xref: /openbmc/qemu/hw/arm/virt.c (revision d044b7c3)
1 /*
2  * ARM mach-virt emulation
3  *
4  * Copyright (c) 2013 Linaro Limited
5  *
6  * This program is free software; you can redistribute it and/or modify it
7  * under the terms and conditions of the GNU General Public License,
8  * version 2 or later, as published by the Free Software Foundation.
9  *
10  * This program is distributed in the hope it will be useful, but WITHOUT
11  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
12  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
13  * more details.
14  *
15  * You should have received a copy of the GNU General Public License along with
16  * this program.  If not, see <http://www.gnu.org/licenses/>.
17  *
18  * Emulate a virtual board which works by passing Linux all the information
19  * it needs about what devices are present via the device tree.
20  * There are some restrictions about what we can do here:
21  *  + we can only present devices whose Linux drivers will work based
22  *    purely on the device tree with no platform data at all
23  *  + we want to present a very stripped-down minimalist platform,
24  *    both because this reduces the security attack surface from the guest
25  *    and also because it reduces our exposure to being broken when
26  *    the kernel updates its device tree bindings and requires further
27  *    information in a device binding that we aren't providing.
28  * This is essentially the same approach kvmtool uses.
29  */
30 
31 #include "qemu/osdep.h"
32 #include "qemu/datadir.h"
33 #include "qemu/units.h"
34 #include "qemu/option.h"
35 #include "monitor/qdev.h"
36 #include "hw/sysbus.h"
37 #include "hw/arm/boot.h"
38 #include "hw/arm/primecell.h"
39 #include "hw/arm/virt.h"
40 #include "hw/block/flash.h"
41 #include "hw/vfio/vfio-calxeda-xgmac.h"
42 #include "hw/vfio/vfio-amd-xgbe.h"
43 #include "hw/display/ramfb.h"
44 #include "net/net.h"
45 #include "sysemu/device_tree.h"
46 #include "sysemu/numa.h"
47 #include "sysemu/runstate.h"
48 #include "sysemu/tpm.h"
49 #include "sysemu/tcg.h"
50 #include "sysemu/kvm.h"
51 #include "sysemu/hvf.h"
52 #include "sysemu/qtest.h"
53 #include "hw/loader.h"
54 #include "qapi/error.h"
55 #include "qemu/bitops.h"
56 #include "qemu/error-report.h"
57 #include "qemu/module.h"
58 #include "hw/pci-host/gpex.h"
59 #include "hw/virtio/virtio-pci.h"
60 #include "hw/core/sysbus-fdt.h"
61 #include "hw/platform-bus.h"
62 #include "hw/qdev-properties.h"
63 #include "hw/arm/fdt.h"
64 #include "hw/intc/arm_gic.h"
65 #include "hw/intc/arm_gicv3_common.h"
66 #include "hw/irq.h"
67 #include "kvm_arm.h"
68 #include "hw/firmware/smbios.h"
69 #include "qapi/visitor.h"
70 #include "qapi/qapi-visit-common.h"
71 #include "standard-headers/linux/input.h"
72 #include "hw/arm/smmuv3.h"
73 #include "hw/acpi/acpi.h"
74 #include "target/arm/internals.h"
75 #include "hw/mem/memory-device.h"
76 #include "hw/mem/pc-dimm.h"
77 #include "hw/mem/nvdimm.h"
78 #include "hw/acpi/generic_event_device.h"
79 #include "hw/virtio/virtio-mem-pci.h"
80 #include "hw/virtio/virtio-iommu.h"
81 #include "hw/char/pl011.h"
82 #include "qemu/guest-random.h"
83 
84 #define DEFINE_VIRT_MACHINE_LATEST(major, minor, latest) \
85     static void virt_##major##_##minor##_class_init(ObjectClass *oc, \
86                                                     void *data) \
87     { \
88         MachineClass *mc = MACHINE_CLASS(oc); \
89         virt_machine_##major##_##minor##_options(mc); \
90         mc->desc = "QEMU " # major "." # minor " ARM Virtual Machine"; \
91         if (latest) { \
92             mc->alias = "virt"; \
93         } \
94     } \
95     static const TypeInfo machvirt_##major##_##minor##_info = { \
96         .name = MACHINE_TYPE_NAME("virt-" # major "." # minor), \
97         .parent = TYPE_VIRT_MACHINE, \
98         .class_init = virt_##major##_##minor##_class_init, \
99     }; \
100     static void machvirt_machine_##major##_##minor##_init(void) \
101     { \
102         type_register_static(&machvirt_##major##_##minor##_info); \
103     } \
104     type_init(machvirt_machine_##major##_##minor##_init);
105 
106 #define DEFINE_VIRT_MACHINE_AS_LATEST(major, minor) \
107     DEFINE_VIRT_MACHINE_LATEST(major, minor, true)
108 #define DEFINE_VIRT_MACHINE(major, minor) \
109     DEFINE_VIRT_MACHINE_LATEST(major, minor, false)
110 
111 
112 /* Number of external interrupt lines to configure the GIC with */
113 #define NUM_IRQS 256
114 
115 #define PLATFORM_BUS_NUM_IRQS 64
116 
117 /* Legacy RAM limit in GB (< version 4.0) */
118 #define LEGACY_RAMLIMIT_GB 255
119 #define LEGACY_RAMLIMIT_BYTES (LEGACY_RAMLIMIT_GB * GiB)
120 
121 /* Addresses and sizes of our components.
122  * 0..128MB is space for a flash device so we can run bootrom code such as UEFI.
123  * 128MB..256MB is used for miscellaneous device I/O.
124  * 256MB..1GB is reserved for possible future PCI support (ie where the
125  * PCI memory window will go if we add a PCI host controller).
126  * 1GB and up is RAM (which may happily spill over into the
127  * high memory region beyond 4GB).
128  * This represents a compromise between how much RAM can be given to
129  * a 32 bit VM and leaving space for expansion and in particular for PCI.
130  * Note that devices should generally be placed at multiples of 0x10000,
131  * to accommodate guests using 64K pages.
132  */
133 static const MemMapEntry base_memmap[] = {
134     /* Space up to 0x8000000 is reserved for a boot ROM */
135     [VIRT_FLASH] =              {          0, 0x08000000 },
136     [VIRT_CPUPERIPHS] =         { 0x08000000, 0x00020000 },
137     /* GIC distributor and CPU interfaces sit inside the CPU peripheral space */
138     [VIRT_GIC_DIST] =           { 0x08000000, 0x00010000 },
139     [VIRT_GIC_CPU] =            { 0x08010000, 0x00010000 },
140     [VIRT_GIC_V2M] =            { 0x08020000, 0x00001000 },
141     [VIRT_GIC_HYP] =            { 0x08030000, 0x00010000 },
142     [VIRT_GIC_VCPU] =           { 0x08040000, 0x00010000 },
143     /* The space in between here is reserved for GICv3 CPU/vCPU/HYP */
144     [VIRT_GIC_ITS] =            { 0x08080000, 0x00020000 },
145     /* This redistributor space allows up to 2*64kB*123 CPUs */
146     [VIRT_GIC_REDIST] =         { 0x080A0000, 0x00F60000 },
147     [VIRT_UART] =               { 0x09000000, 0x00001000 },
148     [VIRT_RTC] =                { 0x09010000, 0x00001000 },
149     [VIRT_FW_CFG] =             { 0x09020000, 0x00000018 },
150     [VIRT_GPIO] =               { 0x09030000, 0x00001000 },
151     [VIRT_SECURE_UART] =        { 0x09040000, 0x00001000 },
152     [VIRT_SMMU] =               { 0x09050000, 0x00020000 },
153     [VIRT_PCDIMM_ACPI] =        { 0x09070000, MEMORY_HOTPLUG_IO_LEN },
154     [VIRT_ACPI_GED] =           { 0x09080000, ACPI_GED_EVT_SEL_LEN },
155     [VIRT_NVDIMM_ACPI] =        { 0x09090000, NVDIMM_ACPI_IO_LEN},
156     [VIRT_PVTIME] =             { 0x090a0000, 0x00010000 },
157     [VIRT_SECURE_GPIO] =        { 0x090b0000, 0x00001000 },
158     [VIRT_MMIO] =               { 0x0a000000, 0x00000200 },
159     /* ...repeating for a total of NUM_VIRTIO_TRANSPORTS, each of that size */
160     [VIRT_PLATFORM_BUS] =       { 0x0c000000, 0x02000000 },
161     [VIRT_SECURE_MEM] =         { 0x0e000000, 0x01000000 },
162     [VIRT_PCIE_MMIO] =          { 0x10000000, 0x2eff0000 },
163     [VIRT_PCIE_PIO] =           { 0x3eff0000, 0x00010000 },
164     [VIRT_PCIE_ECAM] =          { 0x3f000000, 0x01000000 },
165     /* Actual RAM size depends on initial RAM and device memory settings */
166     [VIRT_MEM] =                { GiB, LEGACY_RAMLIMIT_BYTES },
167 };
168 
169 /*
170  * Highmem IO Regions: This memory map is floating, located after the RAM.
171  * Each MemMapEntry base (GPA) will be dynamically computed, depending on the
172  * top of the RAM, so that its base get the same alignment as the size,
173  * ie. a 512GiB entry will be aligned on a 512GiB boundary. If there is
174  * less than 256GiB of RAM, the floating area starts at the 256GiB mark.
175  * Note the extended_memmap is sized so that it eventually also includes the
176  * base_memmap entries (VIRT_HIGH_GIC_REDIST2 index is greater than the last
177  * index of base_memmap).
178  *
179  * The memory map for these Highmem IO Regions can be in legacy or compact
180  * layout, depending on 'compact-highmem' property. With legacy layout, the
181  * PA space for one specific region is always reserved, even if the region
182  * has been disabled or doesn't fit into the PA space. However, the PA space
183  * for the region won't be reserved in these circumstances with compact layout.
184  */
185 static MemMapEntry extended_memmap[] = {
186     /* Additional 64 MB redist region (can contain up to 512 redistributors) */
187     [VIRT_HIGH_GIC_REDIST2] =   { 0x0, 64 * MiB },
188     [VIRT_HIGH_PCIE_ECAM] =     { 0x0, 256 * MiB },
189     /* Second PCIe window */
190     [VIRT_HIGH_PCIE_MMIO] =     { 0x0, 512 * GiB },
191 };
192 
193 static const int a15irqmap[] = {
194     [VIRT_UART] = 1,
195     [VIRT_RTC] = 2,
196     [VIRT_PCIE] = 3, /* ... to 6 */
197     [VIRT_GPIO] = 7,
198     [VIRT_SECURE_UART] = 8,
199     [VIRT_ACPI_GED] = 9,
200     [VIRT_MMIO] = 16, /* ...to 16 + NUM_VIRTIO_TRANSPORTS - 1 */
201     [VIRT_GIC_V2M] = 48, /* ...to 48 + NUM_GICV2M_SPIS - 1 */
202     [VIRT_SMMU] = 74,    /* ...to 74 + NUM_SMMU_IRQS - 1 */
203     [VIRT_PLATFORM_BUS] = 112, /* ...to 112 + PLATFORM_BUS_NUM_IRQS -1 */
204 };
205 
206 static const char *valid_cpus[] = {
207 #ifdef CONFIG_TCG
208     ARM_CPU_TYPE_NAME("cortex-a7"),
209 #endif
210     ARM_CPU_TYPE_NAME("cortex-a15"),
211     ARM_CPU_TYPE_NAME("cortex-a35"),
212     ARM_CPU_TYPE_NAME("cortex-a53"),
213     ARM_CPU_TYPE_NAME("cortex-a55"),
214     ARM_CPU_TYPE_NAME("cortex-a57"),
215     ARM_CPU_TYPE_NAME("cortex-a72"),
216     ARM_CPU_TYPE_NAME("cortex-a76"),
217     ARM_CPU_TYPE_NAME("a64fx"),
218     ARM_CPU_TYPE_NAME("neoverse-n1"),
219     ARM_CPU_TYPE_NAME("host"),
220     ARM_CPU_TYPE_NAME("max"),
221 };
222 
223 static bool cpu_type_valid(const char *cpu)
224 {
225     int i;
226 
227     for (i = 0; i < ARRAY_SIZE(valid_cpus); i++) {
228         if (strcmp(cpu, valid_cpus[i]) == 0) {
229             return true;
230         }
231     }
232     return false;
233 }
234 
235 static void create_randomness(MachineState *ms, const char *node)
236 {
237     struct {
238         uint64_t kaslr;
239         uint8_t rng[32];
240     } seed;
241 
242     if (qemu_guest_getrandom(&seed, sizeof(seed), NULL)) {
243         return;
244     }
245     qemu_fdt_setprop_u64(ms->fdt, node, "kaslr-seed", seed.kaslr);
246     qemu_fdt_setprop(ms->fdt, node, "rng-seed", seed.rng, sizeof(seed.rng));
247 }
248 
249 static void create_fdt(VirtMachineState *vms)
250 {
251     MachineState *ms = MACHINE(vms);
252     int nb_numa_nodes = ms->numa_state->num_nodes;
253     void *fdt = create_device_tree(&vms->fdt_size);
254 
255     if (!fdt) {
256         error_report("create_device_tree() failed");
257         exit(1);
258     }
259 
260     ms->fdt = fdt;
261 
262     /* Header */
263     qemu_fdt_setprop_string(fdt, "/", "compatible", "linux,dummy-virt");
264     qemu_fdt_setprop_cell(fdt, "/", "#address-cells", 0x2);
265     qemu_fdt_setprop_cell(fdt, "/", "#size-cells", 0x2);
266     qemu_fdt_setprop_string(fdt, "/", "model", "linux,dummy-virt");
267 
268     /* /chosen must exist for load_dtb to fill in necessary properties later */
269     qemu_fdt_add_subnode(fdt, "/chosen");
270     if (vms->dtb_randomness) {
271         create_randomness(ms, "/chosen");
272     }
273 
274     if (vms->secure) {
275         qemu_fdt_add_subnode(fdt, "/secure-chosen");
276         if (vms->dtb_randomness) {
277             create_randomness(ms, "/secure-chosen");
278         }
279     }
280 
281     /* Clock node, for the benefit of the UART. The kernel device tree
282      * binding documentation claims the PL011 node clock properties are
283      * optional but in practice if you omit them the kernel refuses to
284      * probe for the device.
285      */
286     vms->clock_phandle = qemu_fdt_alloc_phandle(fdt);
287     qemu_fdt_add_subnode(fdt, "/apb-pclk");
288     qemu_fdt_setprop_string(fdt, "/apb-pclk", "compatible", "fixed-clock");
289     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "#clock-cells", 0x0);
290     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "clock-frequency", 24000000);
291     qemu_fdt_setprop_string(fdt, "/apb-pclk", "clock-output-names",
292                                 "clk24mhz");
293     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "phandle", vms->clock_phandle);
294 
295     if (nb_numa_nodes > 0 && ms->numa_state->have_numa_distance) {
296         int size = nb_numa_nodes * nb_numa_nodes * 3 * sizeof(uint32_t);
297         uint32_t *matrix = g_malloc0(size);
298         int idx, i, j;
299 
300         for (i = 0; i < nb_numa_nodes; i++) {
301             for (j = 0; j < nb_numa_nodes; j++) {
302                 idx = (i * nb_numa_nodes + j) * 3;
303                 matrix[idx + 0] = cpu_to_be32(i);
304                 matrix[idx + 1] = cpu_to_be32(j);
305                 matrix[idx + 2] =
306                     cpu_to_be32(ms->numa_state->nodes[i].distance[j]);
307             }
308         }
309 
310         qemu_fdt_add_subnode(fdt, "/distance-map");
311         qemu_fdt_setprop_string(fdt, "/distance-map", "compatible",
312                                 "numa-distance-map-v1");
313         qemu_fdt_setprop(fdt, "/distance-map", "distance-matrix",
314                          matrix, size);
315         g_free(matrix);
316     }
317 }
318 
319 static void fdt_add_timer_nodes(const VirtMachineState *vms)
320 {
321     /* On real hardware these interrupts are level-triggered.
322      * On KVM they were edge-triggered before host kernel version 4.4,
323      * and level-triggered afterwards.
324      * On emulated QEMU they are level-triggered.
325      *
326      * Getting the DTB info about them wrong is awkward for some
327      * guest kernels:
328      *  pre-4.8 ignore the DT and leave the interrupt configured
329      *   with whatever the GIC reset value (or the bootloader) left it at
330      *  4.8 before rc6 honour the incorrect data by programming it back
331      *   into the GIC, causing problems
332      *  4.8rc6 and later ignore the DT and always write "level triggered"
333      *   into the GIC
334      *
335      * For backwards-compatibility, virt-2.8 and earlier will continue
336      * to say these are edge-triggered, but later machines will report
337      * the correct information.
338      */
339     ARMCPU *armcpu;
340     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
341     uint32_t irqflags = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
342     MachineState *ms = MACHINE(vms);
343 
344     if (vmc->claim_edge_triggered_timers) {
345         irqflags = GIC_FDT_IRQ_FLAGS_EDGE_LO_HI;
346     }
347 
348     if (vms->gic_version == VIRT_GIC_VERSION_2) {
349         irqflags = deposit32(irqflags, GIC_FDT_IRQ_PPI_CPU_START,
350                              GIC_FDT_IRQ_PPI_CPU_WIDTH,
351                              (1 << MACHINE(vms)->smp.cpus) - 1);
352     }
353 
354     qemu_fdt_add_subnode(ms->fdt, "/timer");
355 
356     armcpu = ARM_CPU(qemu_get_cpu(0));
357     if (arm_feature(&armcpu->env, ARM_FEATURE_V8)) {
358         const char compat[] = "arm,armv8-timer\0arm,armv7-timer";
359         qemu_fdt_setprop(ms->fdt, "/timer", "compatible",
360                          compat, sizeof(compat));
361     } else {
362         qemu_fdt_setprop_string(ms->fdt, "/timer", "compatible",
363                                 "arm,armv7-timer");
364     }
365     qemu_fdt_setprop(ms->fdt, "/timer", "always-on", NULL, 0);
366     qemu_fdt_setprop_cells(ms->fdt, "/timer", "interrupts",
367                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_S_EL1_IRQ, irqflags,
368                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_NS_EL1_IRQ, irqflags,
369                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_VIRT_IRQ, irqflags,
370                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_NS_EL2_IRQ, irqflags);
371 }
372 
373 static void fdt_add_cpu_nodes(const VirtMachineState *vms)
374 {
375     int cpu;
376     int addr_cells = 1;
377     const MachineState *ms = MACHINE(vms);
378     const VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
379     int smp_cpus = ms->smp.cpus;
380 
381     /*
382      * See Linux Documentation/devicetree/bindings/arm/cpus.yaml
383      * On ARM v8 64-bit systems value should be set to 2,
384      * that corresponds to the MPIDR_EL1 register size.
385      * If MPIDR_EL1[63:32] value is equal to 0 on all CPUs
386      * in the system, #address-cells can be set to 1, since
387      * MPIDR_EL1[63:32] bits are not used for CPUs
388      * identification.
389      *
390      * Here we actually don't know whether our system is 32- or 64-bit one.
391      * The simplest way to go is to examine affinity IDs of all our CPUs. If
392      * at least one of them has Aff3 populated, we set #address-cells to 2.
393      */
394     for (cpu = 0; cpu < smp_cpus; cpu++) {
395         ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(cpu));
396 
397         if (armcpu->mp_affinity & ARM_AFF3_MASK) {
398             addr_cells = 2;
399             break;
400         }
401     }
402 
403     qemu_fdt_add_subnode(ms->fdt, "/cpus");
404     qemu_fdt_setprop_cell(ms->fdt, "/cpus", "#address-cells", addr_cells);
405     qemu_fdt_setprop_cell(ms->fdt, "/cpus", "#size-cells", 0x0);
406 
407     for (cpu = smp_cpus - 1; cpu >= 0; cpu--) {
408         char *nodename = g_strdup_printf("/cpus/cpu@%d", cpu);
409         ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(cpu));
410         CPUState *cs = CPU(armcpu);
411 
412         qemu_fdt_add_subnode(ms->fdt, nodename);
413         qemu_fdt_setprop_string(ms->fdt, nodename, "device_type", "cpu");
414         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
415                                     armcpu->dtb_compatible);
416 
417         if (vms->psci_conduit != QEMU_PSCI_CONDUIT_DISABLED && smp_cpus > 1) {
418             qemu_fdt_setprop_string(ms->fdt, nodename,
419                                         "enable-method", "psci");
420         }
421 
422         if (addr_cells == 2) {
423             qemu_fdt_setprop_u64(ms->fdt, nodename, "reg",
424                                  armcpu->mp_affinity);
425         } else {
426             qemu_fdt_setprop_cell(ms->fdt, nodename, "reg",
427                                   armcpu->mp_affinity);
428         }
429 
430         if (ms->possible_cpus->cpus[cs->cpu_index].props.has_node_id) {
431             qemu_fdt_setprop_cell(ms->fdt, nodename, "numa-node-id",
432                 ms->possible_cpus->cpus[cs->cpu_index].props.node_id);
433         }
434 
435         if (!vmc->no_cpu_topology) {
436             qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle",
437                                   qemu_fdt_alloc_phandle(ms->fdt));
438         }
439 
440         g_free(nodename);
441     }
442 
443     if (!vmc->no_cpu_topology) {
444         /*
445          * Add vCPU topology description through fdt node cpu-map.
446          *
447          * See Linux Documentation/devicetree/bindings/cpu/cpu-topology.txt
448          * In a SMP system, the hierarchy of CPUs can be defined through
449          * four entities that are used to describe the layout of CPUs in
450          * the system: socket/cluster/core/thread.
451          *
452          * A socket node represents the boundary of system physical package
453          * and its child nodes must be one or more cluster nodes. A system
454          * can contain several layers of clustering within a single physical
455          * package and cluster nodes can be contained in parent cluster nodes.
456          *
457          * Note: currently we only support one layer of clustering within
458          * each physical package.
459          */
460         qemu_fdt_add_subnode(ms->fdt, "/cpus/cpu-map");
461 
462         for (cpu = smp_cpus - 1; cpu >= 0; cpu--) {
463             char *cpu_path = g_strdup_printf("/cpus/cpu@%d", cpu);
464             char *map_path;
465 
466             if (ms->smp.threads > 1) {
467                 map_path = g_strdup_printf(
468                     "/cpus/cpu-map/socket%d/cluster%d/core%d/thread%d",
469                     cpu / (ms->smp.clusters * ms->smp.cores * ms->smp.threads),
470                     (cpu / (ms->smp.cores * ms->smp.threads)) % ms->smp.clusters,
471                     (cpu / ms->smp.threads) % ms->smp.cores,
472                     cpu % ms->smp.threads);
473             } else {
474                 map_path = g_strdup_printf(
475                     "/cpus/cpu-map/socket%d/cluster%d/core%d",
476                     cpu / (ms->smp.clusters * ms->smp.cores),
477                     (cpu / ms->smp.cores) % ms->smp.clusters,
478                     cpu % ms->smp.cores);
479             }
480             qemu_fdt_add_path(ms->fdt, map_path);
481             qemu_fdt_setprop_phandle(ms->fdt, map_path, "cpu", cpu_path);
482 
483             g_free(map_path);
484             g_free(cpu_path);
485         }
486     }
487 }
488 
489 static void fdt_add_its_gic_node(VirtMachineState *vms)
490 {
491     char *nodename;
492     MachineState *ms = MACHINE(vms);
493 
494     vms->msi_phandle = qemu_fdt_alloc_phandle(ms->fdt);
495     nodename = g_strdup_printf("/intc/its@%" PRIx64,
496                                vms->memmap[VIRT_GIC_ITS].base);
497     qemu_fdt_add_subnode(ms->fdt, nodename);
498     qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
499                             "arm,gic-v3-its");
500     qemu_fdt_setprop(ms->fdt, nodename, "msi-controller", NULL, 0);
501     qemu_fdt_setprop_cell(ms->fdt, nodename, "#msi-cells", 1);
502     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
503                                  2, vms->memmap[VIRT_GIC_ITS].base,
504                                  2, vms->memmap[VIRT_GIC_ITS].size);
505     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", vms->msi_phandle);
506     g_free(nodename);
507 }
508 
509 static void fdt_add_v2m_gic_node(VirtMachineState *vms)
510 {
511     MachineState *ms = MACHINE(vms);
512     char *nodename;
513 
514     nodename = g_strdup_printf("/intc/v2m@%" PRIx64,
515                                vms->memmap[VIRT_GIC_V2M].base);
516     vms->msi_phandle = qemu_fdt_alloc_phandle(ms->fdt);
517     qemu_fdt_add_subnode(ms->fdt, nodename);
518     qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
519                             "arm,gic-v2m-frame");
520     qemu_fdt_setprop(ms->fdt, nodename, "msi-controller", NULL, 0);
521     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
522                                  2, vms->memmap[VIRT_GIC_V2M].base,
523                                  2, vms->memmap[VIRT_GIC_V2M].size);
524     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", vms->msi_phandle);
525     g_free(nodename);
526 }
527 
528 static void fdt_add_gic_node(VirtMachineState *vms)
529 {
530     MachineState *ms = MACHINE(vms);
531     char *nodename;
532 
533     vms->gic_phandle = qemu_fdt_alloc_phandle(ms->fdt);
534     qemu_fdt_setprop_cell(ms->fdt, "/", "interrupt-parent", vms->gic_phandle);
535 
536     nodename = g_strdup_printf("/intc@%" PRIx64,
537                                vms->memmap[VIRT_GIC_DIST].base);
538     qemu_fdt_add_subnode(ms->fdt, nodename);
539     qemu_fdt_setprop_cell(ms->fdt, nodename, "#interrupt-cells", 3);
540     qemu_fdt_setprop(ms->fdt, nodename, "interrupt-controller", NULL, 0);
541     qemu_fdt_setprop_cell(ms->fdt, nodename, "#address-cells", 0x2);
542     qemu_fdt_setprop_cell(ms->fdt, nodename, "#size-cells", 0x2);
543     qemu_fdt_setprop(ms->fdt, nodename, "ranges", NULL, 0);
544     if (vms->gic_version != VIRT_GIC_VERSION_2) {
545         int nb_redist_regions = virt_gicv3_redist_region_count(vms);
546 
547         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
548                                 "arm,gic-v3");
549 
550         qemu_fdt_setprop_cell(ms->fdt, nodename,
551                               "#redistributor-regions", nb_redist_regions);
552 
553         if (nb_redist_regions == 1) {
554             qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
555                                          2, vms->memmap[VIRT_GIC_DIST].base,
556                                          2, vms->memmap[VIRT_GIC_DIST].size,
557                                          2, vms->memmap[VIRT_GIC_REDIST].base,
558                                          2, vms->memmap[VIRT_GIC_REDIST].size);
559         } else {
560             qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
561                                  2, vms->memmap[VIRT_GIC_DIST].base,
562                                  2, vms->memmap[VIRT_GIC_DIST].size,
563                                  2, vms->memmap[VIRT_GIC_REDIST].base,
564                                  2, vms->memmap[VIRT_GIC_REDIST].size,
565                                  2, vms->memmap[VIRT_HIGH_GIC_REDIST2].base,
566                                  2, vms->memmap[VIRT_HIGH_GIC_REDIST2].size);
567         }
568 
569         if (vms->virt) {
570             qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
571                                    GIC_FDT_IRQ_TYPE_PPI, ARCH_GIC_MAINT_IRQ,
572                                    GIC_FDT_IRQ_FLAGS_LEVEL_HI);
573         }
574     } else {
575         /* 'cortex-a15-gic' means 'GIC v2' */
576         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
577                                 "arm,cortex-a15-gic");
578         if (!vms->virt) {
579             qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
580                                          2, vms->memmap[VIRT_GIC_DIST].base,
581                                          2, vms->memmap[VIRT_GIC_DIST].size,
582                                          2, vms->memmap[VIRT_GIC_CPU].base,
583                                          2, vms->memmap[VIRT_GIC_CPU].size);
584         } else {
585             qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
586                                          2, vms->memmap[VIRT_GIC_DIST].base,
587                                          2, vms->memmap[VIRT_GIC_DIST].size,
588                                          2, vms->memmap[VIRT_GIC_CPU].base,
589                                          2, vms->memmap[VIRT_GIC_CPU].size,
590                                          2, vms->memmap[VIRT_GIC_HYP].base,
591                                          2, vms->memmap[VIRT_GIC_HYP].size,
592                                          2, vms->memmap[VIRT_GIC_VCPU].base,
593                                          2, vms->memmap[VIRT_GIC_VCPU].size);
594             qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
595                                    GIC_FDT_IRQ_TYPE_PPI, ARCH_GIC_MAINT_IRQ,
596                                    GIC_FDT_IRQ_FLAGS_LEVEL_HI);
597         }
598     }
599 
600     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", vms->gic_phandle);
601     g_free(nodename);
602 }
603 
604 static void fdt_add_pmu_nodes(const VirtMachineState *vms)
605 {
606     ARMCPU *armcpu = ARM_CPU(first_cpu);
607     uint32_t irqflags = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
608     MachineState *ms = MACHINE(vms);
609 
610     if (!arm_feature(&armcpu->env, ARM_FEATURE_PMU)) {
611         assert(!object_property_get_bool(OBJECT(armcpu), "pmu", NULL));
612         return;
613     }
614 
615     if (vms->gic_version == VIRT_GIC_VERSION_2) {
616         irqflags = deposit32(irqflags, GIC_FDT_IRQ_PPI_CPU_START,
617                              GIC_FDT_IRQ_PPI_CPU_WIDTH,
618                              (1 << MACHINE(vms)->smp.cpus) - 1);
619     }
620 
621     qemu_fdt_add_subnode(ms->fdt, "/pmu");
622     if (arm_feature(&armcpu->env, ARM_FEATURE_V8)) {
623         const char compat[] = "arm,armv8-pmuv3";
624         qemu_fdt_setprop(ms->fdt, "/pmu", "compatible",
625                          compat, sizeof(compat));
626         qemu_fdt_setprop_cells(ms->fdt, "/pmu", "interrupts",
627                                GIC_FDT_IRQ_TYPE_PPI, VIRTUAL_PMU_IRQ, irqflags);
628     }
629 }
630 
631 static inline DeviceState *create_acpi_ged(VirtMachineState *vms)
632 {
633     DeviceState *dev;
634     MachineState *ms = MACHINE(vms);
635     int irq = vms->irqmap[VIRT_ACPI_GED];
636     uint32_t event = ACPI_GED_PWR_DOWN_EVT;
637 
638     if (ms->ram_slots) {
639         event |= ACPI_GED_MEM_HOTPLUG_EVT;
640     }
641 
642     if (ms->nvdimms_state->is_enabled) {
643         event |= ACPI_GED_NVDIMM_HOTPLUG_EVT;
644     }
645 
646     dev = qdev_new(TYPE_ACPI_GED);
647     qdev_prop_set_uint32(dev, "ged-event", event);
648 
649     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_ACPI_GED].base);
650     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 1, vms->memmap[VIRT_PCDIMM_ACPI].base);
651     sysbus_connect_irq(SYS_BUS_DEVICE(dev), 0, qdev_get_gpio_in(vms->gic, irq));
652 
653     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
654 
655     return dev;
656 }
657 
658 static void create_its(VirtMachineState *vms)
659 {
660     const char *itsclass = its_class_name();
661     DeviceState *dev;
662 
663     if (!strcmp(itsclass, "arm-gicv3-its")) {
664         if (!vms->tcg_its) {
665             itsclass = NULL;
666         }
667     }
668 
669     if (!itsclass) {
670         /* Do nothing if not supported */
671         return;
672     }
673 
674     dev = qdev_new(itsclass);
675 
676     object_property_set_link(OBJECT(dev), "parent-gicv3", OBJECT(vms->gic),
677                              &error_abort);
678     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
679     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_GIC_ITS].base);
680 
681     fdt_add_its_gic_node(vms);
682     vms->msi_controller = VIRT_MSI_CTRL_ITS;
683 }
684 
685 static void create_v2m(VirtMachineState *vms)
686 {
687     int i;
688     int irq = vms->irqmap[VIRT_GIC_V2M];
689     DeviceState *dev;
690 
691     dev = qdev_new("arm-gicv2m");
692     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_GIC_V2M].base);
693     qdev_prop_set_uint32(dev, "base-spi", irq);
694     qdev_prop_set_uint32(dev, "num-spi", NUM_GICV2M_SPIS);
695     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
696 
697     for (i = 0; i < NUM_GICV2M_SPIS; i++) {
698         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
699                            qdev_get_gpio_in(vms->gic, irq + i));
700     }
701 
702     fdt_add_v2m_gic_node(vms);
703     vms->msi_controller = VIRT_MSI_CTRL_GICV2M;
704 }
705 
706 static void create_gic(VirtMachineState *vms, MemoryRegion *mem)
707 {
708     MachineState *ms = MACHINE(vms);
709     /* We create a standalone GIC */
710     SysBusDevice *gicbusdev;
711     const char *gictype;
712     int i;
713     unsigned int smp_cpus = ms->smp.cpus;
714     uint32_t nb_redist_regions = 0;
715     int revision;
716 
717     if (vms->gic_version == VIRT_GIC_VERSION_2) {
718         gictype = gic_class_name();
719     } else {
720         gictype = gicv3_class_name();
721     }
722 
723     switch (vms->gic_version) {
724     case VIRT_GIC_VERSION_2:
725         revision = 2;
726         break;
727     case VIRT_GIC_VERSION_3:
728         revision = 3;
729         break;
730     case VIRT_GIC_VERSION_4:
731         revision = 4;
732         break;
733     default:
734         g_assert_not_reached();
735     }
736     vms->gic = qdev_new(gictype);
737     qdev_prop_set_uint32(vms->gic, "revision", revision);
738     qdev_prop_set_uint32(vms->gic, "num-cpu", smp_cpus);
739     /* Note that the num-irq property counts both internal and external
740      * interrupts; there are always 32 of the former (mandated by GIC spec).
741      */
742     qdev_prop_set_uint32(vms->gic, "num-irq", NUM_IRQS + 32);
743     if (!kvm_irqchip_in_kernel()) {
744         qdev_prop_set_bit(vms->gic, "has-security-extensions", vms->secure);
745     }
746 
747     if (vms->gic_version != VIRT_GIC_VERSION_2) {
748         uint32_t redist0_capacity = virt_redist_capacity(vms, VIRT_GIC_REDIST);
749         uint32_t redist0_count = MIN(smp_cpus, redist0_capacity);
750 
751         nb_redist_regions = virt_gicv3_redist_region_count(vms);
752 
753         qdev_prop_set_uint32(vms->gic, "len-redist-region-count",
754                              nb_redist_regions);
755         qdev_prop_set_uint32(vms->gic, "redist-region-count[0]", redist0_count);
756 
757         if (!kvm_irqchip_in_kernel()) {
758             if (vms->tcg_its) {
759                 object_property_set_link(OBJECT(vms->gic), "sysmem",
760                                          OBJECT(mem), &error_fatal);
761                 qdev_prop_set_bit(vms->gic, "has-lpi", true);
762             }
763         }
764 
765         if (nb_redist_regions == 2) {
766             uint32_t redist1_capacity =
767                 virt_redist_capacity(vms, VIRT_HIGH_GIC_REDIST2);
768 
769             qdev_prop_set_uint32(vms->gic, "redist-region-count[1]",
770                 MIN(smp_cpus - redist0_count, redist1_capacity));
771         }
772     } else {
773         if (!kvm_irqchip_in_kernel()) {
774             qdev_prop_set_bit(vms->gic, "has-virtualization-extensions",
775                               vms->virt);
776         }
777     }
778     gicbusdev = SYS_BUS_DEVICE(vms->gic);
779     sysbus_realize_and_unref(gicbusdev, &error_fatal);
780     sysbus_mmio_map(gicbusdev, 0, vms->memmap[VIRT_GIC_DIST].base);
781     if (vms->gic_version != VIRT_GIC_VERSION_2) {
782         sysbus_mmio_map(gicbusdev, 1, vms->memmap[VIRT_GIC_REDIST].base);
783         if (nb_redist_regions == 2) {
784             sysbus_mmio_map(gicbusdev, 2,
785                             vms->memmap[VIRT_HIGH_GIC_REDIST2].base);
786         }
787     } else {
788         sysbus_mmio_map(gicbusdev, 1, vms->memmap[VIRT_GIC_CPU].base);
789         if (vms->virt) {
790             sysbus_mmio_map(gicbusdev, 2, vms->memmap[VIRT_GIC_HYP].base);
791             sysbus_mmio_map(gicbusdev, 3, vms->memmap[VIRT_GIC_VCPU].base);
792         }
793     }
794 
795     /* Wire the outputs from each CPU's generic timer and the GICv3
796      * maintenance interrupt signal to the appropriate GIC PPI inputs,
797      * and the GIC's IRQ/FIQ/VIRQ/VFIQ interrupt outputs to the CPU's inputs.
798      */
799     for (i = 0; i < smp_cpus; i++) {
800         DeviceState *cpudev = DEVICE(qemu_get_cpu(i));
801         int ppibase = NUM_IRQS + i * GIC_INTERNAL + GIC_NR_SGIS;
802         int irq;
803         /* Mapping from the output timer irq lines from the CPU to the
804          * GIC PPI inputs we use for the virt board.
805          */
806         const int timer_irq[] = {
807             [GTIMER_PHYS] = ARCH_TIMER_NS_EL1_IRQ,
808             [GTIMER_VIRT] = ARCH_TIMER_VIRT_IRQ,
809             [GTIMER_HYP]  = ARCH_TIMER_NS_EL2_IRQ,
810             [GTIMER_SEC]  = ARCH_TIMER_S_EL1_IRQ,
811         };
812 
813         for (irq = 0; irq < ARRAY_SIZE(timer_irq); irq++) {
814             qdev_connect_gpio_out(cpudev, irq,
815                                   qdev_get_gpio_in(vms->gic,
816                                                    ppibase + timer_irq[irq]));
817         }
818 
819         if (vms->gic_version != VIRT_GIC_VERSION_2) {
820             qemu_irq irq = qdev_get_gpio_in(vms->gic,
821                                             ppibase + ARCH_GIC_MAINT_IRQ);
822             qdev_connect_gpio_out_named(cpudev, "gicv3-maintenance-interrupt",
823                                         0, irq);
824         } else if (vms->virt) {
825             qemu_irq irq = qdev_get_gpio_in(vms->gic,
826                                             ppibase + ARCH_GIC_MAINT_IRQ);
827             sysbus_connect_irq(gicbusdev, i + 4 * smp_cpus, irq);
828         }
829 
830         qdev_connect_gpio_out_named(cpudev, "pmu-interrupt", 0,
831                                     qdev_get_gpio_in(vms->gic, ppibase
832                                                      + VIRTUAL_PMU_IRQ));
833 
834         sysbus_connect_irq(gicbusdev, i, qdev_get_gpio_in(cpudev, ARM_CPU_IRQ));
835         sysbus_connect_irq(gicbusdev, i + smp_cpus,
836                            qdev_get_gpio_in(cpudev, ARM_CPU_FIQ));
837         sysbus_connect_irq(gicbusdev, i + 2 * smp_cpus,
838                            qdev_get_gpio_in(cpudev, ARM_CPU_VIRQ));
839         sysbus_connect_irq(gicbusdev, i + 3 * smp_cpus,
840                            qdev_get_gpio_in(cpudev, ARM_CPU_VFIQ));
841     }
842 
843     fdt_add_gic_node(vms);
844 
845     if (vms->gic_version != VIRT_GIC_VERSION_2 && vms->its) {
846         create_its(vms);
847     } else if (vms->gic_version == VIRT_GIC_VERSION_2) {
848         create_v2m(vms);
849     }
850 }
851 
852 static void create_uart(const VirtMachineState *vms, int uart,
853                         MemoryRegion *mem, Chardev *chr)
854 {
855     char *nodename;
856     hwaddr base = vms->memmap[uart].base;
857     hwaddr size = vms->memmap[uart].size;
858     int irq = vms->irqmap[uart];
859     const char compat[] = "arm,pl011\0arm,primecell";
860     const char clocknames[] = "uartclk\0apb_pclk";
861     DeviceState *dev = qdev_new(TYPE_PL011);
862     SysBusDevice *s = SYS_BUS_DEVICE(dev);
863     MachineState *ms = MACHINE(vms);
864 
865     qdev_prop_set_chr(dev, "chardev", chr);
866     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
867     memory_region_add_subregion(mem, base,
868                                 sysbus_mmio_get_region(s, 0));
869     sysbus_connect_irq(s, 0, qdev_get_gpio_in(vms->gic, irq));
870 
871     nodename = g_strdup_printf("/pl011@%" PRIx64, base);
872     qemu_fdt_add_subnode(ms->fdt, nodename);
873     /* Note that we can't use setprop_string because of the embedded NUL */
874     qemu_fdt_setprop(ms->fdt, nodename, "compatible",
875                          compat, sizeof(compat));
876     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
877                                      2, base, 2, size);
878     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
879                                GIC_FDT_IRQ_TYPE_SPI, irq,
880                                GIC_FDT_IRQ_FLAGS_LEVEL_HI);
881     qemu_fdt_setprop_cells(ms->fdt, nodename, "clocks",
882                                vms->clock_phandle, vms->clock_phandle);
883     qemu_fdt_setprop(ms->fdt, nodename, "clock-names",
884                          clocknames, sizeof(clocknames));
885 
886     if (uart == VIRT_UART) {
887         qemu_fdt_setprop_string(ms->fdt, "/chosen", "stdout-path", nodename);
888     } else {
889         /* Mark as not usable by the normal world */
890         qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
891         qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
892 
893         qemu_fdt_setprop_string(ms->fdt, "/secure-chosen", "stdout-path",
894                                 nodename);
895     }
896 
897     g_free(nodename);
898 }
899 
900 static void create_rtc(const VirtMachineState *vms)
901 {
902     char *nodename;
903     hwaddr base = vms->memmap[VIRT_RTC].base;
904     hwaddr size = vms->memmap[VIRT_RTC].size;
905     int irq = vms->irqmap[VIRT_RTC];
906     const char compat[] = "arm,pl031\0arm,primecell";
907     MachineState *ms = MACHINE(vms);
908 
909     sysbus_create_simple("pl031", base, qdev_get_gpio_in(vms->gic, irq));
910 
911     nodename = g_strdup_printf("/pl031@%" PRIx64, base);
912     qemu_fdt_add_subnode(ms->fdt, nodename);
913     qemu_fdt_setprop(ms->fdt, nodename, "compatible", compat, sizeof(compat));
914     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
915                                  2, base, 2, size);
916     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
917                            GIC_FDT_IRQ_TYPE_SPI, irq,
918                            GIC_FDT_IRQ_FLAGS_LEVEL_HI);
919     qemu_fdt_setprop_cell(ms->fdt, nodename, "clocks", vms->clock_phandle);
920     qemu_fdt_setprop_string(ms->fdt, nodename, "clock-names", "apb_pclk");
921     g_free(nodename);
922 }
923 
924 static DeviceState *gpio_key_dev;
925 static void virt_powerdown_req(Notifier *n, void *opaque)
926 {
927     VirtMachineState *s = container_of(n, VirtMachineState, powerdown_notifier);
928 
929     if (s->acpi_dev) {
930         acpi_send_event(s->acpi_dev, ACPI_POWER_DOWN_STATUS);
931     } else {
932         /* use gpio Pin 3 for power button event */
933         qemu_set_irq(qdev_get_gpio_in(gpio_key_dev, 0), 1);
934     }
935 }
936 
937 static void create_gpio_keys(char *fdt, DeviceState *pl061_dev,
938                              uint32_t phandle)
939 {
940     gpio_key_dev = sysbus_create_simple("gpio-key", -1,
941                                         qdev_get_gpio_in(pl061_dev, 3));
942 
943     qemu_fdt_add_subnode(fdt, "/gpio-keys");
944     qemu_fdt_setprop_string(fdt, "/gpio-keys", "compatible", "gpio-keys");
945 
946     qemu_fdt_add_subnode(fdt, "/gpio-keys/poweroff");
947     qemu_fdt_setprop_string(fdt, "/gpio-keys/poweroff",
948                             "label", "GPIO Key Poweroff");
949     qemu_fdt_setprop_cell(fdt, "/gpio-keys/poweroff", "linux,code",
950                           KEY_POWER);
951     qemu_fdt_setprop_cells(fdt, "/gpio-keys/poweroff",
952                            "gpios", phandle, 3, 0);
953 }
954 
955 #define SECURE_GPIO_POWEROFF 0
956 #define SECURE_GPIO_RESET    1
957 
958 static void create_secure_gpio_pwr(char *fdt, DeviceState *pl061_dev,
959                                    uint32_t phandle)
960 {
961     DeviceState *gpio_pwr_dev;
962 
963     /* gpio-pwr */
964     gpio_pwr_dev = sysbus_create_simple("gpio-pwr", -1, NULL);
965 
966     /* connect secure pl061 to gpio-pwr */
967     qdev_connect_gpio_out(pl061_dev, SECURE_GPIO_RESET,
968                           qdev_get_gpio_in_named(gpio_pwr_dev, "reset", 0));
969     qdev_connect_gpio_out(pl061_dev, SECURE_GPIO_POWEROFF,
970                           qdev_get_gpio_in_named(gpio_pwr_dev, "shutdown", 0));
971 
972     qemu_fdt_add_subnode(fdt, "/gpio-poweroff");
973     qemu_fdt_setprop_string(fdt, "/gpio-poweroff", "compatible",
974                             "gpio-poweroff");
975     qemu_fdt_setprop_cells(fdt, "/gpio-poweroff",
976                            "gpios", phandle, SECURE_GPIO_POWEROFF, 0);
977     qemu_fdt_setprop_string(fdt, "/gpio-poweroff", "status", "disabled");
978     qemu_fdt_setprop_string(fdt, "/gpio-poweroff", "secure-status",
979                             "okay");
980 
981     qemu_fdt_add_subnode(fdt, "/gpio-restart");
982     qemu_fdt_setprop_string(fdt, "/gpio-restart", "compatible",
983                             "gpio-restart");
984     qemu_fdt_setprop_cells(fdt, "/gpio-restart",
985                            "gpios", phandle, SECURE_GPIO_RESET, 0);
986     qemu_fdt_setprop_string(fdt, "/gpio-restart", "status", "disabled");
987     qemu_fdt_setprop_string(fdt, "/gpio-restart", "secure-status",
988                             "okay");
989 }
990 
991 static void create_gpio_devices(const VirtMachineState *vms, int gpio,
992                                 MemoryRegion *mem)
993 {
994     char *nodename;
995     DeviceState *pl061_dev;
996     hwaddr base = vms->memmap[gpio].base;
997     hwaddr size = vms->memmap[gpio].size;
998     int irq = vms->irqmap[gpio];
999     const char compat[] = "arm,pl061\0arm,primecell";
1000     SysBusDevice *s;
1001     MachineState *ms = MACHINE(vms);
1002 
1003     pl061_dev = qdev_new("pl061");
1004     /* Pull lines down to 0 if not driven by the PL061 */
1005     qdev_prop_set_uint32(pl061_dev, "pullups", 0);
1006     qdev_prop_set_uint32(pl061_dev, "pulldowns", 0xff);
1007     s = SYS_BUS_DEVICE(pl061_dev);
1008     sysbus_realize_and_unref(s, &error_fatal);
1009     memory_region_add_subregion(mem, base, sysbus_mmio_get_region(s, 0));
1010     sysbus_connect_irq(s, 0, qdev_get_gpio_in(vms->gic, irq));
1011 
1012     uint32_t phandle = qemu_fdt_alloc_phandle(ms->fdt);
1013     nodename = g_strdup_printf("/pl061@%" PRIx64, base);
1014     qemu_fdt_add_subnode(ms->fdt, nodename);
1015     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1016                                  2, base, 2, size);
1017     qemu_fdt_setprop(ms->fdt, nodename, "compatible", compat, sizeof(compat));
1018     qemu_fdt_setprop_cell(ms->fdt, nodename, "#gpio-cells", 2);
1019     qemu_fdt_setprop(ms->fdt, nodename, "gpio-controller", NULL, 0);
1020     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
1021                            GIC_FDT_IRQ_TYPE_SPI, irq,
1022                            GIC_FDT_IRQ_FLAGS_LEVEL_HI);
1023     qemu_fdt_setprop_cell(ms->fdt, nodename, "clocks", vms->clock_phandle);
1024     qemu_fdt_setprop_string(ms->fdt, nodename, "clock-names", "apb_pclk");
1025     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", phandle);
1026 
1027     if (gpio != VIRT_GPIO) {
1028         /* Mark as not usable by the normal world */
1029         qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
1030         qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
1031     }
1032     g_free(nodename);
1033 
1034     /* Child gpio devices */
1035     if (gpio == VIRT_GPIO) {
1036         create_gpio_keys(ms->fdt, pl061_dev, phandle);
1037     } else {
1038         create_secure_gpio_pwr(ms->fdt, pl061_dev, phandle);
1039     }
1040 }
1041 
1042 static void create_virtio_devices(const VirtMachineState *vms)
1043 {
1044     int i;
1045     hwaddr size = vms->memmap[VIRT_MMIO].size;
1046     MachineState *ms = MACHINE(vms);
1047 
1048     /* We create the transports in forwards order. Since qbus_realize()
1049      * prepends (not appends) new child buses, the incrementing loop below will
1050      * create a list of virtio-mmio buses with decreasing base addresses.
1051      *
1052      * When a -device option is processed from the command line,
1053      * qbus_find_recursive() picks the next free virtio-mmio bus in forwards
1054      * order. The upshot is that -device options in increasing command line
1055      * order are mapped to virtio-mmio buses with decreasing base addresses.
1056      *
1057      * When this code was originally written, that arrangement ensured that the
1058      * guest Linux kernel would give the lowest "name" (/dev/vda, eth0, etc) to
1059      * the first -device on the command line. (The end-to-end order is a
1060      * function of this loop, qbus_realize(), qbus_find_recursive(), and the
1061      * guest kernel's name-to-address assignment strategy.)
1062      *
1063      * Meanwhile, the kernel's traversal seems to have been reversed; see eg.
1064      * the message, if not necessarily the code, of commit 70161ff336.
1065      * Therefore the loop now establishes the inverse of the original intent.
1066      *
1067      * Unfortunately, we can't counteract the kernel change by reversing the
1068      * loop; it would break existing command lines.
1069      *
1070      * In any case, the kernel makes no guarantee about the stability of
1071      * enumeration order of virtio devices (as demonstrated by it changing
1072      * between kernel versions). For reliable and stable identification
1073      * of disks users must use UUIDs or similar mechanisms.
1074      */
1075     for (i = 0; i < NUM_VIRTIO_TRANSPORTS; i++) {
1076         int irq = vms->irqmap[VIRT_MMIO] + i;
1077         hwaddr base = vms->memmap[VIRT_MMIO].base + i * size;
1078 
1079         sysbus_create_simple("virtio-mmio", base,
1080                              qdev_get_gpio_in(vms->gic, irq));
1081     }
1082 
1083     /* We add dtb nodes in reverse order so that they appear in the finished
1084      * device tree lowest address first.
1085      *
1086      * Note that this mapping is independent of the loop above. The previous
1087      * loop influences virtio device to virtio transport assignment, whereas
1088      * this loop controls how virtio transports are laid out in the dtb.
1089      */
1090     for (i = NUM_VIRTIO_TRANSPORTS - 1; i >= 0; i--) {
1091         char *nodename;
1092         int irq = vms->irqmap[VIRT_MMIO] + i;
1093         hwaddr base = vms->memmap[VIRT_MMIO].base + i * size;
1094 
1095         nodename = g_strdup_printf("/virtio_mmio@%" PRIx64, base);
1096         qemu_fdt_add_subnode(ms->fdt, nodename);
1097         qemu_fdt_setprop_string(ms->fdt, nodename,
1098                                 "compatible", "virtio,mmio");
1099         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1100                                      2, base, 2, size);
1101         qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
1102                                GIC_FDT_IRQ_TYPE_SPI, irq,
1103                                GIC_FDT_IRQ_FLAGS_EDGE_LO_HI);
1104         qemu_fdt_setprop(ms->fdt, nodename, "dma-coherent", NULL, 0);
1105         g_free(nodename);
1106     }
1107 }
1108 
1109 #define VIRT_FLASH_SECTOR_SIZE (256 * KiB)
1110 
1111 static PFlashCFI01 *virt_flash_create1(VirtMachineState *vms,
1112                                         const char *name,
1113                                         const char *alias_prop_name)
1114 {
1115     /*
1116      * Create a single flash device.  We use the same parameters as
1117      * the flash devices on the Versatile Express board.
1118      */
1119     DeviceState *dev = qdev_new(TYPE_PFLASH_CFI01);
1120 
1121     qdev_prop_set_uint64(dev, "sector-length", VIRT_FLASH_SECTOR_SIZE);
1122     qdev_prop_set_uint8(dev, "width", 4);
1123     qdev_prop_set_uint8(dev, "device-width", 2);
1124     qdev_prop_set_bit(dev, "big-endian", false);
1125     qdev_prop_set_uint16(dev, "id0", 0x89);
1126     qdev_prop_set_uint16(dev, "id1", 0x18);
1127     qdev_prop_set_uint16(dev, "id2", 0x00);
1128     qdev_prop_set_uint16(dev, "id3", 0x00);
1129     qdev_prop_set_string(dev, "name", name);
1130     object_property_add_child(OBJECT(vms), name, OBJECT(dev));
1131     object_property_add_alias(OBJECT(vms), alias_prop_name,
1132                               OBJECT(dev), "drive");
1133     return PFLASH_CFI01(dev);
1134 }
1135 
1136 static void virt_flash_create(VirtMachineState *vms)
1137 {
1138     vms->flash[0] = virt_flash_create1(vms, "virt.flash0", "pflash0");
1139     vms->flash[1] = virt_flash_create1(vms, "virt.flash1", "pflash1");
1140 }
1141 
1142 static void virt_flash_map1(PFlashCFI01 *flash,
1143                             hwaddr base, hwaddr size,
1144                             MemoryRegion *sysmem)
1145 {
1146     DeviceState *dev = DEVICE(flash);
1147 
1148     assert(QEMU_IS_ALIGNED(size, VIRT_FLASH_SECTOR_SIZE));
1149     assert(size / VIRT_FLASH_SECTOR_SIZE <= UINT32_MAX);
1150     qdev_prop_set_uint32(dev, "num-blocks", size / VIRT_FLASH_SECTOR_SIZE);
1151     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1152 
1153     memory_region_add_subregion(sysmem, base,
1154                                 sysbus_mmio_get_region(SYS_BUS_DEVICE(dev),
1155                                                        0));
1156 }
1157 
1158 static void virt_flash_map(VirtMachineState *vms,
1159                            MemoryRegion *sysmem,
1160                            MemoryRegion *secure_sysmem)
1161 {
1162     /*
1163      * Map two flash devices to fill the VIRT_FLASH space in the memmap.
1164      * sysmem is the system memory space. secure_sysmem is the secure view
1165      * of the system, and the first flash device should be made visible only
1166      * there. The second flash device is visible to both secure and nonsecure.
1167      * If sysmem == secure_sysmem this means there is no separate Secure
1168      * address space and both flash devices are generally visible.
1169      */
1170     hwaddr flashsize = vms->memmap[VIRT_FLASH].size / 2;
1171     hwaddr flashbase = vms->memmap[VIRT_FLASH].base;
1172 
1173     virt_flash_map1(vms->flash[0], flashbase, flashsize,
1174                     secure_sysmem);
1175     virt_flash_map1(vms->flash[1], flashbase + flashsize, flashsize,
1176                     sysmem);
1177 }
1178 
1179 static void virt_flash_fdt(VirtMachineState *vms,
1180                            MemoryRegion *sysmem,
1181                            MemoryRegion *secure_sysmem)
1182 {
1183     hwaddr flashsize = vms->memmap[VIRT_FLASH].size / 2;
1184     hwaddr flashbase = vms->memmap[VIRT_FLASH].base;
1185     MachineState *ms = MACHINE(vms);
1186     char *nodename;
1187 
1188     if (sysmem == secure_sysmem) {
1189         /* Report both flash devices as a single node in the DT */
1190         nodename = g_strdup_printf("/flash@%" PRIx64, flashbase);
1191         qemu_fdt_add_subnode(ms->fdt, nodename);
1192         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible", "cfi-flash");
1193         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1194                                      2, flashbase, 2, flashsize,
1195                                      2, flashbase + flashsize, 2, flashsize);
1196         qemu_fdt_setprop_cell(ms->fdt, nodename, "bank-width", 4);
1197         g_free(nodename);
1198     } else {
1199         /*
1200          * Report the devices as separate nodes so we can mark one as
1201          * only visible to the secure world.
1202          */
1203         nodename = g_strdup_printf("/secflash@%" PRIx64, flashbase);
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, 2, flashsize);
1208         qemu_fdt_setprop_cell(ms->fdt, nodename, "bank-width", 4);
1209         qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
1210         qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
1211         g_free(nodename);
1212 
1213         nodename = g_strdup_printf("/flash@%" PRIx64, flashbase + flashsize);
1214         qemu_fdt_add_subnode(ms->fdt, nodename);
1215         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible", "cfi-flash");
1216         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1217                                      2, flashbase + flashsize, 2, flashsize);
1218         qemu_fdt_setprop_cell(ms->fdt, nodename, "bank-width", 4);
1219         g_free(nodename);
1220     }
1221 }
1222 
1223 static bool virt_firmware_init(VirtMachineState *vms,
1224                                MemoryRegion *sysmem,
1225                                MemoryRegion *secure_sysmem)
1226 {
1227     int i;
1228     const char *bios_name;
1229     BlockBackend *pflash_blk0;
1230 
1231     /* Map legacy -drive if=pflash to machine properties */
1232     for (i = 0; i < ARRAY_SIZE(vms->flash); i++) {
1233         pflash_cfi01_legacy_drive(vms->flash[i],
1234                                   drive_get(IF_PFLASH, 0, i));
1235     }
1236 
1237     virt_flash_map(vms, sysmem, secure_sysmem);
1238 
1239     pflash_blk0 = pflash_cfi01_get_blk(vms->flash[0]);
1240 
1241     bios_name = MACHINE(vms)->firmware;
1242     if (bios_name) {
1243         char *fname;
1244         MemoryRegion *mr;
1245         int image_size;
1246 
1247         if (pflash_blk0) {
1248             error_report("The contents of the first flash device may be "
1249                          "specified with -bios or with -drive if=pflash... "
1250                          "but you cannot use both options at once");
1251             exit(1);
1252         }
1253 
1254         /* Fall back to -bios */
1255 
1256         fname = qemu_find_file(QEMU_FILE_TYPE_BIOS, bios_name);
1257         if (!fname) {
1258             error_report("Could not find ROM image '%s'", bios_name);
1259             exit(1);
1260         }
1261         mr = sysbus_mmio_get_region(SYS_BUS_DEVICE(vms->flash[0]), 0);
1262         image_size = load_image_mr(fname, mr);
1263         g_free(fname);
1264         if (image_size < 0) {
1265             error_report("Could not load ROM image '%s'", bios_name);
1266             exit(1);
1267         }
1268     }
1269 
1270     return pflash_blk0 || bios_name;
1271 }
1272 
1273 static FWCfgState *create_fw_cfg(const VirtMachineState *vms, AddressSpace *as)
1274 {
1275     MachineState *ms = MACHINE(vms);
1276     hwaddr base = vms->memmap[VIRT_FW_CFG].base;
1277     hwaddr size = vms->memmap[VIRT_FW_CFG].size;
1278     FWCfgState *fw_cfg;
1279     char *nodename;
1280 
1281     fw_cfg = fw_cfg_init_mem_wide(base + 8, base, 8, base + 16, as);
1282     fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, (uint16_t)ms->smp.cpus);
1283 
1284     nodename = g_strdup_printf("/fw-cfg@%" PRIx64, base);
1285     qemu_fdt_add_subnode(ms->fdt, nodename);
1286     qemu_fdt_setprop_string(ms->fdt, nodename,
1287                             "compatible", "qemu,fw-cfg-mmio");
1288     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1289                                  2, base, 2, size);
1290     qemu_fdt_setprop(ms->fdt, nodename, "dma-coherent", NULL, 0);
1291     g_free(nodename);
1292     return fw_cfg;
1293 }
1294 
1295 static void create_pcie_irq_map(const MachineState *ms,
1296                                 uint32_t gic_phandle,
1297                                 int first_irq, const char *nodename)
1298 {
1299     int devfn, pin;
1300     uint32_t full_irq_map[4 * 4 * 10] = { 0 };
1301     uint32_t *irq_map = full_irq_map;
1302 
1303     for (devfn = 0; devfn <= 0x18; devfn += 0x8) {
1304         for (pin = 0; pin < 4; pin++) {
1305             int irq_type = GIC_FDT_IRQ_TYPE_SPI;
1306             int irq_nr = first_irq + ((pin + PCI_SLOT(devfn)) % PCI_NUM_PINS);
1307             int irq_level = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
1308             int i;
1309 
1310             uint32_t map[] = {
1311                 devfn << 8, 0, 0,                           /* devfn */
1312                 pin + 1,                                    /* PCI pin */
1313                 gic_phandle, 0, 0, irq_type, irq_nr, irq_level }; /* GIC irq */
1314 
1315             /* Convert map to big endian */
1316             for (i = 0; i < 10; i++) {
1317                 irq_map[i] = cpu_to_be32(map[i]);
1318             }
1319             irq_map += 10;
1320         }
1321     }
1322 
1323     qemu_fdt_setprop(ms->fdt, nodename, "interrupt-map",
1324                      full_irq_map, sizeof(full_irq_map));
1325 
1326     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupt-map-mask",
1327                            cpu_to_be16(PCI_DEVFN(3, 0)), /* Slot 3 */
1328                            0, 0,
1329                            0x7           /* PCI irq */);
1330 }
1331 
1332 static void create_smmu(const VirtMachineState *vms,
1333                         PCIBus *bus)
1334 {
1335     char *node;
1336     const char compat[] = "arm,smmu-v3";
1337     int irq =  vms->irqmap[VIRT_SMMU];
1338     int i;
1339     hwaddr base = vms->memmap[VIRT_SMMU].base;
1340     hwaddr size = vms->memmap[VIRT_SMMU].size;
1341     const char irq_names[] = "eventq\0priq\0cmdq-sync\0gerror";
1342     DeviceState *dev;
1343     MachineState *ms = MACHINE(vms);
1344 
1345     if (vms->iommu != VIRT_IOMMU_SMMUV3 || !vms->iommu_phandle) {
1346         return;
1347     }
1348 
1349     dev = qdev_new(TYPE_ARM_SMMUV3);
1350 
1351     object_property_set_link(OBJECT(dev), "primary-bus", OBJECT(bus),
1352                              &error_abort);
1353     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1354     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, base);
1355     for (i = 0; i < NUM_SMMU_IRQS; i++) {
1356         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
1357                            qdev_get_gpio_in(vms->gic, irq + i));
1358     }
1359 
1360     node = g_strdup_printf("/smmuv3@%" PRIx64, base);
1361     qemu_fdt_add_subnode(ms->fdt, node);
1362     qemu_fdt_setprop(ms->fdt, node, "compatible", compat, sizeof(compat));
1363     qemu_fdt_setprop_sized_cells(ms->fdt, node, "reg", 2, base, 2, size);
1364 
1365     qemu_fdt_setprop_cells(ms->fdt, node, "interrupts",
1366             GIC_FDT_IRQ_TYPE_SPI, irq    , GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1367             GIC_FDT_IRQ_TYPE_SPI, irq + 1, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1368             GIC_FDT_IRQ_TYPE_SPI, irq + 2, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1369             GIC_FDT_IRQ_TYPE_SPI, irq + 3, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI);
1370 
1371     qemu_fdt_setprop(ms->fdt, node, "interrupt-names", irq_names,
1372                      sizeof(irq_names));
1373 
1374     qemu_fdt_setprop(ms->fdt, node, "dma-coherent", NULL, 0);
1375 
1376     qemu_fdt_setprop_cell(ms->fdt, node, "#iommu-cells", 1);
1377 
1378     qemu_fdt_setprop_cell(ms->fdt, node, "phandle", vms->iommu_phandle);
1379     g_free(node);
1380 }
1381 
1382 static void create_virtio_iommu_dt_bindings(VirtMachineState *vms)
1383 {
1384     const char compat[] = "virtio,pci-iommu\0pci1af4,1057";
1385     uint16_t bdf = vms->virtio_iommu_bdf;
1386     MachineState *ms = MACHINE(vms);
1387     char *node;
1388 
1389     vms->iommu_phandle = qemu_fdt_alloc_phandle(ms->fdt);
1390 
1391     node = g_strdup_printf("%s/virtio_iommu@%x,%x", vms->pciehb_nodename,
1392                            PCI_SLOT(bdf), PCI_FUNC(bdf));
1393     qemu_fdt_add_subnode(ms->fdt, node);
1394     qemu_fdt_setprop(ms->fdt, node, "compatible", compat, sizeof(compat));
1395     qemu_fdt_setprop_sized_cells(ms->fdt, node, "reg",
1396                                  1, bdf << 8, 1, 0, 1, 0,
1397                                  1, 0, 1, 0);
1398 
1399     qemu_fdt_setprop_cell(ms->fdt, node, "#iommu-cells", 1);
1400     qemu_fdt_setprop_cell(ms->fdt, node, "phandle", vms->iommu_phandle);
1401     g_free(node);
1402 
1403     qemu_fdt_setprop_cells(ms->fdt, vms->pciehb_nodename, "iommu-map",
1404                            0x0, vms->iommu_phandle, 0x0, bdf,
1405                            bdf + 1, vms->iommu_phandle, bdf + 1, 0xffff - bdf);
1406 }
1407 
1408 static void create_pcie(VirtMachineState *vms)
1409 {
1410     hwaddr base_mmio = vms->memmap[VIRT_PCIE_MMIO].base;
1411     hwaddr size_mmio = vms->memmap[VIRT_PCIE_MMIO].size;
1412     hwaddr base_mmio_high = vms->memmap[VIRT_HIGH_PCIE_MMIO].base;
1413     hwaddr size_mmio_high = vms->memmap[VIRT_HIGH_PCIE_MMIO].size;
1414     hwaddr base_pio = vms->memmap[VIRT_PCIE_PIO].base;
1415     hwaddr size_pio = vms->memmap[VIRT_PCIE_PIO].size;
1416     hwaddr base_ecam, size_ecam;
1417     hwaddr base = base_mmio;
1418     int nr_pcie_buses;
1419     int irq = vms->irqmap[VIRT_PCIE];
1420     MemoryRegion *mmio_alias;
1421     MemoryRegion *mmio_reg;
1422     MemoryRegion *ecam_alias;
1423     MemoryRegion *ecam_reg;
1424     DeviceState *dev;
1425     char *nodename;
1426     int i, ecam_id;
1427     PCIHostState *pci;
1428     MachineState *ms = MACHINE(vms);
1429 
1430     dev = qdev_new(TYPE_GPEX_HOST);
1431     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1432 
1433     ecam_id = VIRT_ECAM_ID(vms->highmem_ecam);
1434     base_ecam = vms->memmap[ecam_id].base;
1435     size_ecam = vms->memmap[ecam_id].size;
1436     nr_pcie_buses = size_ecam / PCIE_MMCFG_SIZE_MIN;
1437     /* Map only the first size_ecam bytes of ECAM space */
1438     ecam_alias = g_new0(MemoryRegion, 1);
1439     ecam_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 0);
1440     memory_region_init_alias(ecam_alias, OBJECT(dev), "pcie-ecam",
1441                              ecam_reg, 0, size_ecam);
1442     memory_region_add_subregion(get_system_memory(), base_ecam, ecam_alias);
1443 
1444     /* Map the MMIO window into system address space so as to expose
1445      * the section of PCI MMIO space which starts at the same base address
1446      * (ie 1:1 mapping for that part of PCI MMIO space visible through
1447      * the window).
1448      */
1449     mmio_alias = g_new0(MemoryRegion, 1);
1450     mmio_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 1);
1451     memory_region_init_alias(mmio_alias, OBJECT(dev), "pcie-mmio",
1452                              mmio_reg, base_mmio, size_mmio);
1453     memory_region_add_subregion(get_system_memory(), base_mmio, mmio_alias);
1454 
1455     if (vms->highmem_mmio) {
1456         /* Map high MMIO space */
1457         MemoryRegion *high_mmio_alias = g_new0(MemoryRegion, 1);
1458 
1459         memory_region_init_alias(high_mmio_alias, OBJECT(dev), "pcie-mmio-high",
1460                                  mmio_reg, base_mmio_high, size_mmio_high);
1461         memory_region_add_subregion(get_system_memory(), base_mmio_high,
1462                                     high_mmio_alias);
1463     }
1464 
1465     /* Map IO port space */
1466     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 2, base_pio);
1467 
1468     for (i = 0; i < GPEX_NUM_IRQS; i++) {
1469         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
1470                            qdev_get_gpio_in(vms->gic, irq + i));
1471         gpex_set_irq_num(GPEX_HOST(dev), i, irq + i);
1472     }
1473 
1474     pci = PCI_HOST_BRIDGE(dev);
1475     pci->bypass_iommu = vms->default_bus_bypass_iommu;
1476     vms->bus = pci->bus;
1477     if (vms->bus) {
1478         for (i = 0; i < nb_nics; i++) {
1479             NICInfo *nd = &nd_table[i];
1480 
1481             if (!nd->model) {
1482                 nd->model = g_strdup("virtio");
1483             }
1484 
1485             pci_nic_init_nofail(nd, pci->bus, nd->model, NULL);
1486         }
1487     }
1488 
1489     nodename = vms->pciehb_nodename = g_strdup_printf("/pcie@%" PRIx64, base);
1490     qemu_fdt_add_subnode(ms->fdt, nodename);
1491     qemu_fdt_setprop_string(ms->fdt, nodename,
1492                             "compatible", "pci-host-ecam-generic");
1493     qemu_fdt_setprop_string(ms->fdt, nodename, "device_type", "pci");
1494     qemu_fdt_setprop_cell(ms->fdt, nodename, "#address-cells", 3);
1495     qemu_fdt_setprop_cell(ms->fdt, nodename, "#size-cells", 2);
1496     qemu_fdt_setprop_cell(ms->fdt, nodename, "linux,pci-domain", 0);
1497     qemu_fdt_setprop_cells(ms->fdt, nodename, "bus-range", 0,
1498                            nr_pcie_buses - 1);
1499     qemu_fdt_setprop(ms->fdt, nodename, "dma-coherent", NULL, 0);
1500 
1501     if (vms->msi_phandle) {
1502         qemu_fdt_setprop_cells(ms->fdt, nodename, "msi-map",
1503                                0, vms->msi_phandle, 0, 0x10000);
1504     }
1505 
1506     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1507                                  2, base_ecam, 2, size_ecam);
1508 
1509     if (vms->highmem_mmio) {
1510         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "ranges",
1511                                      1, FDT_PCI_RANGE_IOPORT, 2, 0,
1512                                      2, base_pio, 2, size_pio,
1513                                      1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
1514                                      2, base_mmio, 2, size_mmio,
1515                                      1, FDT_PCI_RANGE_MMIO_64BIT,
1516                                      2, base_mmio_high,
1517                                      2, base_mmio_high, 2, size_mmio_high);
1518     } else {
1519         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "ranges",
1520                                      1, FDT_PCI_RANGE_IOPORT, 2, 0,
1521                                      2, base_pio, 2, size_pio,
1522                                      1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
1523                                      2, base_mmio, 2, size_mmio);
1524     }
1525 
1526     qemu_fdt_setprop_cell(ms->fdt, nodename, "#interrupt-cells", 1);
1527     create_pcie_irq_map(ms, vms->gic_phandle, irq, nodename);
1528 
1529     if (vms->iommu) {
1530         vms->iommu_phandle = qemu_fdt_alloc_phandle(ms->fdt);
1531 
1532         switch (vms->iommu) {
1533         case VIRT_IOMMU_SMMUV3:
1534             create_smmu(vms, vms->bus);
1535             qemu_fdt_setprop_cells(ms->fdt, nodename, "iommu-map",
1536                                    0x0, vms->iommu_phandle, 0x0, 0x10000);
1537             break;
1538         default:
1539             g_assert_not_reached();
1540         }
1541     }
1542 }
1543 
1544 static void create_platform_bus(VirtMachineState *vms)
1545 {
1546     DeviceState *dev;
1547     SysBusDevice *s;
1548     int i;
1549     MemoryRegion *sysmem = get_system_memory();
1550 
1551     dev = qdev_new(TYPE_PLATFORM_BUS_DEVICE);
1552     dev->id = g_strdup(TYPE_PLATFORM_BUS_DEVICE);
1553     qdev_prop_set_uint32(dev, "num_irqs", PLATFORM_BUS_NUM_IRQS);
1554     qdev_prop_set_uint32(dev, "mmio_size", vms->memmap[VIRT_PLATFORM_BUS].size);
1555     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1556     vms->platform_bus_dev = dev;
1557 
1558     s = SYS_BUS_DEVICE(dev);
1559     for (i = 0; i < PLATFORM_BUS_NUM_IRQS; i++) {
1560         int irq = vms->irqmap[VIRT_PLATFORM_BUS] + i;
1561         sysbus_connect_irq(s, i, qdev_get_gpio_in(vms->gic, irq));
1562     }
1563 
1564     memory_region_add_subregion(sysmem,
1565                                 vms->memmap[VIRT_PLATFORM_BUS].base,
1566                                 sysbus_mmio_get_region(s, 0));
1567 }
1568 
1569 static void create_tag_ram(MemoryRegion *tag_sysmem,
1570                            hwaddr base, hwaddr size,
1571                            const char *name)
1572 {
1573     MemoryRegion *tagram = g_new(MemoryRegion, 1);
1574 
1575     memory_region_init_ram(tagram, NULL, name, size / 32, &error_fatal);
1576     memory_region_add_subregion(tag_sysmem, base / 32, tagram);
1577 }
1578 
1579 static void create_secure_ram(VirtMachineState *vms,
1580                               MemoryRegion *secure_sysmem,
1581                               MemoryRegion *secure_tag_sysmem)
1582 {
1583     MemoryRegion *secram = g_new(MemoryRegion, 1);
1584     char *nodename;
1585     hwaddr base = vms->memmap[VIRT_SECURE_MEM].base;
1586     hwaddr size = vms->memmap[VIRT_SECURE_MEM].size;
1587     MachineState *ms = MACHINE(vms);
1588 
1589     memory_region_init_ram(secram, NULL, "virt.secure-ram", size,
1590                            &error_fatal);
1591     memory_region_add_subregion(secure_sysmem, base, secram);
1592 
1593     nodename = g_strdup_printf("/secram@%" PRIx64, base);
1594     qemu_fdt_add_subnode(ms->fdt, nodename);
1595     qemu_fdt_setprop_string(ms->fdt, nodename, "device_type", "memory");
1596     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg", 2, base, 2, size);
1597     qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
1598     qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
1599 
1600     if (secure_tag_sysmem) {
1601         create_tag_ram(secure_tag_sysmem, base, size, "mach-virt.secure-tag");
1602     }
1603 
1604     g_free(nodename);
1605 }
1606 
1607 static void *machvirt_dtb(const struct arm_boot_info *binfo, int *fdt_size)
1608 {
1609     const VirtMachineState *board = container_of(binfo, VirtMachineState,
1610                                                  bootinfo);
1611     MachineState *ms = MACHINE(board);
1612 
1613 
1614     *fdt_size = board->fdt_size;
1615     return ms->fdt;
1616 }
1617 
1618 static void virt_build_smbios(VirtMachineState *vms)
1619 {
1620     MachineClass *mc = MACHINE_GET_CLASS(vms);
1621     MachineState *ms = MACHINE(vms);
1622     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
1623     uint8_t *smbios_tables, *smbios_anchor;
1624     size_t smbios_tables_len, smbios_anchor_len;
1625     struct smbios_phys_mem_area mem_array;
1626     const char *product = "QEMU Virtual Machine";
1627 
1628     if (kvm_enabled()) {
1629         product = "KVM Virtual Machine";
1630     }
1631 
1632     smbios_set_defaults("QEMU", product,
1633                         vmc->smbios_old_sys_ver ? "1.0" : mc->name, false,
1634                         true, SMBIOS_ENTRY_POINT_TYPE_64);
1635 
1636     /* build the array of physical mem area from base_memmap */
1637     mem_array.address = vms->memmap[VIRT_MEM].base;
1638     mem_array.length = ms->ram_size;
1639 
1640     smbios_get_tables(ms, &mem_array, 1,
1641                       &smbios_tables, &smbios_tables_len,
1642                       &smbios_anchor, &smbios_anchor_len,
1643                       &error_fatal);
1644 
1645     if (smbios_anchor) {
1646         fw_cfg_add_file(vms->fw_cfg, "etc/smbios/smbios-tables",
1647                         smbios_tables, smbios_tables_len);
1648         fw_cfg_add_file(vms->fw_cfg, "etc/smbios/smbios-anchor",
1649                         smbios_anchor, smbios_anchor_len);
1650     }
1651 }
1652 
1653 static
1654 void virt_machine_done(Notifier *notifier, void *data)
1655 {
1656     VirtMachineState *vms = container_of(notifier, VirtMachineState,
1657                                          machine_done);
1658     MachineState *ms = MACHINE(vms);
1659     ARMCPU *cpu = ARM_CPU(first_cpu);
1660     struct arm_boot_info *info = &vms->bootinfo;
1661     AddressSpace *as = arm_boot_address_space(cpu, info);
1662 
1663     /*
1664      * If the user provided a dtb, we assume the dynamic sysbus nodes
1665      * already are integrated there. This corresponds to a use case where
1666      * the dynamic sysbus nodes are complex and their generation is not yet
1667      * supported. In that case the user can take charge of the guest dt
1668      * while qemu takes charge of the qom stuff.
1669      */
1670     if (info->dtb_filename == NULL) {
1671         platform_bus_add_all_fdt_nodes(ms->fdt, "/intc",
1672                                        vms->memmap[VIRT_PLATFORM_BUS].base,
1673                                        vms->memmap[VIRT_PLATFORM_BUS].size,
1674                                        vms->irqmap[VIRT_PLATFORM_BUS]);
1675     }
1676     if (arm_load_dtb(info->dtb_start, info, info->dtb_limit, as, ms) < 0) {
1677         exit(1);
1678     }
1679 
1680     fw_cfg_add_extra_pci_roots(vms->bus, vms->fw_cfg);
1681 
1682     virt_acpi_setup(vms);
1683     virt_build_smbios(vms);
1684 }
1685 
1686 static uint64_t virt_cpu_mp_affinity(VirtMachineState *vms, int idx)
1687 {
1688     uint8_t clustersz = ARM_DEFAULT_CPUS_PER_CLUSTER;
1689     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
1690 
1691     if (!vmc->disallow_affinity_adjustment) {
1692         /* Adjust MPIDR like 64-bit KVM hosts, which incorporate the
1693          * GIC's target-list limitations. 32-bit KVM hosts currently
1694          * always create clusters of 4 CPUs, but that is expected to
1695          * change when they gain support for gicv3. When KVM is enabled
1696          * it will override the changes we make here, therefore our
1697          * purposes are to make TCG consistent (with 64-bit KVM hosts)
1698          * and to improve SGI efficiency.
1699          */
1700         if (vms->gic_version == VIRT_GIC_VERSION_2) {
1701             clustersz = GIC_TARGETLIST_BITS;
1702         } else {
1703             clustersz = GICV3_TARGETLIST_BITS;
1704         }
1705     }
1706     return arm_cpu_mp_affinity(idx, clustersz);
1707 }
1708 
1709 static inline bool *virt_get_high_memmap_enabled(VirtMachineState *vms,
1710                                                  int index)
1711 {
1712     bool *enabled_array[] = {
1713         &vms->highmem_redists,
1714         &vms->highmem_ecam,
1715         &vms->highmem_mmio,
1716     };
1717 
1718     assert(ARRAY_SIZE(extended_memmap) - VIRT_LOWMEMMAP_LAST ==
1719            ARRAY_SIZE(enabled_array));
1720     assert(index - VIRT_LOWMEMMAP_LAST < ARRAY_SIZE(enabled_array));
1721 
1722     return enabled_array[index - VIRT_LOWMEMMAP_LAST];
1723 }
1724 
1725 static void virt_set_high_memmap(VirtMachineState *vms,
1726                                  hwaddr base, int pa_bits)
1727 {
1728     hwaddr region_base, region_size;
1729     bool *region_enabled, fits;
1730     int i;
1731 
1732     for (i = VIRT_LOWMEMMAP_LAST; i < ARRAY_SIZE(extended_memmap); i++) {
1733         region_enabled = virt_get_high_memmap_enabled(vms, i);
1734         region_base = ROUND_UP(base, extended_memmap[i].size);
1735         region_size = extended_memmap[i].size;
1736 
1737         vms->memmap[i].base = region_base;
1738         vms->memmap[i].size = region_size;
1739 
1740         /*
1741          * Check each device to see if it fits in the PA space,
1742          * moving highest_gpa as we go. For compatibility, move
1743          * highest_gpa for disabled fitting devices as well, if
1744          * the compact layout has been disabled.
1745          *
1746          * For each device that doesn't fit, disable it.
1747          */
1748         fits = (region_base + region_size) <= BIT_ULL(pa_bits);
1749         *region_enabled &= fits;
1750         if (vms->highmem_compact && !*region_enabled) {
1751             continue;
1752         }
1753 
1754         base = region_base + region_size;
1755         if (fits) {
1756             vms->highest_gpa = base - 1;
1757         }
1758     }
1759 }
1760 
1761 static void virt_set_memmap(VirtMachineState *vms, int pa_bits)
1762 {
1763     MachineState *ms = MACHINE(vms);
1764     hwaddr base, device_memory_base, device_memory_size, memtop;
1765     int i;
1766 
1767     vms->memmap = extended_memmap;
1768 
1769     for (i = 0; i < ARRAY_SIZE(base_memmap); i++) {
1770         vms->memmap[i] = base_memmap[i];
1771     }
1772 
1773     if (ms->ram_slots > ACPI_MAX_RAM_SLOTS) {
1774         error_report("unsupported number of memory slots: %"PRIu64,
1775                      ms->ram_slots);
1776         exit(EXIT_FAILURE);
1777     }
1778 
1779     /*
1780      * !highmem is exactly the same as limiting the PA space to 32bit,
1781      * irrespective of the underlying capabilities of the HW.
1782      */
1783     if (!vms->highmem) {
1784         pa_bits = 32;
1785     }
1786 
1787     /*
1788      * We compute the base of the high IO region depending on the
1789      * amount of initial and device memory. The device memory start/size
1790      * is aligned on 1GiB. We never put the high IO region below 256GiB
1791      * so that if maxram_size is < 255GiB we keep the legacy memory map.
1792      * The device region size assumes 1GiB page max alignment per slot.
1793      */
1794     device_memory_base =
1795         ROUND_UP(vms->memmap[VIRT_MEM].base + ms->ram_size, GiB);
1796     device_memory_size = ms->maxram_size - ms->ram_size + ms->ram_slots * GiB;
1797 
1798     /* Base address of the high IO region */
1799     memtop = base = device_memory_base + ROUND_UP(device_memory_size, GiB);
1800     if (memtop > BIT_ULL(pa_bits)) {
1801 	    error_report("Addressing limited to %d bits, but memory exceeds it by %llu bytes\n",
1802 			 pa_bits, memtop - BIT_ULL(pa_bits));
1803         exit(EXIT_FAILURE);
1804     }
1805     if (base < device_memory_base) {
1806         error_report("maxmem/slots too huge");
1807         exit(EXIT_FAILURE);
1808     }
1809     if (base < vms->memmap[VIRT_MEM].base + LEGACY_RAMLIMIT_BYTES) {
1810         base = vms->memmap[VIRT_MEM].base + LEGACY_RAMLIMIT_BYTES;
1811     }
1812 
1813     /* We know for sure that at least the memory fits in the PA space */
1814     vms->highest_gpa = memtop - 1;
1815 
1816     virt_set_high_memmap(vms, base, pa_bits);
1817 
1818     if (device_memory_size > 0) {
1819         ms->device_memory = g_malloc0(sizeof(*ms->device_memory));
1820         ms->device_memory->base = device_memory_base;
1821         memory_region_init(&ms->device_memory->mr, OBJECT(vms),
1822                            "device-memory", device_memory_size);
1823     }
1824 }
1825 
1826 static VirtGICType finalize_gic_version_do(const char *accel_name,
1827                                            VirtGICType gic_version,
1828                                            int gics_supported,
1829                                            unsigned int max_cpus)
1830 {
1831     /* Convert host/max/nosel to GIC version number */
1832     switch (gic_version) {
1833     case VIRT_GIC_VERSION_HOST:
1834         if (!kvm_enabled()) {
1835             error_report("gic-version=host requires KVM");
1836             exit(1);
1837         }
1838 
1839         /* For KVM, gic-version=host means gic-version=max */
1840         return finalize_gic_version_do(accel_name, VIRT_GIC_VERSION_MAX,
1841                                        gics_supported, max_cpus);
1842     case VIRT_GIC_VERSION_MAX:
1843         if (gics_supported & VIRT_GIC_VERSION_4_MASK) {
1844             gic_version = VIRT_GIC_VERSION_4;
1845         } else if (gics_supported & VIRT_GIC_VERSION_3_MASK) {
1846             gic_version = VIRT_GIC_VERSION_3;
1847         } else {
1848             gic_version = VIRT_GIC_VERSION_2;
1849         }
1850         break;
1851     case VIRT_GIC_VERSION_NOSEL:
1852         if ((gics_supported & VIRT_GIC_VERSION_2_MASK) &&
1853             max_cpus <= GIC_NCPU) {
1854             gic_version = VIRT_GIC_VERSION_2;
1855         } else if (gics_supported & VIRT_GIC_VERSION_3_MASK) {
1856             /*
1857              * in case the host does not support v2 emulation or
1858              * the end-user requested more than 8 VCPUs we now default
1859              * to v3. In any case defaulting to v2 would be broken.
1860              */
1861             gic_version = VIRT_GIC_VERSION_3;
1862         } else if (max_cpus > GIC_NCPU) {
1863             error_report("%s only supports GICv2 emulation but more than 8 "
1864                          "vcpus are requested", accel_name);
1865             exit(1);
1866         }
1867         break;
1868     case VIRT_GIC_VERSION_2:
1869     case VIRT_GIC_VERSION_3:
1870     case VIRT_GIC_VERSION_4:
1871         break;
1872     }
1873 
1874     /* Check chosen version is effectively supported */
1875     switch (gic_version) {
1876     case VIRT_GIC_VERSION_2:
1877         if (!(gics_supported & VIRT_GIC_VERSION_2_MASK)) {
1878             error_report("%s does not support GICv2 emulation", accel_name);
1879             exit(1);
1880         }
1881         break;
1882     case VIRT_GIC_VERSION_3:
1883         if (!(gics_supported & VIRT_GIC_VERSION_3_MASK)) {
1884             error_report("%s does not support GICv3 emulation", accel_name);
1885             exit(1);
1886         }
1887         break;
1888     case VIRT_GIC_VERSION_4:
1889         if (!(gics_supported & VIRT_GIC_VERSION_4_MASK)) {
1890             error_report("%s does not support GICv4 emulation, is virtualization=on?",
1891                          accel_name);
1892             exit(1);
1893         }
1894         break;
1895     default:
1896         error_report("logic error in finalize_gic_version");
1897         exit(1);
1898         break;
1899     }
1900 
1901     return gic_version;
1902 }
1903 
1904 /*
1905  * finalize_gic_version - Determines the final gic_version
1906  * according to the gic-version property
1907  *
1908  * Default GIC type is v2
1909  */
1910 static void finalize_gic_version(VirtMachineState *vms)
1911 {
1912     const char *accel_name = current_accel_name();
1913     unsigned int max_cpus = MACHINE(vms)->smp.max_cpus;
1914     int gics_supported = 0;
1915 
1916     /* Determine which GIC versions the current environment supports */
1917     if (kvm_enabled() && kvm_irqchip_in_kernel()) {
1918         int probe_bitmap = kvm_arm_vgic_probe();
1919 
1920         if (!probe_bitmap) {
1921             error_report("Unable to determine GIC version supported by host");
1922             exit(1);
1923         }
1924 
1925         if (probe_bitmap & KVM_ARM_VGIC_V2) {
1926             gics_supported |= VIRT_GIC_VERSION_2_MASK;
1927         }
1928         if (probe_bitmap & KVM_ARM_VGIC_V3) {
1929             gics_supported |= VIRT_GIC_VERSION_3_MASK;
1930         }
1931     } else if (kvm_enabled() && !kvm_irqchip_in_kernel()) {
1932         /* KVM w/o kernel irqchip can only deal with GICv2 */
1933         gics_supported |= VIRT_GIC_VERSION_2_MASK;
1934         accel_name = "KVM with kernel-irqchip=off";
1935     } else if (tcg_enabled() || hvf_enabled() || qtest_enabled())  {
1936         gics_supported |= VIRT_GIC_VERSION_2_MASK;
1937         if (module_object_class_by_name("arm-gicv3")) {
1938             gics_supported |= VIRT_GIC_VERSION_3_MASK;
1939             if (vms->virt) {
1940                 /* GICv4 only makes sense if CPU has EL2 */
1941                 gics_supported |= VIRT_GIC_VERSION_4_MASK;
1942             }
1943         }
1944     } else {
1945         error_report("Unsupported accelerator, can not determine GIC support");
1946         exit(1);
1947     }
1948 
1949     /*
1950      * Then convert helpers like host/max to concrete GIC versions and ensure
1951      * the desired version is supported
1952      */
1953     vms->gic_version = finalize_gic_version_do(accel_name, vms->gic_version,
1954                                                gics_supported, max_cpus);
1955 }
1956 
1957 /*
1958  * virt_cpu_post_init() must be called after the CPUs have
1959  * been realized and the GIC has been created.
1960  */
1961 static void virt_cpu_post_init(VirtMachineState *vms, MemoryRegion *sysmem)
1962 {
1963     int max_cpus = MACHINE(vms)->smp.max_cpus;
1964     bool aarch64, pmu, steal_time;
1965     CPUState *cpu;
1966 
1967     aarch64 = object_property_get_bool(OBJECT(first_cpu), "aarch64", NULL);
1968     pmu = object_property_get_bool(OBJECT(first_cpu), "pmu", NULL);
1969     steal_time = object_property_get_bool(OBJECT(first_cpu),
1970                                           "kvm-steal-time", NULL);
1971 
1972     if (kvm_enabled()) {
1973         hwaddr pvtime_reg_base = vms->memmap[VIRT_PVTIME].base;
1974         hwaddr pvtime_reg_size = vms->memmap[VIRT_PVTIME].size;
1975 
1976         if (steal_time) {
1977             MemoryRegion *pvtime = g_new(MemoryRegion, 1);
1978             hwaddr pvtime_size = max_cpus * PVTIME_SIZE_PER_CPU;
1979 
1980             /* The memory region size must be a multiple of host page size. */
1981             pvtime_size = REAL_HOST_PAGE_ALIGN(pvtime_size);
1982 
1983             if (pvtime_size > pvtime_reg_size) {
1984                 error_report("pvtime requires a %" HWADDR_PRId
1985                              " byte memory region for %d CPUs,"
1986                              " but only %" HWADDR_PRId " has been reserved",
1987                              pvtime_size, max_cpus, pvtime_reg_size);
1988                 exit(1);
1989             }
1990 
1991             memory_region_init_ram(pvtime, NULL, "pvtime", pvtime_size, NULL);
1992             memory_region_add_subregion(sysmem, pvtime_reg_base, pvtime);
1993         }
1994 
1995         CPU_FOREACH(cpu) {
1996             if (pmu) {
1997                 assert(arm_feature(&ARM_CPU(cpu)->env, ARM_FEATURE_PMU));
1998                 if (kvm_irqchip_in_kernel()) {
1999                     kvm_arm_pmu_set_irq(cpu, PPI(VIRTUAL_PMU_IRQ));
2000                 }
2001                 kvm_arm_pmu_init(cpu);
2002             }
2003             if (steal_time) {
2004                 kvm_arm_pvtime_init(cpu, pvtime_reg_base +
2005                                          cpu->cpu_index * PVTIME_SIZE_PER_CPU);
2006             }
2007         }
2008     } else {
2009         if (aarch64 && vms->highmem) {
2010             int requested_pa_size = 64 - clz64(vms->highest_gpa);
2011             int pamax = arm_pamax(ARM_CPU(first_cpu));
2012 
2013             if (pamax < requested_pa_size) {
2014                 error_report("VCPU supports less PA bits (%d) than "
2015                              "requested by the memory map (%d)",
2016                              pamax, requested_pa_size);
2017                 exit(1);
2018             }
2019         }
2020     }
2021 }
2022 
2023 static void machvirt_init(MachineState *machine)
2024 {
2025     VirtMachineState *vms = VIRT_MACHINE(machine);
2026     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(machine);
2027     MachineClass *mc = MACHINE_GET_CLASS(machine);
2028     const CPUArchIdList *possible_cpus;
2029     MemoryRegion *sysmem = get_system_memory();
2030     MemoryRegion *secure_sysmem = NULL;
2031     MemoryRegion *tag_sysmem = NULL;
2032     MemoryRegion *secure_tag_sysmem = NULL;
2033     int n, virt_max_cpus;
2034     bool firmware_loaded;
2035     bool aarch64 = true;
2036     bool has_ged = !vmc->no_ged;
2037     unsigned int smp_cpus = machine->smp.cpus;
2038     unsigned int max_cpus = machine->smp.max_cpus;
2039 
2040     if (!cpu_type_valid(machine->cpu_type)) {
2041         error_report("mach-virt: CPU type %s not supported", machine->cpu_type);
2042         exit(1);
2043     }
2044 
2045     possible_cpus = mc->possible_cpu_arch_ids(machine);
2046 
2047     /*
2048      * In accelerated mode, the memory map is computed earlier in kvm_type()
2049      * to create a VM with the right number of IPA bits.
2050      */
2051     if (!vms->memmap) {
2052         Object *cpuobj;
2053         ARMCPU *armcpu;
2054         int pa_bits;
2055 
2056         /*
2057          * Instantiate a temporary CPU object to find out about what
2058          * we are about to deal with. Once this is done, get rid of
2059          * the object.
2060          */
2061         cpuobj = object_new(possible_cpus->cpus[0].type);
2062         armcpu = ARM_CPU(cpuobj);
2063 
2064         pa_bits = arm_pamax(armcpu);
2065 
2066         object_unref(cpuobj);
2067 
2068         virt_set_memmap(vms, pa_bits);
2069     }
2070 
2071     /* We can probe only here because during property set
2072      * KVM is not available yet
2073      */
2074     finalize_gic_version(vms);
2075 
2076     if (vms->secure) {
2077         /*
2078          * The Secure view of the world is the same as the NonSecure,
2079          * but with a few extra devices. Create it as a container region
2080          * containing the system memory at low priority; any secure-only
2081          * devices go in at higher priority and take precedence.
2082          */
2083         secure_sysmem = g_new(MemoryRegion, 1);
2084         memory_region_init(secure_sysmem, OBJECT(machine), "secure-memory",
2085                            UINT64_MAX);
2086         memory_region_add_subregion_overlap(secure_sysmem, 0, sysmem, -1);
2087     }
2088 
2089     firmware_loaded = virt_firmware_init(vms, sysmem,
2090                                          secure_sysmem ?: sysmem);
2091 
2092     /* If we have an EL3 boot ROM then the assumption is that it will
2093      * implement PSCI itself, so disable QEMU's internal implementation
2094      * so it doesn't get in the way. Instead of starting secondary
2095      * CPUs in PSCI powerdown state we will start them all running and
2096      * let the boot ROM sort them out.
2097      * The usual case is that we do use QEMU's PSCI implementation;
2098      * if the guest has EL2 then we will use SMC as the conduit,
2099      * and otherwise we will use HVC (for backwards compatibility and
2100      * because if we're using KVM then we must use HVC).
2101      */
2102     if (vms->secure && firmware_loaded) {
2103         vms->psci_conduit = QEMU_PSCI_CONDUIT_DISABLED;
2104     } else if (vms->virt) {
2105         vms->psci_conduit = QEMU_PSCI_CONDUIT_SMC;
2106     } else {
2107         vms->psci_conduit = QEMU_PSCI_CONDUIT_HVC;
2108     }
2109 
2110     /*
2111      * The maximum number of CPUs depends on the GIC version, or on how
2112      * many redistributors we can fit into the memory map (which in turn
2113      * depends on whether this is a GICv3 or v4).
2114      */
2115     if (vms->gic_version == VIRT_GIC_VERSION_2) {
2116         virt_max_cpus = GIC_NCPU;
2117     } else {
2118         virt_max_cpus = virt_redist_capacity(vms, VIRT_GIC_REDIST);
2119         if (vms->highmem_redists) {
2120             virt_max_cpus += virt_redist_capacity(vms, VIRT_HIGH_GIC_REDIST2);
2121         }
2122     }
2123 
2124     if (max_cpus > virt_max_cpus) {
2125         error_report("Number of SMP CPUs requested (%d) exceeds max CPUs "
2126                      "supported by machine 'mach-virt' (%d)",
2127                      max_cpus, virt_max_cpus);
2128         if (vms->gic_version != VIRT_GIC_VERSION_2 && !vms->highmem_redists) {
2129             error_printf("Try 'highmem-redists=on' for more CPUs\n");
2130         }
2131 
2132         exit(1);
2133     }
2134 
2135     if (vms->secure && (kvm_enabled() || hvf_enabled())) {
2136         error_report("mach-virt: %s does not support providing "
2137                      "Security extensions (TrustZone) to the guest CPU",
2138                      current_accel_name());
2139         exit(1);
2140     }
2141 
2142     if (vms->virt && (kvm_enabled() || hvf_enabled())) {
2143         error_report("mach-virt: %s does not support providing "
2144                      "Virtualization extensions to the guest CPU",
2145                      current_accel_name());
2146         exit(1);
2147     }
2148 
2149     if (vms->mte && (kvm_enabled() || hvf_enabled())) {
2150         error_report("mach-virt: %s does not support providing "
2151                      "MTE to the guest CPU",
2152                      current_accel_name());
2153         exit(1);
2154     }
2155 
2156     create_fdt(vms);
2157 
2158     assert(possible_cpus->len == max_cpus);
2159     for (n = 0; n < possible_cpus->len; n++) {
2160         Object *cpuobj;
2161         CPUState *cs;
2162 
2163         if (n >= smp_cpus) {
2164             break;
2165         }
2166 
2167         cpuobj = object_new(possible_cpus->cpus[n].type);
2168         object_property_set_int(cpuobj, "mp-affinity",
2169                                 possible_cpus->cpus[n].arch_id, NULL);
2170 
2171         cs = CPU(cpuobj);
2172         cs->cpu_index = n;
2173 
2174         numa_cpu_pre_plug(&possible_cpus->cpus[cs->cpu_index], DEVICE(cpuobj),
2175                           &error_fatal);
2176 
2177         aarch64 &= object_property_get_bool(cpuobj, "aarch64", NULL);
2178 
2179         if (!vms->secure) {
2180             object_property_set_bool(cpuobj, "has_el3", false, NULL);
2181         }
2182 
2183         if (!vms->virt && object_property_find(cpuobj, "has_el2")) {
2184             object_property_set_bool(cpuobj, "has_el2", false, NULL);
2185         }
2186 
2187         if (vmc->kvm_no_adjvtime &&
2188             object_property_find(cpuobj, "kvm-no-adjvtime")) {
2189             object_property_set_bool(cpuobj, "kvm-no-adjvtime", true, NULL);
2190         }
2191 
2192         if (vmc->no_kvm_steal_time &&
2193             object_property_find(cpuobj, "kvm-steal-time")) {
2194             object_property_set_bool(cpuobj, "kvm-steal-time", false, NULL);
2195         }
2196 
2197         if (vmc->no_pmu && object_property_find(cpuobj, "pmu")) {
2198             object_property_set_bool(cpuobj, "pmu", false, NULL);
2199         }
2200 
2201         if (vmc->no_tcg_lpa2 && object_property_find(cpuobj, "lpa2")) {
2202             object_property_set_bool(cpuobj, "lpa2", false, NULL);
2203         }
2204 
2205         if (object_property_find(cpuobj, "reset-cbar")) {
2206             object_property_set_int(cpuobj, "reset-cbar",
2207                                     vms->memmap[VIRT_CPUPERIPHS].base,
2208                                     &error_abort);
2209         }
2210 
2211         object_property_set_link(cpuobj, "memory", OBJECT(sysmem),
2212                                  &error_abort);
2213         if (vms->secure) {
2214             object_property_set_link(cpuobj, "secure-memory",
2215                                      OBJECT(secure_sysmem), &error_abort);
2216         }
2217 
2218         if (vms->mte) {
2219             /* Create the memory region only once, but link to all cpus. */
2220             if (!tag_sysmem) {
2221                 /*
2222                  * The property exists only if MemTag is supported.
2223                  * If it is, we must allocate the ram to back that up.
2224                  */
2225                 if (!object_property_find(cpuobj, "tag-memory")) {
2226                     error_report("MTE requested, but not supported "
2227                                  "by the guest CPU");
2228                     exit(1);
2229                 }
2230 
2231                 tag_sysmem = g_new(MemoryRegion, 1);
2232                 memory_region_init(tag_sysmem, OBJECT(machine),
2233                                    "tag-memory", UINT64_MAX / 32);
2234 
2235                 if (vms->secure) {
2236                     secure_tag_sysmem = g_new(MemoryRegion, 1);
2237                     memory_region_init(secure_tag_sysmem, OBJECT(machine),
2238                                        "secure-tag-memory", UINT64_MAX / 32);
2239 
2240                     /* As with ram, secure-tag takes precedence over tag.  */
2241                     memory_region_add_subregion_overlap(secure_tag_sysmem, 0,
2242                                                         tag_sysmem, -1);
2243                 }
2244             }
2245 
2246             object_property_set_link(cpuobj, "tag-memory", OBJECT(tag_sysmem),
2247                                      &error_abort);
2248             if (vms->secure) {
2249                 object_property_set_link(cpuobj, "secure-tag-memory",
2250                                          OBJECT(secure_tag_sysmem),
2251                                          &error_abort);
2252             }
2253         }
2254 
2255         qdev_realize(DEVICE(cpuobj), NULL, &error_fatal);
2256         object_unref(cpuobj);
2257     }
2258     fdt_add_timer_nodes(vms);
2259     fdt_add_cpu_nodes(vms);
2260 
2261     memory_region_add_subregion(sysmem, vms->memmap[VIRT_MEM].base,
2262                                 machine->ram);
2263     if (machine->device_memory) {
2264         memory_region_add_subregion(sysmem, machine->device_memory->base,
2265                                     &machine->device_memory->mr);
2266     }
2267 
2268     virt_flash_fdt(vms, sysmem, secure_sysmem ?: sysmem);
2269 
2270     create_gic(vms, sysmem);
2271 
2272     virt_cpu_post_init(vms, sysmem);
2273 
2274     fdt_add_pmu_nodes(vms);
2275 
2276     create_uart(vms, VIRT_UART, sysmem, serial_hd(0));
2277 
2278     if (vms->secure) {
2279         create_secure_ram(vms, secure_sysmem, secure_tag_sysmem);
2280         create_uart(vms, VIRT_SECURE_UART, secure_sysmem, serial_hd(1));
2281     }
2282 
2283     if (tag_sysmem) {
2284         create_tag_ram(tag_sysmem, vms->memmap[VIRT_MEM].base,
2285                        machine->ram_size, "mach-virt.tag");
2286     }
2287 
2288     vms->highmem_ecam &= (!firmware_loaded || aarch64);
2289 
2290     create_rtc(vms);
2291 
2292     create_pcie(vms);
2293 
2294     if (has_ged && aarch64 && firmware_loaded && virt_is_acpi_enabled(vms)) {
2295         vms->acpi_dev = create_acpi_ged(vms);
2296     } else {
2297         create_gpio_devices(vms, VIRT_GPIO, sysmem);
2298     }
2299 
2300     if (vms->secure && !vmc->no_secure_gpio) {
2301         create_gpio_devices(vms, VIRT_SECURE_GPIO, secure_sysmem);
2302     }
2303 
2304      /* connect powerdown request */
2305      vms->powerdown_notifier.notify = virt_powerdown_req;
2306      qemu_register_powerdown_notifier(&vms->powerdown_notifier);
2307 
2308     /* Create mmio transports, so the user can create virtio backends
2309      * (which will be automatically plugged in to the transports). If
2310      * no backend is created the transport will just sit harmlessly idle.
2311      */
2312     create_virtio_devices(vms);
2313 
2314     vms->fw_cfg = create_fw_cfg(vms, &address_space_memory);
2315     rom_set_fw(vms->fw_cfg);
2316 
2317     create_platform_bus(vms);
2318 
2319     if (machine->nvdimms_state->is_enabled) {
2320         const struct AcpiGenericAddress arm_virt_nvdimm_acpi_dsmio = {
2321             .space_id = AML_AS_SYSTEM_MEMORY,
2322             .address = vms->memmap[VIRT_NVDIMM_ACPI].base,
2323             .bit_width = NVDIMM_ACPI_IO_LEN << 3
2324         };
2325 
2326         nvdimm_init_acpi_state(machine->nvdimms_state, sysmem,
2327                                arm_virt_nvdimm_acpi_dsmio,
2328                                vms->fw_cfg, OBJECT(vms));
2329     }
2330 
2331     vms->bootinfo.ram_size = machine->ram_size;
2332     vms->bootinfo.board_id = -1;
2333     vms->bootinfo.loader_start = vms->memmap[VIRT_MEM].base;
2334     vms->bootinfo.get_dtb = machvirt_dtb;
2335     vms->bootinfo.skip_dtb_autoload = true;
2336     vms->bootinfo.firmware_loaded = firmware_loaded;
2337     vms->bootinfo.psci_conduit = vms->psci_conduit;
2338     arm_load_kernel(ARM_CPU(first_cpu), machine, &vms->bootinfo);
2339 
2340     vms->machine_done.notify = virt_machine_done;
2341     qemu_add_machine_init_done_notifier(&vms->machine_done);
2342 }
2343 
2344 static bool virt_get_secure(Object *obj, Error **errp)
2345 {
2346     VirtMachineState *vms = VIRT_MACHINE(obj);
2347 
2348     return vms->secure;
2349 }
2350 
2351 static void virt_set_secure(Object *obj, bool value, Error **errp)
2352 {
2353     VirtMachineState *vms = VIRT_MACHINE(obj);
2354 
2355     vms->secure = value;
2356 }
2357 
2358 static bool virt_get_virt(Object *obj, Error **errp)
2359 {
2360     VirtMachineState *vms = VIRT_MACHINE(obj);
2361 
2362     return vms->virt;
2363 }
2364 
2365 static void virt_set_virt(Object *obj, bool value, Error **errp)
2366 {
2367     VirtMachineState *vms = VIRT_MACHINE(obj);
2368 
2369     vms->virt = value;
2370 }
2371 
2372 static bool virt_get_highmem(Object *obj, Error **errp)
2373 {
2374     VirtMachineState *vms = VIRT_MACHINE(obj);
2375 
2376     return vms->highmem;
2377 }
2378 
2379 static void virt_set_highmem(Object *obj, bool value, Error **errp)
2380 {
2381     VirtMachineState *vms = VIRT_MACHINE(obj);
2382 
2383     vms->highmem = value;
2384 }
2385 
2386 static bool virt_get_compact_highmem(Object *obj, Error **errp)
2387 {
2388     VirtMachineState *vms = VIRT_MACHINE(obj);
2389 
2390     return vms->highmem_compact;
2391 }
2392 
2393 static void virt_set_compact_highmem(Object *obj, bool value, Error **errp)
2394 {
2395     VirtMachineState *vms = VIRT_MACHINE(obj);
2396 
2397     vms->highmem_compact = value;
2398 }
2399 
2400 static bool virt_get_highmem_redists(Object *obj, Error **errp)
2401 {
2402     VirtMachineState *vms = VIRT_MACHINE(obj);
2403 
2404     return vms->highmem_redists;
2405 }
2406 
2407 static void virt_set_highmem_redists(Object *obj, bool value, Error **errp)
2408 {
2409     VirtMachineState *vms = VIRT_MACHINE(obj);
2410 
2411     vms->highmem_redists = value;
2412 }
2413 
2414 static bool virt_get_highmem_ecam(Object *obj, Error **errp)
2415 {
2416     VirtMachineState *vms = VIRT_MACHINE(obj);
2417 
2418     return vms->highmem_ecam;
2419 }
2420 
2421 static void virt_set_highmem_ecam(Object *obj, bool value, Error **errp)
2422 {
2423     VirtMachineState *vms = VIRT_MACHINE(obj);
2424 
2425     vms->highmem_ecam = value;
2426 }
2427 
2428 static bool virt_get_highmem_mmio(Object *obj, Error **errp)
2429 {
2430     VirtMachineState *vms = VIRT_MACHINE(obj);
2431 
2432     return vms->highmem_mmio;
2433 }
2434 
2435 static void virt_set_highmem_mmio(Object *obj, bool value, Error **errp)
2436 {
2437     VirtMachineState *vms = VIRT_MACHINE(obj);
2438 
2439     vms->highmem_mmio = value;
2440 }
2441 
2442 
2443 static bool virt_get_its(Object *obj, Error **errp)
2444 {
2445     VirtMachineState *vms = VIRT_MACHINE(obj);
2446 
2447     return vms->its;
2448 }
2449 
2450 static void virt_set_its(Object *obj, bool value, Error **errp)
2451 {
2452     VirtMachineState *vms = VIRT_MACHINE(obj);
2453 
2454     vms->its = value;
2455 }
2456 
2457 static bool virt_get_dtb_randomness(Object *obj, Error **errp)
2458 {
2459     VirtMachineState *vms = VIRT_MACHINE(obj);
2460 
2461     return vms->dtb_randomness;
2462 }
2463 
2464 static void virt_set_dtb_randomness(Object *obj, bool value, Error **errp)
2465 {
2466     VirtMachineState *vms = VIRT_MACHINE(obj);
2467 
2468     vms->dtb_randomness = value;
2469 }
2470 
2471 static char *virt_get_oem_id(Object *obj, Error **errp)
2472 {
2473     VirtMachineState *vms = VIRT_MACHINE(obj);
2474 
2475     return g_strdup(vms->oem_id);
2476 }
2477 
2478 static void virt_set_oem_id(Object *obj, const char *value, Error **errp)
2479 {
2480     VirtMachineState *vms = VIRT_MACHINE(obj);
2481     size_t len = strlen(value);
2482 
2483     if (len > 6) {
2484         error_setg(errp,
2485                    "User specified oem-id value is bigger than 6 bytes in size");
2486         return;
2487     }
2488 
2489     strncpy(vms->oem_id, value, 6);
2490 }
2491 
2492 static char *virt_get_oem_table_id(Object *obj, Error **errp)
2493 {
2494     VirtMachineState *vms = VIRT_MACHINE(obj);
2495 
2496     return g_strdup(vms->oem_table_id);
2497 }
2498 
2499 static void virt_set_oem_table_id(Object *obj, const char *value,
2500                                   Error **errp)
2501 {
2502     VirtMachineState *vms = VIRT_MACHINE(obj);
2503     size_t len = strlen(value);
2504 
2505     if (len > 8) {
2506         error_setg(errp,
2507                    "User specified oem-table-id value is bigger than 8 bytes in size");
2508         return;
2509     }
2510     strncpy(vms->oem_table_id, value, 8);
2511 }
2512 
2513 
2514 bool virt_is_acpi_enabled(VirtMachineState *vms)
2515 {
2516     if (vms->acpi == ON_OFF_AUTO_OFF) {
2517         return false;
2518     }
2519     return true;
2520 }
2521 
2522 static void virt_get_acpi(Object *obj, Visitor *v, const char *name,
2523                           void *opaque, Error **errp)
2524 {
2525     VirtMachineState *vms = VIRT_MACHINE(obj);
2526     OnOffAuto acpi = vms->acpi;
2527 
2528     visit_type_OnOffAuto(v, name, &acpi, errp);
2529 }
2530 
2531 static void virt_set_acpi(Object *obj, Visitor *v, const char *name,
2532                           void *opaque, Error **errp)
2533 {
2534     VirtMachineState *vms = VIRT_MACHINE(obj);
2535 
2536     visit_type_OnOffAuto(v, name, &vms->acpi, errp);
2537 }
2538 
2539 static bool virt_get_ras(Object *obj, Error **errp)
2540 {
2541     VirtMachineState *vms = VIRT_MACHINE(obj);
2542 
2543     return vms->ras;
2544 }
2545 
2546 static void virt_set_ras(Object *obj, bool value, Error **errp)
2547 {
2548     VirtMachineState *vms = VIRT_MACHINE(obj);
2549 
2550     vms->ras = value;
2551 }
2552 
2553 static bool virt_get_mte(Object *obj, Error **errp)
2554 {
2555     VirtMachineState *vms = VIRT_MACHINE(obj);
2556 
2557     return vms->mte;
2558 }
2559 
2560 static void virt_set_mte(Object *obj, bool value, Error **errp)
2561 {
2562     VirtMachineState *vms = VIRT_MACHINE(obj);
2563 
2564     vms->mte = value;
2565 }
2566 
2567 static char *virt_get_gic_version(Object *obj, Error **errp)
2568 {
2569     VirtMachineState *vms = VIRT_MACHINE(obj);
2570     const char *val;
2571 
2572     switch (vms->gic_version) {
2573     case VIRT_GIC_VERSION_4:
2574         val = "4";
2575         break;
2576     case VIRT_GIC_VERSION_3:
2577         val = "3";
2578         break;
2579     default:
2580         val = "2";
2581         break;
2582     }
2583     return g_strdup(val);
2584 }
2585 
2586 static void virt_set_gic_version(Object *obj, const char *value, Error **errp)
2587 {
2588     VirtMachineState *vms = VIRT_MACHINE(obj);
2589 
2590     if (!strcmp(value, "4")) {
2591         vms->gic_version = VIRT_GIC_VERSION_4;
2592     } else if (!strcmp(value, "3")) {
2593         vms->gic_version = VIRT_GIC_VERSION_3;
2594     } else if (!strcmp(value, "2")) {
2595         vms->gic_version = VIRT_GIC_VERSION_2;
2596     } else if (!strcmp(value, "host")) {
2597         vms->gic_version = VIRT_GIC_VERSION_HOST; /* Will probe later */
2598     } else if (!strcmp(value, "max")) {
2599         vms->gic_version = VIRT_GIC_VERSION_MAX; /* Will probe later */
2600     } else {
2601         error_setg(errp, "Invalid gic-version value");
2602         error_append_hint(errp, "Valid values are 3, 2, host, max.\n");
2603     }
2604 }
2605 
2606 static char *virt_get_iommu(Object *obj, Error **errp)
2607 {
2608     VirtMachineState *vms = VIRT_MACHINE(obj);
2609 
2610     switch (vms->iommu) {
2611     case VIRT_IOMMU_NONE:
2612         return g_strdup("none");
2613     case VIRT_IOMMU_SMMUV3:
2614         return g_strdup("smmuv3");
2615     default:
2616         g_assert_not_reached();
2617     }
2618 }
2619 
2620 static void virt_set_iommu(Object *obj, const char *value, Error **errp)
2621 {
2622     VirtMachineState *vms = VIRT_MACHINE(obj);
2623 
2624     if (!strcmp(value, "smmuv3")) {
2625         vms->iommu = VIRT_IOMMU_SMMUV3;
2626     } else if (!strcmp(value, "none")) {
2627         vms->iommu = VIRT_IOMMU_NONE;
2628     } else {
2629         error_setg(errp, "Invalid iommu value");
2630         error_append_hint(errp, "Valid values are none, smmuv3.\n");
2631     }
2632 }
2633 
2634 static bool virt_get_default_bus_bypass_iommu(Object *obj, Error **errp)
2635 {
2636     VirtMachineState *vms = VIRT_MACHINE(obj);
2637 
2638     return vms->default_bus_bypass_iommu;
2639 }
2640 
2641 static void virt_set_default_bus_bypass_iommu(Object *obj, bool value,
2642                                               Error **errp)
2643 {
2644     VirtMachineState *vms = VIRT_MACHINE(obj);
2645 
2646     vms->default_bus_bypass_iommu = value;
2647 }
2648 
2649 static CpuInstanceProperties
2650 virt_cpu_index_to_props(MachineState *ms, unsigned cpu_index)
2651 {
2652     MachineClass *mc = MACHINE_GET_CLASS(ms);
2653     const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(ms);
2654 
2655     assert(cpu_index < possible_cpus->len);
2656     return possible_cpus->cpus[cpu_index].props;
2657 }
2658 
2659 static int64_t virt_get_default_cpu_node_id(const MachineState *ms, int idx)
2660 {
2661     int64_t socket_id = ms->possible_cpus->cpus[idx].props.socket_id;
2662 
2663     return socket_id % ms->numa_state->num_nodes;
2664 }
2665 
2666 static const CPUArchIdList *virt_possible_cpu_arch_ids(MachineState *ms)
2667 {
2668     int n;
2669     unsigned int max_cpus = ms->smp.max_cpus;
2670     VirtMachineState *vms = VIRT_MACHINE(ms);
2671     MachineClass *mc = MACHINE_GET_CLASS(vms);
2672 
2673     if (ms->possible_cpus) {
2674         assert(ms->possible_cpus->len == max_cpus);
2675         return ms->possible_cpus;
2676     }
2677 
2678     ms->possible_cpus = g_malloc0(sizeof(CPUArchIdList) +
2679                                   sizeof(CPUArchId) * max_cpus);
2680     ms->possible_cpus->len = max_cpus;
2681     for (n = 0; n < ms->possible_cpus->len; n++) {
2682         ms->possible_cpus->cpus[n].type = ms->cpu_type;
2683         ms->possible_cpus->cpus[n].arch_id =
2684             virt_cpu_mp_affinity(vms, n);
2685 
2686         assert(!mc->smp_props.dies_supported);
2687         ms->possible_cpus->cpus[n].props.has_socket_id = true;
2688         ms->possible_cpus->cpus[n].props.socket_id =
2689             n / (ms->smp.clusters * ms->smp.cores * ms->smp.threads);
2690         ms->possible_cpus->cpus[n].props.has_cluster_id = true;
2691         ms->possible_cpus->cpus[n].props.cluster_id =
2692             (n / (ms->smp.cores * ms->smp.threads)) % ms->smp.clusters;
2693         ms->possible_cpus->cpus[n].props.has_core_id = true;
2694         ms->possible_cpus->cpus[n].props.core_id =
2695             (n / ms->smp.threads) % ms->smp.cores;
2696         ms->possible_cpus->cpus[n].props.has_thread_id = true;
2697         ms->possible_cpus->cpus[n].props.thread_id =
2698             n % ms->smp.threads;
2699     }
2700     return ms->possible_cpus;
2701 }
2702 
2703 static void virt_memory_pre_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
2704                                  Error **errp)
2705 {
2706     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2707     const MachineState *ms = MACHINE(hotplug_dev);
2708     const bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2709 
2710     if (!vms->acpi_dev) {
2711         error_setg(errp,
2712                    "memory hotplug is not enabled: missing acpi-ged device");
2713         return;
2714     }
2715 
2716     if (vms->mte) {
2717         error_setg(errp, "memory hotplug is not enabled: MTE is enabled");
2718         return;
2719     }
2720 
2721     if (is_nvdimm && !ms->nvdimms_state->is_enabled) {
2722         error_setg(errp, "nvdimm is not enabled: add 'nvdimm=on' to '-M'");
2723         return;
2724     }
2725 
2726     pc_dimm_pre_plug(PC_DIMM(dev), MACHINE(hotplug_dev), NULL, errp);
2727 }
2728 
2729 static void virt_memory_plug(HotplugHandler *hotplug_dev,
2730                              DeviceState *dev, Error **errp)
2731 {
2732     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2733     MachineState *ms = MACHINE(hotplug_dev);
2734     bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2735 
2736     pc_dimm_plug(PC_DIMM(dev), MACHINE(vms));
2737 
2738     if (is_nvdimm) {
2739         nvdimm_plug(ms->nvdimms_state);
2740     }
2741 
2742     hotplug_handler_plug(HOTPLUG_HANDLER(vms->acpi_dev),
2743                          dev, &error_abort);
2744 }
2745 
2746 static void virt_virtio_md_pci_pre_plug(HotplugHandler *hotplug_dev,
2747                                         DeviceState *dev, Error **errp)
2748 {
2749     HotplugHandler *hotplug_dev2 = qdev_get_bus_hotplug_handler(dev);
2750     Error *local_err = NULL;
2751 
2752     if (!hotplug_dev2 && dev->hotplugged) {
2753         /*
2754          * Without a bus hotplug handler, we cannot control the plug/unplug
2755          * order. We should never reach this point when hotplugging on ARM.
2756          * However, it's nice to add a safety net, similar to what we have
2757          * on x86.
2758          */
2759         error_setg(errp, "hotplug of virtio based memory devices not supported"
2760                    " on this bus.");
2761         return;
2762     }
2763     /*
2764      * First, see if we can plug this memory device at all. If that
2765      * succeeds, branch of to the actual hotplug handler.
2766      */
2767     memory_device_pre_plug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev), NULL,
2768                            &local_err);
2769     if (!local_err && hotplug_dev2) {
2770         hotplug_handler_pre_plug(hotplug_dev2, dev, &local_err);
2771     }
2772     error_propagate(errp, local_err);
2773 }
2774 
2775 static void virt_virtio_md_pci_plug(HotplugHandler *hotplug_dev,
2776                                     DeviceState *dev, Error **errp)
2777 {
2778     HotplugHandler *hotplug_dev2 = qdev_get_bus_hotplug_handler(dev);
2779     Error *local_err = NULL;
2780 
2781     /*
2782      * Plug the memory device first and then branch off to the actual
2783      * hotplug handler. If that one fails, we can easily undo the memory
2784      * device bits.
2785      */
2786     memory_device_plug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev));
2787     if (hotplug_dev2) {
2788         hotplug_handler_plug(hotplug_dev2, dev, &local_err);
2789         if (local_err) {
2790             memory_device_unplug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev));
2791         }
2792     }
2793     error_propagate(errp, local_err);
2794 }
2795 
2796 static void virt_virtio_md_pci_unplug_request(HotplugHandler *hotplug_dev,
2797                                               DeviceState *dev, Error **errp)
2798 {
2799     /* We don't support hot unplug of virtio based memory devices */
2800     error_setg(errp, "virtio based memory devices cannot be unplugged.");
2801 }
2802 
2803 
2804 static void virt_machine_device_pre_plug_cb(HotplugHandler *hotplug_dev,
2805                                             DeviceState *dev, Error **errp)
2806 {
2807     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2808 
2809     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2810         virt_memory_pre_plug(hotplug_dev, dev, errp);
2811     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI)) {
2812         virt_virtio_md_pci_pre_plug(hotplug_dev, dev, errp);
2813     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2814         hwaddr db_start = 0, db_end = 0;
2815         char *resv_prop_str;
2816 
2817         if (vms->iommu != VIRT_IOMMU_NONE) {
2818             error_setg(errp, "virt machine does not support multiple IOMMUs");
2819             return;
2820         }
2821 
2822         switch (vms->msi_controller) {
2823         case VIRT_MSI_CTRL_NONE:
2824             return;
2825         case VIRT_MSI_CTRL_ITS:
2826             /* GITS_TRANSLATER page */
2827             db_start = base_memmap[VIRT_GIC_ITS].base + 0x10000;
2828             db_end = base_memmap[VIRT_GIC_ITS].base +
2829                      base_memmap[VIRT_GIC_ITS].size - 1;
2830             break;
2831         case VIRT_MSI_CTRL_GICV2M:
2832             /* MSI_SETSPI_NS page */
2833             db_start = base_memmap[VIRT_GIC_V2M].base;
2834             db_end = db_start + base_memmap[VIRT_GIC_V2M].size - 1;
2835             break;
2836         }
2837         resv_prop_str = g_strdup_printf("0x%"PRIx64":0x%"PRIx64":%u",
2838                                         db_start, db_end,
2839                                         VIRTIO_IOMMU_RESV_MEM_T_MSI);
2840 
2841         object_property_set_uint(OBJECT(dev), "len-reserved-regions", 1, errp);
2842         object_property_set_str(OBJECT(dev), "reserved-regions[0]",
2843                                 resv_prop_str, errp);
2844         g_free(resv_prop_str);
2845     }
2846 }
2847 
2848 static void virt_machine_device_plug_cb(HotplugHandler *hotplug_dev,
2849                                         DeviceState *dev, Error **errp)
2850 {
2851     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2852 
2853     if (vms->platform_bus_dev) {
2854         MachineClass *mc = MACHINE_GET_CLASS(vms);
2855 
2856         if (device_is_dynamic_sysbus(mc, dev)) {
2857             platform_bus_link_device(PLATFORM_BUS_DEVICE(vms->platform_bus_dev),
2858                                      SYS_BUS_DEVICE(dev));
2859         }
2860     }
2861     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2862         virt_memory_plug(hotplug_dev, dev, errp);
2863     }
2864 
2865     if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI)) {
2866         virt_virtio_md_pci_plug(hotplug_dev, dev, errp);
2867     }
2868 
2869     if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2870         PCIDevice *pdev = PCI_DEVICE(dev);
2871 
2872         vms->iommu = VIRT_IOMMU_VIRTIO;
2873         vms->virtio_iommu_bdf = pci_get_bdf(pdev);
2874         create_virtio_iommu_dt_bindings(vms);
2875     }
2876 }
2877 
2878 static void virt_dimm_unplug_request(HotplugHandler *hotplug_dev,
2879                                      DeviceState *dev, Error **errp)
2880 {
2881     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2882 
2883     if (!vms->acpi_dev) {
2884         error_setg(errp,
2885                    "memory hotplug is not enabled: missing acpi-ged device");
2886         return;
2887     }
2888 
2889     if (object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM)) {
2890         error_setg(errp, "nvdimm device hot unplug is not supported yet.");
2891         return;
2892     }
2893 
2894     hotplug_handler_unplug_request(HOTPLUG_HANDLER(vms->acpi_dev), dev,
2895                                    errp);
2896 }
2897 
2898 static void virt_dimm_unplug(HotplugHandler *hotplug_dev,
2899                              DeviceState *dev, Error **errp)
2900 {
2901     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2902     Error *local_err = NULL;
2903 
2904     hotplug_handler_unplug(HOTPLUG_HANDLER(vms->acpi_dev), dev, &local_err);
2905     if (local_err) {
2906         goto out;
2907     }
2908 
2909     pc_dimm_unplug(PC_DIMM(dev), MACHINE(vms));
2910     qdev_unrealize(dev);
2911 
2912 out:
2913     error_propagate(errp, local_err);
2914 }
2915 
2916 static void virt_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
2917                                           DeviceState *dev, Error **errp)
2918 {
2919     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2920         virt_dimm_unplug_request(hotplug_dev, dev, errp);
2921     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI)) {
2922         virt_virtio_md_pci_unplug_request(hotplug_dev, dev, errp);
2923     } else {
2924         error_setg(errp, "device unplug request for unsupported device"
2925                    " type: %s", object_get_typename(OBJECT(dev)));
2926     }
2927 }
2928 
2929 static void virt_machine_device_unplug_cb(HotplugHandler *hotplug_dev,
2930                                           DeviceState *dev, Error **errp)
2931 {
2932     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2933         virt_dimm_unplug(hotplug_dev, dev, errp);
2934     } else {
2935         error_setg(errp, "virt: device unplug for unsupported device"
2936                    " type: %s", object_get_typename(OBJECT(dev)));
2937     }
2938 }
2939 
2940 static HotplugHandler *virt_machine_get_hotplug_handler(MachineState *machine,
2941                                                         DeviceState *dev)
2942 {
2943     MachineClass *mc = MACHINE_GET_CLASS(machine);
2944 
2945     if (device_is_dynamic_sysbus(mc, dev) ||
2946         object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) ||
2947         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI) ||
2948         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2949         return HOTPLUG_HANDLER(machine);
2950     }
2951     return NULL;
2952 }
2953 
2954 /*
2955  * for arm64 kvm_type [7-0] encodes the requested number of bits
2956  * in the IPA address space
2957  */
2958 static int virt_kvm_type(MachineState *ms, const char *type_str)
2959 {
2960     VirtMachineState *vms = VIRT_MACHINE(ms);
2961     int max_vm_pa_size, requested_pa_size;
2962     bool fixed_ipa;
2963 
2964     max_vm_pa_size = kvm_arm_get_max_vm_ipa_size(ms, &fixed_ipa);
2965 
2966     /* we freeze the memory map to compute the highest gpa */
2967     virt_set_memmap(vms, max_vm_pa_size);
2968 
2969     requested_pa_size = 64 - clz64(vms->highest_gpa);
2970 
2971     /*
2972      * KVM requires the IPA size to be at least 32 bits.
2973      */
2974     if (requested_pa_size < 32) {
2975         requested_pa_size = 32;
2976     }
2977 
2978     if (requested_pa_size > max_vm_pa_size) {
2979         error_report("-m and ,maxmem option values "
2980                      "require an IPA range (%d bits) larger than "
2981                      "the one supported by the host (%d bits)",
2982                      requested_pa_size, max_vm_pa_size);
2983         exit(1);
2984     }
2985     /*
2986      * We return the requested PA log size, unless KVM only supports
2987      * the implicit legacy 40b IPA setting, in which case the kvm_type
2988      * must be 0.
2989      */
2990     return fixed_ipa ? 0 : requested_pa_size;
2991 }
2992 
2993 static void virt_machine_class_init(ObjectClass *oc, void *data)
2994 {
2995     MachineClass *mc = MACHINE_CLASS(oc);
2996     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
2997 
2998     mc->init = machvirt_init;
2999     /* Start with max_cpus set to 512, which is the maximum supported by KVM.
3000      * The value may be reduced later when we have more information about the
3001      * configuration of the particular instance.
3002      */
3003     mc->max_cpus = 512;
3004     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_CALXEDA_XGMAC);
3005     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_AMD_XGBE);
3006     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_RAMFB_DEVICE);
3007     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_PLATFORM);
3008 #ifdef CONFIG_TPM
3009     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_TPM_TIS_SYSBUS);
3010 #endif
3011     mc->block_default_type = IF_VIRTIO;
3012     mc->no_cdrom = 1;
3013     mc->pci_allow_0_address = true;
3014     /* We know we will never create a pre-ARMv7 CPU which needs 1K pages */
3015     mc->minimum_page_bits = 12;
3016     mc->possible_cpu_arch_ids = virt_possible_cpu_arch_ids;
3017     mc->cpu_index_to_instance_props = virt_cpu_index_to_props;
3018 #ifdef CONFIG_TCG
3019     mc->default_cpu_type = ARM_CPU_TYPE_NAME("cortex-a15");
3020 #else
3021     mc->default_cpu_type = ARM_CPU_TYPE_NAME("max");
3022 #endif
3023     mc->get_default_cpu_node_id = virt_get_default_cpu_node_id;
3024     mc->kvm_type = virt_kvm_type;
3025     assert(!mc->get_hotplug_handler);
3026     mc->get_hotplug_handler = virt_machine_get_hotplug_handler;
3027     hc->pre_plug = virt_machine_device_pre_plug_cb;
3028     hc->plug = virt_machine_device_plug_cb;
3029     hc->unplug_request = virt_machine_device_unplug_request_cb;
3030     hc->unplug = virt_machine_device_unplug_cb;
3031     mc->nvdimm_supported = true;
3032     mc->smp_props.clusters_supported = true;
3033     mc->auto_enable_numa_with_memhp = true;
3034     mc->auto_enable_numa_with_memdev = true;
3035     mc->default_ram_id = "mach-virt.ram";
3036 
3037     object_class_property_add(oc, "acpi", "OnOffAuto",
3038         virt_get_acpi, virt_set_acpi,
3039         NULL, NULL);
3040     object_class_property_set_description(oc, "acpi",
3041         "Enable ACPI");
3042     object_class_property_add_bool(oc, "secure", virt_get_secure,
3043                                    virt_set_secure);
3044     object_class_property_set_description(oc, "secure",
3045                                                 "Set on/off to enable/disable the ARM "
3046                                                 "Security Extensions (TrustZone)");
3047 
3048     object_class_property_add_bool(oc, "virtualization", virt_get_virt,
3049                                    virt_set_virt);
3050     object_class_property_set_description(oc, "virtualization",
3051                                           "Set on/off to enable/disable emulating a "
3052                                           "guest CPU which implements the ARM "
3053                                           "Virtualization Extensions");
3054 
3055     object_class_property_add_bool(oc, "highmem", virt_get_highmem,
3056                                    virt_set_highmem);
3057     object_class_property_set_description(oc, "highmem",
3058                                           "Set on/off to enable/disable using "
3059                                           "physical address space above 32 bits");
3060 
3061     object_class_property_add_bool(oc, "compact-highmem",
3062                                    virt_get_compact_highmem,
3063                                    virt_set_compact_highmem);
3064     object_class_property_set_description(oc, "compact-highmem",
3065                                           "Set on/off to enable/disable compact "
3066                                           "layout for high memory regions");
3067 
3068     object_class_property_add_bool(oc, "highmem-redists",
3069                                    virt_get_highmem_redists,
3070                                    virt_set_highmem_redists);
3071     object_class_property_set_description(oc, "highmem-redists",
3072                                           "Set on/off to enable/disable high "
3073                                           "memory region for GICv3 or GICv4 "
3074                                           "redistributor");
3075 
3076     object_class_property_add_bool(oc, "highmem-ecam",
3077                                    virt_get_highmem_ecam,
3078                                    virt_set_highmem_ecam);
3079     object_class_property_set_description(oc, "highmem-ecam",
3080                                           "Set on/off to enable/disable high "
3081                                           "memory region for PCI ECAM");
3082 
3083     object_class_property_add_bool(oc, "highmem-mmio",
3084                                    virt_get_highmem_mmio,
3085                                    virt_set_highmem_mmio);
3086     object_class_property_set_description(oc, "highmem-mmio",
3087                                           "Set on/off to enable/disable high "
3088                                           "memory region for PCI MMIO");
3089 
3090     object_class_property_add_str(oc, "gic-version", virt_get_gic_version,
3091                                   virt_set_gic_version);
3092     object_class_property_set_description(oc, "gic-version",
3093                                           "Set GIC version. "
3094                                           "Valid values are 2, 3, 4, host and max");
3095 
3096     object_class_property_add_str(oc, "iommu", virt_get_iommu, virt_set_iommu);
3097     object_class_property_set_description(oc, "iommu",
3098                                           "Set the IOMMU type. "
3099                                           "Valid values are none and smmuv3");
3100 
3101     object_class_property_add_bool(oc, "default-bus-bypass-iommu",
3102                                    virt_get_default_bus_bypass_iommu,
3103                                    virt_set_default_bus_bypass_iommu);
3104     object_class_property_set_description(oc, "default-bus-bypass-iommu",
3105                                           "Set on/off to enable/disable "
3106                                           "bypass_iommu for default root bus");
3107 
3108     object_class_property_add_bool(oc, "ras", virt_get_ras,
3109                                    virt_set_ras);
3110     object_class_property_set_description(oc, "ras",
3111                                           "Set on/off to enable/disable reporting host memory errors "
3112                                           "to a KVM guest using ACPI and guest external abort exceptions");
3113 
3114     object_class_property_add_bool(oc, "mte", virt_get_mte, virt_set_mte);
3115     object_class_property_set_description(oc, "mte",
3116                                           "Set on/off to enable/disable emulating a "
3117                                           "guest CPU which implements the ARM "
3118                                           "Memory Tagging Extension");
3119 
3120     object_class_property_add_bool(oc, "its", virt_get_its,
3121                                    virt_set_its);
3122     object_class_property_set_description(oc, "its",
3123                                           "Set on/off to enable/disable "
3124                                           "ITS instantiation");
3125 
3126     object_class_property_add_bool(oc, "dtb-randomness",
3127                                    virt_get_dtb_randomness,
3128                                    virt_set_dtb_randomness);
3129     object_class_property_set_description(oc, "dtb-randomness",
3130                                           "Set off to disable passing random or "
3131                                           "non-deterministic dtb nodes to guest");
3132 
3133     object_class_property_add_bool(oc, "dtb-kaslr-seed",
3134                                    virt_get_dtb_randomness,
3135                                    virt_set_dtb_randomness);
3136     object_class_property_set_description(oc, "dtb-kaslr-seed",
3137                                           "Deprecated synonym of dtb-randomness");
3138 
3139     object_class_property_add_str(oc, "x-oem-id",
3140                                   virt_get_oem_id,
3141                                   virt_set_oem_id);
3142     object_class_property_set_description(oc, "x-oem-id",
3143                                           "Override the default value of field OEMID "
3144                                           "in ACPI table header."
3145                                           "The string may be up to 6 bytes in size");
3146 
3147 
3148     object_class_property_add_str(oc, "x-oem-table-id",
3149                                   virt_get_oem_table_id,
3150                                   virt_set_oem_table_id);
3151     object_class_property_set_description(oc, "x-oem-table-id",
3152                                           "Override the default value of field OEM Table ID "
3153                                           "in ACPI table header."
3154                                           "The string may be up to 8 bytes in size");
3155 
3156 }
3157 
3158 static void virt_instance_init(Object *obj)
3159 {
3160     VirtMachineState *vms = VIRT_MACHINE(obj);
3161     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
3162 
3163     /* EL3 is disabled by default on virt: this makes us consistent
3164      * between KVM and TCG for this board, and it also allows us to
3165      * boot UEFI blobs which assume no TrustZone support.
3166      */
3167     vms->secure = false;
3168 
3169     /* EL2 is also disabled by default, for similar reasons */
3170     vms->virt = false;
3171 
3172     /* High memory is enabled by default */
3173     vms->highmem = true;
3174     vms->highmem_compact = !vmc->no_highmem_compact;
3175     vms->gic_version = VIRT_GIC_VERSION_NOSEL;
3176 
3177     vms->highmem_ecam = !vmc->no_highmem_ecam;
3178     vms->highmem_mmio = true;
3179     vms->highmem_redists = true;
3180 
3181     if (vmc->no_its) {
3182         vms->its = false;
3183     } else {
3184         /* Default allows ITS instantiation */
3185         vms->its = true;
3186 
3187         if (vmc->no_tcg_its) {
3188             vms->tcg_its = false;
3189         } else {
3190             vms->tcg_its = true;
3191         }
3192     }
3193 
3194     /* Default disallows iommu instantiation */
3195     vms->iommu = VIRT_IOMMU_NONE;
3196 
3197     /* The default root bus is attached to iommu by default */
3198     vms->default_bus_bypass_iommu = false;
3199 
3200     /* Default disallows RAS instantiation */
3201     vms->ras = false;
3202 
3203     /* MTE is disabled by default.  */
3204     vms->mte = false;
3205 
3206     /* Supply kaslr-seed and rng-seed by default */
3207     vms->dtb_randomness = true;
3208 
3209     vms->irqmap = a15irqmap;
3210 
3211     virt_flash_create(vms);
3212 
3213     vms->oem_id = g_strndup(ACPI_BUILD_APPNAME6, 6);
3214     vms->oem_table_id = g_strndup(ACPI_BUILD_APPNAME8, 8);
3215 }
3216 
3217 static const TypeInfo virt_machine_info = {
3218     .name          = TYPE_VIRT_MACHINE,
3219     .parent        = TYPE_MACHINE,
3220     .abstract      = true,
3221     .instance_size = sizeof(VirtMachineState),
3222     .class_size    = sizeof(VirtMachineClass),
3223     .class_init    = virt_machine_class_init,
3224     .instance_init = virt_instance_init,
3225     .interfaces = (InterfaceInfo[]) {
3226          { TYPE_HOTPLUG_HANDLER },
3227          { }
3228     },
3229 };
3230 
3231 static void machvirt_machine_init(void)
3232 {
3233     type_register_static(&virt_machine_info);
3234 }
3235 type_init(machvirt_machine_init);
3236 
3237 static void virt_machine_8_1_options(MachineClass *mc)
3238 {
3239 }
3240 DEFINE_VIRT_MACHINE_AS_LATEST(8, 1)
3241 
3242 static void virt_machine_8_0_options(MachineClass *mc)
3243 {
3244     virt_machine_8_1_options(mc);
3245     compat_props_add(mc->compat_props, hw_compat_8_0, hw_compat_8_0_len);
3246 }
3247 DEFINE_VIRT_MACHINE(8, 0)
3248 
3249 static void virt_machine_7_2_options(MachineClass *mc)
3250 {
3251     virt_machine_8_0_options(mc);
3252     compat_props_add(mc->compat_props, hw_compat_7_2, hw_compat_7_2_len);
3253 }
3254 DEFINE_VIRT_MACHINE(7, 2)
3255 
3256 static void virt_machine_7_1_options(MachineClass *mc)
3257 {
3258     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3259 
3260     virt_machine_7_2_options(mc);
3261     compat_props_add(mc->compat_props, hw_compat_7_1, hw_compat_7_1_len);
3262     /* Compact layout for high memory regions was introduced with 7.2 */
3263     vmc->no_highmem_compact = true;
3264 }
3265 DEFINE_VIRT_MACHINE(7, 1)
3266 
3267 static void virt_machine_7_0_options(MachineClass *mc)
3268 {
3269     virt_machine_7_1_options(mc);
3270     compat_props_add(mc->compat_props, hw_compat_7_0, hw_compat_7_0_len);
3271 }
3272 DEFINE_VIRT_MACHINE(7, 0)
3273 
3274 static void virt_machine_6_2_options(MachineClass *mc)
3275 {
3276     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3277 
3278     virt_machine_7_0_options(mc);
3279     compat_props_add(mc->compat_props, hw_compat_6_2, hw_compat_6_2_len);
3280     vmc->no_tcg_lpa2 = true;
3281 }
3282 DEFINE_VIRT_MACHINE(6, 2)
3283 
3284 static void virt_machine_6_1_options(MachineClass *mc)
3285 {
3286     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3287 
3288     virt_machine_6_2_options(mc);
3289     compat_props_add(mc->compat_props, hw_compat_6_1, hw_compat_6_1_len);
3290     mc->smp_props.prefer_sockets = true;
3291     vmc->no_cpu_topology = true;
3292 
3293     /* qemu ITS was introduced with 6.2 */
3294     vmc->no_tcg_its = true;
3295 }
3296 DEFINE_VIRT_MACHINE(6, 1)
3297 
3298 static void virt_machine_6_0_options(MachineClass *mc)
3299 {
3300     virt_machine_6_1_options(mc);
3301     compat_props_add(mc->compat_props, hw_compat_6_0, hw_compat_6_0_len);
3302 }
3303 DEFINE_VIRT_MACHINE(6, 0)
3304 
3305 static void virt_machine_5_2_options(MachineClass *mc)
3306 {
3307     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3308 
3309     virt_machine_6_0_options(mc);
3310     compat_props_add(mc->compat_props, hw_compat_5_2, hw_compat_5_2_len);
3311     vmc->no_secure_gpio = true;
3312 }
3313 DEFINE_VIRT_MACHINE(5, 2)
3314 
3315 static void virt_machine_5_1_options(MachineClass *mc)
3316 {
3317     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3318 
3319     virt_machine_5_2_options(mc);
3320     compat_props_add(mc->compat_props, hw_compat_5_1, hw_compat_5_1_len);
3321     vmc->no_kvm_steal_time = true;
3322 }
3323 DEFINE_VIRT_MACHINE(5, 1)
3324 
3325 static void virt_machine_5_0_options(MachineClass *mc)
3326 {
3327     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3328 
3329     virt_machine_5_1_options(mc);
3330     compat_props_add(mc->compat_props, hw_compat_5_0, hw_compat_5_0_len);
3331     mc->numa_mem_supported = true;
3332     vmc->acpi_expose_flash = true;
3333     mc->auto_enable_numa_with_memdev = false;
3334 }
3335 DEFINE_VIRT_MACHINE(5, 0)
3336 
3337 static void virt_machine_4_2_options(MachineClass *mc)
3338 {
3339     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3340 
3341     virt_machine_5_0_options(mc);
3342     compat_props_add(mc->compat_props, hw_compat_4_2, hw_compat_4_2_len);
3343     vmc->kvm_no_adjvtime = true;
3344 }
3345 DEFINE_VIRT_MACHINE(4, 2)
3346 
3347 static void virt_machine_4_1_options(MachineClass *mc)
3348 {
3349     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3350 
3351     virt_machine_4_2_options(mc);
3352     compat_props_add(mc->compat_props, hw_compat_4_1, hw_compat_4_1_len);
3353     vmc->no_ged = true;
3354     mc->auto_enable_numa_with_memhp = false;
3355 }
3356 DEFINE_VIRT_MACHINE(4, 1)
3357 
3358 static void virt_machine_4_0_options(MachineClass *mc)
3359 {
3360     virt_machine_4_1_options(mc);
3361     compat_props_add(mc->compat_props, hw_compat_4_0, hw_compat_4_0_len);
3362 }
3363 DEFINE_VIRT_MACHINE(4, 0)
3364 
3365 static void virt_machine_3_1_options(MachineClass *mc)
3366 {
3367     virt_machine_4_0_options(mc);
3368     compat_props_add(mc->compat_props, hw_compat_3_1, hw_compat_3_1_len);
3369 }
3370 DEFINE_VIRT_MACHINE(3, 1)
3371 
3372 static void virt_machine_3_0_options(MachineClass *mc)
3373 {
3374     virt_machine_3_1_options(mc);
3375     compat_props_add(mc->compat_props, hw_compat_3_0, hw_compat_3_0_len);
3376 }
3377 DEFINE_VIRT_MACHINE(3, 0)
3378 
3379 static void virt_machine_2_12_options(MachineClass *mc)
3380 {
3381     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3382 
3383     virt_machine_3_0_options(mc);
3384     compat_props_add(mc->compat_props, hw_compat_2_12, hw_compat_2_12_len);
3385     vmc->no_highmem_ecam = true;
3386     mc->max_cpus = 255;
3387 }
3388 DEFINE_VIRT_MACHINE(2, 12)
3389 
3390 static void virt_machine_2_11_options(MachineClass *mc)
3391 {
3392     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3393 
3394     virt_machine_2_12_options(mc);
3395     compat_props_add(mc->compat_props, hw_compat_2_11, hw_compat_2_11_len);
3396     vmc->smbios_old_sys_ver = true;
3397 }
3398 DEFINE_VIRT_MACHINE(2, 11)
3399 
3400 static void virt_machine_2_10_options(MachineClass *mc)
3401 {
3402     virt_machine_2_11_options(mc);
3403     compat_props_add(mc->compat_props, hw_compat_2_10, hw_compat_2_10_len);
3404     /* before 2.11 we never faulted accesses to bad addresses */
3405     mc->ignore_memory_transaction_failures = true;
3406 }
3407 DEFINE_VIRT_MACHINE(2, 10)
3408 
3409 static void virt_machine_2_9_options(MachineClass *mc)
3410 {
3411     virt_machine_2_10_options(mc);
3412     compat_props_add(mc->compat_props, hw_compat_2_9, hw_compat_2_9_len);
3413 }
3414 DEFINE_VIRT_MACHINE(2, 9)
3415 
3416 static void virt_machine_2_8_options(MachineClass *mc)
3417 {
3418     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3419 
3420     virt_machine_2_9_options(mc);
3421     compat_props_add(mc->compat_props, hw_compat_2_8, hw_compat_2_8_len);
3422     /* For 2.8 and earlier we falsely claimed in the DT that
3423      * our timers were edge-triggered, not level-triggered.
3424      */
3425     vmc->claim_edge_triggered_timers = true;
3426 }
3427 DEFINE_VIRT_MACHINE(2, 8)
3428 
3429 static void virt_machine_2_7_options(MachineClass *mc)
3430 {
3431     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3432 
3433     virt_machine_2_8_options(mc);
3434     compat_props_add(mc->compat_props, hw_compat_2_7, hw_compat_2_7_len);
3435     /* ITS was introduced with 2.8 */
3436     vmc->no_its = true;
3437     /* Stick with 1K pages for migration compatibility */
3438     mc->minimum_page_bits = 0;
3439 }
3440 DEFINE_VIRT_MACHINE(2, 7)
3441 
3442 static void virt_machine_2_6_options(MachineClass *mc)
3443 {
3444     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3445 
3446     virt_machine_2_7_options(mc);
3447     compat_props_add(mc->compat_props, hw_compat_2_6, hw_compat_2_6_len);
3448     vmc->disallow_affinity_adjustment = true;
3449     /* Disable PMU for 2.6 as PMU support was first introduced in 2.7 */
3450     vmc->no_pmu = true;
3451 }
3452 DEFINE_VIRT_MACHINE(2, 6)
3453