xref: /qemu/hw/arm/virt.c (revision 60f782b6)
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     ARM_CPU_TYPE_NAME("cortex-a15"),
210     ARM_CPU_TYPE_NAME("cortex-a35"),
211     ARM_CPU_TYPE_NAME("cortex-a55"),
212     ARM_CPU_TYPE_NAME("cortex-a72"),
213     ARM_CPU_TYPE_NAME("cortex-a76"),
214     ARM_CPU_TYPE_NAME("a64fx"),
215     ARM_CPU_TYPE_NAME("neoverse-n1"),
216 #endif
217     ARM_CPU_TYPE_NAME("cortex-a53"),
218     ARM_CPU_TYPE_NAME("cortex-a57"),
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     MachineClass *mc = MACHINE_GET_CLASS(ms);
1430 
1431     dev = qdev_new(TYPE_GPEX_HOST);
1432     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1433 
1434     ecam_id = VIRT_ECAM_ID(vms->highmem_ecam);
1435     base_ecam = vms->memmap[ecam_id].base;
1436     size_ecam = vms->memmap[ecam_id].size;
1437     nr_pcie_buses = size_ecam / PCIE_MMCFG_SIZE_MIN;
1438     /* Map only the first size_ecam bytes of ECAM space */
1439     ecam_alias = g_new0(MemoryRegion, 1);
1440     ecam_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 0);
1441     memory_region_init_alias(ecam_alias, OBJECT(dev), "pcie-ecam",
1442                              ecam_reg, 0, size_ecam);
1443     memory_region_add_subregion(get_system_memory(), base_ecam, ecam_alias);
1444 
1445     /* Map the MMIO window into system address space so as to expose
1446      * the section of PCI MMIO space which starts at the same base address
1447      * (ie 1:1 mapping for that part of PCI MMIO space visible through
1448      * the window).
1449      */
1450     mmio_alias = g_new0(MemoryRegion, 1);
1451     mmio_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 1);
1452     memory_region_init_alias(mmio_alias, OBJECT(dev), "pcie-mmio",
1453                              mmio_reg, base_mmio, size_mmio);
1454     memory_region_add_subregion(get_system_memory(), base_mmio, mmio_alias);
1455 
1456     if (vms->highmem_mmio) {
1457         /* Map high MMIO space */
1458         MemoryRegion *high_mmio_alias = g_new0(MemoryRegion, 1);
1459 
1460         memory_region_init_alias(high_mmio_alias, OBJECT(dev), "pcie-mmio-high",
1461                                  mmio_reg, base_mmio_high, size_mmio_high);
1462         memory_region_add_subregion(get_system_memory(), base_mmio_high,
1463                                     high_mmio_alias);
1464     }
1465 
1466     /* Map IO port space */
1467     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 2, base_pio);
1468 
1469     for (i = 0; i < GPEX_NUM_IRQS; i++) {
1470         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
1471                            qdev_get_gpio_in(vms->gic, irq + i));
1472         gpex_set_irq_num(GPEX_HOST(dev), i, irq + i);
1473     }
1474 
1475     pci = PCI_HOST_BRIDGE(dev);
1476     pci->bypass_iommu = vms->default_bus_bypass_iommu;
1477     vms->bus = pci->bus;
1478     if (vms->bus) {
1479         for (i = 0; i < nb_nics; i++) {
1480             NICInfo *nd = &nd_table[i];
1481 
1482             if (!nd->model) {
1483                 nd->model = g_strdup(mc->default_nic);
1484             }
1485 
1486             pci_nic_init_nofail(nd, pci->bus, nd->model, NULL);
1487         }
1488     }
1489 
1490     nodename = vms->pciehb_nodename = g_strdup_printf("/pcie@%" PRIx64, base);
1491     qemu_fdt_add_subnode(ms->fdt, nodename);
1492     qemu_fdt_setprop_string(ms->fdt, nodename,
1493                             "compatible", "pci-host-ecam-generic");
1494     qemu_fdt_setprop_string(ms->fdt, nodename, "device_type", "pci");
1495     qemu_fdt_setprop_cell(ms->fdt, nodename, "#address-cells", 3);
1496     qemu_fdt_setprop_cell(ms->fdt, nodename, "#size-cells", 2);
1497     qemu_fdt_setprop_cell(ms->fdt, nodename, "linux,pci-domain", 0);
1498     qemu_fdt_setprop_cells(ms->fdt, nodename, "bus-range", 0,
1499                            nr_pcie_buses - 1);
1500     qemu_fdt_setprop(ms->fdt, nodename, "dma-coherent", NULL, 0);
1501 
1502     if (vms->msi_phandle) {
1503         qemu_fdt_setprop_cells(ms->fdt, nodename, "msi-map",
1504                                0, vms->msi_phandle, 0, 0x10000);
1505     }
1506 
1507     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1508                                  2, base_ecam, 2, size_ecam);
1509 
1510     if (vms->highmem_mmio) {
1511         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "ranges",
1512                                      1, FDT_PCI_RANGE_IOPORT, 2, 0,
1513                                      2, base_pio, 2, size_pio,
1514                                      1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
1515                                      2, base_mmio, 2, size_mmio,
1516                                      1, FDT_PCI_RANGE_MMIO_64BIT,
1517                                      2, base_mmio_high,
1518                                      2, base_mmio_high, 2, size_mmio_high);
1519     } else {
1520         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "ranges",
1521                                      1, FDT_PCI_RANGE_IOPORT, 2, 0,
1522                                      2, base_pio, 2, size_pio,
1523                                      1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
1524                                      2, base_mmio, 2, size_mmio);
1525     }
1526 
1527     qemu_fdt_setprop_cell(ms->fdt, nodename, "#interrupt-cells", 1);
1528     create_pcie_irq_map(ms, vms->gic_phandle, irq, nodename);
1529 
1530     if (vms->iommu) {
1531         vms->iommu_phandle = qemu_fdt_alloc_phandle(ms->fdt);
1532 
1533         switch (vms->iommu) {
1534         case VIRT_IOMMU_SMMUV3:
1535             create_smmu(vms, vms->bus);
1536             qemu_fdt_setprop_cells(ms->fdt, nodename, "iommu-map",
1537                                    0x0, vms->iommu_phandle, 0x0, 0x10000);
1538             break;
1539         default:
1540             g_assert_not_reached();
1541         }
1542     }
1543 }
1544 
1545 static void create_platform_bus(VirtMachineState *vms)
1546 {
1547     DeviceState *dev;
1548     SysBusDevice *s;
1549     int i;
1550     MemoryRegion *sysmem = get_system_memory();
1551 
1552     dev = qdev_new(TYPE_PLATFORM_BUS_DEVICE);
1553     dev->id = g_strdup(TYPE_PLATFORM_BUS_DEVICE);
1554     qdev_prop_set_uint32(dev, "num_irqs", PLATFORM_BUS_NUM_IRQS);
1555     qdev_prop_set_uint32(dev, "mmio_size", vms->memmap[VIRT_PLATFORM_BUS].size);
1556     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1557     vms->platform_bus_dev = dev;
1558 
1559     s = SYS_BUS_DEVICE(dev);
1560     for (i = 0; i < PLATFORM_BUS_NUM_IRQS; i++) {
1561         int irq = vms->irqmap[VIRT_PLATFORM_BUS] + i;
1562         sysbus_connect_irq(s, i, qdev_get_gpio_in(vms->gic, irq));
1563     }
1564 
1565     memory_region_add_subregion(sysmem,
1566                                 vms->memmap[VIRT_PLATFORM_BUS].base,
1567                                 sysbus_mmio_get_region(s, 0));
1568 }
1569 
1570 static void create_tag_ram(MemoryRegion *tag_sysmem,
1571                            hwaddr base, hwaddr size,
1572                            const char *name)
1573 {
1574     MemoryRegion *tagram = g_new(MemoryRegion, 1);
1575 
1576     memory_region_init_ram(tagram, NULL, name, size / 32, &error_fatal);
1577     memory_region_add_subregion(tag_sysmem, base / 32, tagram);
1578 }
1579 
1580 static void create_secure_ram(VirtMachineState *vms,
1581                               MemoryRegion *secure_sysmem,
1582                               MemoryRegion *secure_tag_sysmem)
1583 {
1584     MemoryRegion *secram = g_new(MemoryRegion, 1);
1585     char *nodename;
1586     hwaddr base = vms->memmap[VIRT_SECURE_MEM].base;
1587     hwaddr size = vms->memmap[VIRT_SECURE_MEM].size;
1588     MachineState *ms = MACHINE(vms);
1589 
1590     memory_region_init_ram(secram, NULL, "virt.secure-ram", size,
1591                            &error_fatal);
1592     memory_region_add_subregion(secure_sysmem, base, secram);
1593 
1594     nodename = g_strdup_printf("/secram@%" PRIx64, base);
1595     qemu_fdt_add_subnode(ms->fdt, nodename);
1596     qemu_fdt_setprop_string(ms->fdt, nodename, "device_type", "memory");
1597     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg", 2, base, 2, size);
1598     qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
1599     qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
1600 
1601     if (secure_tag_sysmem) {
1602         create_tag_ram(secure_tag_sysmem, base, size, "mach-virt.secure-tag");
1603     }
1604 
1605     g_free(nodename);
1606 }
1607 
1608 static void *machvirt_dtb(const struct arm_boot_info *binfo, int *fdt_size)
1609 {
1610     const VirtMachineState *board = container_of(binfo, VirtMachineState,
1611                                                  bootinfo);
1612     MachineState *ms = MACHINE(board);
1613 
1614 
1615     *fdt_size = board->fdt_size;
1616     return ms->fdt;
1617 }
1618 
1619 static void virt_build_smbios(VirtMachineState *vms)
1620 {
1621     MachineClass *mc = MACHINE_GET_CLASS(vms);
1622     MachineState *ms = MACHINE(vms);
1623     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
1624     uint8_t *smbios_tables, *smbios_anchor;
1625     size_t smbios_tables_len, smbios_anchor_len;
1626     struct smbios_phys_mem_area mem_array;
1627     const char *product = "QEMU Virtual Machine";
1628 
1629     if (kvm_enabled()) {
1630         product = "KVM Virtual Machine";
1631     }
1632 
1633     smbios_set_defaults("QEMU", product,
1634                         vmc->smbios_old_sys_ver ? "1.0" : mc->name, false,
1635                         true, SMBIOS_ENTRY_POINT_TYPE_64);
1636 
1637     /* build the array of physical mem area from base_memmap */
1638     mem_array.address = vms->memmap[VIRT_MEM].base;
1639     mem_array.length = ms->ram_size;
1640 
1641     smbios_get_tables(ms, &mem_array, 1,
1642                       &smbios_tables, &smbios_tables_len,
1643                       &smbios_anchor, &smbios_anchor_len,
1644                       &error_fatal);
1645 
1646     if (smbios_anchor) {
1647         fw_cfg_add_file(vms->fw_cfg, "etc/smbios/smbios-tables",
1648                         smbios_tables, smbios_tables_len);
1649         fw_cfg_add_file(vms->fw_cfg, "etc/smbios/smbios-anchor",
1650                         smbios_anchor, smbios_anchor_len);
1651     }
1652 }
1653 
1654 static
1655 void virt_machine_done(Notifier *notifier, void *data)
1656 {
1657     VirtMachineState *vms = container_of(notifier, VirtMachineState,
1658                                          machine_done);
1659     MachineState *ms = MACHINE(vms);
1660     ARMCPU *cpu = ARM_CPU(first_cpu);
1661     struct arm_boot_info *info = &vms->bootinfo;
1662     AddressSpace *as = arm_boot_address_space(cpu, info);
1663 
1664     /*
1665      * If the user provided a dtb, we assume the dynamic sysbus nodes
1666      * already are integrated there. This corresponds to a use case where
1667      * the dynamic sysbus nodes are complex and their generation is not yet
1668      * supported. In that case the user can take charge of the guest dt
1669      * while qemu takes charge of the qom stuff.
1670      */
1671     if (info->dtb_filename == NULL) {
1672         platform_bus_add_all_fdt_nodes(ms->fdt, "/intc",
1673                                        vms->memmap[VIRT_PLATFORM_BUS].base,
1674                                        vms->memmap[VIRT_PLATFORM_BUS].size,
1675                                        vms->irqmap[VIRT_PLATFORM_BUS]);
1676     }
1677     if (arm_load_dtb(info->dtb_start, info, info->dtb_limit, as, ms) < 0) {
1678         exit(1);
1679     }
1680 
1681     fw_cfg_add_extra_pci_roots(vms->bus, vms->fw_cfg);
1682 
1683     virt_acpi_setup(vms);
1684     virt_build_smbios(vms);
1685 }
1686 
1687 static uint64_t virt_cpu_mp_affinity(VirtMachineState *vms, int idx)
1688 {
1689     uint8_t clustersz = ARM_DEFAULT_CPUS_PER_CLUSTER;
1690     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
1691 
1692     if (!vmc->disallow_affinity_adjustment) {
1693         /* Adjust MPIDR like 64-bit KVM hosts, which incorporate the
1694          * GIC's target-list limitations. 32-bit KVM hosts currently
1695          * always create clusters of 4 CPUs, but that is expected to
1696          * change when they gain support for gicv3. When KVM is enabled
1697          * it will override the changes we make here, therefore our
1698          * purposes are to make TCG consistent (with 64-bit KVM hosts)
1699          * and to improve SGI efficiency.
1700          */
1701         if (vms->gic_version == VIRT_GIC_VERSION_2) {
1702             clustersz = GIC_TARGETLIST_BITS;
1703         } else {
1704             clustersz = GICV3_TARGETLIST_BITS;
1705         }
1706     }
1707     return arm_cpu_mp_affinity(idx, clustersz);
1708 }
1709 
1710 static inline bool *virt_get_high_memmap_enabled(VirtMachineState *vms,
1711                                                  int index)
1712 {
1713     bool *enabled_array[] = {
1714         &vms->highmem_redists,
1715         &vms->highmem_ecam,
1716         &vms->highmem_mmio,
1717     };
1718 
1719     assert(ARRAY_SIZE(extended_memmap) - VIRT_LOWMEMMAP_LAST ==
1720            ARRAY_SIZE(enabled_array));
1721     assert(index - VIRT_LOWMEMMAP_LAST < ARRAY_SIZE(enabled_array));
1722 
1723     return enabled_array[index - VIRT_LOWMEMMAP_LAST];
1724 }
1725 
1726 static void virt_set_high_memmap(VirtMachineState *vms,
1727                                  hwaddr base, int pa_bits)
1728 {
1729     hwaddr region_base, region_size;
1730     bool *region_enabled, fits;
1731     int i;
1732 
1733     for (i = VIRT_LOWMEMMAP_LAST; i < ARRAY_SIZE(extended_memmap); i++) {
1734         region_enabled = virt_get_high_memmap_enabled(vms, i);
1735         region_base = ROUND_UP(base, extended_memmap[i].size);
1736         region_size = extended_memmap[i].size;
1737 
1738         vms->memmap[i].base = region_base;
1739         vms->memmap[i].size = region_size;
1740 
1741         /*
1742          * Check each device to see if it fits in the PA space,
1743          * moving highest_gpa as we go. For compatibility, move
1744          * highest_gpa for disabled fitting devices as well, if
1745          * the compact layout has been disabled.
1746          *
1747          * For each device that doesn't fit, disable it.
1748          */
1749         fits = (region_base + region_size) <= BIT_ULL(pa_bits);
1750         *region_enabled &= fits;
1751         if (vms->highmem_compact && !*region_enabled) {
1752             continue;
1753         }
1754 
1755         base = region_base + region_size;
1756         if (fits) {
1757             vms->highest_gpa = base - 1;
1758         }
1759     }
1760 }
1761 
1762 static void virt_set_memmap(VirtMachineState *vms, int pa_bits)
1763 {
1764     MachineState *ms = MACHINE(vms);
1765     hwaddr base, device_memory_base, device_memory_size, memtop;
1766     int i;
1767 
1768     vms->memmap = extended_memmap;
1769 
1770     for (i = 0; i < ARRAY_SIZE(base_memmap); i++) {
1771         vms->memmap[i] = base_memmap[i];
1772     }
1773 
1774     if (ms->ram_slots > ACPI_MAX_RAM_SLOTS) {
1775         error_report("unsupported number of memory slots: %"PRIu64,
1776                      ms->ram_slots);
1777         exit(EXIT_FAILURE);
1778     }
1779 
1780     /*
1781      * !highmem is exactly the same as limiting the PA space to 32bit,
1782      * irrespective of the underlying capabilities of the HW.
1783      */
1784     if (!vms->highmem) {
1785         pa_bits = 32;
1786     }
1787 
1788     /*
1789      * We compute the base of the high IO region depending on the
1790      * amount of initial and device memory. The device memory start/size
1791      * is aligned on 1GiB. We never put the high IO region below 256GiB
1792      * so that if maxram_size is < 255GiB we keep the legacy memory map.
1793      * The device region size assumes 1GiB page max alignment per slot.
1794      */
1795     device_memory_base =
1796         ROUND_UP(vms->memmap[VIRT_MEM].base + ms->ram_size, GiB);
1797     device_memory_size = ms->maxram_size - ms->ram_size + ms->ram_slots * GiB;
1798 
1799     /* Base address of the high IO region */
1800     memtop = base = device_memory_base + ROUND_UP(device_memory_size, GiB);
1801     if (memtop > BIT_ULL(pa_bits)) {
1802 	    error_report("Addressing limited to %d bits, but memory exceeds it by %llu bytes\n",
1803 			 pa_bits, memtop - BIT_ULL(pa_bits));
1804         exit(EXIT_FAILURE);
1805     }
1806     if (base < device_memory_base) {
1807         error_report("maxmem/slots too huge");
1808         exit(EXIT_FAILURE);
1809     }
1810     if (base < vms->memmap[VIRT_MEM].base + LEGACY_RAMLIMIT_BYTES) {
1811         base = vms->memmap[VIRT_MEM].base + LEGACY_RAMLIMIT_BYTES;
1812     }
1813 
1814     /* We know for sure that at least the memory fits in the PA space */
1815     vms->highest_gpa = memtop - 1;
1816 
1817     virt_set_high_memmap(vms, base, pa_bits);
1818 
1819     if (device_memory_size > 0) {
1820         ms->device_memory = g_malloc0(sizeof(*ms->device_memory));
1821         ms->device_memory->base = device_memory_base;
1822         memory_region_init(&ms->device_memory->mr, OBJECT(vms),
1823                            "device-memory", device_memory_size);
1824     }
1825 }
1826 
1827 static VirtGICType finalize_gic_version_do(const char *accel_name,
1828                                            VirtGICType gic_version,
1829                                            int gics_supported,
1830                                            unsigned int max_cpus)
1831 {
1832     /* Convert host/max/nosel to GIC version number */
1833     switch (gic_version) {
1834     case VIRT_GIC_VERSION_HOST:
1835         if (!kvm_enabled()) {
1836             error_report("gic-version=host requires KVM");
1837             exit(1);
1838         }
1839 
1840         /* For KVM, gic-version=host means gic-version=max */
1841         return finalize_gic_version_do(accel_name, VIRT_GIC_VERSION_MAX,
1842                                        gics_supported, max_cpus);
1843     case VIRT_GIC_VERSION_MAX:
1844         if (gics_supported & VIRT_GIC_VERSION_4_MASK) {
1845             gic_version = VIRT_GIC_VERSION_4;
1846         } else if (gics_supported & VIRT_GIC_VERSION_3_MASK) {
1847             gic_version = VIRT_GIC_VERSION_3;
1848         } else {
1849             gic_version = VIRT_GIC_VERSION_2;
1850         }
1851         break;
1852     case VIRT_GIC_VERSION_NOSEL:
1853         if ((gics_supported & VIRT_GIC_VERSION_2_MASK) &&
1854             max_cpus <= GIC_NCPU) {
1855             gic_version = VIRT_GIC_VERSION_2;
1856         } else if (gics_supported & VIRT_GIC_VERSION_3_MASK) {
1857             /*
1858              * in case the host does not support v2 emulation or
1859              * the end-user requested more than 8 VCPUs we now default
1860              * to v3. In any case defaulting to v2 would be broken.
1861              */
1862             gic_version = VIRT_GIC_VERSION_3;
1863         } else if (max_cpus > GIC_NCPU) {
1864             error_report("%s only supports GICv2 emulation but more than 8 "
1865                          "vcpus are requested", accel_name);
1866             exit(1);
1867         }
1868         break;
1869     case VIRT_GIC_VERSION_2:
1870     case VIRT_GIC_VERSION_3:
1871     case VIRT_GIC_VERSION_4:
1872         break;
1873     }
1874 
1875     /* Check chosen version is effectively supported */
1876     switch (gic_version) {
1877     case VIRT_GIC_VERSION_2:
1878         if (!(gics_supported & VIRT_GIC_VERSION_2_MASK)) {
1879             error_report("%s does not support GICv2 emulation", accel_name);
1880             exit(1);
1881         }
1882         break;
1883     case VIRT_GIC_VERSION_3:
1884         if (!(gics_supported & VIRT_GIC_VERSION_3_MASK)) {
1885             error_report("%s does not support GICv3 emulation", accel_name);
1886             exit(1);
1887         }
1888         break;
1889     case VIRT_GIC_VERSION_4:
1890         if (!(gics_supported & VIRT_GIC_VERSION_4_MASK)) {
1891             error_report("%s does not support GICv4 emulation, is virtualization=on?",
1892                          accel_name);
1893             exit(1);
1894         }
1895         break;
1896     default:
1897         error_report("logic error in finalize_gic_version");
1898         exit(1);
1899         break;
1900     }
1901 
1902     return gic_version;
1903 }
1904 
1905 /*
1906  * finalize_gic_version - Determines the final gic_version
1907  * according to the gic-version property
1908  *
1909  * Default GIC type is v2
1910  */
1911 static void finalize_gic_version(VirtMachineState *vms)
1912 {
1913     const char *accel_name = current_accel_name();
1914     unsigned int max_cpus = MACHINE(vms)->smp.max_cpus;
1915     int gics_supported = 0;
1916 
1917     /* Determine which GIC versions the current environment supports */
1918     if (kvm_enabled() && kvm_irqchip_in_kernel()) {
1919         int probe_bitmap = kvm_arm_vgic_probe();
1920 
1921         if (!probe_bitmap) {
1922             error_report("Unable to determine GIC version supported by host");
1923             exit(1);
1924         }
1925 
1926         if (probe_bitmap & KVM_ARM_VGIC_V2) {
1927             gics_supported |= VIRT_GIC_VERSION_2_MASK;
1928         }
1929         if (probe_bitmap & KVM_ARM_VGIC_V3) {
1930             gics_supported |= VIRT_GIC_VERSION_3_MASK;
1931         }
1932     } else if (kvm_enabled() && !kvm_irqchip_in_kernel()) {
1933         /* KVM w/o kernel irqchip can only deal with GICv2 */
1934         gics_supported |= VIRT_GIC_VERSION_2_MASK;
1935         accel_name = "KVM with kernel-irqchip=off";
1936     } else if (tcg_enabled() || hvf_enabled() || qtest_enabled())  {
1937         gics_supported |= VIRT_GIC_VERSION_2_MASK;
1938         if (module_object_class_by_name("arm-gicv3")) {
1939             gics_supported |= VIRT_GIC_VERSION_3_MASK;
1940             if (vms->virt) {
1941                 /* GICv4 only makes sense if CPU has EL2 */
1942                 gics_supported |= VIRT_GIC_VERSION_4_MASK;
1943             }
1944         }
1945     } else {
1946         error_report("Unsupported accelerator, can not determine GIC support");
1947         exit(1);
1948     }
1949 
1950     /*
1951      * Then convert helpers like host/max to concrete GIC versions and ensure
1952      * the desired version is supported
1953      */
1954     vms->gic_version = finalize_gic_version_do(accel_name, vms->gic_version,
1955                                                gics_supported, max_cpus);
1956 }
1957 
1958 /*
1959  * virt_cpu_post_init() must be called after the CPUs have
1960  * been realized and the GIC has been created.
1961  */
1962 static void virt_cpu_post_init(VirtMachineState *vms, MemoryRegion *sysmem)
1963 {
1964     int max_cpus = MACHINE(vms)->smp.max_cpus;
1965     bool aarch64, pmu, steal_time;
1966     CPUState *cpu;
1967 
1968     aarch64 = object_property_get_bool(OBJECT(first_cpu), "aarch64", NULL);
1969     pmu = object_property_get_bool(OBJECT(first_cpu), "pmu", NULL);
1970     steal_time = object_property_get_bool(OBJECT(first_cpu),
1971                                           "kvm-steal-time", NULL);
1972 
1973     if (kvm_enabled()) {
1974         hwaddr pvtime_reg_base = vms->memmap[VIRT_PVTIME].base;
1975         hwaddr pvtime_reg_size = vms->memmap[VIRT_PVTIME].size;
1976 
1977         if (steal_time) {
1978             MemoryRegion *pvtime = g_new(MemoryRegion, 1);
1979             hwaddr pvtime_size = max_cpus * PVTIME_SIZE_PER_CPU;
1980 
1981             /* The memory region size must be a multiple of host page size. */
1982             pvtime_size = REAL_HOST_PAGE_ALIGN(pvtime_size);
1983 
1984             if (pvtime_size > pvtime_reg_size) {
1985                 error_report("pvtime requires a %" HWADDR_PRId
1986                              " byte memory region for %d CPUs,"
1987                              " but only %" HWADDR_PRId " has been reserved",
1988                              pvtime_size, max_cpus, pvtime_reg_size);
1989                 exit(1);
1990             }
1991 
1992             memory_region_init_ram(pvtime, NULL, "pvtime", pvtime_size, NULL);
1993             memory_region_add_subregion(sysmem, pvtime_reg_base, pvtime);
1994         }
1995 
1996         CPU_FOREACH(cpu) {
1997             if (pmu) {
1998                 assert(arm_feature(&ARM_CPU(cpu)->env, ARM_FEATURE_PMU));
1999                 if (kvm_irqchip_in_kernel()) {
2000                     kvm_arm_pmu_set_irq(cpu, PPI(VIRTUAL_PMU_IRQ));
2001                 }
2002                 kvm_arm_pmu_init(cpu);
2003             }
2004             if (steal_time) {
2005                 kvm_arm_pvtime_init(cpu, pvtime_reg_base +
2006                                          cpu->cpu_index * PVTIME_SIZE_PER_CPU);
2007             }
2008         }
2009     } else {
2010         if (aarch64 && vms->highmem) {
2011             int requested_pa_size = 64 - clz64(vms->highest_gpa);
2012             int pamax = arm_pamax(ARM_CPU(first_cpu));
2013 
2014             if (pamax < requested_pa_size) {
2015                 error_report("VCPU supports less PA bits (%d) than "
2016                              "requested by the memory map (%d)",
2017                              pamax, requested_pa_size);
2018                 exit(1);
2019             }
2020         }
2021     }
2022 }
2023 
2024 static void machvirt_init(MachineState *machine)
2025 {
2026     VirtMachineState *vms = VIRT_MACHINE(machine);
2027     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(machine);
2028     MachineClass *mc = MACHINE_GET_CLASS(machine);
2029     const CPUArchIdList *possible_cpus;
2030     MemoryRegion *sysmem = get_system_memory();
2031     MemoryRegion *secure_sysmem = NULL;
2032     MemoryRegion *tag_sysmem = NULL;
2033     MemoryRegion *secure_tag_sysmem = NULL;
2034     int n, virt_max_cpus;
2035     bool firmware_loaded;
2036     bool aarch64 = true;
2037     bool has_ged = !vmc->no_ged;
2038     unsigned int smp_cpus = machine->smp.cpus;
2039     unsigned int max_cpus = machine->smp.max_cpus;
2040 
2041     if (!cpu_type_valid(machine->cpu_type)) {
2042         error_report("mach-virt: CPU type %s not supported", machine->cpu_type);
2043         exit(1);
2044     }
2045 
2046     possible_cpus = mc->possible_cpu_arch_ids(machine);
2047 
2048     /*
2049      * In accelerated mode, the memory map is computed earlier in kvm_type()
2050      * to create a VM with the right number of IPA bits.
2051      */
2052     if (!vms->memmap) {
2053         Object *cpuobj;
2054         ARMCPU *armcpu;
2055         int pa_bits;
2056 
2057         /*
2058          * Instantiate a temporary CPU object to find out about what
2059          * we are about to deal with. Once this is done, get rid of
2060          * the object.
2061          */
2062         cpuobj = object_new(possible_cpus->cpus[0].type);
2063         armcpu = ARM_CPU(cpuobj);
2064 
2065         pa_bits = arm_pamax(armcpu);
2066 
2067         object_unref(cpuobj);
2068 
2069         virt_set_memmap(vms, pa_bits);
2070     }
2071 
2072     /* We can probe only here because during property set
2073      * KVM is not available yet
2074      */
2075     finalize_gic_version(vms);
2076 
2077     if (vms->secure) {
2078         /*
2079          * The Secure view of the world is the same as the NonSecure,
2080          * but with a few extra devices. Create it as a container region
2081          * containing the system memory at low priority; any secure-only
2082          * devices go in at higher priority and take precedence.
2083          */
2084         secure_sysmem = g_new(MemoryRegion, 1);
2085         memory_region_init(secure_sysmem, OBJECT(machine), "secure-memory",
2086                            UINT64_MAX);
2087         memory_region_add_subregion_overlap(secure_sysmem, 0, sysmem, -1);
2088     }
2089 
2090     firmware_loaded = virt_firmware_init(vms, sysmem,
2091                                          secure_sysmem ?: sysmem);
2092 
2093     /* If we have an EL3 boot ROM then the assumption is that it will
2094      * implement PSCI itself, so disable QEMU's internal implementation
2095      * so it doesn't get in the way. Instead of starting secondary
2096      * CPUs in PSCI powerdown state we will start them all running and
2097      * let the boot ROM sort them out.
2098      * The usual case is that we do use QEMU's PSCI implementation;
2099      * if the guest has EL2 then we will use SMC as the conduit,
2100      * and otherwise we will use HVC (for backwards compatibility and
2101      * because if we're using KVM then we must use HVC).
2102      */
2103     if (vms->secure && firmware_loaded) {
2104         vms->psci_conduit = QEMU_PSCI_CONDUIT_DISABLED;
2105     } else if (vms->virt) {
2106         vms->psci_conduit = QEMU_PSCI_CONDUIT_SMC;
2107     } else {
2108         vms->psci_conduit = QEMU_PSCI_CONDUIT_HVC;
2109     }
2110 
2111     /*
2112      * The maximum number of CPUs depends on the GIC version, or on how
2113      * many redistributors we can fit into the memory map (which in turn
2114      * depends on whether this is a GICv3 or v4).
2115      */
2116     if (vms->gic_version == VIRT_GIC_VERSION_2) {
2117         virt_max_cpus = GIC_NCPU;
2118     } else {
2119         virt_max_cpus = virt_redist_capacity(vms, VIRT_GIC_REDIST);
2120         if (vms->highmem_redists) {
2121             virt_max_cpus += virt_redist_capacity(vms, VIRT_HIGH_GIC_REDIST2);
2122         }
2123     }
2124 
2125     if (max_cpus > virt_max_cpus) {
2126         error_report("Number of SMP CPUs requested (%d) exceeds max CPUs "
2127                      "supported by machine 'mach-virt' (%d)",
2128                      max_cpus, virt_max_cpus);
2129         if (vms->gic_version != VIRT_GIC_VERSION_2 && !vms->highmem_redists) {
2130             error_printf("Try 'highmem-redists=on' for more CPUs\n");
2131         }
2132 
2133         exit(1);
2134     }
2135 
2136     if (vms->secure && (kvm_enabled() || hvf_enabled())) {
2137         error_report("mach-virt: %s does not support providing "
2138                      "Security extensions (TrustZone) to the guest CPU",
2139                      current_accel_name());
2140         exit(1);
2141     }
2142 
2143     if (vms->virt && (kvm_enabled() || hvf_enabled())) {
2144         error_report("mach-virt: %s does not support providing "
2145                      "Virtualization extensions to the guest CPU",
2146                      current_accel_name());
2147         exit(1);
2148     }
2149 
2150     if (vms->mte && (kvm_enabled() || hvf_enabled())) {
2151         error_report("mach-virt: %s does not support providing "
2152                      "MTE to the guest CPU",
2153                      current_accel_name());
2154         exit(1);
2155     }
2156 
2157     create_fdt(vms);
2158 
2159     assert(possible_cpus->len == max_cpus);
2160     for (n = 0; n < possible_cpus->len; n++) {
2161         Object *cpuobj;
2162         CPUState *cs;
2163 
2164         if (n >= smp_cpus) {
2165             break;
2166         }
2167 
2168         cpuobj = object_new(possible_cpus->cpus[n].type);
2169         object_property_set_int(cpuobj, "mp-affinity",
2170                                 possible_cpus->cpus[n].arch_id, NULL);
2171 
2172         cs = CPU(cpuobj);
2173         cs->cpu_index = n;
2174 
2175         numa_cpu_pre_plug(&possible_cpus->cpus[cs->cpu_index], DEVICE(cpuobj),
2176                           &error_fatal);
2177 
2178         aarch64 &= object_property_get_bool(cpuobj, "aarch64", NULL);
2179 
2180         if (!vms->secure) {
2181             object_property_set_bool(cpuobj, "has_el3", false, NULL);
2182         }
2183 
2184         if (!vms->virt && object_property_find(cpuobj, "has_el2")) {
2185             object_property_set_bool(cpuobj, "has_el2", false, NULL);
2186         }
2187 
2188         if (vmc->kvm_no_adjvtime &&
2189             object_property_find(cpuobj, "kvm-no-adjvtime")) {
2190             object_property_set_bool(cpuobj, "kvm-no-adjvtime", true, NULL);
2191         }
2192 
2193         if (vmc->no_kvm_steal_time &&
2194             object_property_find(cpuobj, "kvm-steal-time")) {
2195             object_property_set_bool(cpuobj, "kvm-steal-time", false, NULL);
2196         }
2197 
2198         if (vmc->no_pmu && object_property_find(cpuobj, "pmu")) {
2199             object_property_set_bool(cpuobj, "pmu", false, NULL);
2200         }
2201 
2202         if (vmc->no_tcg_lpa2 && object_property_find(cpuobj, "lpa2")) {
2203             object_property_set_bool(cpuobj, "lpa2", false, NULL);
2204         }
2205 
2206         if (object_property_find(cpuobj, "reset-cbar")) {
2207             object_property_set_int(cpuobj, "reset-cbar",
2208                                     vms->memmap[VIRT_CPUPERIPHS].base,
2209                                     &error_abort);
2210         }
2211 
2212         object_property_set_link(cpuobj, "memory", OBJECT(sysmem),
2213                                  &error_abort);
2214         if (vms->secure) {
2215             object_property_set_link(cpuobj, "secure-memory",
2216                                      OBJECT(secure_sysmem), &error_abort);
2217         }
2218 
2219         if (vms->mte) {
2220             /* Create the memory region only once, but link to all cpus. */
2221             if (!tag_sysmem) {
2222                 /*
2223                  * The property exists only if MemTag is supported.
2224                  * If it is, we must allocate the ram to back that up.
2225                  */
2226                 if (!object_property_find(cpuobj, "tag-memory")) {
2227                     error_report("MTE requested, but not supported "
2228                                  "by the guest CPU");
2229                     exit(1);
2230                 }
2231 
2232                 tag_sysmem = g_new(MemoryRegion, 1);
2233                 memory_region_init(tag_sysmem, OBJECT(machine),
2234                                    "tag-memory", UINT64_MAX / 32);
2235 
2236                 if (vms->secure) {
2237                     secure_tag_sysmem = g_new(MemoryRegion, 1);
2238                     memory_region_init(secure_tag_sysmem, OBJECT(machine),
2239                                        "secure-tag-memory", UINT64_MAX / 32);
2240 
2241                     /* As with ram, secure-tag takes precedence over tag.  */
2242                     memory_region_add_subregion_overlap(secure_tag_sysmem, 0,
2243                                                         tag_sysmem, -1);
2244                 }
2245             }
2246 
2247             object_property_set_link(cpuobj, "tag-memory", OBJECT(tag_sysmem),
2248                                      &error_abort);
2249             if (vms->secure) {
2250                 object_property_set_link(cpuobj, "secure-tag-memory",
2251                                          OBJECT(secure_tag_sysmem),
2252                                          &error_abort);
2253             }
2254         }
2255 
2256         qdev_realize(DEVICE(cpuobj), NULL, &error_fatal);
2257         object_unref(cpuobj);
2258     }
2259     fdt_add_timer_nodes(vms);
2260     fdt_add_cpu_nodes(vms);
2261 
2262     memory_region_add_subregion(sysmem, vms->memmap[VIRT_MEM].base,
2263                                 machine->ram);
2264     if (machine->device_memory) {
2265         memory_region_add_subregion(sysmem, machine->device_memory->base,
2266                                     &machine->device_memory->mr);
2267     }
2268 
2269     virt_flash_fdt(vms, sysmem, secure_sysmem ?: sysmem);
2270 
2271     create_gic(vms, sysmem);
2272 
2273     virt_cpu_post_init(vms, sysmem);
2274 
2275     fdt_add_pmu_nodes(vms);
2276 
2277     create_uart(vms, VIRT_UART, sysmem, serial_hd(0));
2278 
2279     if (vms->secure) {
2280         create_secure_ram(vms, secure_sysmem, secure_tag_sysmem);
2281         create_uart(vms, VIRT_SECURE_UART, secure_sysmem, serial_hd(1));
2282     }
2283 
2284     if (tag_sysmem) {
2285         create_tag_ram(tag_sysmem, vms->memmap[VIRT_MEM].base,
2286                        machine->ram_size, "mach-virt.tag");
2287     }
2288 
2289     vms->highmem_ecam &= (!firmware_loaded || aarch64);
2290 
2291     create_rtc(vms);
2292 
2293     create_pcie(vms);
2294 
2295     if (has_ged && aarch64 && firmware_loaded && virt_is_acpi_enabled(vms)) {
2296         vms->acpi_dev = create_acpi_ged(vms);
2297     } else {
2298         create_gpio_devices(vms, VIRT_GPIO, sysmem);
2299     }
2300 
2301     if (vms->secure && !vmc->no_secure_gpio) {
2302         create_gpio_devices(vms, VIRT_SECURE_GPIO, secure_sysmem);
2303     }
2304 
2305      /* connect powerdown request */
2306      vms->powerdown_notifier.notify = virt_powerdown_req;
2307      qemu_register_powerdown_notifier(&vms->powerdown_notifier);
2308 
2309     /* Create mmio transports, so the user can create virtio backends
2310      * (which will be automatically plugged in to the transports). If
2311      * no backend is created the transport will just sit harmlessly idle.
2312      */
2313     create_virtio_devices(vms);
2314 
2315     vms->fw_cfg = create_fw_cfg(vms, &address_space_memory);
2316     rom_set_fw(vms->fw_cfg);
2317 
2318     create_platform_bus(vms);
2319 
2320     if (machine->nvdimms_state->is_enabled) {
2321         const struct AcpiGenericAddress arm_virt_nvdimm_acpi_dsmio = {
2322             .space_id = AML_AS_SYSTEM_MEMORY,
2323             .address = vms->memmap[VIRT_NVDIMM_ACPI].base,
2324             .bit_width = NVDIMM_ACPI_IO_LEN << 3
2325         };
2326 
2327         nvdimm_init_acpi_state(machine->nvdimms_state, sysmem,
2328                                arm_virt_nvdimm_acpi_dsmio,
2329                                vms->fw_cfg, OBJECT(vms));
2330     }
2331 
2332     vms->bootinfo.ram_size = machine->ram_size;
2333     vms->bootinfo.board_id = -1;
2334     vms->bootinfo.loader_start = vms->memmap[VIRT_MEM].base;
2335     vms->bootinfo.get_dtb = machvirt_dtb;
2336     vms->bootinfo.skip_dtb_autoload = true;
2337     vms->bootinfo.firmware_loaded = firmware_loaded;
2338     vms->bootinfo.psci_conduit = vms->psci_conduit;
2339     arm_load_kernel(ARM_CPU(first_cpu), machine, &vms->bootinfo);
2340 
2341     vms->machine_done.notify = virt_machine_done;
2342     qemu_add_machine_init_done_notifier(&vms->machine_done);
2343 }
2344 
2345 static bool virt_get_secure(Object *obj, Error **errp)
2346 {
2347     VirtMachineState *vms = VIRT_MACHINE(obj);
2348 
2349     return vms->secure;
2350 }
2351 
2352 static void virt_set_secure(Object *obj, bool value, Error **errp)
2353 {
2354     VirtMachineState *vms = VIRT_MACHINE(obj);
2355 
2356     vms->secure = value;
2357 }
2358 
2359 static bool virt_get_virt(Object *obj, Error **errp)
2360 {
2361     VirtMachineState *vms = VIRT_MACHINE(obj);
2362 
2363     return vms->virt;
2364 }
2365 
2366 static void virt_set_virt(Object *obj, bool value, Error **errp)
2367 {
2368     VirtMachineState *vms = VIRT_MACHINE(obj);
2369 
2370     vms->virt = value;
2371 }
2372 
2373 static bool virt_get_highmem(Object *obj, Error **errp)
2374 {
2375     VirtMachineState *vms = VIRT_MACHINE(obj);
2376 
2377     return vms->highmem;
2378 }
2379 
2380 static void virt_set_highmem(Object *obj, bool value, Error **errp)
2381 {
2382     VirtMachineState *vms = VIRT_MACHINE(obj);
2383 
2384     vms->highmem = value;
2385 }
2386 
2387 static bool virt_get_compact_highmem(Object *obj, Error **errp)
2388 {
2389     VirtMachineState *vms = VIRT_MACHINE(obj);
2390 
2391     return vms->highmem_compact;
2392 }
2393 
2394 static void virt_set_compact_highmem(Object *obj, bool value, Error **errp)
2395 {
2396     VirtMachineState *vms = VIRT_MACHINE(obj);
2397 
2398     vms->highmem_compact = value;
2399 }
2400 
2401 static bool virt_get_highmem_redists(Object *obj, Error **errp)
2402 {
2403     VirtMachineState *vms = VIRT_MACHINE(obj);
2404 
2405     return vms->highmem_redists;
2406 }
2407 
2408 static void virt_set_highmem_redists(Object *obj, bool value, Error **errp)
2409 {
2410     VirtMachineState *vms = VIRT_MACHINE(obj);
2411 
2412     vms->highmem_redists = value;
2413 }
2414 
2415 static bool virt_get_highmem_ecam(Object *obj, Error **errp)
2416 {
2417     VirtMachineState *vms = VIRT_MACHINE(obj);
2418 
2419     return vms->highmem_ecam;
2420 }
2421 
2422 static void virt_set_highmem_ecam(Object *obj, bool value, Error **errp)
2423 {
2424     VirtMachineState *vms = VIRT_MACHINE(obj);
2425 
2426     vms->highmem_ecam = value;
2427 }
2428 
2429 static bool virt_get_highmem_mmio(Object *obj, Error **errp)
2430 {
2431     VirtMachineState *vms = VIRT_MACHINE(obj);
2432 
2433     return vms->highmem_mmio;
2434 }
2435 
2436 static void virt_set_highmem_mmio(Object *obj, bool value, Error **errp)
2437 {
2438     VirtMachineState *vms = VIRT_MACHINE(obj);
2439 
2440     vms->highmem_mmio = value;
2441 }
2442 
2443 
2444 static bool virt_get_its(Object *obj, Error **errp)
2445 {
2446     VirtMachineState *vms = VIRT_MACHINE(obj);
2447 
2448     return vms->its;
2449 }
2450 
2451 static void virt_set_its(Object *obj, bool value, Error **errp)
2452 {
2453     VirtMachineState *vms = VIRT_MACHINE(obj);
2454 
2455     vms->its = value;
2456 }
2457 
2458 static bool virt_get_dtb_randomness(Object *obj, Error **errp)
2459 {
2460     VirtMachineState *vms = VIRT_MACHINE(obj);
2461 
2462     return vms->dtb_randomness;
2463 }
2464 
2465 static void virt_set_dtb_randomness(Object *obj, bool value, Error **errp)
2466 {
2467     VirtMachineState *vms = VIRT_MACHINE(obj);
2468 
2469     vms->dtb_randomness = value;
2470 }
2471 
2472 static char *virt_get_oem_id(Object *obj, Error **errp)
2473 {
2474     VirtMachineState *vms = VIRT_MACHINE(obj);
2475 
2476     return g_strdup(vms->oem_id);
2477 }
2478 
2479 static void virt_set_oem_id(Object *obj, const char *value, Error **errp)
2480 {
2481     VirtMachineState *vms = VIRT_MACHINE(obj);
2482     size_t len = strlen(value);
2483 
2484     if (len > 6) {
2485         error_setg(errp,
2486                    "User specified oem-id value is bigger than 6 bytes in size");
2487         return;
2488     }
2489 
2490     strncpy(vms->oem_id, value, 6);
2491 }
2492 
2493 static char *virt_get_oem_table_id(Object *obj, Error **errp)
2494 {
2495     VirtMachineState *vms = VIRT_MACHINE(obj);
2496 
2497     return g_strdup(vms->oem_table_id);
2498 }
2499 
2500 static void virt_set_oem_table_id(Object *obj, const char *value,
2501                                   Error **errp)
2502 {
2503     VirtMachineState *vms = VIRT_MACHINE(obj);
2504     size_t len = strlen(value);
2505 
2506     if (len > 8) {
2507         error_setg(errp,
2508                    "User specified oem-table-id value is bigger than 8 bytes in size");
2509         return;
2510     }
2511     strncpy(vms->oem_table_id, value, 8);
2512 }
2513 
2514 
2515 bool virt_is_acpi_enabled(VirtMachineState *vms)
2516 {
2517     if (vms->acpi == ON_OFF_AUTO_OFF) {
2518         return false;
2519     }
2520     return true;
2521 }
2522 
2523 static void virt_get_acpi(Object *obj, Visitor *v, const char *name,
2524                           void *opaque, Error **errp)
2525 {
2526     VirtMachineState *vms = VIRT_MACHINE(obj);
2527     OnOffAuto acpi = vms->acpi;
2528 
2529     visit_type_OnOffAuto(v, name, &acpi, errp);
2530 }
2531 
2532 static void virt_set_acpi(Object *obj, Visitor *v, const char *name,
2533                           void *opaque, Error **errp)
2534 {
2535     VirtMachineState *vms = VIRT_MACHINE(obj);
2536 
2537     visit_type_OnOffAuto(v, name, &vms->acpi, errp);
2538 }
2539 
2540 static bool virt_get_ras(Object *obj, Error **errp)
2541 {
2542     VirtMachineState *vms = VIRT_MACHINE(obj);
2543 
2544     return vms->ras;
2545 }
2546 
2547 static void virt_set_ras(Object *obj, bool value, Error **errp)
2548 {
2549     VirtMachineState *vms = VIRT_MACHINE(obj);
2550 
2551     vms->ras = value;
2552 }
2553 
2554 static bool virt_get_mte(Object *obj, Error **errp)
2555 {
2556     VirtMachineState *vms = VIRT_MACHINE(obj);
2557 
2558     return vms->mte;
2559 }
2560 
2561 static void virt_set_mte(Object *obj, bool value, Error **errp)
2562 {
2563     VirtMachineState *vms = VIRT_MACHINE(obj);
2564 
2565     vms->mte = value;
2566 }
2567 
2568 static char *virt_get_gic_version(Object *obj, Error **errp)
2569 {
2570     VirtMachineState *vms = VIRT_MACHINE(obj);
2571     const char *val;
2572 
2573     switch (vms->gic_version) {
2574     case VIRT_GIC_VERSION_4:
2575         val = "4";
2576         break;
2577     case VIRT_GIC_VERSION_3:
2578         val = "3";
2579         break;
2580     default:
2581         val = "2";
2582         break;
2583     }
2584     return g_strdup(val);
2585 }
2586 
2587 static void virt_set_gic_version(Object *obj, const char *value, Error **errp)
2588 {
2589     VirtMachineState *vms = VIRT_MACHINE(obj);
2590 
2591     if (!strcmp(value, "4")) {
2592         vms->gic_version = VIRT_GIC_VERSION_4;
2593     } else if (!strcmp(value, "3")) {
2594         vms->gic_version = VIRT_GIC_VERSION_3;
2595     } else if (!strcmp(value, "2")) {
2596         vms->gic_version = VIRT_GIC_VERSION_2;
2597     } else if (!strcmp(value, "host")) {
2598         vms->gic_version = VIRT_GIC_VERSION_HOST; /* Will probe later */
2599     } else if (!strcmp(value, "max")) {
2600         vms->gic_version = VIRT_GIC_VERSION_MAX; /* Will probe later */
2601     } else {
2602         error_setg(errp, "Invalid gic-version value");
2603         error_append_hint(errp, "Valid values are 3, 2, host, max.\n");
2604     }
2605 }
2606 
2607 static char *virt_get_iommu(Object *obj, Error **errp)
2608 {
2609     VirtMachineState *vms = VIRT_MACHINE(obj);
2610 
2611     switch (vms->iommu) {
2612     case VIRT_IOMMU_NONE:
2613         return g_strdup("none");
2614     case VIRT_IOMMU_SMMUV3:
2615         return g_strdup("smmuv3");
2616     default:
2617         g_assert_not_reached();
2618     }
2619 }
2620 
2621 static void virt_set_iommu(Object *obj, const char *value, Error **errp)
2622 {
2623     VirtMachineState *vms = VIRT_MACHINE(obj);
2624 
2625     if (!strcmp(value, "smmuv3")) {
2626         vms->iommu = VIRT_IOMMU_SMMUV3;
2627     } else if (!strcmp(value, "none")) {
2628         vms->iommu = VIRT_IOMMU_NONE;
2629     } else {
2630         error_setg(errp, "Invalid iommu value");
2631         error_append_hint(errp, "Valid values are none, smmuv3.\n");
2632     }
2633 }
2634 
2635 static bool virt_get_default_bus_bypass_iommu(Object *obj, Error **errp)
2636 {
2637     VirtMachineState *vms = VIRT_MACHINE(obj);
2638 
2639     return vms->default_bus_bypass_iommu;
2640 }
2641 
2642 static void virt_set_default_bus_bypass_iommu(Object *obj, bool value,
2643                                               Error **errp)
2644 {
2645     VirtMachineState *vms = VIRT_MACHINE(obj);
2646 
2647     vms->default_bus_bypass_iommu = value;
2648 }
2649 
2650 static CpuInstanceProperties
2651 virt_cpu_index_to_props(MachineState *ms, unsigned cpu_index)
2652 {
2653     MachineClass *mc = MACHINE_GET_CLASS(ms);
2654     const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(ms);
2655 
2656     assert(cpu_index < possible_cpus->len);
2657     return possible_cpus->cpus[cpu_index].props;
2658 }
2659 
2660 static int64_t virt_get_default_cpu_node_id(const MachineState *ms, int idx)
2661 {
2662     int64_t socket_id = ms->possible_cpus->cpus[idx].props.socket_id;
2663 
2664     return socket_id % ms->numa_state->num_nodes;
2665 }
2666 
2667 static const CPUArchIdList *virt_possible_cpu_arch_ids(MachineState *ms)
2668 {
2669     int n;
2670     unsigned int max_cpus = ms->smp.max_cpus;
2671     VirtMachineState *vms = VIRT_MACHINE(ms);
2672     MachineClass *mc = MACHINE_GET_CLASS(vms);
2673 
2674     if (ms->possible_cpus) {
2675         assert(ms->possible_cpus->len == max_cpus);
2676         return ms->possible_cpus;
2677     }
2678 
2679     ms->possible_cpus = g_malloc0(sizeof(CPUArchIdList) +
2680                                   sizeof(CPUArchId) * max_cpus);
2681     ms->possible_cpus->len = max_cpus;
2682     for (n = 0; n < ms->possible_cpus->len; n++) {
2683         ms->possible_cpus->cpus[n].type = ms->cpu_type;
2684         ms->possible_cpus->cpus[n].arch_id =
2685             virt_cpu_mp_affinity(vms, n);
2686 
2687         assert(!mc->smp_props.dies_supported);
2688         ms->possible_cpus->cpus[n].props.has_socket_id = true;
2689         ms->possible_cpus->cpus[n].props.socket_id =
2690             n / (ms->smp.clusters * ms->smp.cores * ms->smp.threads);
2691         ms->possible_cpus->cpus[n].props.has_cluster_id = true;
2692         ms->possible_cpus->cpus[n].props.cluster_id =
2693             (n / (ms->smp.cores * ms->smp.threads)) % ms->smp.clusters;
2694         ms->possible_cpus->cpus[n].props.has_core_id = true;
2695         ms->possible_cpus->cpus[n].props.core_id =
2696             (n / ms->smp.threads) % ms->smp.cores;
2697         ms->possible_cpus->cpus[n].props.has_thread_id = true;
2698         ms->possible_cpus->cpus[n].props.thread_id =
2699             n % ms->smp.threads;
2700     }
2701     return ms->possible_cpus;
2702 }
2703 
2704 static void virt_memory_pre_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
2705                                  Error **errp)
2706 {
2707     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2708     const MachineState *ms = MACHINE(hotplug_dev);
2709     const bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2710 
2711     if (!vms->acpi_dev) {
2712         error_setg(errp,
2713                    "memory hotplug is not enabled: missing acpi-ged device");
2714         return;
2715     }
2716 
2717     if (vms->mte) {
2718         error_setg(errp, "memory hotplug is not enabled: MTE is enabled");
2719         return;
2720     }
2721 
2722     if (is_nvdimm && !ms->nvdimms_state->is_enabled) {
2723         error_setg(errp, "nvdimm is not enabled: add 'nvdimm=on' to '-M'");
2724         return;
2725     }
2726 
2727     pc_dimm_pre_plug(PC_DIMM(dev), MACHINE(hotplug_dev), NULL, errp);
2728 }
2729 
2730 static void virt_memory_plug(HotplugHandler *hotplug_dev,
2731                              DeviceState *dev, Error **errp)
2732 {
2733     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2734     MachineState *ms = MACHINE(hotplug_dev);
2735     bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2736 
2737     pc_dimm_plug(PC_DIMM(dev), MACHINE(vms));
2738 
2739     if (is_nvdimm) {
2740         nvdimm_plug(ms->nvdimms_state);
2741     }
2742 
2743     hotplug_handler_plug(HOTPLUG_HANDLER(vms->acpi_dev),
2744                          dev, &error_abort);
2745 }
2746 
2747 static void virt_virtio_md_pci_pre_plug(HotplugHandler *hotplug_dev,
2748                                         DeviceState *dev, Error **errp)
2749 {
2750     HotplugHandler *hotplug_dev2 = qdev_get_bus_hotplug_handler(dev);
2751     Error *local_err = NULL;
2752 
2753     if (!hotplug_dev2 && dev->hotplugged) {
2754         /*
2755          * Without a bus hotplug handler, we cannot control the plug/unplug
2756          * order. We should never reach this point when hotplugging on ARM.
2757          * However, it's nice to add a safety net, similar to what we have
2758          * on x86.
2759          */
2760         error_setg(errp, "hotplug of virtio based memory devices not supported"
2761                    " on this bus.");
2762         return;
2763     }
2764     /*
2765      * First, see if we can plug this memory device at all. If that
2766      * succeeds, branch of to the actual hotplug handler.
2767      */
2768     memory_device_pre_plug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev), NULL,
2769                            &local_err);
2770     if (!local_err && hotplug_dev2) {
2771         hotplug_handler_pre_plug(hotplug_dev2, dev, &local_err);
2772     }
2773     error_propagate(errp, local_err);
2774 }
2775 
2776 static void virt_virtio_md_pci_plug(HotplugHandler *hotplug_dev,
2777                                     DeviceState *dev, Error **errp)
2778 {
2779     HotplugHandler *hotplug_dev2 = qdev_get_bus_hotplug_handler(dev);
2780     Error *local_err = NULL;
2781 
2782     /*
2783      * Plug the memory device first and then branch off to the actual
2784      * hotplug handler. If that one fails, we can easily undo the memory
2785      * device bits.
2786      */
2787     memory_device_plug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev));
2788     if (hotplug_dev2) {
2789         hotplug_handler_plug(hotplug_dev2, dev, &local_err);
2790         if (local_err) {
2791             memory_device_unplug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev));
2792         }
2793     }
2794     error_propagate(errp, local_err);
2795 }
2796 
2797 static void virt_virtio_md_pci_unplug_request(HotplugHandler *hotplug_dev,
2798                                               DeviceState *dev, Error **errp)
2799 {
2800     /* We don't support hot unplug of virtio based memory devices */
2801     error_setg(errp, "virtio based memory devices cannot be unplugged.");
2802 }
2803 
2804 
2805 static void virt_machine_device_pre_plug_cb(HotplugHandler *hotplug_dev,
2806                                             DeviceState *dev, Error **errp)
2807 {
2808     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2809 
2810     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2811         virt_memory_pre_plug(hotplug_dev, dev, errp);
2812     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI)) {
2813         virt_virtio_md_pci_pre_plug(hotplug_dev, dev, errp);
2814     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2815         hwaddr db_start = 0, db_end = 0;
2816         char *resv_prop_str;
2817 
2818         if (vms->iommu != VIRT_IOMMU_NONE) {
2819             error_setg(errp, "virt machine does not support multiple IOMMUs");
2820             return;
2821         }
2822 
2823         switch (vms->msi_controller) {
2824         case VIRT_MSI_CTRL_NONE:
2825             return;
2826         case VIRT_MSI_CTRL_ITS:
2827             /* GITS_TRANSLATER page */
2828             db_start = base_memmap[VIRT_GIC_ITS].base + 0x10000;
2829             db_end = base_memmap[VIRT_GIC_ITS].base +
2830                      base_memmap[VIRT_GIC_ITS].size - 1;
2831             break;
2832         case VIRT_MSI_CTRL_GICV2M:
2833             /* MSI_SETSPI_NS page */
2834             db_start = base_memmap[VIRT_GIC_V2M].base;
2835             db_end = db_start + base_memmap[VIRT_GIC_V2M].size - 1;
2836             break;
2837         }
2838         resv_prop_str = g_strdup_printf("0x%"PRIx64":0x%"PRIx64":%u",
2839                                         db_start, db_end,
2840                                         VIRTIO_IOMMU_RESV_MEM_T_MSI);
2841 
2842         object_property_set_uint(OBJECT(dev), "len-reserved-regions", 1, errp);
2843         object_property_set_str(OBJECT(dev), "reserved-regions[0]",
2844                                 resv_prop_str, errp);
2845         g_free(resv_prop_str);
2846     }
2847 }
2848 
2849 static void virt_machine_device_plug_cb(HotplugHandler *hotplug_dev,
2850                                         DeviceState *dev, Error **errp)
2851 {
2852     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2853 
2854     if (vms->platform_bus_dev) {
2855         MachineClass *mc = MACHINE_GET_CLASS(vms);
2856 
2857         if (device_is_dynamic_sysbus(mc, dev)) {
2858             platform_bus_link_device(PLATFORM_BUS_DEVICE(vms->platform_bus_dev),
2859                                      SYS_BUS_DEVICE(dev));
2860         }
2861     }
2862     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2863         virt_memory_plug(hotplug_dev, dev, errp);
2864     }
2865 
2866     if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI)) {
2867         virt_virtio_md_pci_plug(hotplug_dev, dev, errp);
2868     }
2869 
2870     if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2871         PCIDevice *pdev = PCI_DEVICE(dev);
2872 
2873         vms->iommu = VIRT_IOMMU_VIRTIO;
2874         vms->virtio_iommu_bdf = pci_get_bdf(pdev);
2875         create_virtio_iommu_dt_bindings(vms);
2876     }
2877 }
2878 
2879 static void virt_dimm_unplug_request(HotplugHandler *hotplug_dev,
2880                                      DeviceState *dev, Error **errp)
2881 {
2882     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2883 
2884     if (!vms->acpi_dev) {
2885         error_setg(errp,
2886                    "memory hotplug is not enabled: missing acpi-ged device");
2887         return;
2888     }
2889 
2890     if (object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM)) {
2891         error_setg(errp, "nvdimm device hot unplug is not supported yet.");
2892         return;
2893     }
2894 
2895     hotplug_handler_unplug_request(HOTPLUG_HANDLER(vms->acpi_dev), dev,
2896                                    errp);
2897 }
2898 
2899 static void virt_dimm_unplug(HotplugHandler *hotplug_dev,
2900                              DeviceState *dev, Error **errp)
2901 {
2902     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2903     Error *local_err = NULL;
2904 
2905     hotplug_handler_unplug(HOTPLUG_HANDLER(vms->acpi_dev), dev, &local_err);
2906     if (local_err) {
2907         goto out;
2908     }
2909 
2910     pc_dimm_unplug(PC_DIMM(dev), MACHINE(vms));
2911     qdev_unrealize(dev);
2912 
2913 out:
2914     error_propagate(errp, local_err);
2915 }
2916 
2917 static void virt_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
2918                                           DeviceState *dev, Error **errp)
2919 {
2920     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2921         virt_dimm_unplug_request(hotplug_dev, dev, errp);
2922     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI)) {
2923         virt_virtio_md_pci_unplug_request(hotplug_dev, dev, errp);
2924     } else {
2925         error_setg(errp, "device unplug request for unsupported device"
2926                    " type: %s", object_get_typename(OBJECT(dev)));
2927     }
2928 }
2929 
2930 static void virt_machine_device_unplug_cb(HotplugHandler *hotplug_dev,
2931                                           DeviceState *dev, Error **errp)
2932 {
2933     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2934         virt_dimm_unplug(hotplug_dev, dev, errp);
2935     } else {
2936         error_setg(errp, "virt: device unplug for unsupported device"
2937                    " type: %s", object_get_typename(OBJECT(dev)));
2938     }
2939 }
2940 
2941 static HotplugHandler *virt_machine_get_hotplug_handler(MachineState *machine,
2942                                                         DeviceState *dev)
2943 {
2944     MachineClass *mc = MACHINE_GET_CLASS(machine);
2945 
2946     if (device_is_dynamic_sysbus(mc, dev) ||
2947         object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) ||
2948         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI) ||
2949         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2950         return HOTPLUG_HANDLER(machine);
2951     }
2952     return NULL;
2953 }
2954 
2955 /*
2956  * for arm64 kvm_type [7-0] encodes the requested number of bits
2957  * in the IPA address space
2958  */
2959 static int virt_kvm_type(MachineState *ms, const char *type_str)
2960 {
2961     VirtMachineState *vms = VIRT_MACHINE(ms);
2962     int max_vm_pa_size, requested_pa_size;
2963     bool fixed_ipa;
2964 
2965     max_vm_pa_size = kvm_arm_get_max_vm_ipa_size(ms, &fixed_ipa);
2966 
2967     /* we freeze the memory map to compute the highest gpa */
2968     virt_set_memmap(vms, max_vm_pa_size);
2969 
2970     requested_pa_size = 64 - clz64(vms->highest_gpa);
2971 
2972     /*
2973      * KVM requires the IPA size to be at least 32 bits.
2974      */
2975     if (requested_pa_size < 32) {
2976         requested_pa_size = 32;
2977     }
2978 
2979     if (requested_pa_size > max_vm_pa_size) {
2980         error_report("-m and ,maxmem option values "
2981                      "require an IPA range (%d bits) larger than "
2982                      "the one supported by the host (%d bits)",
2983                      requested_pa_size, max_vm_pa_size);
2984         exit(1);
2985     }
2986     /*
2987      * We return the requested PA log size, unless KVM only supports
2988      * the implicit legacy 40b IPA setting, in which case the kvm_type
2989      * must be 0.
2990      */
2991     return fixed_ipa ? 0 : requested_pa_size;
2992 }
2993 
2994 static void virt_machine_class_init(ObjectClass *oc, void *data)
2995 {
2996     MachineClass *mc = MACHINE_CLASS(oc);
2997     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
2998 
2999     mc->init = machvirt_init;
3000     /* Start with max_cpus set to 512, which is the maximum supported by KVM.
3001      * The value may be reduced later when we have more information about the
3002      * configuration of the particular instance.
3003      */
3004     mc->max_cpus = 512;
3005     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_CALXEDA_XGMAC);
3006     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_AMD_XGBE);
3007     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_RAMFB_DEVICE);
3008     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_PLATFORM);
3009 #ifdef CONFIG_TPM
3010     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_TPM_TIS_SYSBUS);
3011 #endif
3012     mc->block_default_type = IF_VIRTIO;
3013     mc->no_cdrom = 1;
3014     mc->pci_allow_0_address = true;
3015     /* We know we will never create a pre-ARMv7 CPU which needs 1K pages */
3016     mc->minimum_page_bits = 12;
3017     mc->possible_cpu_arch_ids = virt_possible_cpu_arch_ids;
3018     mc->cpu_index_to_instance_props = virt_cpu_index_to_props;
3019 #ifdef CONFIG_TCG
3020     mc->default_cpu_type = ARM_CPU_TYPE_NAME("cortex-a15");
3021 #else
3022     mc->default_cpu_type = ARM_CPU_TYPE_NAME("max");
3023 #endif
3024     mc->get_default_cpu_node_id = virt_get_default_cpu_node_id;
3025     mc->kvm_type = virt_kvm_type;
3026     assert(!mc->get_hotplug_handler);
3027     mc->get_hotplug_handler = virt_machine_get_hotplug_handler;
3028     hc->pre_plug = virt_machine_device_pre_plug_cb;
3029     hc->plug = virt_machine_device_plug_cb;
3030     hc->unplug_request = virt_machine_device_unplug_request_cb;
3031     hc->unplug = virt_machine_device_unplug_cb;
3032     mc->nvdimm_supported = true;
3033     mc->smp_props.clusters_supported = true;
3034     mc->auto_enable_numa_with_memhp = true;
3035     mc->auto_enable_numa_with_memdev = true;
3036     mc->default_ram_id = "mach-virt.ram";
3037     mc->default_nic = "virtio-net-pci";
3038 
3039     object_class_property_add(oc, "acpi", "OnOffAuto",
3040         virt_get_acpi, virt_set_acpi,
3041         NULL, NULL);
3042     object_class_property_set_description(oc, "acpi",
3043         "Enable ACPI");
3044     object_class_property_add_bool(oc, "secure", virt_get_secure,
3045                                    virt_set_secure);
3046     object_class_property_set_description(oc, "secure",
3047                                                 "Set on/off to enable/disable the ARM "
3048                                                 "Security Extensions (TrustZone)");
3049 
3050     object_class_property_add_bool(oc, "virtualization", virt_get_virt,
3051                                    virt_set_virt);
3052     object_class_property_set_description(oc, "virtualization",
3053                                           "Set on/off to enable/disable emulating a "
3054                                           "guest CPU which implements the ARM "
3055                                           "Virtualization Extensions");
3056 
3057     object_class_property_add_bool(oc, "highmem", virt_get_highmem,
3058                                    virt_set_highmem);
3059     object_class_property_set_description(oc, "highmem",
3060                                           "Set on/off to enable/disable using "
3061                                           "physical address space above 32 bits");
3062 
3063     object_class_property_add_bool(oc, "compact-highmem",
3064                                    virt_get_compact_highmem,
3065                                    virt_set_compact_highmem);
3066     object_class_property_set_description(oc, "compact-highmem",
3067                                           "Set on/off to enable/disable compact "
3068                                           "layout for high memory regions");
3069 
3070     object_class_property_add_bool(oc, "highmem-redists",
3071                                    virt_get_highmem_redists,
3072                                    virt_set_highmem_redists);
3073     object_class_property_set_description(oc, "highmem-redists",
3074                                           "Set on/off to enable/disable high "
3075                                           "memory region for GICv3 or GICv4 "
3076                                           "redistributor");
3077 
3078     object_class_property_add_bool(oc, "highmem-ecam",
3079                                    virt_get_highmem_ecam,
3080                                    virt_set_highmem_ecam);
3081     object_class_property_set_description(oc, "highmem-ecam",
3082                                           "Set on/off to enable/disable high "
3083                                           "memory region for PCI ECAM");
3084 
3085     object_class_property_add_bool(oc, "highmem-mmio",
3086                                    virt_get_highmem_mmio,
3087                                    virt_set_highmem_mmio);
3088     object_class_property_set_description(oc, "highmem-mmio",
3089                                           "Set on/off to enable/disable high "
3090                                           "memory region for PCI MMIO");
3091 
3092     object_class_property_add_str(oc, "gic-version", virt_get_gic_version,
3093                                   virt_set_gic_version);
3094     object_class_property_set_description(oc, "gic-version",
3095                                           "Set GIC version. "
3096                                           "Valid values are 2, 3, 4, host and max");
3097 
3098     object_class_property_add_str(oc, "iommu", virt_get_iommu, virt_set_iommu);
3099     object_class_property_set_description(oc, "iommu",
3100                                           "Set the IOMMU type. "
3101                                           "Valid values are none and smmuv3");
3102 
3103     object_class_property_add_bool(oc, "default-bus-bypass-iommu",
3104                                    virt_get_default_bus_bypass_iommu,
3105                                    virt_set_default_bus_bypass_iommu);
3106     object_class_property_set_description(oc, "default-bus-bypass-iommu",
3107                                           "Set on/off to enable/disable "
3108                                           "bypass_iommu for default root bus");
3109 
3110     object_class_property_add_bool(oc, "ras", virt_get_ras,
3111                                    virt_set_ras);
3112     object_class_property_set_description(oc, "ras",
3113                                           "Set on/off to enable/disable reporting host memory errors "
3114                                           "to a KVM guest using ACPI and guest external abort exceptions");
3115 
3116     object_class_property_add_bool(oc, "mte", virt_get_mte, virt_set_mte);
3117     object_class_property_set_description(oc, "mte",
3118                                           "Set on/off to enable/disable emulating a "
3119                                           "guest CPU which implements the ARM "
3120                                           "Memory Tagging Extension");
3121 
3122     object_class_property_add_bool(oc, "its", virt_get_its,
3123                                    virt_set_its);
3124     object_class_property_set_description(oc, "its",
3125                                           "Set on/off to enable/disable "
3126                                           "ITS instantiation");
3127 
3128     object_class_property_add_bool(oc, "dtb-randomness",
3129                                    virt_get_dtb_randomness,
3130                                    virt_set_dtb_randomness);
3131     object_class_property_set_description(oc, "dtb-randomness",
3132                                           "Set off to disable passing random or "
3133                                           "non-deterministic dtb nodes to guest");
3134 
3135     object_class_property_add_bool(oc, "dtb-kaslr-seed",
3136                                    virt_get_dtb_randomness,
3137                                    virt_set_dtb_randomness);
3138     object_class_property_set_description(oc, "dtb-kaslr-seed",
3139                                           "Deprecated synonym of dtb-randomness");
3140 
3141     object_class_property_add_str(oc, "x-oem-id",
3142                                   virt_get_oem_id,
3143                                   virt_set_oem_id);
3144     object_class_property_set_description(oc, "x-oem-id",
3145                                           "Override the default value of field OEMID "
3146                                           "in ACPI table header."
3147                                           "The string may be up to 6 bytes in size");
3148 
3149 
3150     object_class_property_add_str(oc, "x-oem-table-id",
3151                                   virt_get_oem_table_id,
3152                                   virt_set_oem_table_id);
3153     object_class_property_set_description(oc, "x-oem-table-id",
3154                                           "Override the default value of field OEM Table ID "
3155                                           "in ACPI table header."
3156                                           "The string may be up to 8 bytes in size");
3157 
3158 }
3159 
3160 static void virt_instance_init(Object *obj)
3161 {
3162     VirtMachineState *vms = VIRT_MACHINE(obj);
3163     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
3164 
3165     /* EL3 is disabled by default on virt: this makes us consistent
3166      * between KVM and TCG for this board, and it also allows us to
3167      * boot UEFI blobs which assume no TrustZone support.
3168      */
3169     vms->secure = false;
3170 
3171     /* EL2 is also disabled by default, for similar reasons */
3172     vms->virt = false;
3173 
3174     /* High memory is enabled by default */
3175     vms->highmem = true;
3176     vms->highmem_compact = !vmc->no_highmem_compact;
3177     vms->gic_version = VIRT_GIC_VERSION_NOSEL;
3178 
3179     vms->highmem_ecam = !vmc->no_highmem_ecam;
3180     vms->highmem_mmio = true;
3181     vms->highmem_redists = true;
3182 
3183     if (vmc->no_its) {
3184         vms->its = false;
3185     } else {
3186         /* Default allows ITS instantiation */
3187         vms->its = true;
3188 
3189         if (vmc->no_tcg_its) {
3190             vms->tcg_its = false;
3191         } else {
3192             vms->tcg_its = true;
3193         }
3194     }
3195 
3196     /* Default disallows iommu instantiation */
3197     vms->iommu = VIRT_IOMMU_NONE;
3198 
3199     /* The default root bus is attached to iommu by default */
3200     vms->default_bus_bypass_iommu = false;
3201 
3202     /* Default disallows RAS instantiation */
3203     vms->ras = false;
3204 
3205     /* MTE is disabled by default.  */
3206     vms->mte = false;
3207 
3208     /* Supply kaslr-seed and rng-seed by default */
3209     vms->dtb_randomness = true;
3210 
3211     vms->irqmap = a15irqmap;
3212 
3213     virt_flash_create(vms);
3214 
3215     vms->oem_id = g_strndup(ACPI_BUILD_APPNAME6, 6);
3216     vms->oem_table_id = g_strndup(ACPI_BUILD_APPNAME8, 8);
3217 }
3218 
3219 static const TypeInfo virt_machine_info = {
3220     .name          = TYPE_VIRT_MACHINE,
3221     .parent        = TYPE_MACHINE,
3222     .abstract      = true,
3223     .instance_size = sizeof(VirtMachineState),
3224     .class_size    = sizeof(VirtMachineClass),
3225     .class_init    = virt_machine_class_init,
3226     .instance_init = virt_instance_init,
3227     .interfaces = (InterfaceInfo[]) {
3228          { TYPE_HOTPLUG_HANDLER },
3229          { }
3230     },
3231 };
3232 
3233 static void machvirt_machine_init(void)
3234 {
3235     type_register_static(&virt_machine_info);
3236 }
3237 type_init(machvirt_machine_init);
3238 
3239 static void virt_machine_8_1_options(MachineClass *mc)
3240 {
3241 }
3242 DEFINE_VIRT_MACHINE_AS_LATEST(8, 1)
3243 
3244 static void virt_machine_8_0_options(MachineClass *mc)
3245 {
3246     virt_machine_8_1_options(mc);
3247     compat_props_add(mc->compat_props, hw_compat_8_0, hw_compat_8_0_len);
3248 }
3249 DEFINE_VIRT_MACHINE(8, 0)
3250 
3251 static void virt_machine_7_2_options(MachineClass *mc)
3252 {
3253     virt_machine_8_0_options(mc);
3254     compat_props_add(mc->compat_props, hw_compat_7_2, hw_compat_7_2_len);
3255 }
3256 DEFINE_VIRT_MACHINE(7, 2)
3257 
3258 static void virt_machine_7_1_options(MachineClass *mc)
3259 {
3260     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3261 
3262     virt_machine_7_2_options(mc);
3263     compat_props_add(mc->compat_props, hw_compat_7_1, hw_compat_7_1_len);
3264     /* Compact layout for high memory regions was introduced with 7.2 */
3265     vmc->no_highmem_compact = true;
3266 }
3267 DEFINE_VIRT_MACHINE(7, 1)
3268 
3269 static void virt_machine_7_0_options(MachineClass *mc)
3270 {
3271     virt_machine_7_1_options(mc);
3272     compat_props_add(mc->compat_props, hw_compat_7_0, hw_compat_7_0_len);
3273 }
3274 DEFINE_VIRT_MACHINE(7, 0)
3275 
3276 static void virt_machine_6_2_options(MachineClass *mc)
3277 {
3278     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3279 
3280     virt_machine_7_0_options(mc);
3281     compat_props_add(mc->compat_props, hw_compat_6_2, hw_compat_6_2_len);
3282     vmc->no_tcg_lpa2 = true;
3283 }
3284 DEFINE_VIRT_MACHINE(6, 2)
3285 
3286 static void virt_machine_6_1_options(MachineClass *mc)
3287 {
3288     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3289 
3290     virt_machine_6_2_options(mc);
3291     compat_props_add(mc->compat_props, hw_compat_6_1, hw_compat_6_1_len);
3292     mc->smp_props.prefer_sockets = true;
3293     vmc->no_cpu_topology = true;
3294 
3295     /* qemu ITS was introduced with 6.2 */
3296     vmc->no_tcg_its = true;
3297 }
3298 DEFINE_VIRT_MACHINE(6, 1)
3299 
3300 static void virt_machine_6_0_options(MachineClass *mc)
3301 {
3302     virt_machine_6_1_options(mc);
3303     compat_props_add(mc->compat_props, hw_compat_6_0, hw_compat_6_0_len);
3304 }
3305 DEFINE_VIRT_MACHINE(6, 0)
3306 
3307 static void virt_machine_5_2_options(MachineClass *mc)
3308 {
3309     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3310 
3311     virt_machine_6_0_options(mc);
3312     compat_props_add(mc->compat_props, hw_compat_5_2, hw_compat_5_2_len);
3313     vmc->no_secure_gpio = true;
3314 }
3315 DEFINE_VIRT_MACHINE(5, 2)
3316 
3317 static void virt_machine_5_1_options(MachineClass *mc)
3318 {
3319     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3320 
3321     virt_machine_5_2_options(mc);
3322     compat_props_add(mc->compat_props, hw_compat_5_1, hw_compat_5_1_len);
3323     vmc->no_kvm_steal_time = true;
3324 }
3325 DEFINE_VIRT_MACHINE(5, 1)
3326 
3327 static void virt_machine_5_0_options(MachineClass *mc)
3328 {
3329     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3330 
3331     virt_machine_5_1_options(mc);
3332     compat_props_add(mc->compat_props, hw_compat_5_0, hw_compat_5_0_len);
3333     mc->numa_mem_supported = true;
3334     vmc->acpi_expose_flash = true;
3335     mc->auto_enable_numa_with_memdev = false;
3336 }
3337 DEFINE_VIRT_MACHINE(5, 0)
3338 
3339 static void virt_machine_4_2_options(MachineClass *mc)
3340 {
3341     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3342 
3343     virt_machine_5_0_options(mc);
3344     compat_props_add(mc->compat_props, hw_compat_4_2, hw_compat_4_2_len);
3345     vmc->kvm_no_adjvtime = true;
3346 }
3347 DEFINE_VIRT_MACHINE(4, 2)
3348 
3349 static void virt_machine_4_1_options(MachineClass *mc)
3350 {
3351     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3352 
3353     virt_machine_4_2_options(mc);
3354     compat_props_add(mc->compat_props, hw_compat_4_1, hw_compat_4_1_len);
3355     vmc->no_ged = true;
3356     mc->auto_enable_numa_with_memhp = false;
3357 }
3358 DEFINE_VIRT_MACHINE(4, 1)
3359 
3360 static void virt_machine_4_0_options(MachineClass *mc)
3361 {
3362     virt_machine_4_1_options(mc);
3363     compat_props_add(mc->compat_props, hw_compat_4_0, hw_compat_4_0_len);
3364 }
3365 DEFINE_VIRT_MACHINE(4, 0)
3366 
3367 static void virt_machine_3_1_options(MachineClass *mc)
3368 {
3369     virt_machine_4_0_options(mc);
3370     compat_props_add(mc->compat_props, hw_compat_3_1, hw_compat_3_1_len);
3371 }
3372 DEFINE_VIRT_MACHINE(3, 1)
3373 
3374 static void virt_machine_3_0_options(MachineClass *mc)
3375 {
3376     virt_machine_3_1_options(mc);
3377     compat_props_add(mc->compat_props, hw_compat_3_0, hw_compat_3_0_len);
3378 }
3379 DEFINE_VIRT_MACHINE(3, 0)
3380 
3381 static void virt_machine_2_12_options(MachineClass *mc)
3382 {
3383     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3384 
3385     virt_machine_3_0_options(mc);
3386     compat_props_add(mc->compat_props, hw_compat_2_12, hw_compat_2_12_len);
3387     vmc->no_highmem_ecam = true;
3388     mc->max_cpus = 255;
3389 }
3390 DEFINE_VIRT_MACHINE(2, 12)
3391 
3392 static void virt_machine_2_11_options(MachineClass *mc)
3393 {
3394     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3395 
3396     virt_machine_2_12_options(mc);
3397     compat_props_add(mc->compat_props, hw_compat_2_11, hw_compat_2_11_len);
3398     vmc->smbios_old_sys_ver = true;
3399 }
3400 DEFINE_VIRT_MACHINE(2, 11)
3401 
3402 static void virt_machine_2_10_options(MachineClass *mc)
3403 {
3404     virt_machine_2_11_options(mc);
3405     compat_props_add(mc->compat_props, hw_compat_2_10, hw_compat_2_10_len);
3406     /* before 2.11 we never faulted accesses to bad addresses */
3407     mc->ignore_memory_transaction_failures = true;
3408 }
3409 DEFINE_VIRT_MACHINE(2, 10)
3410 
3411 static void virt_machine_2_9_options(MachineClass *mc)
3412 {
3413     virt_machine_2_10_options(mc);
3414     compat_props_add(mc->compat_props, hw_compat_2_9, hw_compat_2_9_len);
3415 }
3416 DEFINE_VIRT_MACHINE(2, 9)
3417 
3418 static void virt_machine_2_8_options(MachineClass *mc)
3419 {
3420     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3421 
3422     virt_machine_2_9_options(mc);
3423     compat_props_add(mc->compat_props, hw_compat_2_8, hw_compat_2_8_len);
3424     /* For 2.8 and earlier we falsely claimed in the DT that
3425      * our timers were edge-triggered, not level-triggered.
3426      */
3427     vmc->claim_edge_triggered_timers = true;
3428 }
3429 DEFINE_VIRT_MACHINE(2, 8)
3430 
3431 static void virt_machine_2_7_options(MachineClass *mc)
3432 {
3433     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3434 
3435     virt_machine_2_8_options(mc);
3436     compat_props_add(mc->compat_props, hw_compat_2_7, hw_compat_2_7_len);
3437     /* ITS was introduced with 2.8 */
3438     vmc->no_its = true;
3439     /* Stick with 1K pages for migration compatibility */
3440     mc->minimum_page_bits = 0;
3441 }
3442 DEFINE_VIRT_MACHINE(2, 7)
3443 
3444 static void virt_machine_2_6_options(MachineClass *mc)
3445 {
3446     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3447 
3448     virt_machine_2_7_options(mc);
3449     compat_props_add(mc->compat_props, hw_compat_2_6, hw_compat_2_6_len);
3450     vmc->disallow_affinity_adjustment = true;
3451     /* Disable PMU for 2.6 as PMU support was first introduced in 2.7 */
3452     vmc->no_pmu = true;
3453 }
3454 DEFINE_VIRT_MACHINE(2, 6)
3455