CONDUIT_SMC;
2061     } else {
2062         vms->psci_conduit = QEMU_PSCI_CONDUIT_HVC;
2063     }
2064 
2065     /*
2066      * The maximum number of CPUs depends on the GIC version, or on how
2067      * many redistributors we can fit into the memory map (which in turn
2068      * depends on whether this is a GICv3 or v4).
2069      */
2070     if (vms->gic_version == VIRT_GIC_VERSION_2) {
2071         virt_max_cpus = GIC_NCPU;
2072     } else {
2073         virt_max_cpus = virt_redist_capacity(vms, VIRT_GIC_REDIST) +
2074             virt_redist_capacity(vms, VIRT_HIGH_GIC_REDIST2);
2075     }
2076 
2077     if (max_cpus > virt_max_cpus) {
2078         error_report("Number of SMP CPUs requested (%d) exceeds max CPUs "
2079                      "supported by machine 'mach-virt' (%d)",
2080                      max_cpus, virt_max_cpus);
2081         exit(1);
2082     }
2083 
2084     if (vms->secure && (kvm_enabled() || hvf_enabled())) {
2085         error_report("mach-virt: %s does not support providing "
2086                      "Security extensions (TrustZone) to the guest CPU",
2087                      kvm_enabled() ? "KVM" : "HVF");
2088         exit(1);
2089     }
2090 
2091     if (vms->virt && (kvm_enabled() || hvf_enabled())) {
2092         error_report("mach-virt: %s does not support providing "
2093                      "Virtualization extensions to the guest CPU",
2094                      kvm_enabled() ? "KVM" : "HVF");
2095         exit(1);
2096     }
2097 
2098     if (vms->mte && (kvm_enabled() || hvf_enabled())) {
2099         error_report("mach-virt: %s does not support providing "
2100                      "MTE to the guest CPU",
2101                      kvm_enabled() ? "KVM" : "HVF");
2102         exit(1);
2103     }
2104 
2105     create_fdt(vms);
2106 
2107     assert(possible_cpus->len == max_cpus);
2108     for (n = 0; n < possible_cpus->len; n++) {
2109         Object *cpuobj;
2110         CPUState *cs;
2111 
2112         if (n >= smp_cpus) {
2113             break;
2114         }
2115 
2116         cpuobj = object_new(possible_cpus->cpus[n].type);
2117         object_property_set_int(cpuobj, "mp-affinity",
2118                                 possible_cpus->cpus[n].arch_id, NULL);
2119 
2120         cs = CPU(cpuobj);
2121         cs->cpu_index = n;
2122 
2123         numa_cpu_pre_plug(&possible_cpus->cpus[cs->cpu_index], DEVICE(cpuobj),
2124                           &error_fatal);
2125 
2126         aarch64 &= object_property_get_bool(cpuobj, "aarch64", NULL);
2127 
2128         if (!vms->secure) {
2129             object_property_set_bool(cpuobj, "has_el3", false, NULL);
2130         }
2131 
2132         if (!vms->virt && object_property_find(cpuobj, "has_el2")) {
2133             object_property_set_bool(cpuobj, "has_el2", false, NULL);
2134         }
2135 
2136         if (vmc->kvm_no_adjvtime &&
2137             object_property_find(cpuobj, "kvm-no-adjvtime")) {
2138             object_property_set_bool(cpuobj, "kvm-no-adjvtime", true, NULL);
2139         }
2140 
2141         if (vmc->no_kvm_steal_time &&
2142             object_property_find(cpuobj, "kvm-steal-time")) {
2143             object_property_set_bool(cpuobj, "kvm-steal-time", false, NULL);
2144         }
2145 
2146         if (vmc->no_pmu && object_property_find(cpuobj, "pmu")) {
2147             object_property_set_bool(cpuobj, "pmu", false, NULL);
2148         }
2149 
2150         if (vmc->no_tcg_lpa2 && object_property_find(cpuobj, "lpa2")) {
2151             object_property_set_bool(cpuobj, "lpa2", false, NULL);
2152         }
2153 
2154         if (object_property_find(cpuobj, "reset-cbar")) {
2155             object_property_set_int(cpuobj, "reset-cbar",
2156                                     vms->memmap[VIRT_CPUPERIPHS].base,
2157                                     &error_abort);
2158         }
2159 
2160         object_property_set_link(cpuobj, "memory", OBJECT(sysmem),
2161                                  &error_abort);
2162         if (vms->secure) {
2163             object_property_set_link(cpuobj, "secure-memory",
2164                                      OBJECT(secure_sysmem), &error_abort);
2165         }
2166 
2167         if (vms->mte) {
2168             /* Create the memory region only once, but link to all cpus. */
2169             if (!tag_sysmem) {
2170                 /*
2171                  * The property exists only if MemTag is supported.
2172                  * If it is, we must allocate the ram to back that up.
2173                  */
2174                 if (!object_property_find(cpuobj, "tag-memory")) {
2175                     error_report("MTE requested, but not supported "
2176                                  "by the guest CPU");
2177                     exit(1);
2178                 }
2179 
2180                 tag_sysmem = g_new(MemoryRegion, 1);
2181                 memory_region_init(tag_sysmem, OBJECT(machine),
2182                                    "tag-memory", UINT64_MAX / 32);
2183 
2184                 if (vms->secure) {
2185                     secure_tag_sysmem = g_new(MemoryRegion, 1);
2186                     memory_region_init(secure_tag_sysmem, OBJECT(machine),
2187                                        "secure-tag-memory", UINT64_MAX / 32);
2188 
2189                     /* As with ram, secure-tag takes precedence over tag.  */
2190                     memory_region_add_subregion_overlap(secure_tag_sysmem, 0,
2191                                                         tag_sysmem, -1);
2192                 }
2193             }
2194 
2195             object_property_set_link(cpuobj, "tag-memory", OBJECT(tag_sysmem),
2196                                      &error_abort);
2197             if (vms->secure) {
2198                 object_property_set_link(cpuobj, "secure-tag-memory",
2199                                          OBJECT(secure_tag_sysmem),
2200                                          &error_abort);
2201             }
2202         }
2203 
2204         qdev_realize(DEVICE(cpuobj), NULL, &error_fatal);
2205         object_unref(cpuobj);
2206     }
2207     fdt_add_timer_nodes(vms);
2208     fdt_add_cpu_nodes(vms);
2209 
2210     memory_region_add_subregion(sysmem, vms->memmap[VIRT_MEM].base,
2211                                 machine->ram);
2212     if (machine->device_memory) {
2213         memory_region_add_subregion(sysmem, machine->device_memory->base,
2214                                     &machine->device_memory->mr);
2215     }
2216 
2217     virt_flash_fdt(vms, sysmem, secure_sysmem ?: sysmem);
2218 
2219     create_gic(vms, sysmem);
2220 
2221     virt_cpu_post_init(vms, sysmem);
2222 
2223     fdt_add_pmu_nodes(vms);
2224 
2225     create_uart(vms, VIRT_UART, sysmem, serial_hd(0));
2226 
2227     if (vms->secure) {
2228         create_secure_ram(vms, secure_sysmem, secure_tag_sysmem);
2229         create_uart(vms, VIRT_SECURE_UART, secure_sysmem, serial_hd(1));
2230     }
2231 
2232     if (tag_sysmem) {
2233         create_tag_ram(tag_sysmem, vms->memmap[VIRT_MEM].base,
2234                        machine->ram_size, "mach-virt.tag");
2235     }
2236 
2237     vms->highmem_ecam &= (!firmware_loaded || aarch64);
2238 
2239     create_rtc(vms);
2240 
2241     create_pcie(vms);
2242 
2243     if (has_ged && aarch64 && firmware_loaded && virt_is_acpi_enabled(vms)) {
2244         vms->acpi_dev = create_acpi_ged(vms);
2245     } else {
2246         create_gpio_devices(vms, VIRT_GPIO, sysmem);
2247     }
2248 
2249     if (vms->secure && !vmc->no_secure_gpio) {
2250         create_gpio_devices(vms, VIRT_SECURE_GPIO, secure_sysmem);
2251     }
2252 
2253      /* connect powerdown request */
2254      vms->powerdown_notifier.notify = virt_powerdown_req;
2255      qemu_register_powerdown_notifier(&vms->powerdown_notifier);
2256 
2257     /* Create mmio transports, so the user can create virtio backends
2258      * (which will be automatically plugged in to the transports). If
2259      * no backend is created the transport will just sit harmlessly idle.
2260      */
2261     create_virtio_devices(vms);
2262 
2263     vms->fw_cfg = create_fw_cfg(vms, &address_space_memory);
2264     rom_set_fw(vms->fw_cfg);
2265 
2266     create_platform_bus(vms);
2267 
2268     if (machine->nvdimms_state->is_enabled) {
2269         const struct AcpiGenericAddress arm_virt_nvdimm_acpi_dsmio = {
2270             .space_id = AML_AS_SYSTEM_MEMORY,
2271             .address = vms->memmap[VIRT_NVDIMM_ACPI].base,
2272             .bit_width = NVDIMM_ACPI_IO_LEN << 3
2273         };
2274 
2275         nvdimm_init_acpi_state(machine->nvdimms_state, sysmem,
2276                                arm_virt_nvdimm_acpi_dsmio,
2277                                vms->fw_cfg, OBJECT(vms));
2278     }
2279 
2280     vms->bootinfo.ram_size = machine->ram_size;
2281     vms->bootinfo.board_id = -1;
2282     vms->bootinfo.loader_start = vms->memmap[VIRT_MEM].base;
2283     vms->bootinfo.get_dtb = machvirt_dtb;
2284     vms->bootinfo.skip_dtb_autoload = true;
2285     vms->bootinfo.firmware_loaded = firmware_loaded;
2286     vms->bootinfo.psci_conduit = vms->psci_conduit;
2287     arm_load_kernel(ARM_CPU(first_cpu), machine, &vms->bootinfo);
2288 
2289     vms->machine_done.notify = virt_machine_done;
2290     qemu_add_machine_init_done_notifier(&vms->machine_done);
2291 }
2292 
2293 static bool virt_get_secure(Object *obj, Error **errp)
2294 {
2295     VirtMachineState *vms = VIRT_MACHINE(obj);
2296 
2297     return vms->secure;
2298 }
2299 
2300 static void virt_set_secure(Object *obj, bool value, Error **errp)
2301 {
2302     VirtMachineState *vms = VIRT_MACHINE(obj);
2303 
2304     vms->secure = value;
2305 }
2306 
2307 static bool virt_get_virt(Object *obj, Error **errp)
2308 {
2309     VirtMachineState *vms = VIRT_MACHINE(obj);
2310 
2311     return vms->virt;
2312 }
2313 
2314 static void virt_set_virt(Object *obj, bool value, Error **errp)
2315 {
2316     VirtMachineState *vms = VIRT_MACHINE(obj);
2317 
2318     vms->virt = value;
2319 }
2320 
2321 static bool virt_get_highmem(Object *obj, Error **errp)
2322 {
2323     VirtMachineState *vms = VIRT_MACHINE(obj);
2324 
2325     return vms->highmem;
2326 }
2327 
2328 static void virt_set_highmem(Object *obj, bool value, Error **errp)
2329 {
2330     VirtMachineState *vms = VIRT_MACHINE(obj);
2331 
2332     vms->highmem = value;
2333 }
2334 
2335 static bool virt_get_its(Object *obj, Error **errp)
2336 {
2337     VirtMachineState *vms = VIRT_MACHINE(obj);
2338 
2339     return vms->its;
2340 }
2341 
2342 static void virt_set_its(Object *obj, bool value, Error **errp)
2343 {
2344     VirtMachineState *vms = VIRT_MACHINE(obj);
2345 
2346     vms->its = value;
2347 }
2348 
2349 static bool virt_get_dtb_randomness(Object *obj, Error **errp)
2350 {
2351     VirtMachineState *vms = VIRT_MACHINE(obj);
2352 
2353     return vms->dtb_randomness;
2354 }
2355 
2356 static void virt_set_dtb_randomness(Object *obj, bool value, Error **errp)
2357 {
2358     VirtMachineState *vms = VIRT_MACHINE(obj);
2359 
2360     vms->dtb_randomness = value;
2361 }
2362 
2363 static char *virt_get_oem_id(Object *obj, Error **errp)
2364 {
2365     VirtMachineState *vms = VIRT_MACHINE(obj);
2366 
2367     return g_strdup(vms->oem_id);
2368 }
2369 
2370 static void virt_set_oem_id(Object *obj, const char *value, Error **errp)
2371 {
2372     VirtMachineState *vms = VIRT_MACHINE(obj);
2373     size_t len = strlen(value);
2374 
2375     if (len > 6) {
2376         error_setg(errp,
2377                    "User specified oem-id value is bigger than 6 bytes in size");
2378         return;
2379     }
2380 
2381     strncpy(vms->oem_id, value, 6);
2382 }
2383 
2384 static char *virt_get_oem_table_id(Object *obj, Error **errp)
2385 {
2386     VirtMachineState *vms = VIRT_MACHINE(obj);
2387 
2388     return g_strdup(vms->oem_table_id);
2389 }
2390 
2391 static void virt_set_oem_table_id(Object *obj, const char *value,
2392                                   Error **errp)
2393 {
2394     VirtMachineState *vms = VIRT_MACHINE(obj);
2395     size_t len = strlen(value);
2396 
2397     if (len > 8) {
2398         error_setg(errp,
2399                    "User specified oem-table-id value is bigger than 8 bytes in size");
2400         return;
2401     }
2402     strncpy(vms->oem_table_id, value, 8);
2403 }
2404 
2405 
2406 bool virt_is_acpi_enabled(VirtMachineState *vms)
2407 {
2408     if (vms->acpi == ON_OFF_AUTO_OFF) {
2409         return false;
2410     }
2411     return true;
2412 }
2413 
2414 static void virt_get_acpi(Object *obj, Visitor *v, const char *name,
2415                           void *opaque, Error **errp)
2416 {
2417     VirtMachineState *vms = VIRT_MACHINE(obj);
2418     OnOffAuto acpi = vms->acpi;
2419 
2420     visit_type_OnOffAuto(v, name, &acpi, errp);
2421 }
2422 
2423 static void virt_set_acpi(Object *obj, Visitor *v, const char *name,
2424                           void *opaque, Error **errp)
2425 {
2426     VirtMachineState *vms = VIRT_MACHINE(obj);
2427 
2428     visit_type_OnOffAuto(v, name, &vms->acpi, errp);
2429 }
2430 
2431 static bool virt_get_ras(Object *obj, Error **errp)
2432 {
2433     VirtMachineState *vms = VIRT_MACHINE(obj);
2434 
2435     return vms->ras;
2436 }
2437 
2438 static void virt_set_ras(Object *obj, bool value, Error **errp)
2439 {
2440     VirtMachineState *vms = VIRT_MACHINE(obj);
2441 
2442     vms->ras = value;
2443 }
2444 
2445 static bool virt_get_mte(Object *obj, Error **errp)
2446 {
2447     VirtMachineState *vms = VIRT_MACHINE(obj);
2448 
2449     return vms->mte;
2450 }
2451 
2452 static void virt_set_mte(Object *obj, bool value, Error **errp)
2453 {
2454     VirtMachineState *vms = VIRT_MACHINE(obj);
2455 
2456     vms->mte = value;
2457 }
2458 
2459 static char *virt_get_gic_version(Object *obj, Error **errp)
2460 {
2461     VirtMachineState *vms = VIRT_MACHINE(obj);
2462     const char *val;
2463 
2464     switch (vms->gic_version) {
2465     case VIRT_GIC_VERSION_4:
2466         val = "4";
2467         break;
2468     case VIRT_GIC_VERSION_3:
2469         val = "3";
2470         break;
2471     default:
2472         val = "2";
2473         break;
2474     }
2475     return g_strdup(val);
2476 }
2477 
2478 static void virt_set_gic_version(Object *obj, const char *value, Error **errp)
2479 {
2480     VirtMachineState *vms = VIRT_MACHINE(obj);
2481 
2482     if (!strcmp(value, "4")) {
2483         vms->gic_version = VIRT_GIC_VERSION_4;
2484     } else if (!strcmp(value, "3")) {
2485         vms->gic_version = VIRT_GIC_VERSION_3;
2486     } else if (!strcmp(value, "2")) {
2487         vms->gic_version = VIRT_GIC_VERSION_2;
2488     } else if (!strcmp(value, "host")) {
2489         vms->gic_version = VIRT_GIC_VERSION_HOST; /* Will probe later */
2490     } else if (!strcmp(value, "max")) {
2491         vms->gic_version = VIRT_GIC_VERSION_MAX; /* Will probe later */
2492     } else {
2493         error_setg(errp, "Invalid gic-version value");
2494         error_append_hint(errp, "Valid values are 3, 2, host, max.\n");
2495     }
2496 }
2497 
2498 static char *virt_get_iommu(Object *obj, Error **errp)
2499 {
2500     VirtMachineState *vms = VIRT_MACHINE(obj);
2501 
2502     switch (vms->iommu) {
2503     case VIRT_IOMMU_NONE:
2504         return g_strdup("none");
2505     case VIRT_IOMMU_SMMUV3:
2506         return g_strdup("smmuv3");
2507     default:
2508         g_assert_not_reached();
2509     }
2510 }
2511 
2512 static void virt_set_iommu(Object *obj, const char *value, Error **errp)
2513 {
2514     VirtMachineState *vms = VIRT_MACHINE(obj);
2515 
2516     if (!strcmp(value, "smmuv3")) {
2517         vms->iommu = VIRT_IOMMU_SMMUV3;
2518     } else if (!strcmp(value, "none")) {
2519         vms->iommu = VIRT_IOMMU_NONE;
2520     } else {
2521         error_setg(errp, "Invalid iommu value");
2522         error_append_hint(errp, "Valid values are none, smmuv3.\n");
2523     }
2524 }
2525 
2526 static bool virt_get_default_bus_bypass_iommu(Object *obj, Error **errp)
2527 {
2528     VirtMachineState *vms = VIRT_MACHINE(obj);
2529 
2530     return vms->default_bus_bypass_iommu;
2531 }
2532 
2533 static void virt_set_default_bus_bypass_iommu(Object *obj, bool value,
2534                                               Error **errp)
2535 {
2536     VirtMachineState *vms = VIRT_MACHINE(obj);
2537 
2538     vms->default_bus_bypass_iommu = value;
2539 }
2540 
2541 static CpuInstanceProperties
2542 virt_cpu_index_to_props(MachineState *ms, unsigned cpu_index)
2543 {
2544     MachineClass *mc = MACHINE_GET_CLASS(ms);
2545     const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(ms);
2546 
2547     assert(cpu_index < possible_cpus->len);
2548     return possible_cpus->cpus[cpu_index].props;
2549 }
2550 
2551 static int64_t virt_get_default_cpu_node_id(const MachineState *ms, int idx)
2552 {
2553     int64_t socket_id = ms->possible_cpus->cpus[idx].props.socket_id;
2554 
2555     return socket_id % ms->numa_state->num_nodes;
2556 }
2557 
2558 static const CPUArchIdList *virt_possible_cpu_arch_ids(MachineState *ms)
2559 {
2560     int n;
2561     unsigned int max_cpus = ms->smp.max_cpus;
2562     VirtMachineState *vms = VIRT_MACHINE(ms);
2563     MachineClass *mc = MACHINE_GET_CLASS(vms);
2564 
2565     if (ms->possible_cpus) {
2566         assert(ms->possible_cpus->len == max_cpus);
2567         return ms->possible_cpus;
2568     }
2569 
2570     ms->possible_cpus = g_malloc0(sizeof(CPUArchIdList) +
2571                                   sizeof(CPUArchId) * max_cpus);
2572     ms->possible_cpus->len = max_cpus;
2573     for (n = 0; n < ms->possible_cpus->len; n++) {
2574         ms->possible_cpus->cpus[n].type = ms->cpu_type;
2575         ms->possible_cpus->cpus[n].arch_id =
2576             virt_cpu_mp_affinity(vms, n);
2577 
2578         assert(!mc->smp_props.dies_supported);
2579         ms->possible_cpus->cpus[n].props.has_socket_id = true;
2580         ms->possible_cpus->cpus[n].props.socket_id =
2581             n / (ms->smp.clusters * ms->smp.cores * ms->smp.threads);
2582         ms->possible_cpus->cpus[n].props.has_cluster_id = true;
2583         ms->possible_cpus->cpus[n].props.cluster_id =
2584             (n / (ms->smp.cores * ms->smp.threads)) % ms->smp.clusters;
2585         ms->possible_cpus->cpus[n].props.has_core_id = true;
2586         ms->possible_cpus->cpus[n].props.core_id =
2587             (n / ms->smp.threads) % ms->smp.cores;
2588         ms->possible_cpus->cpus[n].props.has_thread_id = true;
2589         ms->possible_cpus->cpus[n].props.thread_id =
2590             n % ms->smp.threads;
2591     }
2592     return ms->possible_cpus;
2593 }
2594 
2595 static void virt_memory_pre_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
2596                                  Error **errp)
2597 {
2598     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2599     const MachineState *ms = MACHINE(hotplug_dev);
2600     const bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2601 
2602     if (!vms->acpi_dev) {
2603         error_setg(errp,
2604                    "memory hotplug is not enabled: missing acpi-ged device");
2605         return;
2606     }
2607 
2608     if (vms->mte) {
2609         error_setg(errp, "memory hotplug is not enabled: MTE is enabled");
2610         return;
2611     }
2612 
2613     if (is_nvdimm && !ms->nvdimms_state->is_enabled) {
2614         error_setg(errp, "nvdimm is not enabled: add 'nvdimm=on' to '-M'");
2615         return;
2616     }
2617 
2618     pc_dimm_pre_plug(PC_DIMM(dev), MACHINE(hotplug_dev), NULL, errp);
2619 }
2620 
2621 static void virt_memory_plug(HotplugHandler *hotplug_dev,
2622                              DeviceState *dev, Error **errp)
2623 {
2624     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2625     MachineState *ms = MACHINE(hotplug_dev);
2626     bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2627 
2628     pc_dimm_plug(PC_DIMM(dev), MACHINE(vms));
2629 
2630     if (is_nvdimm) {
2631         nvdimm_plug(ms->nvdimms_state);
2632     }
2633 
2634     hotplug_handler_plug(HOTPLUG_HANDLER(vms->acpi_dev),
2635                          dev, &error_abort);
2636 }
2637 
2638 static void virt_virtio_md_pci_pre_plug(HotplugHandler *hotplug_dev,
2639                                         DeviceState *dev, Error **errp)
2640 {
2641     HotplugHandler *hotplug_dev2 = qdev_get_bus_hotplug_handler(dev);
2642     Error *local_err = NULL;
2643 
2644     if (!hotplug_dev2 && dev->hotplugged) {
2645         /*
2646          * Without a bus hotplug handler, we cannot control the plug/unplug
2647          * order. We should never reach this point when hotplugging on ARM.
2648          * However, it's nice to add a safety net, similar to what we have
2649          * on x86.
2650          */
2651         error_setg(errp, "hotplug of virtio based memory devices not supported"
2652                    " on this bus.");
2653         return;
2654     }
2655     /*
2656      * First, see if we can plug this memory device at all. If that
2657      * succeeds, branch of to the actual hotplug handler.
2658      */
2659     memory_device_pre_plug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev), NULL,
2660                            &local_err);
2661     if (!local_err && hotplug_dev2) {
2662         hotplug_handler_pre_plug(hotplug_dev2, dev, &local_err);
2663     }
2664     error_propagate(errp, local_err);
2665 }
2666 
2667 static void virt_virtio_md_pci_plug(HotplugHandler *hotplug_dev,
2668                                     DeviceState *dev, Error **errp)
2669 {
2670     HotplugHandler *hotplug_dev2 = qdev_get_bus_hotplug_handler(dev);
2671     Error *local_err = NULL;
2672 
2673     /*
2674      * Plug the memory device first and then branch off to the actual
2675      * hotplug handler. If that one fails, we can easily undo the memory
2676      * device bits.
2677      */
2678     memory_device_plug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev));
2679     if (hotplug_dev2) {
2680         hotplug_handler_plug(hotplug_dev2, dev, &local_err);
2681         if (local_err) {
2682             memory_device_unplug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev));
2683         }
2684     }
2685     error_propagate(errp, local_err);
2686 }
2687 
2688 static void virt_virtio_md_pci_unplug_request(HotplugHandler *hotplug_dev,
2689                                               DeviceState *dev, Error **errp)
2690 {
2691     /* We don't support hot unplug of virtio based memory devices */
2692     error_setg(errp, "virtio based memory devices cannot be unplugged.");
2693 }
2694 
2695 
2696 static void virt_machine_device_pre_plug_cb(HotplugHandler *hotplug_dev,
2697                                             DeviceState *dev, Error **errp)
2698 {
2699     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2700 
2701     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2702         virt_memory_pre_plug(hotplug_dev, dev, errp);
2703     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI)) {
2704         virt_virtio_md_pci_pre_plug(hotplug_dev, dev, errp);
2705     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2706         hwaddr db_start = 0, db_end = 0;
2707         char *resv_prop_str;
2708 
2709         if (vms->iommu != VIRT_IOMMU_NONE) {
2710             error_setg(errp, "virt machine does not support multiple IOMMUs");
2711             return;
2712         }
2713 
2714         switch (vms->msi_controller) {
2715         case VIRT_MSI_CTRL_NONE:
2716             return;
2717         case VIRT_MSI_CTRL_ITS:
2718             /* GITS_TRANSLATER page */
2719             db_start = base_memmap[VIRT_GIC_ITS].base + 0x10000;
2720             db_end = base_memmap[VIRT_GIC_ITS].base +
2721                      base_memmap[VIRT_GIC_ITS].size - 1;
2722             break;
2723         case VIRT_MSI_CTRL_GICV2M:
2724             /* MSI_SETSPI_NS page */
2725             db_start = base_memmap[VIRT_GIC_V2M].base;
2726             db_end = db_start + base_memmap[VIRT_GIC_V2M].size - 1;
2727             break;
2728         }
2729         resv_prop_str = g_strdup_printf("0x%"PRIx64":0x%"PRIx64":%u",
2730                                         db_start, db_end,
2731                                         VIRTIO_IOMMU_RESV_MEM_T_MSI);
2732 
2733         object_property_set_uint(OBJECT(dev), "len-reserved-regions", 1, errp);
2734         object_property_set_str(OBJECT(dev), "reserved-regions[0]",
2735                                 resv_prop_str, errp);
2736         g_free(resv_prop_str);
2737     }
2738 }
2739 
2740 static void virt_machine_device_plug_cb(HotplugHandler *hotplug_dev,
2741                                         DeviceState *dev, Error **errp)
2742 {
2743     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2744 
2745     if (vms->platform_bus_dev) {
2746         MachineClass *mc = MACHINE_GET_CLASS(vms);
2747 
2748         if (device_is_dynamic_sysbus(mc, dev)) {
2749             platform_bus_link_device(PLATFORM_BUS_DEVICE(vms->platform_bus_dev),
2750                                      SYS_BUS_DEVICE(dev));
2751         }
2752     }
2753     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2754         virt_memory_plug(hotplug_dev, dev, errp);
2755     }
2756 
2757     if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI)) {
2758         virt_virtio_md_pci_plug(hotplug_dev, dev, errp);
2759     }
2760 
2761     if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2762         PCIDevice *pdev = PCI_DEVICE(dev);
2763 
2764         vms->iommu = VIRT_IOMMU_VIRTIO;
2765         vms->virtio_iommu_bdf = pci_get_bdf(pdev);
2766         create_virtio_iommu_dt_bindings(vms);
2767     }
2768 }
2769 
2770 static void virt_dimm_unplug_request(HotplugHandler *hotplug_dev,
2771                                      DeviceState *dev, Error **errp)
2772 {
2773     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2774 
2775     if (!vms->acpi_dev) {
2776         error_setg(errp,
2777                    "memory hotplug is not enabled: missing acpi-ged device");
2778         return;
2779     }
2780 
2781     if (object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM)) {
2782         error_setg(errp, "nvdimm device hot unplug is not supported yet.");
2783         return;
2784     }
2785 
2786     hotplug_handler_unplug_request(HOTPLUG_HANDLER(vms->acpi_dev), dev,
2787                                    errp);
2788 }
2789 
2790 static void virt_dimm_unplug(HotplugHandler *hotplug_dev,
2791                              DeviceState *dev, Error **errp)
2792 {
2793     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2794     Error *local_err = NULL;
2795 
2796     hotplug_handler_unplug(HOTPLUG_HANDLER(vms->acpi_dev), dev, &local_err);
2797     if (local_err) {
2798         goto out;
2799     }
2800 
2801     pc_dimm_unplug(PC_DIMM(dev), MACHINE(vms));
2802     qdev_unrealize(dev);
2803 
2804 out:
2805     error_propagate(errp, local_err);
2806 }
2807 
2808 static void virt_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
2809                                           DeviceState *dev, Error **errp)
2810 {
2811     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2812         virt_dimm_unplug_request(hotplug_dev, dev, errp);
2813     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI)) {
2814         virt_virtio_md_pci_unplug_request(hotplug_dev, dev, errp);
2815     } else {
2816         error_setg(errp, "device unplug request for unsupported device"
2817                    " type: %s", object_get_typename(OBJECT(dev)));
2818     }
2819 }
2820 
2821 static void virt_machine_device_unplug_cb(HotplugHandler *hotplug_dev,
2822                                           DeviceState *dev, Error **errp)
2823 {
2824     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2825         virt_dimm_unplug(hotplug_dev, dev, errp);
2826     } else {
2827         error_setg(errp, "virt: device unplug for unsupported device"
2828                    " type: %s", object_get_typename(OBJECT(dev)));
2829     }
2830 }
2831 
2832 static HotplugHandler *virt_machine_get_hotplug_handler(MachineState *machine,
2833                                                         DeviceState *dev)
2834 {
2835     MachineClass *mc = MACHINE_GET_CLASS(machine);
2836 
2837     if (device_is_dynamic_sysbus(mc, dev) ||
2838         object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) ||
2839         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI) ||
2840         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2841         return HOTPLUG_HANDLER(machine);
2842     }
2843     return NULL;
2844 }
2845 
2846 /*
2847  * for arm64 kvm_type [7-0] encodes the requested number of bits
2848  * in the IPA address space
2849  */
2850 static int virt_kvm_type(MachineState *ms, const char *type_str)
2851 {
2852     VirtMachineState *vms = VIRT_MACHINE(ms);
2853     int max_vm_pa_size, requested_pa_size;
2854     bool fixed_ipa;
2855 
2856     max_vm_pa_size = kvm_arm_get_max_vm_ipa_size(ms, &fixed_ipa);
2857 
2858     /* we freeze the memory map to compute the highest gpa */
2859     virt_set_memmap(vms, max_vm_pa_size);
2860 
2861     requested_pa_size = 64 - clz64(vms->highest_gpa);
2862 
2863     /*
2864      * KVM requires the IPA size to be at least 32 bits.
2865      */
2866     if (requested_pa_size < 32) {
2867         requested_pa_size = 32;
2868     }
2869 
2870     if (requested_pa_size > max_vm_pa_size) {
2871         error_report("-m and ,maxmem option values "
2872                      "require an IPA range (%d bits) larger than "
2873                      "the one supported by the host (%d bits)",
2874                      requested_pa_size, max_vm_pa_size);
2875         exit(1);
2876     }
2877     /*
2878      * We return the requested PA log size, unless KVM only supports
2879      * the implicit legacy 40b IPA setting, in which case the kvm_type
2880      * must be 0.
2881      */
2882     return fixed_ipa ? 0 : requested_pa_size;
2883 }
2884 
2885 static void virt_machine_class_init(ObjectClass *oc, void *data)
2886 {
2887     MachineClass *mc = MACHINE_CLASS(oc);
2888     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
2889 
2890     mc->init = machvirt_init;
2891     /* Start with max_cpus set to 512, which is the maximum supported by KVM.
2892      * The value may be reduced later when we have more information about the
2893      * configuration of the particular instance.
2894      */
2895     mc->max_cpus = 512;
2896     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_CALXEDA_XGMAC);
2897     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_AMD_XGBE);
2898     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_RAMFB_DEVICE);
2899     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_PLATFORM);
2900 #ifdef CONFIG_TPM
2901     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_TPM_TIS_SYSBUS);
2902 #endif
2903     mc->block_default_type = IF_VIRTIO;
2904     mc->no_cdrom = 1;
2905     mc->pci_allow_0_address = true;
2906     /* We know we will never create a pre-ARMv7 CPU which needs 1K pages */
2907     mc->minimum_page_bits = 12;
2908     mc->possible_cpu_arch_ids = virt_possible_cpu_arch_ids;
2909     mc->cpu_index_to_instance_props = virt_cpu_index_to_props;
2910     mc->default_cpu_type = ARM_CPU_TYPE_NAME("cortex-a15");
2911     mc->get_default_cpu_node_id = virt_get_default_cpu_node_id;
2912     mc->kvm_type = virt_kvm_type;
2913     assert(!mc->get_hotplug_handler);
2914     mc->get_hotplug_handler = virt_machine_get_hotplug_handler;
2915     hc->pre_plug = virt_machine_device_pre_plug_cb;
2916     hc->plug = virt_machine_device_plug_cb;
2917     hc->unplug_request = virt_machine_device_unplug_request_cb;
2918     hc->unplug = virt_machine_device_unplug_cb;
2919     mc->nvdimm_supported = true;
2920     mc->smp_props.clusters_supported = true;
2921     mc->auto_enable_numa_with_memhp = true;
2922     mc->auto_enable_numa_with_memdev = true;
2923     mc->default_ram_id = "mach-virt.ram";
2924 
2925     object_class_property_add(oc, "acpi", "OnOffAuto",
2926         virt_get_acpi, virt_set_acpi,
2927         NULL, NULL);
2928     object_class_property_set_description(oc, "acpi",
2929         "Enable ACPI");
2930     object_class_property_add_bool(oc, "secure", virt_get_secure,
2931                                    virt_set_secure);
2932     object_class_property_set_description(oc, "secure",
2933                                                 "Set on/off to enable/disable the ARM "
2934                                                 "Security Extensions (TrustZone)");
2935 
2936     object_class_property_add_bool(oc, "virtualization", virt_get_virt,
2937                                    virt_set_virt);
2938     object_class_property_set_description(oc, "virtualization",
2939                                           "Set on/off to enable/disable emulating a "
2940                                           "guest CPU which implements the ARM "
2941                                           "Virtualization Extensions");
2942 
2943     object_class_property_add_bool(oc, "highmem", virt_get_highmem,
2944                                    virt_set_highmem);
2945     object_class_property_set_description(oc, "highmem",
2946                                           "Set on/off to enable/disable using "
2947                                           "physical address space above 32 bits");
2948 
2949     object_class_property_add_str(oc, "gic-version", virt_get_gic_version,
2950                                   virt_set_gic_version);
2951     object_class_property_set_description(oc, "gic-version",
2952                                           "Set GIC version. "
2953                                           "Valid values are 2, 3, 4, host and max");
2954 
2955     object_class_property_add_str(oc, "iommu", virt_get_iommu, virt_set_iommu);
2956     object_class_property_set_description(oc, "iommu",
2957                                           "Set the IOMMU type. "
2958                                           "Valid values are none and smmuv3");
2959 
2960     object_class_property_add_bool(oc, "default-bus-bypass-iommu",
2961                                    virt_get_default_bus_bypass_iommu,
2962                                    virt_set_default_bus_bypass_iommu);
2963     object_class_property_set_description(oc, "default-bus-bypass-iommu",
2964                                           "Set on/off to enable/disable "
2965                                           "bypass_iommu for default root bus");
2966 
2967     object_class_property_add_bool(oc, "ras", virt_get_ras,
2968                                    virt_set_ras);
2969     object_class_property_set_description(oc, "ras",
2970                                           "Set on/off to enable/disable reporting host memory errors "
2971                                           "to a KVM guest using ACPI and guest external abort exceptions");
2972 
2973     object_class_property_add_bool(oc, "mte", virt_get_mte, virt_set_mte);
2974     object_class_property_set_description(oc, "mte",
2975                                           "Set on/off to enable/disable emulating a "
2976                                           "guest CPU which implements the ARM "
2977                                           "Memory Tagging Extension");
2978 
2979     object_class_property_add_bool(oc, "its", virt_get_its,
2980                                    virt_set_its);
2981     object_class_property_set_description(oc, "its",
2982                                           "Set on/off to enable/disable "
2983                                           "ITS instantiation");
2984 
2985     object_class_property_add_bool(oc, "dtb-randomness",
2986                                    virt_get_dtb_randomness,
2987                                    virt_set_dtb_randomness);
2988     object_class_property_set_description(oc, "dtb-randomness",
2989                                           "Set off to disable passing random or "
2990                                           "non-deterministic dtb nodes to guest");
2991 
2992     object_class_property_add_bool(oc, "dtb-kaslr-seed",
2993                                    virt_get_dtb_randomness,
2994                                    virt_set_dtb_randomness);
2995     object_class_property_set_description(oc, "dtb-kaslr-seed",
2996                                           "Deprecated synonym of dtb-randomness");
2997 
2998     object_class_property_add_str(oc, "x-oem-id",
2999                                   virt_get_oem_id,
3000                                   virt_set_oem_id);
3001     object_class_property_set_description(oc, "x-oem-id",
3002                                           "Override the default value of field OEMID "
3003                                           "in ACPI table header."
3004                                           "The string may be up to 6 bytes in size");
3005 
3006 
3007     object_class_property_add_str(oc, "x-oem-table-id",
3008                                   virt_get_oem_table_id,
3009                                   virt_set_oem_table_id);
3010     object_class_property_set_description(oc, "x-oem-table-id",
3011                                           "Override the default value of field OEM Table ID "
3012                                           "in ACPI table header."
3013                                           "The string may be up to 8 bytes in size");
3014 
3015 }
3016 
3017 static void virt_instance_init(Object *obj)
3018 {
3019     VirtMachineState *vms = VIRT_MACHINE(obj);
3020     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
3021 
3022     /* EL3 is disabled by default on virt: this makes us consistent
3023      * between KVM and TCG for this board, and it also allows us to
3024      * boot UEFI blobs which assume no TrustZone support.
3025      */
3026     vms->secure = false;
3027 
3028     /* EL2 is also disabled by default, for similar reasons */
3029     vms->virt = false;
3030 
3031     /* High memory is enabled by default */
3032     vms->highmem = true;
3033     vms->gic_version = VIRT_GIC_VERSION_NOSEL;
3034 
3035     vms->highmem_ecam = !vmc->no_highmem_ecam;
3036     vms->highmem_mmio = true;
3037     vms->highmem_redists = true;
3038 
3039     if (vmc->no_its) {
3040         vms->its = false;
3041     } else {
3042         /* Default allows ITS instantiation */
3043         vms->its = true;
3044 
3045         if (vmc->no_tcg_its) {
3046             vms->tcg_its = false;
3047         } else {
3048             vms->tcg_its = true;
3049         }
3050     }
3051 
3052     /* Default disallows iommu instantiation */
3053     vms->iommu = VIRT_IOMMU_NONE;
3054 
3055     /* The default root bus is attached to iommu by default */
3056     vms->default_bus_bypass_iommu = false;
3057 
3058     /* Default disallows RAS instantiation */
3059     vms->ras = false;
3060 
3061     /* MTE is disabled by default.  */
3062     vms->mte = false;
3063 
3064     /* Supply kaslr-seed and rng-seed by default */
3065     vms->dtb_randomness = true;
3066 
3067     vms->irqmap = a15irqmap;
3068 
3069     virt_flash_create(vms);
3070 
3071     vms->oem_id = g_strndup(ACPI_BUILD_APPNAME6, 6);
3072     vms->oem_table_id = g_strndup(ACPI_BUILD_APPNAME8, 8);
3073 }
3074 
3075 static const TypeInfo virt_machine_info = {
3076     .name          = TYPE_VIRT_MACHINE,
3077     .parent        = TYPE_MACHINE,
3078     .abstract      = true,
3079     .instance_size = sizeof(VirtMachineState),
3080     .class_size    = sizeof(VirtMachineClass),
3081     .class_init    = virt_machine_class_init,
3082     .instance_init = virt_instance_init,
3083     .interfaces = (InterfaceInfo[]) {
3084          { TYPE_HOTPLUG_HANDLER },
3085          { }
3086     },
3087 };
3088 
3089 static void machvirt_machine_init(void)
3090 {
3091     type_register_static(&virt_machine_info);
3092 }
3093 type_init(machvirt_machine_init);
3094 
3095 static void virt_machine_7_2_options(MachineClass *mc)
3096 {
3097 }
3098 DEFINE_VIRT_MACHINE_AS_LATEST(7, 2)
3099 
3100 static void virt_machine_7_1_options(MachineClass *mc)
3101 {
3102     virt_machine_7_2_options(mc);
3103     compat_props_add(mc->compat_props, hw_compat_7_1, hw_compat_7_1_len);
3104 }
3105 DEFINE_VIRT_MACHINE(7, 1)
3106 
3107 static void virt_machine_7_0_options(MachineClass *mc)
3108 {
3109     virt_machine_7_1_options(mc);
3110     compat_props_add(mc->compat_props, hw_compat_7_0, hw_compat_7_0_len);
3111 }
3112 DEFINE_VIRT_MACHINE(7, 0)
3113 
3114 static void virt_machine_6_2_options(MachineClass *mc)
3115 {
3116     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3117 
3118     virt_machine_7_0_options(mc);
3119     compat_props_add(mc->compat_props, hw_compat_6_2, hw_compat_6_2_len);
3120     vmc->no_tcg_lpa2 = true;
3121 }
3122 DEFINE_VIRT_MACHINE(6, 2)
3123 
3124 static void virt_machine_6_1_options(MachineClass *mc)
3125 {
3126     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3127 
3128     virt_machine_6_2_options(mc);
3129     compat_props_add(mc->compat_props, hw_compat_6_1, hw_compat_6_1_len);
3130     mc->smp_props.prefer_sockets = true;
3131     vmc->no_cpu_topology = true;
3132 
3133     /* qemu ITS was introduced with 6.2 */
3134     vmc->no_tcg_its = true;
3135 }
3136 DEFINE_VIRT_MACHINE(6, 1)
3137 
3138 static void virt_machine_6_0_options(MachineClass *mc)
3139 {
3140     virt_machine_6_1_options(mc);
3141     compat_props_add(mc->compat_props, hw_compat_6_0, hw_compat_6_0_len);
3142 }
3143 DEFINE_VIRT_MACHINE(6, 0)
3144 
3145 static void virt_machine_5_2_options(MachineClass *mc)
3146 {
3147     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3148 
3149     virt_machine_6_0_options(mc);
3150     compat_props_add(mc->compat_props, hw_compat_5_2, hw_compat_5_2_len);
3151     vmc->no_secure_gpio = true;
3152 }
3153 DEFINE_VIRT_MACHINE(5, 2)
3154 
3155 static void virt_machine_5_1_options(MachineClass *mc)
3156 {
3157     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3158 
3159     virt_machine_5_2_options(mc);
3160     compat_props_add(mc->compat_props, hw_compat_5_1, hw_compat_5_1_len);
3161     vmc->no_kvm_steal_time = true;
3162 }
3163 DEFINE_VIRT_MACHINE(5, 1)
3164 
3165 static void virt_machine_5_0_options(MachineClass *mc)
3166 {
3167     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3168 
3169     virt_machine_5_1_options(mc);
3170     compat_props_add(mc->compat_props, hw_compat_5_0, hw_compat_5_0_len);
3171     mc->numa_mem_supported = true;
3172     vmc->acpi_expose_flash = true;
3173     mc->auto_enable_numa_with_memdev = false;
3174 }
3175 DEFINE_VIRT_MACHINE(5, 0)
3176 
3177 static void virt_machine_4_2_options(MachineClass *mc)
3178 {
3179     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3180 
3181     virt_machine_5_0_options(mc);
3182     compat_props_add(mc->compat_props, hw_compat_4_2, hw_compat_4_2_len);
3183     vmc->kvm_no_adjvtime = true;
3184 }
3185 DEFINE_VIRT_MACHINE(4, 2)
3186 
3187 static void virt_machine_4_1_options(MachineClass *mc)
3188 {
3189     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3190 
3191     virt_machine_4_2_options(mc);
3192     compat_props_add(mc->compat_props, hw_compat_4_1, hw_compat_4_1_len);
3193     vmc->no_ged = true;
3194     mc->auto_enable_numa_with_memhp = false;
3195 }
3196 DEFINE_VIRT_MACHINE(4, 1)
3197 
3198 static void virt_machine_4_0_options(MachineClass *mc)
3199 {
3200     virt_machine_4_1_options(mc);
3201     compat_props_add(mc->compat_props, hw_compat_4_0, hw_compat_4_0_len);
3202 }
3203 DEFINE_VIRT_MACHINE(4, 0)
3204 
3205 static void virt_machine_3_1_options(MachineClass *mc)
3206 {
3207     virt_machine_4_0_options(mc);
3208     compat_props_add(mc->compat_props, hw_compat_3_1, hw_compat_3_1_len);
3209 }
3210 DEFINE_VIRT_MACHINE(3, 1)
3211 
3212 static void virt_machine_3_0_options(MachineClass *mc)
3213 {
3214     virt_machine_3_1_options(mc);
3215     compat_props_add(mc->compat_props, hw_compat_3_0, hw_compat_3_0_len);
3216 }
3217 DEFINE_VIRT_MACHINE(3, 0)
3218 
3219 static void virt_machine_2_12_options(MachineClass *mc)
3220 {
3221     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3222 
3223     virt_machine_3_0_options(mc);
3224     compat_props_add(mc->compat_props, hw_compat_2_12, hw_compat_2_12_len);
3225     vmc->no_highmem_ecam = true;
3226     mc->max_cpus = 255;
3227 }
3228 DEFINE_VIRT_MACHINE(2, 12)
3229 
3230 static void virt_machine_2_11_options(MachineClass *mc)
3231 {
3232     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3233 
3234     virt_machine_2_12_options(mc);
3235     compat_props_add(mc->compat_props, hw_compat_2_11, hw_compat_2_11_len);
3236     vmc->smbios_old_sys_ver = true;
3237 }
3238 DEFINE_VIRT_MACHINE(2, 11)
3239 
3240 static void virt_machine_2_10_options(MachineClass *mc)
3241 {
3242     virt_machine_2_11_options(mc);
3243     compat_props_add(mc->compat_props, hw_compat_2_10, hw_compat_2_10_len);
3244     /* before 2.11 we never faulted accesses to bad addresses */
3245     mc->ignore_memory_transaction_failures = true;
3246 }
3247 DEFINE_VIRT_MACHINE(2, 10)
3248 
3249 static void virt_machine_2_9_options(MachineClass *mc)
3250 {
3251     virt_machine_2_10_options(mc);
3252     compat_props_add(mc->compat_props, hw_compat_2_9, hw_compat_2_9_len);
3253 }
3254 DEFINE_VIRT_MACHINE(2, 9)
3255 
3256 static void virt_machine_2_8_options(MachineClass *mc)
3257 {
3258     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3259 
3260     virt_machine_2_9_options(mc);
3261     compat_props_add(mc->compat_props, hw_compat_2_8, hw_compat_2_8_len);
3262     /* For 2.8 and earlier we falsely claimed in the DT that
3263      * our timers were edge-triggered, not level-triggered.
3264      */
3265     vmc->claim_edge_triggered_timers = true;
3266 }
3267 DEFINE_VIRT_MACHINE(2, 8)
3268 
3269 static void virt_machine_2_7_options(MachineClass *mc)
3270 {
3271     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3272 
3273     virt_machine_2_8_options(mc);
3274     compat_props_add(mc->compat_props, hw_compat_2_7, hw_compat_2_7_len);
3275     /* ITS was introduced with 2.8 */
3276     vmc->no_its = true;
3277     /* Stick with 1K pages for migration compatibility */
3278     mc->minimum_page_bits = 0;
3279 }
3280 DEFINE_VIRT_MACHINE(2, 7)
3281 
3282 static void virt_machine_2_6_options(MachineClass *mc)
3283 {
3284     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3285 
3286     virt_machine_2_7_options(mc);
3287     compat_props_add(mc->compat_props, hw_compat_2_6, hw_compat_2_6_len);
3288     vmc->disallow_affinity_adjustment = true;
3289     /* Disable PMU for 2.6 as PMU support was first introduced in 2.7 */
3290     vmc->no_pmu = true;
3291 }
3292 DEFINE_VIRT_MACHINE(2, 6)
3293