xref: /openbmc/qemu/hw/arm/virt.c (revision f2b90109)
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     ARM_CPU_TYPE_NAME("cortex-a7"),
208     ARM_CPU_TYPE_NAME("cortex-a15"),
209     ARM_CPU_TYPE_NAME("cortex-a35"),
210     ARM_CPU_TYPE_NAME("cortex-a53"),
211     ARM_CPU_TYPE_NAME("cortex-a55"),
212     ARM_CPU_TYPE_NAME("cortex-a57"),
213     ARM_CPU_TYPE_NAME("cortex-a72"),
214     ARM_CPU_TYPE_NAME("cortex-a76"),
215     ARM_CPU_TYPE_NAME("a64fx"),
216     ARM_CPU_TYPE_NAME("neoverse-n1"),
217     ARM_CPU_TYPE_NAME("host"),
218     ARM_CPU_TYPE_NAME("max"),
219 };
220 
221 static bool cpu_type_valid(const char *cpu)
222 {
223     int i;
224 
225     for (i = 0; i < ARRAY_SIZE(valid_cpus); i++) {
226         if (strcmp(cpu, valid_cpus[i]) == 0) {
227             return true;
228         }
229     }
230     return false;
231 }
232 
233 static void create_randomness(MachineState *ms, const char *node)
234 {
235     struct {
236         uint64_t kaslr;
237         uint8_t rng[32];
238     } seed;
239 
240     if (qemu_guest_getrandom(&seed, sizeof(seed), NULL)) {
241         return;
242     }
243     qemu_fdt_setprop_u64(ms->fdt, node, "kaslr-seed", seed.kaslr);
244     qemu_fdt_setprop(ms->fdt, node, "rng-seed", seed.rng, sizeof(seed.rng));
245 }
246 
247 static void create_fdt(VirtMachineState *vms)
248 {
249     MachineState *ms = MACHINE(vms);
250     int nb_numa_nodes = ms->numa_state->num_nodes;
251     void *fdt = create_device_tree(&vms->fdt_size);
252 
253     if (!fdt) {
254         error_report("create_device_tree() failed");
255         exit(1);
256     }
257 
258     ms->fdt = fdt;
259 
260     /* Header */
261     qemu_fdt_setprop_string(fdt, "/", "compatible", "linux,dummy-virt");
262     qemu_fdt_setprop_cell(fdt, "/", "#address-cells", 0x2);
263     qemu_fdt_setprop_cell(fdt, "/", "#size-cells", 0x2);
264     qemu_fdt_setprop_string(fdt, "/", "model", "linux,dummy-virt");
265 
266     /* /chosen must exist for load_dtb to fill in necessary properties later */
267     qemu_fdt_add_subnode(fdt, "/chosen");
268     if (vms->dtb_randomness) {
269         create_randomness(ms, "/chosen");
270     }
271 
272     if (vms->secure) {
273         qemu_fdt_add_subnode(fdt, "/secure-chosen");
274         if (vms->dtb_randomness) {
275             create_randomness(ms, "/secure-chosen");
276         }
277     }
278 
279     /* Clock node, for the benefit of the UART. The kernel device tree
280      * binding documentation claims the PL011 node clock properties are
281      * optional but in practice if you omit them the kernel refuses to
282      * probe for the device.
283      */
284     vms->clock_phandle = qemu_fdt_alloc_phandle(fdt);
285     qemu_fdt_add_subnode(fdt, "/apb-pclk");
286     qemu_fdt_setprop_string(fdt, "/apb-pclk", "compatible", "fixed-clock");
287     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "#clock-cells", 0x0);
288     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "clock-frequency", 24000000);
289     qemu_fdt_setprop_string(fdt, "/apb-pclk", "clock-output-names",
290                                 "clk24mhz");
291     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "phandle", vms->clock_phandle);
292 
293     if (nb_numa_nodes > 0 && ms->numa_state->have_numa_distance) {
294         int size = nb_numa_nodes * nb_numa_nodes * 3 * sizeof(uint32_t);
295         uint32_t *matrix = g_malloc0(size);
296         int idx, i, j;
297 
298         for (i = 0; i < nb_numa_nodes; i++) {
299             for (j = 0; j < nb_numa_nodes; j++) {
300                 idx = (i * nb_numa_nodes + j) * 3;
301                 matrix[idx + 0] = cpu_to_be32(i);
302                 matrix[idx + 1] = cpu_to_be32(j);
303                 matrix[idx + 2] =
304                     cpu_to_be32(ms->numa_state->nodes[i].distance[j]);
305             }
306         }
307 
308         qemu_fdt_add_subnode(fdt, "/distance-map");
309         qemu_fdt_setprop_string(fdt, "/distance-map", "compatible",
310                                 "numa-distance-map-v1");
311         qemu_fdt_setprop(fdt, "/distance-map", "distance-matrix",
312                          matrix, size);
313         g_free(matrix);
314     }
315 }
316 
317 static void fdt_add_timer_nodes(const VirtMachineState *vms)
318 {
319     /* On real hardware these interrupts are level-triggered.
320      * On KVM they were edge-triggered before host kernel version 4.4,
321      * and level-triggered afterwards.
322      * On emulated QEMU they are level-triggered.
323      *
324      * Getting the DTB info about them wrong is awkward for some
325      * guest kernels:
326      *  pre-4.8 ignore the DT and leave the interrupt configured
327      *   with whatever the GIC reset value (or the bootloader) left it at
328      *  4.8 before rc6 honour the incorrect data by programming it back
329      *   into the GIC, causing problems
330      *  4.8rc6 and later ignore the DT and always write "level triggered"
331      *   into the GIC
332      *
333      * For backwards-compatibility, virt-2.8 and earlier will continue
334      * to say these are edge-triggered, but later machines will report
335      * the correct information.
336      */
337     ARMCPU *armcpu;
338     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
339     uint32_t irqflags = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
340     MachineState *ms = MACHINE(vms);
341 
342     if (vmc->claim_edge_triggered_timers) {
343         irqflags = GIC_FDT_IRQ_FLAGS_EDGE_LO_HI;
344     }
345 
346     if (vms->gic_version == VIRT_GIC_VERSION_2) {
347         irqflags = deposit32(irqflags, GIC_FDT_IRQ_PPI_CPU_START,
348                              GIC_FDT_IRQ_PPI_CPU_WIDTH,
349                              (1 << MACHINE(vms)->smp.cpus) - 1);
350     }
351 
352     qemu_fdt_add_subnode(ms->fdt, "/timer");
353 
354     armcpu = ARM_CPU(qemu_get_cpu(0));
355     if (arm_feature(&armcpu->env, ARM_FEATURE_V8)) {
356         const char compat[] = "arm,armv8-timer\0arm,armv7-timer";
357         qemu_fdt_setprop(ms->fdt, "/timer", "compatible",
358                          compat, sizeof(compat));
359     } else {
360         qemu_fdt_setprop_string(ms->fdt, "/timer", "compatible",
361                                 "arm,armv7-timer");
362     }
363     qemu_fdt_setprop(ms->fdt, "/timer", "always-on", NULL, 0);
364     qemu_fdt_setprop_cells(ms->fdt, "/timer", "interrupts",
365                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_S_EL1_IRQ, irqflags,
366                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_NS_EL1_IRQ, irqflags,
367                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_VIRT_IRQ, irqflags,
368                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_NS_EL2_IRQ, irqflags);
369 }
370 
371 static void fdt_add_cpu_nodes(const VirtMachineState *vms)
372 {
373     int cpu;
374     int addr_cells = 1;
375     const MachineState *ms = MACHINE(vms);
376     const VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
377     int smp_cpus = ms->smp.cpus;
378 
379     /*
380      * See Linux Documentation/devicetree/bindings/arm/cpus.yaml
381      * On ARM v8 64-bit systems value should be set to 2,
382      * that corresponds to the MPIDR_EL1 register size.
383      * If MPIDR_EL1[63:32] value is equal to 0 on all CPUs
384      * in the system, #address-cells can be set to 1, since
385      * MPIDR_EL1[63:32] bits are not used for CPUs
386      * identification.
387      *
388      * Here we actually don't know whether our system is 32- or 64-bit one.
389      * The simplest way to go is to examine affinity IDs of all our CPUs. If
390      * at least one of them has Aff3 populated, we set #address-cells to 2.
391      */
392     for (cpu = 0; cpu < smp_cpus; cpu++) {
393         ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(cpu));
394 
395         if (armcpu->mp_affinity & ARM_AFF3_MASK) {
396             addr_cells = 2;
397             break;
398         }
399     }
400 
401     qemu_fdt_add_subnode(ms->fdt, "/cpus");
402     qemu_fdt_setprop_cell(ms->fdt, "/cpus", "#address-cells", addr_cells);
403     qemu_fdt_setprop_cell(ms->fdt, "/cpus", "#size-cells", 0x0);
404 
405     for (cpu = smp_cpus - 1; cpu >= 0; cpu--) {
406         char *nodename = g_strdup_printf("/cpus/cpu@%d", cpu);
407         ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(cpu));
408         CPUState *cs = CPU(armcpu);
409 
410         qemu_fdt_add_subnode(ms->fdt, nodename);
411         qemu_fdt_setprop_string(ms->fdt, nodename, "device_type", "cpu");
412         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
413                                     armcpu->dtb_compatible);
414 
415         if (vms->psci_conduit != QEMU_PSCI_CONDUIT_DISABLED && smp_cpus > 1) {
416             qemu_fdt_setprop_string(ms->fdt, nodename,
417                                         "enable-method", "psci");
418         }
419 
420         if (addr_cells == 2) {
421             qemu_fdt_setprop_u64(ms->fdt, nodename, "reg",
422                                  armcpu->mp_affinity);
423         } else {
424             qemu_fdt_setprop_cell(ms->fdt, nodename, "reg",
425                                   armcpu->mp_affinity);
426         }
427 
428         if (ms->possible_cpus->cpus[cs->cpu_index].props.has_node_id) {
429             qemu_fdt_setprop_cell(ms->fdt, nodename, "numa-node-id",
430                 ms->possible_cpus->cpus[cs->cpu_index].props.node_id);
431         }
432 
433         if (!vmc->no_cpu_topology) {
434             qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle",
435                                   qemu_fdt_alloc_phandle(ms->fdt));
436         }
437 
438         g_free(nodename);
439     }
440 
441     if (!vmc->no_cpu_topology) {
442         /*
443          * Add vCPU topology description through fdt node cpu-map.
444          *
445          * See Linux Documentation/devicetree/bindings/cpu/cpu-topology.txt
446          * In a SMP system, the hierarchy of CPUs can be defined through
447          * four entities that are used to describe the layout of CPUs in
448          * the system: socket/cluster/core/thread.
449          *
450          * A socket node represents the boundary of system physical package
451          * and its child nodes must be one or more cluster nodes. A system
452          * can contain several layers of clustering within a single physical
453          * package and cluster nodes can be contained in parent cluster nodes.
454          *
455          * Note: currently we only support one layer of clustering within
456          * each physical package.
457          */
458         qemu_fdt_add_subnode(ms->fdt, "/cpus/cpu-map");
459 
460         for (cpu = smp_cpus - 1; cpu >= 0; cpu--) {
461             char *cpu_path = g_strdup_printf("/cpus/cpu@%d", cpu);
462             char *map_path;
463 
464             if (ms->smp.threads > 1) {
465                 map_path = g_strdup_printf(
466                     "/cpus/cpu-map/socket%d/cluster%d/core%d/thread%d",
467                     cpu / (ms->smp.clusters * ms->smp.cores * ms->smp.threads),
468                     (cpu / (ms->smp.cores * ms->smp.threads)) % ms->smp.clusters,
469                     (cpu / ms->smp.threads) % ms->smp.cores,
470                     cpu % ms->smp.threads);
471             } else {
472                 map_path = g_strdup_printf(
473                     "/cpus/cpu-map/socket%d/cluster%d/core%d",
474                     cpu / (ms->smp.clusters * ms->smp.cores),
475                     (cpu / ms->smp.cores) % ms->smp.clusters,
476                     cpu % ms->smp.cores);
477             }
478             qemu_fdt_add_path(ms->fdt, map_path);
479             qemu_fdt_setprop_phandle(ms->fdt, map_path, "cpu", cpu_path);
480 
481             g_free(map_path);
482             g_free(cpu_path);
483         }
484     }
485 }
486 
487 static void fdt_add_its_gic_node(VirtMachineState *vms)
488 {
489     char *nodename;
490     MachineState *ms = MACHINE(vms);
491 
492     vms->msi_phandle = qemu_fdt_alloc_phandle(ms->fdt);
493     nodename = g_strdup_printf("/intc/its@%" PRIx64,
494                                vms->memmap[VIRT_GIC_ITS].base);
495     qemu_fdt_add_subnode(ms->fdt, nodename);
496     qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
497                             "arm,gic-v3-its");
498     qemu_fdt_setprop(ms->fdt, nodename, "msi-controller", NULL, 0);
499     qemu_fdt_setprop_cell(ms->fdt, nodename, "#msi-cells", 1);
500     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
501                                  2, vms->memmap[VIRT_GIC_ITS].base,
502                                  2, vms->memmap[VIRT_GIC_ITS].size);
503     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", vms->msi_phandle);
504     g_free(nodename);
505 }
506 
507 static void fdt_add_v2m_gic_node(VirtMachineState *vms)
508 {
509     MachineState *ms = MACHINE(vms);
510     char *nodename;
511 
512     nodename = g_strdup_printf("/intc/v2m@%" PRIx64,
513                                vms->memmap[VIRT_GIC_V2M].base);
514     vms->msi_phandle = qemu_fdt_alloc_phandle(ms->fdt);
515     qemu_fdt_add_subnode(ms->fdt, nodename);
516     qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
517                             "arm,gic-v2m-frame");
518     qemu_fdt_setprop(ms->fdt, nodename, "msi-controller", NULL, 0);
519     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
520                                  2, vms->memmap[VIRT_GIC_V2M].base,
521                                  2, vms->memmap[VIRT_GIC_V2M].size);
522     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", vms->msi_phandle);
523     g_free(nodename);
524 }
525 
526 static void fdt_add_gic_node(VirtMachineState *vms)
527 {
528     MachineState *ms = MACHINE(vms);
529     char *nodename;
530 
531     vms->gic_phandle = qemu_fdt_alloc_phandle(ms->fdt);
532     qemu_fdt_setprop_cell(ms->fdt, "/", "interrupt-parent", vms->gic_phandle);
533 
534     nodename = g_strdup_printf("/intc@%" PRIx64,
535                                vms->memmap[VIRT_GIC_DIST].base);
536     qemu_fdt_add_subnode(ms->fdt, nodename);
537     qemu_fdt_setprop_cell(ms->fdt, nodename, "#interrupt-cells", 3);
538     qemu_fdt_setprop(ms->fdt, nodename, "interrupt-controller", NULL, 0);
539     qemu_fdt_setprop_cell(ms->fdt, nodename, "#address-cells", 0x2);
540     qemu_fdt_setprop_cell(ms->fdt, nodename, "#size-cells", 0x2);
541     qemu_fdt_setprop(ms->fdt, nodename, "ranges", NULL, 0);
542     if (vms->gic_version != VIRT_GIC_VERSION_2) {
543         int nb_redist_regions = virt_gicv3_redist_region_count(vms);
544 
545         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
546                                 "arm,gic-v3");
547 
548         qemu_fdt_setprop_cell(ms->fdt, nodename,
549                               "#redistributor-regions", nb_redist_regions);
550 
551         if (nb_redist_regions == 1) {
552             qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
553                                          2, vms->memmap[VIRT_GIC_DIST].base,
554                                          2, vms->memmap[VIRT_GIC_DIST].size,
555                                          2, vms->memmap[VIRT_GIC_REDIST].base,
556                                          2, vms->memmap[VIRT_GIC_REDIST].size);
557         } else {
558             qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
559                                  2, vms->memmap[VIRT_GIC_DIST].base,
560                                  2, vms->memmap[VIRT_GIC_DIST].size,
561                                  2, vms->memmap[VIRT_GIC_REDIST].base,
562                                  2, vms->memmap[VIRT_GIC_REDIST].size,
563                                  2, vms->memmap[VIRT_HIGH_GIC_REDIST2].base,
564                                  2, vms->memmap[VIRT_HIGH_GIC_REDIST2].size);
565         }
566 
567         if (vms->virt) {
568             qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
569                                    GIC_FDT_IRQ_TYPE_PPI, ARCH_GIC_MAINT_IRQ,
570                                    GIC_FDT_IRQ_FLAGS_LEVEL_HI);
571         }
572     } else {
573         /* 'cortex-a15-gic' means 'GIC v2' */
574         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
575                                 "arm,cortex-a15-gic");
576         if (!vms->virt) {
577             qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
578                                          2, vms->memmap[VIRT_GIC_DIST].base,
579                                          2, vms->memmap[VIRT_GIC_DIST].size,
580                                          2, vms->memmap[VIRT_GIC_CPU].base,
581                                          2, vms->memmap[VIRT_GIC_CPU].size);
582         } else {
583             qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
584                                          2, vms->memmap[VIRT_GIC_DIST].base,
585                                          2, vms->memmap[VIRT_GIC_DIST].size,
586                                          2, vms->memmap[VIRT_GIC_CPU].base,
587                                          2, vms->memmap[VIRT_GIC_CPU].size,
588                                          2, vms->memmap[VIRT_GIC_HYP].base,
589                                          2, vms->memmap[VIRT_GIC_HYP].size,
590                                          2, vms->memmap[VIRT_GIC_VCPU].base,
591                                          2, vms->memmap[VIRT_GIC_VCPU].size);
592             qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
593                                    GIC_FDT_IRQ_TYPE_PPI, ARCH_GIC_MAINT_IRQ,
594                                    GIC_FDT_IRQ_FLAGS_LEVEL_HI);
595         }
596     }
597 
598     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", vms->gic_phandle);
599     g_free(nodename);
600 }
601 
602 static void fdt_add_pmu_nodes(const VirtMachineState *vms)
603 {
604     ARMCPU *armcpu = ARM_CPU(first_cpu);
605     uint32_t irqflags = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
606     MachineState *ms = MACHINE(vms);
607 
608     if (!arm_feature(&armcpu->env, ARM_FEATURE_PMU)) {
609         assert(!object_property_get_bool(OBJECT(armcpu), "pmu", NULL));
610         return;
611     }
612 
613     if (vms->gic_version == VIRT_GIC_VERSION_2) {
614         irqflags = deposit32(irqflags, GIC_FDT_IRQ_PPI_CPU_START,
615                              GIC_FDT_IRQ_PPI_CPU_WIDTH,
616                              (1 << MACHINE(vms)->smp.cpus) - 1);
617     }
618 
619     qemu_fdt_add_subnode(ms->fdt, "/pmu");
620     if (arm_feature(&armcpu->env, ARM_FEATURE_V8)) {
621         const char compat[] = "arm,armv8-pmuv3";
622         qemu_fdt_setprop(ms->fdt, "/pmu", "compatible",
623                          compat, sizeof(compat));
624         qemu_fdt_setprop_cells(ms->fdt, "/pmu", "interrupts",
625                                GIC_FDT_IRQ_TYPE_PPI, VIRTUAL_PMU_IRQ, irqflags);
626     }
627 }
628 
629 static inline DeviceState *create_acpi_ged(VirtMachineState *vms)
630 {
631     DeviceState *dev;
632     MachineState *ms = MACHINE(vms);
633     int irq = vms->irqmap[VIRT_ACPI_GED];
634     uint32_t event = ACPI_GED_PWR_DOWN_EVT;
635 
636     if (ms->ram_slots) {
637         event |= ACPI_GED_MEM_HOTPLUG_EVT;
638     }
639 
640     if (ms->nvdimms_state->is_enabled) {
641         event |= ACPI_GED_NVDIMM_HOTPLUG_EVT;
642     }
643 
644     dev = qdev_new(TYPE_ACPI_GED);
645     qdev_prop_set_uint32(dev, "ged-event", event);
646 
647     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_ACPI_GED].base);
648     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 1, vms->memmap[VIRT_PCDIMM_ACPI].base);
649     sysbus_connect_irq(SYS_BUS_DEVICE(dev), 0, qdev_get_gpio_in(vms->gic, irq));
650 
651     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
652 
653     return dev;
654 }
655 
656 static void create_its(VirtMachineState *vms)
657 {
658     const char *itsclass = its_class_name();
659     DeviceState *dev;
660 
661     if (!strcmp(itsclass, "arm-gicv3-its")) {
662         if (!vms->tcg_its) {
663             itsclass = NULL;
664         }
665     }
666 
667     if (!itsclass) {
668         /* Do nothing if not supported */
669         return;
670     }
671 
672     dev = qdev_new(itsclass);
673 
674     object_property_set_link(OBJECT(dev), "parent-gicv3", OBJECT(vms->gic),
675                              &error_abort);
676     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
677     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_GIC_ITS].base);
678 
679     fdt_add_its_gic_node(vms);
680     vms->msi_controller = VIRT_MSI_CTRL_ITS;
681 }
682 
683 static void create_v2m(VirtMachineState *vms)
684 {
685     int i;
686     int irq = vms->irqmap[VIRT_GIC_V2M];
687     DeviceState *dev;
688 
689     dev = qdev_new("arm-gicv2m");
690     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_GIC_V2M].base);
691     qdev_prop_set_uint32(dev, "base-spi", irq);
692     qdev_prop_set_uint32(dev, "num-spi", NUM_GICV2M_SPIS);
693     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
694 
695     for (i = 0; i < NUM_GICV2M_SPIS; i++) {
696         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
697                            qdev_get_gpio_in(vms->gic, irq + i));
698     }
699 
700     fdt_add_v2m_gic_node(vms);
701     vms->msi_controller = VIRT_MSI_CTRL_GICV2M;
702 }
703 
704 static void create_gic(VirtMachineState *vms, MemoryRegion *mem)
705 {
706     MachineState *ms = MACHINE(vms);
707     /* We create a standalone GIC */
708     SysBusDevice *gicbusdev;
709     const char *gictype;
710     int i;
711     unsigned int smp_cpus = ms->smp.cpus;
712     uint32_t nb_redist_regions = 0;
713     int revision;
714 
715     if (vms->gic_version == VIRT_GIC_VERSION_2) {
716         gictype = gic_class_name();
717     } else {
718         gictype = gicv3_class_name();
719     }
720 
721     switch (vms->gic_version) {
722     case VIRT_GIC_VERSION_2:
723         revision = 2;
724         break;
725     case VIRT_GIC_VERSION_3:
726         revision = 3;
727         break;
728     case VIRT_GIC_VERSION_4:
729         revision = 4;
730         break;
731     default:
732         g_assert_not_reached();
733     }
734     vms->gic = qdev_new(gictype);
735     qdev_prop_set_uint32(vms->gic, "revision", revision);
736     qdev_prop_set_uint32(vms->gic, "num-cpu", smp_cpus);
737     /* Note that the num-irq property counts both internal and external
738      * interrupts; there are always 32 of the former (mandated by GIC spec).
739      */
740     qdev_prop_set_uint32(vms->gic, "num-irq", NUM_IRQS + 32);
741     if (!kvm_irqchip_in_kernel()) {
742         qdev_prop_set_bit(vms->gic, "has-security-extensions", vms->secure);
743     }
744 
745     if (vms->gic_version != VIRT_GIC_VERSION_2) {
746         uint32_t redist0_capacity = virt_redist_capacity(vms, VIRT_GIC_REDIST);
747         uint32_t redist0_count = MIN(smp_cpus, redist0_capacity);
748 
749         nb_redist_regions = virt_gicv3_redist_region_count(vms);
750 
751         qdev_prop_set_uint32(vms->gic, "len-redist-region-count",
752                              nb_redist_regions);
753         qdev_prop_set_uint32(vms->gic, "redist-region-count[0]", redist0_count);
754 
755         if (!kvm_irqchip_in_kernel()) {
756             if (vms->tcg_its) {
757                 object_property_set_link(OBJECT(vms->gic), "sysmem",
758                                          OBJECT(mem), &error_fatal);
759                 qdev_prop_set_bit(vms->gic, "has-lpi", true);
760             }
761         }
762 
763         if (nb_redist_regions == 2) {
764             uint32_t redist1_capacity =
765                 virt_redist_capacity(vms, VIRT_HIGH_GIC_REDIST2);
766 
767             qdev_prop_set_uint32(vms->gic, "redist-region-count[1]",
768                 MIN(smp_cpus - redist0_count, redist1_capacity));
769         }
770     } else {
771         if (!kvm_irqchip_in_kernel()) {
772             qdev_prop_set_bit(vms->gic, "has-virtualization-extensions",
773                               vms->virt);
774         }
775     }
776     gicbusdev = SYS_BUS_DEVICE(vms->gic);
777     sysbus_realize_and_unref(gicbusdev, &error_fatal);
778     sysbus_mmio_map(gicbusdev, 0, vms->memmap[VIRT_GIC_DIST].base);
779     if (vms->gic_version != VIRT_GIC_VERSION_2) {
780         sysbus_mmio_map(gicbusdev, 1, vms->memmap[VIRT_GIC_REDIST].base);
781         if (nb_redist_regions == 2) {
782             sysbus_mmio_map(gicbusdev, 2,
783                             vms->memmap[VIRT_HIGH_GIC_REDIST2].base);
784         }
785     } else {
786         sysbus_mmio_map(gicbusdev, 1, vms->memmap[VIRT_GIC_CPU].base);
787         if (vms->virt) {
788             sysbus_mmio_map(gicbusdev, 2, vms->memmap[VIRT_GIC_HYP].base);
789             sysbus_mmio_map(gicbusdev, 3, vms->memmap[VIRT_GIC_VCPU].base);
790         }
791     }
792 
793     /* Wire the outputs from each CPU's generic timer and the GICv3
794      * maintenance interrupt signal to the appropriate GIC PPI inputs,
795      * and the GIC's IRQ/FIQ/VIRQ/VFIQ interrupt outputs to the CPU's inputs.
796      */
797     for (i = 0; i < smp_cpus; i++) {
798         DeviceState *cpudev = DEVICE(qemu_get_cpu(i));
799         int ppibase = NUM_IRQS + i * GIC_INTERNAL + GIC_NR_SGIS;
800         int irq;
801         /* Mapping from the output timer irq lines from the CPU to the
802          * GIC PPI inputs we use for the virt board.
803          */
804         const int timer_irq[] = {
805             [GTIMER_PHYS] = ARCH_TIMER_NS_EL1_IRQ,
806             [GTIMER_VIRT] = ARCH_TIMER_VIRT_IRQ,
807             [GTIMER_HYP]  = ARCH_TIMER_NS_EL2_IRQ,
808             [GTIMER_SEC]  = ARCH_TIMER_S_EL1_IRQ,
809         };
810 
811         for (irq = 0; irq < ARRAY_SIZE(timer_irq); irq++) {
812             qdev_connect_gpio_out(cpudev, irq,
813                                   qdev_get_gpio_in(vms->gic,
814                                                    ppibase + timer_irq[irq]));
815         }
816 
817         if (vms->gic_version != VIRT_GIC_VERSION_2) {
818             qemu_irq irq = qdev_get_gpio_in(vms->gic,
819                                             ppibase + ARCH_GIC_MAINT_IRQ);
820             qdev_connect_gpio_out_named(cpudev, "gicv3-maintenance-interrupt",
821                                         0, irq);
822         } else if (vms->virt) {
823             qemu_irq irq = qdev_get_gpio_in(vms->gic,
824                                             ppibase + ARCH_GIC_MAINT_IRQ);
825             sysbus_connect_irq(gicbusdev, i + 4 * smp_cpus, irq);
826         }
827 
828         qdev_connect_gpio_out_named(cpudev, "pmu-interrupt", 0,
829                                     qdev_get_gpio_in(vms->gic, ppibase
830                                                      + VIRTUAL_PMU_IRQ));
831 
832         sysbus_connect_irq(gicbusdev, i, qdev_get_gpio_in(cpudev, ARM_CPU_IRQ));
833         sysbus_connect_irq(gicbusdev, i + smp_cpus,
834                            qdev_get_gpio_in(cpudev, ARM_CPU_FIQ));
835         sysbus_connect_irq(gicbusdev, i + 2 * smp_cpus,
836                            qdev_get_gpio_in(cpudev, ARM_CPU_VIRQ));
837         sysbus_connect_irq(gicbusdev, i + 3 * smp_cpus,
838                            qdev_get_gpio_in(cpudev, ARM_CPU_VFIQ));
839     }
840 
841     fdt_add_gic_node(vms);
842 
843     if (vms->gic_version != VIRT_GIC_VERSION_2 && vms->its) {
844         create_its(vms);
845     } else if (vms->gic_version == VIRT_GIC_VERSION_2) {
846         create_v2m(vms);
847     }
848 }
849 
850 static void create_uart(const VirtMachineState *vms, int uart,
851                         MemoryRegion *mem, Chardev *chr)
852 {
853     char *nodename;
854     hwaddr base = vms->memmap[uart].base;
855     hwaddr size = vms->memmap[uart].size;
856     int irq = vms->irqmap[uart];
857     const char compat[] = "arm,pl011\0arm,primecell";
858     const char clocknames[] = "uartclk\0apb_pclk";
859     DeviceState *dev = qdev_new(TYPE_PL011);
860     SysBusDevice *s = SYS_BUS_DEVICE(dev);
861     MachineState *ms = MACHINE(vms);
862 
863     qdev_prop_set_chr(dev, "chardev", chr);
864     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
865     memory_region_add_subregion(mem, base,
866                                 sysbus_mmio_get_region(s, 0));
867     sysbus_connect_irq(s, 0, qdev_get_gpio_in(vms->gic, irq));
868 
869     nodename = g_strdup_printf("/pl011@%" PRIx64, base);
870     qemu_fdt_add_subnode(ms->fdt, nodename);
871     /* Note that we can't use setprop_string because of the embedded NUL */
872     qemu_fdt_setprop(ms->fdt, nodename, "compatible",
873                          compat, sizeof(compat));
874     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
875                                      2, base, 2, size);
876     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
877                                GIC_FDT_IRQ_TYPE_SPI, irq,
878                                GIC_FDT_IRQ_FLAGS_LEVEL_HI);
879     qemu_fdt_setprop_cells(ms->fdt, nodename, "clocks",
880                                vms->clock_phandle, vms->clock_phandle);
881     qemu_fdt_setprop(ms->fdt, nodename, "clock-names",
882                          clocknames, sizeof(clocknames));
883 
884     if (uart == VIRT_UART) {
885         qemu_fdt_setprop_string(ms->fdt, "/chosen", "stdout-path", nodename);
886     } else {
887         /* Mark as not usable by the normal world */
888         qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
889         qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
890 
891         qemu_fdt_setprop_string(ms->fdt, "/secure-chosen", "stdout-path",
892                                 nodename);
893     }
894 
895     g_free(nodename);
896 }
897 
898 static void create_rtc(const VirtMachineState *vms)
899 {
900     char *nodename;
901     hwaddr base = vms->memmap[VIRT_RTC].base;
902     hwaddr size = vms->memmap[VIRT_RTC].size;
903     int irq = vms->irqmap[VIRT_RTC];
904     const char compat[] = "arm,pl031\0arm,primecell";
905     MachineState *ms = MACHINE(vms);
906 
907     sysbus_create_simple("pl031", base, qdev_get_gpio_in(vms->gic, irq));
908 
909     nodename = g_strdup_printf("/pl031@%" PRIx64, base);
910     qemu_fdt_add_subnode(ms->fdt, nodename);
911     qemu_fdt_setprop(ms->fdt, nodename, "compatible", compat, sizeof(compat));
912     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
913                                  2, base, 2, size);
914     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
915                            GIC_FDT_IRQ_TYPE_SPI, irq,
916                            GIC_FDT_IRQ_FLAGS_LEVEL_HI);
917     qemu_fdt_setprop_cell(ms->fdt, nodename, "clocks", vms->clock_phandle);
918     qemu_fdt_setprop_string(ms->fdt, nodename, "clock-names", "apb_pclk");
919     g_free(nodename);
920 }
921 
922 static DeviceState *gpio_key_dev;
923 static void virt_powerdown_req(Notifier *n, void *opaque)
924 {
925     VirtMachineState *s = container_of(n, VirtMachineState, powerdown_notifier);
926 
927     if (s->acpi_dev) {
928         acpi_send_event(s->acpi_dev, ACPI_POWER_DOWN_STATUS);
929     } else {
930         /* use gpio Pin 3 for power button event */
931         qemu_set_irq(qdev_get_gpio_in(gpio_key_dev, 0), 1);
932     }
933 }
934 
935 static void create_gpio_keys(char *fdt, DeviceState *pl061_dev,
936                              uint32_t phandle)
937 {
938     gpio_key_dev = sysbus_create_simple("gpio-key", -1,
939                                         qdev_get_gpio_in(pl061_dev, 3));
940 
941     qemu_fdt_add_subnode(fdt, "/gpio-keys");
942     qemu_fdt_setprop_string(fdt, "/gpio-keys", "compatible", "gpio-keys");
943 
944     qemu_fdt_add_subnode(fdt, "/gpio-keys/poweroff");
945     qemu_fdt_setprop_string(fdt, "/gpio-keys/poweroff",
946                             "label", "GPIO Key Poweroff");
947     qemu_fdt_setprop_cell(fdt, "/gpio-keys/poweroff", "linux,code",
948                           KEY_POWER);
949     qemu_fdt_setprop_cells(fdt, "/gpio-keys/poweroff",
950                            "gpios", phandle, 3, 0);
951 }
952 
953 #define SECURE_GPIO_POWEROFF 0
954 #define SECURE_GPIO_RESET    1
955 
956 static void create_secure_gpio_pwr(char *fdt, DeviceState *pl061_dev,
957                                    uint32_t phandle)
958 {
959     DeviceState *gpio_pwr_dev;
960 
961     /* gpio-pwr */
962     gpio_pwr_dev = sysbus_create_simple("gpio-pwr", -1, NULL);
963 
964     /* connect secure pl061 to gpio-pwr */
965     qdev_connect_gpio_out(pl061_dev, SECURE_GPIO_RESET,
966                           qdev_get_gpio_in_named(gpio_pwr_dev, "reset", 0));
967     qdev_connect_gpio_out(pl061_dev, SECURE_GPIO_POWEROFF,
968                           qdev_get_gpio_in_named(gpio_pwr_dev, "shutdown", 0));
969 
970     qemu_fdt_add_subnode(fdt, "/gpio-poweroff");
971     qemu_fdt_setprop_string(fdt, "/gpio-poweroff", "compatible",
972                             "gpio-poweroff");
973     qemu_fdt_setprop_cells(fdt, "/gpio-poweroff",
974                            "gpios", phandle, SECURE_GPIO_POWEROFF, 0);
975     qemu_fdt_setprop_string(fdt, "/gpio-poweroff", "status", "disabled");
976     qemu_fdt_setprop_string(fdt, "/gpio-poweroff", "secure-status",
977                             "okay");
978 
979     qemu_fdt_add_subnode(fdt, "/gpio-restart");
980     qemu_fdt_setprop_string(fdt, "/gpio-restart", "compatible",
981                             "gpio-restart");
982     qemu_fdt_setprop_cells(fdt, "/gpio-restart",
983                            "gpios", phandle, SECURE_GPIO_RESET, 0);
984     qemu_fdt_setprop_string(fdt, "/gpio-restart", "status", "disabled");
985     qemu_fdt_setprop_string(fdt, "/gpio-restart", "secure-status",
986                             "okay");
987 }
988 
989 static void create_gpio_devices(const VirtMachineState *vms, int gpio,
990                                 MemoryRegion *mem)
991 {
992     char *nodename;
993     DeviceState *pl061_dev;
994     hwaddr base = vms->memmap[gpio].base;
995     hwaddr size = vms->memmap[gpio].size;
996     int irq = vms->irqmap[gpio];
997     const char compat[] = "arm,pl061\0arm,primecell";
998     SysBusDevice *s;
999     MachineState *ms = MACHINE(vms);
1000 
1001     pl061_dev = qdev_new("pl061");
1002     /* Pull lines down to 0 if not driven by the PL061 */
1003     qdev_prop_set_uint32(pl061_dev, "pullups", 0);
1004     qdev_prop_set_uint32(pl061_dev, "pulldowns", 0xff);
1005     s = SYS_BUS_DEVICE(pl061_dev);
1006     sysbus_realize_and_unref(s, &error_fatal);
1007     memory_region_add_subregion(mem, base, sysbus_mmio_get_region(s, 0));
1008     sysbus_connect_irq(s, 0, qdev_get_gpio_in(vms->gic, irq));
1009 
1010     uint32_t phandle = qemu_fdt_alloc_phandle(ms->fdt);
1011     nodename = g_strdup_printf("/pl061@%" PRIx64, base);
1012     qemu_fdt_add_subnode(ms->fdt, nodename);
1013     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1014                                  2, base, 2, size);
1015     qemu_fdt_setprop(ms->fdt, nodename, "compatible", compat, sizeof(compat));
1016     qemu_fdt_setprop_cell(ms->fdt, nodename, "#gpio-cells", 2);
1017     qemu_fdt_setprop(ms->fdt, nodename, "gpio-controller", NULL, 0);
1018     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
1019                            GIC_FDT_IRQ_TYPE_SPI, irq,
1020                            GIC_FDT_IRQ_FLAGS_LEVEL_HI);
1021     qemu_fdt_setprop_cell(ms->fdt, nodename, "clocks", vms->clock_phandle);
1022     qemu_fdt_setprop_string(ms->fdt, nodename, "clock-names", "apb_pclk");
1023     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", phandle);
1024 
1025     if (gpio != VIRT_GPIO) {
1026         /* Mark as not usable by the normal world */
1027         qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
1028         qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
1029     }
1030     g_free(nodename);
1031 
1032     /* Child gpio devices */
1033     if (gpio == VIRT_GPIO) {
1034         create_gpio_keys(ms->fdt, pl061_dev, phandle);
1035     } else {
1036         create_secure_gpio_pwr(ms->fdt, pl061_dev, phandle);
1037     }
1038 }
1039 
1040 static void create_virtio_devices(const VirtMachineState *vms)
1041 {
1042     int i;
1043     hwaddr size = vms->memmap[VIRT_MMIO].size;
1044     MachineState *ms = MACHINE(vms);
1045 
1046     /* We create the transports in forwards order. Since qbus_realize()
1047      * prepends (not appends) new child buses, the incrementing loop below will
1048      * create a list of virtio-mmio buses with decreasing base addresses.
1049      *
1050      * When a -device option is processed from the command line,
1051      * qbus_find_recursive() picks the next free virtio-mmio bus in forwards
1052      * order. The upshot is that -device options in increasing command line
1053      * order are mapped to virtio-mmio buses with decreasing base addresses.
1054      *
1055      * When this code was originally written, that arrangement ensured that the
1056      * guest Linux kernel would give the lowest "name" (/dev/vda, eth0, etc) to
1057      * the first -device on the command line. (The end-to-end order is a
1058      * function of this loop, qbus_realize(), qbus_find_recursive(), and the
1059      * guest kernel's name-to-address assignment strategy.)
1060      *
1061      * Meanwhile, the kernel's traversal seems to have been reversed; see eg.
1062      * the message, if not necessarily the code, of commit 70161ff336.
1063      * Therefore the loop now establishes the inverse of the original intent.
1064      *
1065      * Unfortunately, we can't counteract the kernel change by reversing the
1066      * loop; it would break existing command lines.
1067      *
1068      * In any case, the kernel makes no guarantee about the stability of
1069      * enumeration order of virtio devices (as demonstrated by it changing
1070      * between kernel versions). For reliable and stable identification
1071      * of disks users must use UUIDs or similar mechanisms.
1072      */
1073     for (i = 0; i < NUM_VIRTIO_TRANSPORTS; i++) {
1074         int irq = vms->irqmap[VIRT_MMIO] + i;
1075         hwaddr base = vms->memmap[VIRT_MMIO].base + i * size;
1076 
1077         sysbus_create_simple("virtio-mmio", base,
1078                              qdev_get_gpio_in(vms->gic, irq));
1079     }
1080 
1081     /* We add dtb nodes in reverse order so that they appear in the finished
1082      * device tree lowest address first.
1083      *
1084      * Note that this mapping is independent of the loop above. The previous
1085      * loop influences virtio device to virtio transport assignment, whereas
1086      * this loop controls how virtio transports are laid out in the dtb.
1087      */
1088     for (i = NUM_VIRTIO_TRANSPORTS - 1; i >= 0; i--) {
1089         char *nodename;
1090         int irq = vms->irqmap[VIRT_MMIO] + i;
1091         hwaddr base = vms->memmap[VIRT_MMIO].base + i * size;
1092 
1093         nodename = g_strdup_printf("/virtio_mmio@%" PRIx64, base);
1094         qemu_fdt_add_subnode(ms->fdt, nodename);
1095         qemu_fdt_setprop_string(ms->fdt, nodename,
1096                                 "compatible", "virtio,mmio");
1097         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1098                                      2, base, 2, size);
1099         qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
1100                                GIC_FDT_IRQ_TYPE_SPI, irq,
1101                                GIC_FDT_IRQ_FLAGS_EDGE_LO_HI);
1102         qemu_fdt_setprop(ms->fdt, nodename, "dma-coherent", NULL, 0);
1103         g_free(nodename);
1104     }
1105 }
1106 
1107 #define VIRT_FLASH_SECTOR_SIZE (256 * KiB)
1108 
1109 static PFlashCFI01 *virt_flash_create1(VirtMachineState *vms,
1110                                         const char *name,
1111                                         const char *alias_prop_name)
1112 {
1113     /*
1114      * Create a single flash device.  We use the same parameters as
1115      * the flash devices on the Versatile Express board.
1116      */
1117     DeviceState *dev = qdev_new(TYPE_PFLASH_CFI01);
1118 
1119     qdev_prop_set_uint64(dev, "sector-length", VIRT_FLASH_SECTOR_SIZE);
1120     qdev_prop_set_uint8(dev, "width", 4);
1121     qdev_prop_set_uint8(dev, "device-width", 2);
1122     qdev_prop_set_bit(dev, "big-endian", false);
1123     qdev_prop_set_uint16(dev, "id0", 0x89);
1124     qdev_prop_set_uint16(dev, "id1", 0x18);
1125     qdev_prop_set_uint16(dev, "id2", 0x00);
1126     qdev_prop_set_uint16(dev, "id3", 0x00);
1127     qdev_prop_set_string(dev, "name", name);
1128     object_property_add_child(OBJECT(vms), name, OBJECT(dev));
1129     object_property_add_alias(OBJECT(vms), alias_prop_name,
1130                               OBJECT(dev), "drive");
1131     return PFLASH_CFI01(dev);
1132 }
1133 
1134 static void virt_flash_create(VirtMachineState *vms)
1135 {
1136     vms->flash[0] = virt_flash_create1(vms, "virt.flash0", "pflash0");
1137     vms->flash[1] = virt_flash_create1(vms, "virt.flash1", "pflash1");
1138 }
1139 
1140 static void virt_flash_map1(PFlashCFI01 *flash,
1141                             hwaddr base, hwaddr size,
1142                             MemoryRegion *sysmem)
1143 {
1144     DeviceState *dev = DEVICE(flash);
1145 
1146     assert(QEMU_IS_ALIGNED(size, VIRT_FLASH_SECTOR_SIZE));
1147     assert(size / VIRT_FLASH_SECTOR_SIZE <= UINT32_MAX);
1148     qdev_prop_set_uint32(dev, "num-blocks", size / VIRT_FLASH_SECTOR_SIZE);
1149     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1150 
1151     memory_region_add_subregion(sysmem, base,
1152                                 sysbus_mmio_get_region(SYS_BUS_DEVICE(dev),
1153                                                        0));
1154 }
1155 
1156 static void virt_flash_map(VirtMachineState *vms,
1157                            MemoryRegion *sysmem,
1158                            MemoryRegion *secure_sysmem)
1159 {
1160     /*
1161      * Map two flash devices to fill the VIRT_FLASH space in the memmap.
1162      * sysmem is the system memory space. secure_sysmem is the secure view
1163      * of the system, and the first flash device should be made visible only
1164      * there. The second flash device is visible to both secure and nonsecure.
1165      * If sysmem == secure_sysmem this means there is no separate Secure
1166      * address space and both flash devices are generally visible.
1167      */
1168     hwaddr flashsize = vms->memmap[VIRT_FLASH].size / 2;
1169     hwaddr flashbase = vms->memmap[VIRT_FLASH].base;
1170 
1171     virt_flash_map1(vms->flash[0], flashbase, flashsize,
1172                     secure_sysmem);
1173     virt_flash_map1(vms->flash[1], flashbase + flashsize, flashsize,
1174                     sysmem);
1175 }
1176 
1177 static void virt_flash_fdt(VirtMachineState *vms,
1178                            MemoryRegion *sysmem,
1179                            MemoryRegion *secure_sysmem)
1180 {
1181     hwaddr flashsize = vms->memmap[VIRT_FLASH].size / 2;
1182     hwaddr flashbase = vms->memmap[VIRT_FLASH].base;
1183     MachineState *ms = MACHINE(vms);
1184     char *nodename;
1185 
1186     if (sysmem == secure_sysmem) {
1187         /* Report both flash devices as a single node in the DT */
1188         nodename = g_strdup_printf("/flash@%" PRIx64, flashbase);
1189         qemu_fdt_add_subnode(ms->fdt, nodename);
1190         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible", "cfi-flash");
1191         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1192                                      2, flashbase, 2, flashsize,
1193                                      2, flashbase + flashsize, 2, flashsize);
1194         qemu_fdt_setprop_cell(ms->fdt, nodename, "bank-width", 4);
1195         g_free(nodename);
1196     } else {
1197         /*
1198          * Report the devices as separate nodes so we can mark one as
1199          * only visible to the secure world.
1200          */
1201         nodename = g_strdup_printf("/secflash@%" PRIx64, flashbase);
1202         qemu_fdt_add_subnode(ms->fdt, nodename);
1203         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible", "cfi-flash");
1204         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1205                                      2, flashbase, 2, flashsize);
1206         qemu_fdt_setprop_cell(ms->fdt, nodename, "bank-width", 4);
1207         qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
1208         qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
1209         g_free(nodename);
1210 
1211         nodename = g_strdup_printf("/flash@%" PRIx64, flashbase + flashsize);
1212         qemu_fdt_add_subnode(ms->fdt, nodename);
1213         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible", "cfi-flash");
1214         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1215                                      2, flashbase + flashsize, 2, flashsize);
1216         qemu_fdt_setprop_cell(ms->fdt, nodename, "bank-width", 4);
1217         g_free(nodename);
1218     }
1219 }
1220 
1221 static bool virt_firmware_init(VirtMachineState *vms,
1222                                MemoryRegion *sysmem,
1223                                MemoryRegion *secure_sysmem)
1224 {
1225     int i;
1226     const char *bios_name;
1227     BlockBackend *pflash_blk0;
1228 
1229     /* Map legacy -drive if=pflash to machine properties */
1230     for (i = 0; i < ARRAY_SIZE(vms->flash); i++) {
1231         pflash_cfi01_legacy_drive(vms->flash[i],
1232                                   drive_get(IF_PFLASH, 0, i));
1233     }
1234 
1235     virt_flash_map(vms, sysmem, secure_sysmem);
1236 
1237     pflash_blk0 = pflash_cfi01_get_blk(vms->flash[0]);
1238 
1239     bios_name = MACHINE(vms)->firmware;
1240     if (bios_name) {
1241         char *fname;
1242         MemoryRegion *mr;
1243         int image_size;
1244 
1245         if (pflash_blk0) {
1246             error_report("The contents of the first flash device may be "
1247                          "specified with -bios or with -drive if=pflash... "
1248                          "but you cannot use both options at once");
1249             exit(1);
1250         }
1251 
1252         /* Fall back to -bios */
1253 
1254         fname = qemu_find_file(QEMU_FILE_TYPE_BIOS, bios_name);
1255         if (!fname) {
1256             error_report("Could not find ROM image '%s'", bios_name);
1257             exit(1);
1258         }
1259         mr = sysbus_mmio_get_region(SYS_BUS_DEVICE(vms->flash[0]), 0);
1260         image_size = load_image_mr(fname, mr);
1261         g_free(fname);
1262         if (image_size < 0) {
1263             error_report("Could not load ROM image '%s'", bios_name);
1264             exit(1);
1265         }
1266     }
1267 
1268     return pflash_blk0 || bios_name;
1269 }
1270 
1271 static FWCfgState *create_fw_cfg(const VirtMachineState *vms, AddressSpace *as)
1272 {
1273     MachineState *ms = MACHINE(vms);
1274     hwaddr base = vms->memmap[VIRT_FW_CFG].base;
1275     hwaddr size = vms->memmap[VIRT_FW_CFG].size;
1276     FWCfgState *fw_cfg;
1277     char *nodename;
1278 
1279     fw_cfg = fw_cfg_init_mem_wide(base + 8, base, 8, base + 16, as);
1280     fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, (uint16_t)ms->smp.cpus);
1281 
1282     nodename = g_strdup_printf("/fw-cfg@%" PRIx64, base);
1283     qemu_fdt_add_subnode(ms->fdt, nodename);
1284     qemu_fdt_setprop_string(ms->fdt, nodename,
1285                             "compatible", "qemu,fw-cfg-mmio");
1286     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1287                                  2, base, 2, size);
1288     qemu_fdt_setprop(ms->fdt, nodename, "dma-coherent", NULL, 0);
1289     g_free(nodename);
1290     return fw_cfg;
1291 }
1292 
1293 static void create_pcie_irq_map(const MachineState *ms,
1294                                 uint32_t gic_phandle,
1295                                 int first_irq, const char *nodename)
1296 {
1297     int devfn, pin;
1298     uint32_t full_irq_map[4 * 4 * 10] = { 0 };
1299     uint32_t *irq_map = full_irq_map;
1300 
1301     for (devfn = 0; devfn <= 0x18; devfn += 0x8) {
1302         for (pin = 0; pin < 4; pin++) {
1303             int irq_type = GIC_FDT_IRQ_TYPE_SPI;
1304             int irq_nr = first_irq + ((pin + PCI_SLOT(devfn)) % PCI_NUM_PINS);
1305             int irq_level = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
1306             int i;
1307 
1308             uint32_t map[] = {
1309                 devfn << 8, 0, 0,                           /* devfn */
1310                 pin + 1,                                    /* PCI pin */
1311                 gic_phandle, 0, 0, irq_type, irq_nr, irq_level }; /* GIC irq */
1312 
1313             /* Convert map to big endian */
1314             for (i = 0; i < 10; i++) {
1315                 irq_map[i] = cpu_to_be32(map[i]);
1316             }
1317             irq_map += 10;
1318         }
1319     }
1320 
1321     qemu_fdt_setprop(ms->fdt, nodename, "interrupt-map",
1322                      full_irq_map, sizeof(full_irq_map));
1323 
1324     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupt-map-mask",
1325                            cpu_to_be16(PCI_DEVFN(3, 0)), /* Slot 3 */
1326                            0, 0,
1327                            0x7           /* PCI irq */);
1328 }
1329 
1330 static void create_smmu(const VirtMachineState *vms,
1331                         PCIBus *bus)
1332 {
1333     char *node;
1334     const char compat[] = "arm,smmu-v3";
1335     int irq =  vms->irqmap[VIRT_SMMU];
1336     int i;
1337     hwaddr base = vms->memmap[VIRT_SMMU].base;
1338     hwaddr size = vms->memmap[VIRT_SMMU].size;
1339     const char irq_names[] = "eventq\0priq\0cmdq-sync\0gerror";
1340     DeviceState *dev;
1341     MachineState *ms = MACHINE(vms);
1342 
1343     if (vms->iommu != VIRT_IOMMU_SMMUV3 || !vms->iommu_phandle) {
1344         return;
1345     }
1346 
1347     dev = qdev_new(TYPE_ARM_SMMUV3);
1348 
1349     object_property_set_link(OBJECT(dev), "primary-bus", OBJECT(bus),
1350                              &error_abort);
1351     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1352     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, base);
1353     for (i = 0; i < NUM_SMMU_IRQS; i++) {
1354         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
1355                            qdev_get_gpio_in(vms->gic, irq + i));
1356     }
1357 
1358     node = g_strdup_printf("/smmuv3@%" PRIx64, base);
1359     qemu_fdt_add_subnode(ms->fdt, node);
1360     qemu_fdt_setprop(ms->fdt, node, "compatible", compat, sizeof(compat));
1361     qemu_fdt_setprop_sized_cells(ms->fdt, node, "reg", 2, base, 2, size);
1362 
1363     qemu_fdt_setprop_cells(ms->fdt, node, "interrupts",
1364             GIC_FDT_IRQ_TYPE_SPI, irq    , GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1365             GIC_FDT_IRQ_TYPE_SPI, irq + 1, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1366             GIC_FDT_IRQ_TYPE_SPI, irq + 2, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1367             GIC_FDT_IRQ_TYPE_SPI, irq + 3, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI);
1368 
1369     qemu_fdt_setprop(ms->fdt, node, "interrupt-names", irq_names,
1370                      sizeof(irq_names));
1371 
1372     qemu_fdt_setprop(ms->fdt, node, "dma-coherent", NULL, 0);
1373 
1374     qemu_fdt_setprop_cell(ms->fdt, node, "#iommu-cells", 1);
1375 
1376     qemu_fdt_setprop_cell(ms->fdt, node, "phandle", vms->iommu_phandle);
1377     g_free(node);
1378 }
1379 
1380 static void create_virtio_iommu_dt_bindings(VirtMachineState *vms)
1381 {
1382     const char compat[] = "virtio,pci-iommu\0pci1af4,1057";
1383     uint16_t bdf = vms->virtio_iommu_bdf;
1384     MachineState *ms = MACHINE(vms);
1385     char *node;
1386 
1387     vms->iommu_phandle = qemu_fdt_alloc_phandle(ms->fdt);
1388 
1389     node = g_strdup_printf("%s/virtio_iommu@%x,%x", vms->pciehb_nodename,
1390                            PCI_SLOT(bdf), PCI_FUNC(bdf));
1391     qemu_fdt_add_subnode(ms->fdt, node);
1392     qemu_fdt_setprop(ms->fdt, node, "compatible", compat, sizeof(compat));
1393     qemu_fdt_setprop_sized_cells(ms->fdt, node, "reg",
1394                                  1, bdf << 8, 1, 0, 1, 0,
1395                                  1, 0, 1, 0);
1396 
1397     qemu_fdt_setprop_cell(ms->fdt, node, "#iommu-cells", 1);
1398     qemu_fdt_setprop_cell(ms->fdt, node, "phandle", vms->iommu_phandle);
1399     g_free(node);
1400 
1401     qemu_fdt_setprop_cells(ms->fdt, vms->pciehb_nodename, "iommu-map",
1402                            0x0, vms->iommu_phandle, 0x0, bdf,
1403                            bdf + 1, vms->iommu_phandle, bdf + 1, 0xffff - bdf);
1404 }
1405 
1406 static void create_pcie(VirtMachineState *vms)
1407 {
1408     hwaddr base_mmio = vms->memmap[VIRT_PCIE_MMIO].base;
1409     hwaddr size_mmio = vms->memmap[VIRT_PCIE_MMIO].size;
1410     hwaddr base_mmio_high = vms->memmap[VIRT_HIGH_PCIE_MMIO].base;
1411     hwaddr size_mmio_high = vms->memmap[VIRT_HIGH_PCIE_MMIO].size;
1412     hwaddr base_pio = vms->memmap[VIRT_PCIE_PIO].base;
1413     hwaddr size_pio = vms->memmap[VIRT_PCIE_PIO].size;
1414     hwaddr base_ecam, size_ecam;
1415     hwaddr base = base_mmio;
1416     int nr_pcie_buses;
1417     int irq = vms->irqmap[VIRT_PCIE];
1418     MemoryRegion *mmio_alias;
1419     MemoryRegion *mmio_reg;
1420     MemoryRegion *ecam_alias;
1421     MemoryRegion *ecam_reg;
1422     DeviceState *dev;
1423     char *nodename;
1424     int i, ecam_id;
1425     PCIHostState *pci;
1426     MachineState *ms = MACHINE(vms);
1427 
1428     dev = qdev_new(TYPE_GPEX_HOST);
1429     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1430 
1431     ecam_id = VIRT_ECAM_ID(vms->highmem_ecam);
1432     base_ecam = vms->memmap[ecam_id].base;
1433     size_ecam = vms->memmap[ecam_id].size;
1434     nr_pcie_buses = size_ecam / PCIE_MMCFG_SIZE_MIN;
1435     /* Map only the first size_ecam bytes of ECAM space */
1436     ecam_alias = g_new0(MemoryRegion, 1);
1437     ecam_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 0);
1438     memory_region_init_alias(ecam_alias, OBJECT(dev), "pcie-ecam",
1439                              ecam_reg, 0, size_ecam);
1440     memory_region_add_subregion(get_system_memory(), base_ecam, ecam_alias);
1441 
1442     /* Map the MMIO window into system address space so as to expose
1443      * the section of PCI MMIO space which starts at the same base address
1444      * (ie 1:1 mapping for that part of PCI MMIO space visible through
1445      * the window).
1446      */
1447     mmio_alias = g_new0(MemoryRegion, 1);
1448     mmio_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 1);
1449     memory_region_init_alias(mmio_alias, OBJECT(dev), "pcie-mmio",
1450                              mmio_reg, base_mmio, size_mmio);
1451     memory_region_add_subregion(get_system_memory(), base_mmio, mmio_alias);
1452 
1453     if (vms->highmem_mmio) {
1454         /* Map high MMIO space */
1455         MemoryRegion *high_mmio_alias = g_new0(MemoryRegion, 1);
1456 
1457         memory_region_init_alias(high_mmio_alias, OBJECT(dev), "pcie-mmio-high",
1458                                  mmio_reg, base_mmio_high, size_mmio_high);
1459         memory_region_add_subregion(get_system_memory(), base_mmio_high,
1460                                     high_mmio_alias);
1461     }
1462 
1463     /* Map IO port space */
1464     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 2, base_pio);
1465 
1466     for (i = 0; i < GPEX_NUM_IRQS; i++) {
1467         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
1468                            qdev_get_gpio_in(vms->gic, irq + i));
1469         gpex_set_irq_num(GPEX_HOST(dev), i, irq + i);
1470     }
1471 
1472     pci = PCI_HOST_BRIDGE(dev);
1473     pci->bypass_iommu = vms->default_bus_bypass_iommu;
1474     vms->bus = pci->bus;
1475     if (vms->bus) {
1476         for (i = 0; i < nb_nics; i++) {
1477             NICInfo *nd = &nd_table[i];
1478 
1479             if (!nd->model) {
1480                 nd->model = g_strdup("virtio");
1481             }
1482 
1483             pci_nic_init_nofail(nd, pci->bus, nd->model, NULL);
1484         }
1485     }
1486 
1487     nodename = vms->pciehb_nodename = g_strdup_printf("/pcie@%" PRIx64, base);
1488     qemu_fdt_add_subnode(ms->fdt, nodename);
1489     qemu_fdt_setprop_string(ms->fdt, nodename,
1490                             "compatible", "pci-host-ecam-generic");
1491     qemu_fdt_setprop_string(ms->fdt, nodename, "device_type", "pci");
1492     qemu_fdt_setprop_cell(ms->fdt, nodename, "#address-cells", 3);
1493     qemu_fdt_setprop_cell(ms->fdt, nodename, "#size-cells", 2);
1494     qemu_fdt_setprop_cell(ms->fdt, nodename, "linux,pci-domain", 0);
1495     qemu_fdt_setprop_cells(ms->fdt, nodename, "bus-range", 0,
1496                            nr_pcie_buses - 1);
1497     qemu_fdt_setprop(ms->fdt, nodename, "dma-coherent", NULL, 0);
1498 
1499     if (vms->msi_phandle) {
1500         qemu_fdt_setprop_cells(ms->fdt, nodename, "msi-map",
1501                                0, vms->msi_phandle, 0, 0x10000);
1502     }
1503 
1504     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1505                                  2, base_ecam, 2, size_ecam);
1506 
1507     if (vms->highmem_mmio) {
1508         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "ranges",
1509                                      1, FDT_PCI_RANGE_IOPORT, 2, 0,
1510                                      2, base_pio, 2, size_pio,
1511                                      1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
1512                                      2, base_mmio, 2, size_mmio,
1513                                      1, FDT_PCI_RANGE_MMIO_64BIT,
1514                                      2, base_mmio_high,
1515                                      2, base_mmio_high, 2, size_mmio_high);
1516     } else {
1517         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "ranges",
1518                                      1, FDT_PCI_RANGE_IOPORT, 2, 0,
1519                                      2, base_pio, 2, size_pio,
1520                                      1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
1521                                      2, base_mmio, 2, size_mmio);
1522     }
1523 
1524     qemu_fdt_setprop_cell(ms->fdt, nodename, "#interrupt-cells", 1);
1525     create_pcie_irq_map(ms, vms->gic_phandle, irq, nodename);
1526 
1527     if (vms->iommu) {
1528         vms->iommu_phandle = qemu_fdt_alloc_phandle(ms->fdt);
1529 
1530         switch (vms->iommu) {
1531         case VIRT_IOMMU_SMMUV3:
1532             create_smmu(vms, vms->bus);
1533             qemu_fdt_setprop_cells(ms->fdt, nodename, "iommu-map",
1534                                    0x0, vms->iommu_phandle, 0x0, 0x10000);
1535             break;
1536         default:
1537             g_assert_not_reached();
1538         }
1539     }
1540 }
1541 
1542 static void create_platform_bus(VirtMachineState *vms)
1543 {
1544     DeviceState *dev;
1545     SysBusDevice *s;
1546     int i;
1547     MemoryRegion *sysmem = get_system_memory();
1548 
1549     dev = qdev_new(TYPE_PLATFORM_BUS_DEVICE);
1550     dev->id = g_strdup(TYPE_PLATFORM_BUS_DEVICE);
1551     qdev_prop_set_uint32(dev, "num_irqs", PLATFORM_BUS_NUM_IRQS);
1552     qdev_prop_set_uint32(dev, "mmio_size", vms->memmap[VIRT_PLATFORM_BUS].size);
1553     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1554     vms->platform_bus_dev = dev;
1555 
1556     s = SYS_BUS_DEVICE(dev);
1557     for (i = 0; i < PLATFORM_BUS_NUM_IRQS; i++) {
1558         int irq = vms->irqmap[VIRT_PLATFORM_BUS] + i;
1559         sysbus_connect_irq(s, i, qdev_get_gpio_in(vms->gic, irq));
1560     }
1561 
1562     memory_region_add_subregion(sysmem,
1563                                 vms->memmap[VIRT_PLATFORM_BUS].base,
1564                                 sysbus_mmio_get_region(s, 0));
1565 }
1566 
1567 static void create_tag_ram(MemoryRegion *tag_sysmem,
1568                            hwaddr base, hwaddr size,
1569                            const char *name)
1570 {
1571     MemoryRegion *tagram = g_new(MemoryRegion, 1);
1572 
1573     memory_region_init_ram(tagram, NULL, name, size / 32, &error_fatal);
1574     memory_region_add_subregion(tag_sysmem, base / 32, tagram);
1575 }
1576 
1577 static void create_secure_ram(VirtMachineState *vms,
1578                               MemoryRegion *secure_sysmem,
1579                               MemoryRegion *secure_tag_sysmem)
1580 {
1581     MemoryRegion *secram = g_new(MemoryRegion, 1);
1582     char *nodename;
1583     hwaddr base = vms->memmap[VIRT_SECURE_MEM].base;
1584     hwaddr size = vms->memmap[VIRT_SECURE_MEM].size;
1585     MachineState *ms = MACHINE(vms);
1586 
1587     memory_region_init_ram(secram, NULL, "virt.secure-ram", size,
1588                            &error_fatal);
1589     memory_region_add_subregion(secure_sysmem, base, secram);
1590 
1591     nodename = g_strdup_printf("/secram@%" PRIx64, base);
1592     qemu_fdt_add_subnode(ms->fdt, nodename);
1593     qemu_fdt_setprop_string(ms->fdt, nodename, "device_type", "memory");
1594     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg", 2, base, 2, size);
1595     qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
1596     qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
1597 
1598     if (secure_tag_sysmem) {
1599         create_tag_ram(secure_tag_sysmem, base, size, "mach-virt.secure-tag");
1600     }
1601 
1602     g_free(nodename);
1603 }
1604 
1605 static void *machvirt_dtb(const struct arm_boot_info *binfo, int *fdt_size)
1606 {
1607     const VirtMachineState *board = container_of(binfo, VirtMachineState,
1608                                                  bootinfo);
1609     MachineState *ms = MACHINE(board);
1610 
1611 
1612     *fdt_size = board->fdt_size;
1613     return ms->fdt;
1614 }
1615 
1616 static void virt_build_smbios(VirtMachineState *vms)
1617 {
1618     MachineClass *mc = MACHINE_GET_CLASS(vms);
1619     MachineState *ms = MACHINE(vms);
1620     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
1621     uint8_t *smbios_tables, *smbios_anchor;
1622     size_t smbios_tables_len, smbios_anchor_len;
1623     struct smbios_phys_mem_area mem_array;
1624     const char *product = "QEMU Virtual Machine";
1625 
1626     if (kvm_enabled()) {
1627         product = "KVM Virtual Machine";
1628     }
1629 
1630     smbios_set_defaults("QEMU", product,
1631                         vmc->smbios_old_sys_ver ? "1.0" : mc->name, false,
1632                         true, SMBIOS_ENTRY_POINT_TYPE_64);
1633 
1634     /* build the array of physical mem area from base_memmap */
1635     mem_array.address = vms->memmap[VIRT_MEM].base;
1636     mem_array.length = ms->ram_size;
1637 
1638     smbios_get_tables(ms, &mem_array, 1,
1639                       &smbios_tables, &smbios_tables_len,
1640                       &smbios_anchor, &smbios_anchor_len,
1641                       &error_fatal);
1642 
1643     if (smbios_anchor) {
1644         fw_cfg_add_file(vms->fw_cfg, "etc/smbios/smbios-tables",
1645                         smbios_tables, smbios_tables_len);
1646         fw_cfg_add_file(vms->fw_cfg, "etc/smbios/smbios-anchor",
1647                         smbios_anchor, smbios_anchor_len);
1648     }
1649 }
1650 
1651 static
1652 void virt_machine_done(Notifier *notifier, void *data)
1653 {
1654     VirtMachineState *vms = container_of(notifier, VirtMachineState,
1655                                          machine_done);
1656     MachineState *ms = MACHINE(vms);
1657     ARMCPU *cpu = ARM_CPU(first_cpu);
1658     struct arm_boot_info *info = &vms->bootinfo;
1659     AddressSpace *as = arm_boot_address_space(cpu, info);
1660 
1661     /*
1662      * If the user provided a dtb, we assume the dynamic sysbus nodes
1663      * already are integrated there. This corresponds to a use case where
1664      * the dynamic sysbus nodes are complex and their generation is not yet
1665      * supported. In that case the user can take charge of the guest dt
1666      * while qemu takes charge of the qom stuff.
1667      */
1668     if (info->dtb_filename == NULL) {
1669         platform_bus_add_all_fdt_nodes(ms->fdt, "/intc",
1670                                        vms->memmap[VIRT_PLATFORM_BUS].base,
1671                                        vms->memmap[VIRT_PLATFORM_BUS].size,
1672                                        vms->irqmap[VIRT_PLATFORM_BUS]);
1673     }
1674     if (arm_load_dtb(info->dtb_start, info, info->dtb_limit, as, ms) < 0) {
1675         exit(1);
1676     }
1677 
1678     fw_cfg_add_extra_pci_roots(vms->bus, vms->fw_cfg);
1679 
1680     virt_acpi_setup(vms);
1681     virt_build_smbios(vms);
1682 }
1683 
1684 static uint64_t virt_cpu_mp_affinity(VirtMachineState *vms, int idx)
1685 {
1686     uint8_t clustersz = ARM_DEFAULT_CPUS_PER_CLUSTER;
1687     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
1688 
1689     if (!vmc->disallow_affinity_adjustment) {
1690         /* Adjust MPIDR like 64-bit KVM hosts, which incorporate the
1691          * GIC's target-list limitations. 32-bit KVM hosts currently
1692          * always create clusters of 4 CPUs, but that is expected to
1693          * change when they gain support for gicv3. When KVM is enabled
1694          * it will override the changes we make here, therefore our
1695          * purposes are to make TCG consistent (with 64-bit KVM hosts)
1696          * and to improve SGI efficiency.
1697          */
1698         if (vms->gic_version == VIRT_GIC_VERSION_2) {
1699             clustersz = GIC_TARGETLIST_BITS;
1700         } else {
1701             clustersz = GICV3_TARGETLIST_BITS;
1702         }
1703     }
1704     return arm_cpu_mp_affinity(idx, clustersz);
1705 }
1706 
1707 static inline bool *virt_get_high_memmap_enabled(VirtMachineState *vms,
1708                                                  int index)
1709 {
1710     bool *enabled_array[] = {
1711         &vms->highmem_redists,
1712         &vms->highmem_ecam,
1713         &vms->highmem_mmio,
1714     };
1715 
1716     assert(ARRAY_SIZE(extended_memmap) - VIRT_LOWMEMMAP_LAST ==
1717            ARRAY_SIZE(enabled_array));
1718     assert(index - VIRT_LOWMEMMAP_LAST < ARRAY_SIZE(enabled_array));
1719 
1720     return enabled_array[index - VIRT_LOWMEMMAP_LAST];
1721 }
1722 
1723 static void virt_set_high_memmap(VirtMachineState *vms,
1724                                  hwaddr base, int pa_bits)
1725 {
1726     hwaddr region_base, region_size;
1727     bool *region_enabled, fits;
1728     int i;
1729 
1730     for (i = VIRT_LOWMEMMAP_LAST; i < ARRAY_SIZE(extended_memmap); i++) {
1731         region_enabled = virt_get_high_memmap_enabled(vms, i);
1732         region_base = ROUND_UP(base, extended_memmap[i].size);
1733         region_size = extended_memmap[i].size;
1734 
1735         vms->memmap[i].base = region_base;
1736         vms->memmap[i].size = region_size;
1737 
1738         /*
1739          * Check each device to see if it fits in the PA space,
1740          * moving highest_gpa as we go. For compatibility, move
1741          * highest_gpa for disabled fitting devices as well, if
1742          * the compact layout has been disabled.
1743          *
1744          * For each device that doesn't fit, disable it.
1745          */
1746         fits = (region_base + region_size) <= BIT_ULL(pa_bits);
1747         *region_enabled &= fits;
1748         if (vms->highmem_compact && !*region_enabled) {
1749             continue;
1750         }
1751 
1752         base = region_base + region_size;
1753         if (fits) {
1754             vms->highest_gpa = base - 1;
1755         }
1756     }
1757 }
1758 
1759 static void virt_set_memmap(VirtMachineState *vms, int pa_bits)
1760 {
1761     MachineState *ms = MACHINE(vms);
1762     hwaddr base, device_memory_base, device_memory_size, memtop;
1763     int i;
1764 
1765     vms->memmap = extended_memmap;
1766 
1767     for (i = 0; i < ARRAY_SIZE(base_memmap); i++) {
1768         vms->memmap[i] = base_memmap[i];
1769     }
1770 
1771     if (ms->ram_slots > ACPI_MAX_RAM_SLOTS) {
1772         error_report("unsupported number of memory slots: %"PRIu64,
1773                      ms->ram_slots);
1774         exit(EXIT_FAILURE);
1775     }
1776 
1777     /*
1778      * !highmem is exactly the same as limiting the PA space to 32bit,
1779      * irrespective of the underlying capabilities of the HW.
1780      */
1781     if (!vms->highmem) {
1782         pa_bits = 32;
1783     }
1784 
1785     /*
1786      * We compute the base of the high IO region depending on the
1787      * amount of initial and device memory. The device memory start/size
1788      * is aligned on 1GiB. We never put the high IO region below 256GiB
1789      * so that if maxram_size is < 255GiB we keep the legacy memory map.
1790      * The device region size assumes 1GiB page max alignment per slot.
1791      */
1792     device_memory_base =
1793         ROUND_UP(vms->memmap[VIRT_MEM].base + ms->ram_size, GiB);
1794     device_memory_size = ms->maxram_size - ms->ram_size + ms->ram_slots * GiB;
1795 
1796     /* Base address of the high IO region */
1797     memtop = base = device_memory_base + ROUND_UP(device_memory_size, GiB);
1798     if (memtop > BIT_ULL(pa_bits)) {
1799 	    error_report("Addressing limited to %d bits, but memory exceeds it by %llu bytes\n",
1800 			 pa_bits, memtop - BIT_ULL(pa_bits));
1801         exit(EXIT_FAILURE);
1802     }
1803     if (base < device_memory_base) {
1804         error_report("maxmem/slots too huge");
1805         exit(EXIT_FAILURE);
1806     }
1807     if (base < vms->memmap[VIRT_MEM].base + LEGACY_RAMLIMIT_BYTES) {
1808         base = vms->memmap[VIRT_MEM].base + LEGACY_RAMLIMIT_BYTES;
1809     }
1810 
1811     /* We know for sure that at least the memory fits in the PA space */
1812     vms->highest_gpa = memtop - 1;
1813 
1814     virt_set_high_memmap(vms, base, pa_bits);
1815 
1816     if (device_memory_size > 0) {
1817         ms->device_memory = g_malloc0(sizeof(*ms->device_memory));
1818         ms->device_memory->base = device_memory_base;
1819         memory_region_init(&ms->device_memory->mr, OBJECT(vms),
1820                            "device-memory", device_memory_size);
1821     }
1822 }
1823 
1824 static VirtGICType finalize_gic_version_do(const char *accel_name,
1825                                            VirtGICType gic_version,
1826                                            int gics_supported,
1827                                            unsigned int max_cpus)
1828 {
1829     /* Convert host/max/nosel to GIC version number */
1830     switch (gic_version) {
1831     case VIRT_GIC_VERSION_HOST:
1832         if (!kvm_enabled()) {
1833             error_report("gic-version=host requires KVM");
1834             exit(1);
1835         }
1836 
1837         /* For KVM, gic-version=host means gic-version=max */
1838         return finalize_gic_version_do(accel_name, VIRT_GIC_VERSION_MAX,
1839                                        gics_supported, max_cpus);
1840     case VIRT_GIC_VERSION_MAX:
1841         if (gics_supported & VIRT_GIC_VERSION_4_MASK) {
1842             gic_version = VIRT_GIC_VERSION_4;
1843         } else if (gics_supported & VIRT_GIC_VERSION_3_MASK) {
1844             gic_version = VIRT_GIC_VERSION_3;
1845         } else {
1846             gic_version = VIRT_GIC_VERSION_2;
1847         }
1848         break;
1849     case VIRT_GIC_VERSION_NOSEL:
1850         if ((gics_supported & VIRT_GIC_VERSION_2_MASK) &&
1851             max_cpus <= GIC_NCPU) {
1852             gic_version = VIRT_GIC_VERSION_2;
1853         } else if (gics_supported & VIRT_GIC_VERSION_3_MASK) {
1854             /*
1855              * in case the host does not support v2 emulation or
1856              * the end-user requested more than 8 VCPUs we now default
1857              * to v3. In any case defaulting to v2 would be broken.
1858              */
1859             gic_version = VIRT_GIC_VERSION_3;
1860         } else if (max_cpus > GIC_NCPU) {
1861             error_report("%s only supports GICv2 emulation but more than 8 "
1862                          "vcpus are requested", accel_name);
1863             exit(1);
1864         }
1865         break;
1866     case VIRT_GIC_VERSION_2:
1867     case VIRT_GIC_VERSION_3:
1868     case VIRT_GIC_VERSION_4:
1869         break;
1870     }
1871 
1872     /* Check chosen version is effectively supported */
1873     switch (gic_version) {
1874     case VIRT_GIC_VERSION_2:
1875         if (!(gics_supported & VIRT_GIC_VERSION_2_MASK)) {
1876             error_report("%s does not support GICv2 emulation", accel_name);
1877             exit(1);
1878         }
1879         break;
1880     case VIRT_GIC_VERSION_3:
1881         if (!(gics_supported & VIRT_GIC_VERSION_3_MASK)) {
1882             error_report("%s does not support GICv3 emulation", accel_name);
1883             exit(1);
1884         }
1885         break;
1886     case VIRT_GIC_VERSION_4:
1887         if (!(gics_supported & VIRT_GIC_VERSION_4_MASK)) {
1888             error_report("%s does not support GICv4 emulation, is virtualization=on?",
1889                          accel_name);
1890             exit(1);
1891         }
1892         break;
1893     default:
1894         error_report("logic error in finalize_gic_version");
1895         exit(1);
1896         break;
1897     }
1898 
1899     return gic_version;
1900 }
1901 
1902 /*
1903  * finalize_gic_version - Determines the final gic_version
1904  * according to the gic-version property
1905  *
1906  * Default GIC type is v2
1907  */
1908 static void finalize_gic_version(VirtMachineState *vms)
1909 {
1910     const char *accel_name = current_accel_name();
1911     unsigned int max_cpus = MACHINE(vms)->smp.max_cpus;
1912     int gics_supported = 0;
1913 
1914     /* Determine which GIC versions the current environment supports */
1915     if (kvm_enabled() && kvm_irqchip_in_kernel()) {
1916         int probe_bitmap = kvm_arm_vgic_probe();
1917 
1918         if (!probe_bitmap) {
1919             error_report("Unable to determine GIC version supported by host");
1920             exit(1);
1921         }
1922 
1923         if (probe_bitmap & KVM_ARM_VGIC_V2) {
1924             gics_supported |= VIRT_GIC_VERSION_2_MASK;
1925         }
1926         if (probe_bitmap & KVM_ARM_VGIC_V3) {
1927             gics_supported |= VIRT_GIC_VERSION_3_MASK;
1928         }
1929     } else if (kvm_enabled() && !kvm_irqchip_in_kernel()) {
1930         /* KVM w/o kernel irqchip can only deal with GICv2 */
1931         gics_supported |= VIRT_GIC_VERSION_2_MASK;
1932         accel_name = "KVM with kernel-irqchip=off";
1933     } else if (tcg_enabled() || hvf_enabled() || qtest_enabled())  {
1934         gics_supported |= VIRT_GIC_VERSION_2_MASK;
1935         if (module_object_class_by_name("arm-gicv3")) {
1936             gics_supported |= VIRT_GIC_VERSION_3_MASK;
1937             if (vms->virt) {
1938                 /* GICv4 only makes sense if CPU has EL2 */
1939                 gics_supported |= VIRT_GIC_VERSION_4_MASK;
1940             }
1941         }
1942     } else {
1943         error_report("Unsupported accelerator, can not determine GIC support");
1944         exit(1);
1945     }
1946 
1947     /*
1948      * Then convert helpers like host/max to concrete GIC versions and ensure
1949      * the desired version is supported
1950      */
1951     vms->gic_version = finalize_gic_version_do(accel_name, vms->gic_version,
1952                                                gics_supported, max_cpus);
1953 }
1954 
1955 /*
1956  * virt_cpu_post_init() must be called after the CPUs have
1957  * been realized and the GIC has been created.
1958  */
1959 static void virt_cpu_post_init(VirtMachineState *vms, MemoryRegion *sysmem)
1960 {
1961     int max_cpus = MACHINE(vms)->smp.max_cpus;
1962     bool aarch64, pmu, steal_time;
1963     CPUState *cpu;
1964 
1965     aarch64 = object_property_get_bool(OBJECT(first_cpu), "aarch64", NULL);
1966     pmu = object_property_get_bool(OBJECT(first_cpu), "pmu", NULL);
1967     steal_time = object_property_get_bool(OBJECT(first_cpu),
1968                                           "kvm-steal-time", NULL);
1969 
1970     if (kvm_enabled()) {
1971         hwaddr pvtime_reg_base = vms->memmap[VIRT_PVTIME].base;
1972         hwaddr pvtime_reg_size = vms->memmap[VIRT_PVTIME].size;
1973 
1974         if (steal_time) {
1975             MemoryRegion *pvtime = g_new(MemoryRegion, 1);
1976             hwaddr pvtime_size = max_cpus * PVTIME_SIZE_PER_CPU;
1977 
1978             /* The memory region size must be a multiple of host page size. */
1979             pvtime_size = REAL_HOST_PAGE_ALIGN(pvtime_size);
1980 
1981             if (pvtime_size > pvtime_reg_size) {
1982                 error_report("pvtime requires a %" HWADDR_PRId
1983                              " byte memory region for %d CPUs,"
1984                              " but only %" HWADDR_PRId " has been reserved",
1985                              pvtime_size, max_cpus, pvtime_reg_size);
1986                 exit(1);
1987             }
1988 
1989             memory_region_init_ram(pvtime, NULL, "pvtime", pvtime_size, NULL);
1990             memory_region_add_subregion(sysmem, pvtime_reg_base, pvtime);
1991         }
1992 
1993         CPU_FOREACH(cpu) {
1994             if (pmu) {
1995                 assert(arm_feature(&ARM_CPU(cpu)->env, ARM_FEATURE_PMU));
1996                 if (kvm_irqchip_in_kernel()) {
1997                     kvm_arm_pmu_set_irq(cpu, PPI(VIRTUAL_PMU_IRQ));
1998                 }
1999                 kvm_arm_pmu_init(cpu);
2000             }
2001             if (steal_time) {
2002                 kvm_arm_pvtime_init(cpu, pvtime_reg_base +
2003                                          cpu->cpu_index * PVTIME_SIZE_PER_CPU);
2004             }
2005         }
2006     } else {
2007         if (aarch64 && vms->highmem) {
2008             int requested_pa_size = 64 - clz64(vms->highest_gpa);
2009             int pamax = arm_pamax(ARM_CPU(first_cpu));
2010 
2011             if (pamax < requested_pa_size) {
2012                 error_report("VCPU supports less PA bits (%d) than "
2013                              "requested by the memory map (%d)",
2014                              pamax, requested_pa_size);
2015                 exit(1);
2016             }
2017         }
2018     }
2019 }
2020 
2021 static void machvirt_init(MachineState *machine)
2022 {
2023     VirtMachineState *vms = VIRT_MACHINE(machine);
2024     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(machine);
2025     MachineClass *mc = MACHINE_GET_CLASS(machine);
2026     const CPUArchIdList *possible_cpus;
2027     MemoryRegion *sysmem = get_system_memory();
2028     MemoryRegion *secure_sysmem = NULL;
2029     MemoryRegion *tag_sysmem = NULL;
2030     MemoryRegion *secure_tag_sysmem = NULL;
2031     int n, virt_max_cpus;
2032     bool firmware_loaded;
2033     bool aarch64 = true;
2034     bool has_ged = !vmc->no_ged;
2035     unsigned int smp_cpus = machine->smp.cpus;
2036     unsigned int max_cpus = machine->smp.max_cpus;
2037 
2038     if (!cpu_type_valid(machine->cpu_type)) {
2039         error_report("mach-virt: CPU type %s not supported", machine->cpu_type);
2040         exit(1);
2041     }
2042 
2043     possible_cpus = mc->possible_cpu_arch_ids(machine);
2044 
2045     /*
2046      * In accelerated mode, the memory map is computed earlier in kvm_type()
2047      * to create a VM with the right number of IPA bits.
2048      */
2049     if (!vms->memmap) {
2050         Object *cpuobj;
2051         ARMCPU *armcpu;
2052         int pa_bits;
2053 
2054         /*
2055          * Instanciate a temporary CPU object to find out about what
2056          * we are about to deal with. Once this is done, get rid of
2057          * the object.
2058          */
2059         cpuobj = object_new(possible_cpus->cpus[0].type);
2060         armcpu = ARM_CPU(cpuobj);
2061 
2062         pa_bits = arm_pamax(armcpu);
2063 
2064         object_unref(cpuobj);
2065 
2066         virt_set_memmap(vms, pa_bits);
2067     }
2068 
2069     /* We can probe only here because during property set
2070      * KVM is not available yet
2071      */
2072     finalize_gic_version(vms);
2073 
2074     if (vms->secure) {
2075         /*
2076          * The Secure view of the world is the same as the NonSecure,
2077          * but with a few extra devices. Create it as a container region
2078          * containing the system memory at low priority; any secure-only
2079          * devices go in at higher priority and take precedence.
2080          */
2081         secure_sysmem = g_new(MemoryRegion, 1);
2082         memory_region_init(secure_sysmem, OBJECT(machine), "secure-memory",
2083                            UINT64_MAX);
2084         memory_region_add_subregion_overlap(secure_sysmem, 0, sysmem, -1);
2085     }
2086 
2087     firmware_loaded = virt_firmware_init(vms, sysmem,
2088                                          secure_sysmem ?: sysmem);
2089 
2090     /* If we have an EL3 boot ROM then the assumption is that it will
2091      * implement PSCI itself, so disable QEMU's internal implementation
2092      * so it doesn't get in the way. Instead of starting secondary
2093      * CPUs in PSCI powerdown state we will start them all running and
2094      * let the boot ROM sort them out.
2095      * The usual case is that we do use QEMU's PSCI implementation;
2096      * if the guest has EL2 then we will use SMC as the conduit,
2097      * and otherwise we will use HVC (for backwards compatibility and
2098      * because if we're using KVM then we must use HVC).
2099      */
2100     if (vms->secure && firmware_loaded) {
2101         vms->psci_conduit = QEMU_PSCI_CONDUIT_DISABLED;
2102     } else if (vms->virt) {
2103         vms->psci_conduit = QEMU_PSCI_CONDUIT_SMC;
2104     } else {
2105         vms->psci_conduit = QEMU_PSCI_CONDUIT_HVC;
2106     }
2107 
2108     /*
2109      * The maximum number of CPUs depends on the GIC version, or on how
2110      * many redistributors we can fit into the memory map (which in turn
2111      * depends on whether this is a GICv3 or v4).
2112      */
2113     if (vms->gic_version == VIRT_GIC_VERSION_2) {
2114         virt_max_cpus = GIC_NCPU;
2115     } else {
2116         virt_max_cpus = virt_redist_capacity(vms, VIRT_GIC_REDIST);
2117         if (vms->highmem_redists) {
2118             virt_max_cpus += virt_redist_capacity(vms, VIRT_HIGH_GIC_REDIST2);
2119         }
2120     }
2121 
2122     if (max_cpus > virt_max_cpus) {
2123         error_report("Number of SMP CPUs requested (%d) exceeds max CPUs "
2124                      "supported by machine 'mach-virt' (%d)",
2125                      max_cpus, virt_max_cpus);
2126         if (vms->gic_version != VIRT_GIC_VERSION_2 && !vms->highmem_redists) {
2127             error_printf("Try 'highmem-redists=on' for more CPUs\n");
2128         }
2129 
2130         exit(1);
2131     }
2132 
2133     if (vms->secure && (kvm_enabled() || hvf_enabled())) {
2134         error_report("mach-virt: %s does not support providing "
2135                      "Security extensions (TrustZone) to the guest CPU",
2136                      current_accel_name());
2137         exit(1);
2138     }
2139 
2140     if (vms->virt && (kvm_enabled() || hvf_enabled())) {
2141         error_report("mach-virt: %s does not support providing "
2142                      "Virtualization extensions to the guest CPU",
2143                      current_accel_name());
2144         exit(1);
2145     }
2146 
2147     if (vms->mte && (kvm_enabled() || hvf_enabled())) {
2148         error_report("mach-virt: %s does not support providing "
2149                      "MTE to the guest CPU",
2150                      current_accel_name());
2151         exit(1);
2152     }
2153 
2154     create_fdt(vms);
2155 
2156     assert(possible_cpus->len == max_cpus);
2157     for (n = 0; n < possible_cpus->len; n++) {
2158         Object *cpuobj;
2159         CPUState *cs;
2160 
2161         if (n >= smp_cpus) {
2162             break;
2163         }
2164 
2165         cpuobj = object_new(possible_cpus->cpus[n].type);
2166         object_property_set_int(cpuobj, "mp-affinity",
2167                                 possible_cpus->cpus[n].arch_id, NULL);
2168 
2169         cs = CPU(cpuobj);
2170         cs->cpu_index = n;
2171 
2172         numa_cpu_pre_plug(&possible_cpus->cpus[cs->cpu_index], DEVICE(cpuobj),
2173                           &error_fatal);
2174 
2175         aarch64 &= object_property_get_bool(cpuobj, "aarch64", NULL);
2176 
2177         if (!vms->secure) {
2178             object_property_set_bool(cpuobj, "has_el3", false, NULL);
2179         }
2180 
2181         if (!vms->virt && object_property_find(cpuobj, "has_el2")) {
2182             object_property_set_bool(cpuobj, "has_el2", false, NULL);
2183         }
2184 
2185         if (vmc->kvm_no_adjvtime &&
2186             object_property_find(cpuobj, "kvm-no-adjvtime")) {
2187             object_property_set_bool(cpuobj, "kvm-no-adjvtime", true, NULL);
2188         }
2189 
2190         if (vmc->no_kvm_steal_time &&
2191             object_property_find(cpuobj, "kvm-steal-time")) {
2192             object_property_set_bool(cpuobj, "kvm-steal-time", false, NULL);
2193         }
2194 
2195         if (vmc->no_pmu && object_property_find(cpuobj, "pmu")) {
2196             object_property_set_bool(cpuobj, "pmu", false, NULL);
2197         }
2198 
2199         if (vmc->no_tcg_lpa2 && object_property_find(cpuobj, "lpa2")) {
2200             object_property_set_bool(cpuobj, "lpa2", false, NULL);
2201         }
2202 
2203         if (object_property_find(cpuobj, "reset-cbar")) {
2204             object_property_set_int(cpuobj, "reset-cbar",
2205                                     vms->memmap[VIRT_CPUPERIPHS].base,
2206                                     &error_abort);
2207         }
2208 
2209         object_property_set_link(cpuobj, "memory", OBJECT(sysmem),
2210                                  &error_abort);
2211         if (vms->secure) {
2212             object_property_set_link(cpuobj, "secure-memory",
2213                                      OBJECT(secure_sysmem), &error_abort);
2214         }
2215 
2216         if (vms->mte) {
2217             /* Create the memory region only once, but link to all cpus. */
2218             if (!tag_sysmem) {
2219                 /*
2220                  * The property exists only if MemTag is supported.
2221                  * If it is, we must allocate the ram to back that up.
2222                  */
2223                 if (!object_property_find(cpuobj, "tag-memory")) {
2224                     error_report("MTE requested, but not supported "
2225                                  "by the guest CPU");
2226                     exit(1);
2227                 }
2228 
2229                 tag_sysmem = g_new(MemoryRegion, 1);
2230                 memory_region_init(tag_sysmem, OBJECT(machine),
2231                                    "tag-memory", UINT64_MAX / 32);
2232 
2233                 if (vms->secure) {
2234                     secure_tag_sysmem = g_new(MemoryRegion, 1);
2235                     memory_region_init(secure_tag_sysmem, OBJECT(machine),
2236                                        "secure-tag-memory", UINT64_MAX / 32);
2237 
2238                     /* As with ram, secure-tag takes precedence over tag.  */
2239                     memory_region_add_subregion_overlap(secure_tag_sysmem, 0,
2240                                                         tag_sysmem, -1);
2241                 }
2242             }
2243 
2244             object_property_set_link(cpuobj, "tag-memory", OBJECT(tag_sysmem),
2245                                      &error_abort);
2246             if (vms->secure) {
2247                 object_property_set_link(cpuobj, "secure-tag-memory",
2248                                          OBJECT(secure_tag_sysmem),
2249                                          &error_abort);
2250             }
2251         }
2252 
2253         qdev_realize(DEVICE(cpuobj), NULL, &error_fatal);
2254         object_unref(cpuobj);
2255     }
2256     fdt_add_timer_nodes(vms);
2257     fdt_add_cpu_nodes(vms);
2258 
2259     memory_region_add_subregion(sysmem, vms->memmap[VIRT_MEM].base,
2260                                 machine->ram);
2261     if (machine->device_memory) {
2262         memory_region_add_subregion(sysmem, machine->device_memory->base,
2263                                     &machine->device_memory->mr);
2264     }
2265 
2266     virt_flash_fdt(vms, sysmem, secure_sysmem ?: sysmem);
2267 
2268     create_gic(vms, sysmem);
2269 
2270     virt_cpu_post_init(vms, sysmem);
2271 
2272     fdt_add_pmu_nodes(vms);
2273 
2274     create_uart(vms, VIRT_UART, sysmem, serial_hd(0));
2275 
2276     if (vms->secure) {
2277         create_secure_ram(vms, secure_sysmem, secure_tag_sysmem);
2278         create_uart(vms, VIRT_SECURE_UART, secure_sysmem, serial_hd(1));
2279     }
2280 
2281     if (tag_sysmem) {
2282         create_tag_ram(tag_sysmem, vms->memmap[VIRT_MEM].base,
2283                        machine->ram_size, "mach-virt.tag");
2284     }
2285 
2286     vms->highmem_ecam &= (!firmware_loaded || aarch64);
2287 
2288     create_rtc(vms);
2289 
2290     create_pcie(vms);
2291 
2292     if (has_ged && aarch64 && firmware_loaded && virt_is_acpi_enabled(vms)) {
2293         vms->acpi_dev = create_acpi_ged(vms);
2294     } else {
2295         create_gpio_devices(vms, VIRT_GPIO, sysmem);
2296     }
2297 
2298     if (vms->secure && !vmc->no_secure_gpio) {
2299         create_gpio_devices(vms, VIRT_SECURE_GPIO, secure_sysmem);
2300     }
2301 
2302      /* connect powerdown request */
2303      vms->powerdown_notifier.notify = virt_powerdown_req;
2304      qemu_register_powerdown_notifier(&vms->powerdown_notifier);
2305 
2306     /* Create mmio transports, so the user can create virtio backends
2307      * (which will be automatically plugged in to the transports). If
2308      * no backend is created the transport will just sit harmlessly idle.
2309      */
2310     create_virtio_devices(vms);
2311 
2312     vms->fw_cfg = create_fw_cfg(vms, &address_space_memory);
2313     rom_set_fw(vms->fw_cfg);
2314 
2315     create_platform_bus(vms);
2316 
2317     if (machine->nvdimms_state->is_enabled) {
2318         const struct AcpiGenericAddress arm_virt_nvdimm_acpi_dsmio = {
2319             .space_id = AML_AS_SYSTEM_MEMORY,
2320             .address = vms->memmap[VIRT_NVDIMM_ACPI].base,
2321             .bit_width = NVDIMM_ACPI_IO_LEN << 3
2322         };
2323 
2324         nvdimm_init_acpi_state(machine->nvdimms_state, sysmem,
2325                                arm_virt_nvdimm_acpi_dsmio,
2326                                vms->fw_cfg, OBJECT(vms));
2327     }
2328 
2329     vms->bootinfo.ram_size = machine->ram_size;
2330     vms->bootinfo.board_id = -1;
2331     vms->bootinfo.loader_start = vms->memmap[VIRT_MEM].base;
2332     vms->bootinfo.get_dtb = machvirt_dtb;
2333     vms->bootinfo.skip_dtb_autoload = true;
2334     vms->bootinfo.firmware_loaded = firmware_loaded;
2335     vms->bootinfo.psci_conduit = vms->psci_conduit;
2336     arm_load_kernel(ARM_CPU(first_cpu), machine, &vms->bootinfo);
2337 
2338     vms->machine_done.notify = virt_machine_done;
2339     qemu_add_machine_init_done_notifier(&vms->machine_done);
2340 }
2341 
2342 static bool virt_get_secure(Object *obj, Error **errp)
2343 {
2344     VirtMachineState *vms = VIRT_MACHINE(obj);
2345 
2346     return vms->secure;
2347 }
2348 
2349 static void virt_set_secure(Object *obj, bool value, Error **errp)
2350 {
2351     VirtMachineState *vms = VIRT_MACHINE(obj);
2352 
2353     vms->secure = value;
2354 }
2355 
2356 static bool virt_get_virt(Object *obj, Error **errp)
2357 {
2358     VirtMachineState *vms = VIRT_MACHINE(obj);
2359 
2360     return vms->virt;
2361 }
2362 
2363 static void virt_set_virt(Object *obj, bool value, Error **errp)
2364 {
2365     VirtMachineState *vms = VIRT_MACHINE(obj);
2366 
2367     vms->virt = value;
2368 }
2369 
2370 static bool virt_get_highmem(Object *obj, Error **errp)
2371 {
2372     VirtMachineState *vms = VIRT_MACHINE(obj);
2373 
2374     return vms->highmem;
2375 }
2376 
2377 static void virt_set_highmem(Object *obj, bool value, Error **errp)
2378 {
2379     VirtMachineState *vms = VIRT_MACHINE(obj);
2380 
2381     vms->highmem = value;
2382 }
2383 
2384 static bool virt_get_compact_highmem(Object *obj, Error **errp)
2385 {
2386     VirtMachineState *vms = VIRT_MACHINE(obj);
2387 
2388     return vms->highmem_compact;
2389 }
2390 
2391 static void virt_set_compact_highmem(Object *obj, bool value, Error **errp)
2392 {
2393     VirtMachineState *vms = VIRT_MACHINE(obj);
2394 
2395     vms->highmem_compact = value;
2396 }
2397 
2398 static bool virt_get_highmem_redists(Object *obj, Error **errp)
2399 {
2400     VirtMachineState *vms = VIRT_MACHINE(obj);
2401 
2402     return vms->highmem_redists;
2403 }
2404 
2405 static void virt_set_highmem_redists(Object *obj, bool value, Error **errp)
2406 {
2407     VirtMachineState *vms = VIRT_MACHINE(obj);
2408 
2409     vms->highmem_redists = value;
2410 }
2411 
2412 static bool virt_get_highmem_ecam(Object *obj, Error **errp)
2413 {
2414     VirtMachineState *vms = VIRT_MACHINE(obj);
2415 
2416     return vms->highmem_ecam;
2417 }
2418 
2419 static void virt_set_highmem_ecam(Object *obj, bool value, Error **errp)
2420 {
2421     VirtMachineState *vms = VIRT_MACHINE(obj);
2422 
2423     vms->highmem_ecam = value;
2424 }
2425 
2426 static bool virt_get_highmem_mmio(Object *obj, Error **errp)
2427 {
2428     VirtMachineState *vms = VIRT_MACHINE(obj);
2429 
2430     return vms->highmem_mmio;
2431 }
2432 
2433 static void virt_set_highmem_mmio(Object *obj, bool value, Error **errp)
2434 {
2435     VirtMachineState *vms = VIRT_MACHINE(obj);
2436 
2437     vms->highmem_mmio = value;
2438 }
2439 
2440 
2441 static bool virt_get_its(Object *obj, Error **errp)
2442 {
2443     VirtMachineState *vms = VIRT_MACHINE(obj);
2444 
2445     return vms->its;
2446 }
2447 
2448 static void virt_set_its(Object *obj, bool value, Error **errp)
2449 {
2450     VirtMachineState *vms = VIRT_MACHINE(obj);
2451 
2452     vms->its = value;
2453 }
2454 
2455 static bool virt_get_dtb_randomness(Object *obj, Error **errp)
2456 {
2457     VirtMachineState *vms = VIRT_MACHINE(obj);
2458 
2459     return vms->dtb_randomness;
2460 }
2461 
2462 static void virt_set_dtb_randomness(Object *obj, bool value, Error **errp)
2463 {
2464     VirtMachineState *vms = VIRT_MACHINE(obj);
2465 
2466     vms->dtb_randomness = value;
2467 }
2468 
2469 static char *virt_get_oem_id(Object *obj, Error **errp)
2470 {
2471     VirtMachineState *vms = VIRT_MACHINE(obj);
2472 
2473     return g_strdup(vms->oem_id);
2474 }
2475 
2476 static void virt_set_oem_id(Object *obj, const char *value, Error **errp)
2477 {
2478     VirtMachineState *vms = VIRT_MACHINE(obj);
2479     size_t len = strlen(value);
2480 
2481     if (len > 6) {
2482         error_setg(errp,
2483                    "User specified oem-id value is bigger than 6 bytes in size");
2484         return;
2485     }
2486 
2487     strncpy(vms->oem_id, value, 6);
2488 }
2489 
2490 static char *virt_get_oem_table_id(Object *obj, Error **errp)
2491 {
2492     VirtMachineState *vms = VIRT_MACHINE(obj);
2493 
2494     return g_strdup(vms->oem_table_id);
2495 }
2496 
2497 static void virt_set_oem_table_id(Object *obj, const char *value,
2498                                   Error **errp)
2499 {
2500     VirtMachineState *vms = VIRT_MACHINE(obj);
2501     size_t len = strlen(value);
2502 
2503     if (len > 8) {
2504         error_setg(errp,
2505                    "User specified oem-table-id value is bigger than 8 bytes in size");
2506         return;
2507     }
2508     strncpy(vms->oem_table_id, value, 8);
2509 }
2510 
2511 
2512 bool virt_is_acpi_enabled(VirtMachineState *vms)
2513 {
2514     if (vms->acpi == ON_OFF_AUTO_OFF) {
2515         return false;
2516     }
2517     return true;
2518 }
2519 
2520 static void virt_get_acpi(Object *obj, Visitor *v, const char *name,
2521                           void *opaque, Error **errp)
2522 {
2523     VirtMachineState *vms = VIRT_MACHINE(obj);
2524     OnOffAuto acpi = vms->acpi;
2525 
2526     visit_type_OnOffAuto(v, name, &acpi, errp);
2527 }
2528 
2529 static void virt_set_acpi(Object *obj, Visitor *v, const char *name,
2530                           void *opaque, Error **errp)
2531 {
2532     VirtMachineState *vms = VIRT_MACHINE(obj);
2533 
2534     visit_type_OnOffAuto(v, name, &vms->acpi, errp);
2535 }
2536 
2537 static bool virt_get_ras(Object *obj, Error **errp)
2538 {
2539     VirtMachineState *vms = VIRT_MACHINE(obj);
2540 
2541     return vms->ras;
2542 }
2543 
2544 static void virt_set_ras(Object *obj, bool value, Error **errp)
2545 {
2546     VirtMachineState *vms = VIRT_MACHINE(obj);
2547 
2548     vms->ras = value;
2549 }
2550 
2551 static bool virt_get_mte(Object *obj, Error **errp)
2552 {
2553     VirtMachineState *vms = VIRT_MACHINE(obj);
2554 
2555     return vms->mte;
2556 }
2557 
2558 static void virt_set_mte(Object *obj, bool value, Error **errp)
2559 {
2560     VirtMachineState *vms = VIRT_MACHINE(obj);
2561 
2562     vms->mte = value;
2563 }
2564 
2565 static char *virt_get_gic_version(Object *obj, Error **errp)
2566 {
2567     VirtMachineState *vms = VIRT_MACHINE(obj);
2568     const char *val;
2569 
2570     switch (vms->gic_version) {
2571     case VIRT_GIC_VERSION_4:
2572         val = "4";
2573         break;
2574     case VIRT_GIC_VERSION_3:
2575         val = "3";
2576         break;
2577     default:
2578         val = "2";
2579         break;
2580     }
2581     return g_strdup(val);
2582 }
2583 
2584 static void virt_set_gic_version(Object *obj, const char *value, Error **errp)
2585 {
2586     VirtMachineState *vms = VIRT_MACHINE(obj);
2587 
2588     if (!strcmp(value, "4")) {
2589         vms->gic_version = VIRT_GIC_VERSION_4;
2590     } else if (!strcmp(value, "3")) {
2591         vms->gic_version = VIRT_GIC_VERSION_3;
2592     } else if (!strcmp(value, "2")) {
2593         vms->gic_version = VIRT_GIC_VERSION_2;
2594     } else if (!strcmp(value, "host")) {
2595         vms->gic_version = VIRT_GIC_VERSION_HOST; /* Will probe later */
2596     } else if (!strcmp(value, "max")) {
2597         vms->gic_version = VIRT_GIC_VERSION_MAX; /* Will probe later */
2598     } else {
2599         error_setg(errp, "Invalid gic-version value");
2600         error_append_hint(errp, "Valid values are 3, 2, host, max.\n");
2601     }
2602 }
2603 
2604 static char *virt_get_iommu(Object *obj, Error **errp)
2605 {
2606     VirtMachineState *vms = VIRT_MACHINE(obj);
2607 
2608     switch (vms->iommu) {
2609     case VIRT_IOMMU_NONE:
2610         return g_strdup("none");
2611     case VIRT_IOMMU_SMMUV3:
2612         return g_strdup("smmuv3");
2613     default:
2614         g_assert_not_reached();
2615     }
2616 }
2617 
2618 static void virt_set_iommu(Object *obj, const char *value, Error **errp)
2619 {
2620     VirtMachineState *vms = VIRT_MACHINE(obj);
2621 
2622     if (!strcmp(value, "smmuv3")) {
2623         vms->iommu = VIRT_IOMMU_SMMUV3;
2624     } else if (!strcmp(value, "none")) {
2625         vms->iommu = VIRT_IOMMU_NONE;
2626     } else {
2627         error_setg(errp, "Invalid iommu value");
2628         error_append_hint(errp, "Valid values are none, smmuv3.\n");
2629     }
2630 }
2631 
2632 static bool virt_get_default_bus_bypass_iommu(Object *obj, Error **errp)
2633 {
2634     VirtMachineState *vms = VIRT_MACHINE(obj);
2635 
2636     return vms->default_bus_bypass_iommu;
2637 }
2638 
2639 static void virt_set_default_bus_bypass_iommu(Object *obj, bool value,
2640                                               Error **errp)
2641 {
2642     VirtMachineState *vms = VIRT_MACHINE(obj);
2643 
2644     vms->default_bus_bypass_iommu = value;
2645 }
2646 
2647 static CpuInstanceProperties
2648 virt_cpu_index_to_props(MachineState *ms, unsigned cpu_index)
2649 {
2650     MachineClass *mc = MACHINE_GET_CLASS(ms);
2651     const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(ms);
2652 
2653     assert(cpu_index < possible_cpus->len);
2654     return possible_cpus->cpus[cpu_index].props;
2655 }
2656 
2657 static int64_t virt_get_default_cpu_node_id(const MachineState *ms, int idx)
2658 {
2659     int64_t socket_id = ms->possible_cpus->cpus[idx].props.socket_id;
2660 
2661     return socket_id % ms->numa_state->num_nodes;
2662 }
2663 
2664 static const CPUArchIdList *virt_possible_cpu_arch_ids(MachineState *ms)
2665 {
2666     int n;
2667     unsigned int max_cpus = ms->smp.max_cpus;
2668     VirtMachineState *vms = VIRT_MACHINE(ms);
2669     MachineClass *mc = MACHINE_GET_CLASS(vms);
2670 
2671     if (ms->possible_cpus) {
2672         assert(ms->possible_cpus->len == max_cpus);
2673         return ms->possible_cpus;
2674     }
2675 
2676     ms->possible_cpus = g_malloc0(sizeof(CPUArchIdList) +
2677                                   sizeof(CPUArchId) * max_cpus);
2678     ms->possible_cpus->len = max_cpus;
2679     for (n = 0; n < ms->possible_cpus->len; n++) {
2680         ms->possible_cpus->cpus[n].type = ms->cpu_type;
2681         ms->possible_cpus->cpus[n].arch_id =
2682             virt_cpu_mp_affinity(vms, n);
2683 
2684         assert(!mc->smp_props.dies_supported);
2685         ms->possible_cpus->cpus[n].props.has_socket_id = true;
2686         ms->possible_cpus->cpus[n].props.socket_id =
2687             n / (ms->smp.clusters * ms->smp.cores * ms->smp.threads);
2688         ms->possible_cpus->cpus[n].props.has_cluster_id = true;
2689         ms->possible_cpus->cpus[n].props.cluster_id =
2690             (n / (ms->smp.cores * ms->smp.threads)) % ms->smp.clusters;
2691         ms->possible_cpus->cpus[n].props.has_core_id = true;
2692         ms->possible_cpus->cpus[n].props.core_id =
2693             (n / ms->smp.threads) % ms->smp.cores;
2694         ms->possible_cpus->cpus[n].props.has_thread_id = true;
2695         ms->possible_cpus->cpus[n].props.thread_id =
2696             n % ms->smp.threads;
2697     }
2698     return ms->possible_cpus;
2699 }
2700 
2701 static void virt_memory_pre_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
2702                                  Error **errp)
2703 {
2704     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2705     const MachineState *ms = MACHINE(hotplug_dev);
2706     const bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2707 
2708     if (!vms->acpi_dev) {
2709         error_setg(errp,
2710                    "memory hotplug is not enabled: missing acpi-ged device");
2711         return;
2712     }
2713 
2714     if (vms->mte) {
2715         error_setg(errp, "memory hotplug is not enabled: MTE is enabled");
2716         return;
2717     }
2718 
2719     if (is_nvdimm && !ms->nvdimms_state->is_enabled) {
2720         error_setg(errp, "nvdimm is not enabled: add 'nvdimm=on' to '-M'");
2721         return;
2722     }
2723 
2724     pc_dimm_pre_plug(PC_DIMM(dev), MACHINE(hotplug_dev), NULL, errp);
2725 }
2726 
2727 static void virt_memory_plug(HotplugHandler *hotplug_dev,
2728                              DeviceState *dev, Error **errp)
2729 {
2730     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2731     MachineState *ms = MACHINE(hotplug_dev);
2732     bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2733 
2734     pc_dimm_plug(PC_DIMM(dev), MACHINE(vms));
2735 
2736     if (is_nvdimm) {
2737         nvdimm_plug(ms->nvdimms_state);
2738     }
2739 
2740     hotplug_handler_plug(HOTPLUG_HANDLER(vms->acpi_dev),
2741                          dev, &error_abort);
2742 }
2743 
2744 static void virt_virtio_md_pci_pre_plug(HotplugHandler *hotplug_dev,
2745                                         DeviceState *dev, Error **errp)
2746 {
2747     HotplugHandler *hotplug_dev2 = qdev_get_bus_hotplug_handler(dev);
2748     Error *local_err = NULL;
2749 
2750     if (!hotplug_dev2 && dev->hotplugged) {
2751         /*
2752          * Without a bus hotplug handler, we cannot control the plug/unplug
2753          * order. We should never reach this point when hotplugging on ARM.
2754          * However, it's nice to add a safety net, similar to what we have
2755          * on x86.
2756          */
2757         error_setg(errp, "hotplug of virtio based memory devices not supported"
2758                    " on this bus.");
2759         return;
2760     }
2761     /*
2762      * First, see if we can plug this memory device at all. If that
2763      * succeeds, branch of to the actual hotplug handler.
2764      */
2765     memory_device_pre_plug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev), NULL,
2766                            &local_err);
2767     if (!local_err && hotplug_dev2) {
2768         hotplug_handler_pre_plug(hotplug_dev2, dev, &local_err);
2769     }
2770     error_propagate(errp, local_err);
2771 }
2772 
2773 static void virt_virtio_md_pci_plug(HotplugHandler *hotplug_dev,
2774                                     DeviceState *dev, Error **errp)
2775 {
2776     HotplugHandler *hotplug_dev2 = qdev_get_bus_hotplug_handler(dev);
2777     Error *local_err = NULL;
2778 
2779     /*
2780      * Plug the memory device first and then branch off to the actual
2781      * hotplug handler. If that one fails, we can easily undo the memory
2782      * device bits.
2783      */
2784     memory_device_plug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev));
2785     if (hotplug_dev2) {
2786         hotplug_handler_plug(hotplug_dev2, dev, &local_err);
2787         if (local_err) {
2788             memory_device_unplug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev));
2789         }
2790     }
2791     error_propagate(errp, local_err);
2792 }
2793 
2794 static void virt_virtio_md_pci_unplug_request(HotplugHandler *hotplug_dev,
2795                                               DeviceState *dev, Error **errp)
2796 {
2797     /* We don't support hot unplug of virtio based memory devices */
2798     error_setg(errp, "virtio based memory devices cannot be unplugged.");
2799 }
2800 
2801 
2802 static void virt_machine_device_pre_plug_cb(HotplugHandler *hotplug_dev,
2803                                             DeviceState *dev, Error **errp)
2804 {
2805     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2806 
2807     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2808         virt_memory_pre_plug(hotplug_dev, dev, errp);
2809     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI)) {
2810         virt_virtio_md_pci_pre_plug(hotplug_dev, dev, errp);
2811     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2812         hwaddr db_start = 0, db_end = 0;
2813         char *resv_prop_str;
2814 
2815         if (vms->iommu != VIRT_IOMMU_NONE) {
2816             error_setg(errp, "virt machine does not support multiple IOMMUs");
2817             return;
2818         }
2819 
2820         switch (vms->msi_controller) {
2821         case VIRT_MSI_CTRL_NONE:
2822             return;
2823         case VIRT_MSI_CTRL_ITS:
2824             /* GITS_TRANSLATER page */
2825             db_start = base_memmap[VIRT_GIC_ITS].base + 0x10000;
2826             db_end = base_memmap[VIRT_GIC_ITS].base +
2827                      base_memmap[VIRT_GIC_ITS].size - 1;
2828             break;
2829         case VIRT_MSI_CTRL_GICV2M:
2830             /* MSI_SETSPI_NS page */
2831             db_start = base_memmap[VIRT_GIC_V2M].base;
2832             db_end = db_start + base_memmap[VIRT_GIC_V2M].size - 1;
2833             break;
2834         }
2835         resv_prop_str = g_strdup_printf("0x%"PRIx64":0x%"PRIx64":%u",
2836                                         db_start, db_end,
2837                                         VIRTIO_IOMMU_RESV_MEM_T_MSI);
2838 
2839         object_property_set_uint(OBJECT(dev), "len-reserved-regions", 1, errp);
2840         object_property_set_str(OBJECT(dev), "reserved-regions[0]",
2841                                 resv_prop_str, errp);
2842         g_free(resv_prop_str);
2843     }
2844 }
2845 
2846 static void virt_machine_device_plug_cb(HotplugHandler *hotplug_dev,
2847                                         DeviceState *dev, Error **errp)
2848 {
2849     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2850 
2851     if (vms->platform_bus_dev) {
2852         MachineClass *mc = MACHINE_GET_CLASS(vms);
2853 
2854         if (device_is_dynamic_sysbus(mc, dev)) {
2855             platform_bus_link_device(PLATFORM_BUS_DEVICE(vms->platform_bus_dev),
2856                                      SYS_BUS_DEVICE(dev));
2857         }
2858     }
2859     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2860         virt_memory_plug(hotplug_dev, dev, errp);
2861     }
2862 
2863     if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI)) {
2864         virt_virtio_md_pci_plug(hotplug_dev, dev, errp);
2865     }
2866 
2867     if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2868         PCIDevice *pdev = PCI_DEVICE(dev);
2869 
2870         vms->iommu = VIRT_IOMMU_VIRTIO;
2871         vms->virtio_iommu_bdf = pci_get_bdf(pdev);
2872         create_virtio_iommu_dt_bindings(vms);
2873     }
2874 }
2875 
2876 static void virt_dimm_unplug_request(HotplugHandler *hotplug_dev,
2877                                      DeviceState *dev, Error **errp)
2878 {
2879     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2880 
2881     if (!vms->acpi_dev) {
2882         error_setg(errp,
2883                    "memory hotplug is not enabled: missing acpi-ged device");
2884         return;
2885     }
2886 
2887     if (object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM)) {
2888         error_setg(errp, "nvdimm device hot unplug is not supported yet.");
2889         return;
2890     }
2891 
2892     hotplug_handler_unplug_request(HOTPLUG_HANDLER(vms->acpi_dev), dev,
2893                                    errp);
2894 }
2895 
2896 static void virt_dimm_unplug(HotplugHandler *hotplug_dev,
2897                              DeviceState *dev, Error **errp)
2898 {
2899     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2900     Error *local_err = NULL;
2901 
2902     hotplug_handler_unplug(HOTPLUG_HANDLER(vms->acpi_dev), dev, &local_err);
2903     if (local_err) {
2904         goto out;
2905     }
2906 
2907     pc_dimm_unplug(PC_DIMM(dev), MACHINE(vms));
2908     qdev_unrealize(dev);
2909 
2910 out:
2911     error_propagate(errp, local_err);
2912 }
2913 
2914 static void virt_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
2915                                           DeviceState *dev, Error **errp)
2916 {
2917     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2918         virt_dimm_unplug_request(hotplug_dev, dev, errp);
2919     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI)) {
2920         virt_virtio_md_pci_unplug_request(hotplug_dev, dev, errp);
2921     } else {
2922         error_setg(errp, "device unplug request for unsupported device"
2923                    " type: %s", object_get_typename(OBJECT(dev)));
2924     }
2925 }
2926 
2927 static void virt_machine_device_unplug_cb(HotplugHandler *hotplug_dev,
2928                                           DeviceState *dev, Error **errp)
2929 {
2930     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2931         virt_dimm_unplug(hotplug_dev, dev, errp);
2932     } else {
2933         error_setg(errp, "virt: device unplug for unsupported device"
2934                    " type: %s", object_get_typename(OBJECT(dev)));
2935     }
2936 }
2937 
2938 static HotplugHandler *virt_machine_get_hotplug_handler(MachineState *machine,
2939                                                         DeviceState *dev)
2940 {
2941     MachineClass *mc = MACHINE_GET_CLASS(machine);
2942 
2943     if (device_is_dynamic_sysbus(mc, dev) ||
2944         object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) ||
2945         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI) ||
2946         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2947         return HOTPLUG_HANDLER(machine);
2948     }
2949     return NULL;
2950 }
2951 
2952 /*
2953  * for arm64 kvm_type [7-0] encodes the requested number of bits
2954  * in the IPA address space
2955  */
2956 static int virt_kvm_type(MachineState *ms, const char *type_str)
2957 {
2958     VirtMachineState *vms = VIRT_MACHINE(ms);
2959     int max_vm_pa_size, requested_pa_size;
2960     bool fixed_ipa;
2961 
2962     max_vm_pa_size = kvm_arm_get_max_vm_ipa_size(ms, &fixed_ipa);
2963 
2964     /* we freeze the memory map to compute the highest gpa */
2965     virt_set_memmap(vms, max_vm_pa_size);
2966 
2967     requested_pa_size = 64 - clz64(vms->highest_gpa);
2968 
2969     /*
2970      * KVM requires the IPA size to be at least 32 bits.
2971      */
2972     if (requested_pa_size < 32) {
2973         requested_pa_size = 32;
2974     }
2975 
2976     if (requested_pa_size > max_vm_pa_size) {
2977         error_report("-m and ,maxmem option values "
2978                      "require an IPA range (%d bits) larger than "
2979                      "the one supported by the host (%d bits)",
2980                      requested_pa_size, max_vm_pa_size);
2981         exit(1);
2982     }
2983     /*
2984      * We return the requested PA log size, unless KVM only supports
2985      * the implicit legacy 40b IPA setting, in which case the kvm_type
2986      * must be 0.
2987      */
2988     return fixed_ipa ? 0 : requested_pa_size;
2989 }
2990 
2991 static void virt_machine_class_init(ObjectClass *oc, void *data)
2992 {
2993     MachineClass *mc = MACHINE_CLASS(oc);
2994     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
2995 
2996     mc->init = machvirt_init;
2997     /* Start with max_cpus set to 512, which is the maximum supported by KVM.
2998      * The value may be reduced later when we have more information about the
2999      * configuration of the particular instance.
3000      */
3001     mc->max_cpus = 512;
3002     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_CALXEDA_XGMAC);
3003     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_AMD_XGBE);
3004     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_RAMFB_DEVICE);
3005     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_PLATFORM);
3006 #ifdef CONFIG_TPM
3007     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_TPM_TIS_SYSBUS);
3008 #endif
3009     mc->block_default_type = IF_VIRTIO;
3010     mc->no_cdrom = 1;
3011     mc->pci_allow_0_address = true;
3012     /* We know we will never create a pre-ARMv7 CPU which needs 1K pages */
3013     mc->minimum_page_bits = 12;
3014     mc->possible_cpu_arch_ids = virt_possible_cpu_arch_ids;
3015     mc->cpu_index_to_instance_props = virt_cpu_index_to_props;
3016 #ifdef CONFIG_TCG
3017     mc->default_cpu_type = ARM_CPU_TYPE_NAME("cortex-a15");
3018 #else
3019     mc->default_cpu_type = ARM_CPU_TYPE_NAME("max");
3020 #endif
3021     mc->get_default_cpu_node_id = virt_get_default_cpu_node_id;
3022     mc->kvm_type = virt_kvm_type;
3023     assert(!mc->get_hotplug_handler);
3024     mc->get_hotplug_handler = virt_machine_get_hotplug_handler;
3025     hc->pre_plug = virt_machine_device_pre_plug_cb;
3026     hc->plug = virt_machine_device_plug_cb;
3027     hc->unplug_request = virt_machine_device_unplug_request_cb;
3028     hc->unplug = virt_machine_device_unplug_cb;
3029     mc->nvdimm_supported = true;
3030     mc->smp_props.clusters_supported = true;
3031     mc->auto_enable_numa_with_memhp = true;
3032     mc->auto_enable_numa_with_memdev = true;
3033     mc->default_ram_id = "mach-virt.ram";
3034 
3035     object_class_property_add(oc, "acpi", "OnOffAuto",
3036         virt_get_acpi, virt_set_acpi,
3037         NULL, NULL);
3038     object_class_property_set_description(oc, "acpi",
3039         "Enable ACPI");
3040     object_class_property_add_bool(oc, "secure", virt_get_secure,
3041                                    virt_set_secure);
3042     object_class_property_set_description(oc, "secure",
3043                                                 "Set on/off to enable/disable the ARM "
3044                                                 "Security Extensions (TrustZone)");
3045 
3046     object_class_property_add_bool(oc, "virtualization", virt_get_virt,
3047                                    virt_set_virt);
3048     object_class_property_set_description(oc, "virtualization",
3049                                           "Set on/off to enable/disable emulating a "
3050                                           "guest CPU which implements the ARM "
3051                                           "Virtualization Extensions");
3052 
3053     object_class_property_add_bool(oc, "highmem", virt_get_highmem,
3054                                    virt_set_highmem);
3055     object_class_property_set_description(oc, "highmem",
3056                                           "Set on/off to enable/disable using "
3057                                           "physical address space above 32 bits");
3058 
3059     object_class_property_add_bool(oc, "compact-highmem",
3060                                    virt_get_compact_highmem,
3061                                    virt_set_compact_highmem);
3062     object_class_property_set_description(oc, "compact-highmem",
3063                                           "Set on/off to enable/disable compact "
3064                                           "layout for high memory regions");
3065 
3066     object_class_property_add_bool(oc, "highmem-redists",
3067                                    virt_get_highmem_redists,
3068                                    virt_set_highmem_redists);
3069     object_class_property_set_description(oc, "highmem-redists",
3070                                           "Set on/off to enable/disable high "
3071                                           "memory region for GICv3 or GICv4 "
3072                                           "redistributor");
3073 
3074     object_class_property_add_bool(oc, "highmem-ecam",
3075                                    virt_get_highmem_ecam,
3076                                    virt_set_highmem_ecam);
3077     object_class_property_set_description(oc, "highmem-ecam",
3078                                           "Set on/off to enable/disable high "
3079                                           "memory region for PCI ECAM");
3080 
3081     object_class_property_add_bool(oc, "highmem-mmio",
3082                                    virt_get_highmem_mmio,
3083                                    virt_set_highmem_mmio);
3084     object_class_property_set_description(oc, "highmem-mmio",
3085                                           "Set on/off to enable/disable high "
3086                                           "memory region for PCI MMIO");
3087 
3088     object_class_property_add_str(oc, "gic-version", virt_get_gic_version,
3089                                   virt_set_gic_version);
3090     object_class_property_set_description(oc, "gic-version",
3091                                           "Set GIC version. "
3092                                           "Valid values are 2, 3, 4, host and max");
3093 
3094     object_class_property_add_str(oc, "iommu", virt_get_iommu, virt_set_iommu);
3095     object_class_property_set_description(oc, "iommu",
3096                                           "Set the IOMMU type. "
3097                                           "Valid values are none and smmuv3");
3098 
3099     object_class_property_add_bool(oc, "default-bus-bypass-iommu",
3100                                    virt_get_default_bus_bypass_iommu,
3101                                    virt_set_default_bus_bypass_iommu);
3102     object_class_property_set_description(oc, "default-bus-bypass-iommu",
3103                                           "Set on/off to enable/disable "
3104                                           "bypass_iommu for default root bus");
3105 
3106     object_class_property_add_bool(oc, "ras", virt_get_ras,
3107                                    virt_set_ras);
3108     object_class_property_set_description(oc, "ras",
3109                                           "Set on/off to enable/disable reporting host memory errors "
3110                                           "to a KVM guest using ACPI and guest external abort exceptions");
3111 
3112     object_class_property_add_bool(oc, "mte", virt_get_mte, virt_set_mte);
3113     object_class_property_set_description(oc, "mte",
3114                                           "Set on/off to enable/disable emulating a "
3115                                           "guest CPU which implements the ARM "
3116                                           "Memory Tagging Extension");
3117 
3118     object_class_property_add_bool(oc, "its", virt_get_its,
3119                                    virt_set_its);
3120     object_class_property_set_description(oc, "its",
3121                                           "Set on/off to enable/disable "
3122                                           "ITS instantiation");
3123 
3124     object_class_property_add_bool(oc, "dtb-randomness",
3125                                    virt_get_dtb_randomness,
3126                                    virt_set_dtb_randomness);
3127     object_class_property_set_description(oc, "dtb-randomness",
3128                                           "Set off to disable passing random or "
3129                                           "non-deterministic dtb nodes to guest");
3130 
3131     object_class_property_add_bool(oc, "dtb-kaslr-seed",
3132                                    virt_get_dtb_randomness,
3133                                    virt_set_dtb_randomness);
3134     object_class_property_set_description(oc, "dtb-kaslr-seed",
3135                                           "Deprecated synonym of dtb-randomness");
3136 
3137     object_class_property_add_str(oc, "x-oem-id",
3138                                   virt_get_oem_id,
3139                                   virt_set_oem_id);
3140     object_class_property_set_description(oc, "x-oem-id",
3141                                           "Override the default value of field OEMID "
3142                                           "in ACPI table header."
3143                                           "The string may be up to 6 bytes in size");
3144 
3145 
3146     object_class_property_add_str(oc, "x-oem-table-id",
3147                                   virt_get_oem_table_id,
3148                                   virt_set_oem_table_id);
3149     object_class_property_set_description(oc, "x-oem-table-id",
3150                                           "Override the default value of field OEM Table ID "
3151                                           "in ACPI table header."
3152                                           "The string may be up to 8 bytes in size");
3153 
3154 }
3155 
3156 static void virt_instance_init(Object *obj)
3157 {
3158     VirtMachineState *vms = VIRT_MACHINE(obj);
3159     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
3160 
3161     /* EL3 is disabled by default on virt: this makes us consistent
3162      * between KVM and TCG for this board, and it also allows us to
3163      * boot UEFI blobs which assume no TrustZone support.
3164      */
3165     vms->secure = false;
3166 
3167     /* EL2 is also disabled by default, for similar reasons */
3168     vms->virt = false;
3169 
3170     /* High memory is enabled by default */
3171     vms->highmem = true;
3172     vms->highmem_compact = !vmc->no_highmem_compact;
3173     vms->gic_version = VIRT_GIC_VERSION_NOSEL;
3174 
3175     vms->highmem_ecam = !vmc->no_highmem_ecam;
3176     vms->highmem_mmio = true;
3177     vms->highmem_redists = true;
3178 
3179     if (vmc->no_its) {
3180         vms->its = false;
3181     } else {
3182         /* Default allows ITS instantiation */
3183         vms->its = true;
3184 
3185         if (vmc->no_tcg_its) {
3186             vms->tcg_its = false;
3187         } else {
3188             vms->tcg_its = true;
3189         }
3190     }
3191 
3192     /* Default disallows iommu instantiation */
3193     vms->iommu = VIRT_IOMMU_NONE;
3194 
3195     /* The default root bus is attached to iommu by default */
3196     vms->default_bus_bypass_iommu = false;
3197 
3198     /* Default disallows RAS instantiation */
3199     vms->ras = false;
3200 
3201     /* MTE is disabled by default.  */
3202     vms->mte = false;
3203 
3204     /* Supply kaslr-seed and rng-seed by default */
3205     vms->dtb_randomness = true;
3206 
3207     vms->irqmap = a15irqmap;
3208 
3209     virt_flash_create(vms);
3210 
3211     vms->oem_id = g_strndup(ACPI_BUILD_APPNAME6, 6);
3212     vms->oem_table_id = g_strndup(ACPI_BUILD_APPNAME8, 8);
3213 }
3214 
3215 static const TypeInfo virt_machine_info = {
3216     .name          = TYPE_VIRT_MACHINE,
3217     .parent        = TYPE_MACHINE,
3218     .abstract      = true,
3219     .instance_size = sizeof(VirtMachineState),
3220     .class_size    = sizeof(VirtMachineClass),
3221     .class_init    = virt_machine_class_init,
3222     .instance_init = virt_instance_init,
3223     .interfaces = (InterfaceInfo[]) {
3224          { TYPE_HOTPLUG_HANDLER },
3225          { }
3226     },
3227 };
3228 
3229 static void machvirt_machine_init(void)
3230 {
3231     type_register_static(&virt_machine_info);
3232 }
3233 type_init(machvirt_machine_init);
3234 
3235 static void virt_machine_8_0_options(MachineClass *mc)
3236 {
3237 }
3238 DEFINE_VIRT_MACHINE_AS_LATEST(8, 0)
3239 
3240 static void virt_machine_7_2_options(MachineClass *mc)
3241 {
3242     virt_machine_8_0_options(mc);
3243     compat_props_add(mc->compat_props, hw_compat_7_2, hw_compat_7_2_len);
3244 }
3245 DEFINE_VIRT_MACHINE(7, 2)
3246 
3247 static void virt_machine_7_1_options(MachineClass *mc)
3248 {
3249     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3250 
3251     virt_machine_7_2_options(mc);
3252     compat_props_add(mc->compat_props, hw_compat_7_1, hw_compat_7_1_len);
3253     /* Compact layout for high memory regions was introduced with 7.2 */
3254     vmc->no_highmem_compact = true;
3255 }
3256 DEFINE_VIRT_MACHINE(7, 1)
3257 
3258 static void virt_machine_7_0_options(MachineClass *mc)
3259 {
3260     virt_machine_7_1_options(mc);
3261     compat_props_add(mc->compat_props, hw_compat_7_0, hw_compat_7_0_len);
3262 }
3263 DEFINE_VIRT_MACHINE(7, 0)
3264 
3265 static void virt_machine_6_2_options(MachineClass *mc)
3266 {
3267     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3268 
3269     virt_machine_7_0_options(mc);
3270     compat_props_add(mc->compat_props, hw_compat_6_2, hw_compat_6_2_len);
3271     vmc->no_tcg_lpa2 = true;
3272 }
3273 DEFINE_VIRT_MACHINE(6, 2)
3274 
3275 static void virt_machine_6_1_options(MachineClass *mc)
3276 {
3277     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3278 
3279     virt_machine_6_2_options(mc);
3280     compat_props_add(mc->compat_props, hw_compat_6_1, hw_compat_6_1_len);
3281     mc->smp_props.prefer_sockets = true;
3282     vmc->no_cpu_topology = true;
3283 
3284     /* qemu ITS was introduced with 6.2 */
3285     vmc->no_tcg_its = true;
3286 }
3287 DEFINE_VIRT_MACHINE(6, 1)
3288 
3289 static void virt_machine_6_0_options(MachineClass *mc)
3290 {
3291     virt_machine_6_1_options(mc);
3292     compat_props_add(mc->compat_props, hw_compat_6_0, hw_compat_6_0_len);
3293 }
3294 DEFINE_VIRT_MACHINE(6, 0)
3295 
3296 static void virt_machine_5_2_options(MachineClass *mc)
3297 {
3298     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3299 
3300     virt_machine_6_0_options(mc);
3301     compat_props_add(mc->compat_props, hw_compat_5_2, hw_compat_5_2_len);
3302     vmc->no_secure_gpio = true;
3303 }
3304 DEFINE_VIRT_MACHINE(5, 2)
3305 
3306 static void virt_machine_5_1_options(MachineClass *mc)
3307 {
3308     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3309 
3310     virt_machine_5_2_options(mc);
3311     compat_props_add(mc->compat_props, hw_compat_5_1, hw_compat_5_1_len);
3312     vmc->no_kvm_steal_time = true;
3313 }
3314 DEFINE_VIRT_MACHINE(5, 1)
3315 
3316 static void virt_machine_5_0_options(MachineClass *mc)
3317 {
3318     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3319 
3320     virt_machine_5_1_options(mc);
3321     compat_props_add(mc->compat_props, hw_compat_5_0, hw_compat_5_0_len);
3322     mc->numa_mem_supported = true;
3323     vmc->acpi_expose_flash = true;
3324     mc->auto_enable_numa_with_memdev = false;
3325 }
3326 DEFINE_VIRT_MACHINE(5, 0)
3327 
3328 static void virt_machine_4_2_options(MachineClass *mc)
3329 {
3330     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3331 
3332     virt_machine_5_0_options(mc);
3333     compat_props_add(mc->compat_props, hw_compat_4_2, hw_compat_4_2_len);
3334     vmc->kvm_no_adjvtime = true;
3335 }
3336 DEFINE_VIRT_MACHINE(4, 2)
3337 
3338 static void virt_machine_4_1_options(MachineClass *mc)
3339 {
3340     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3341 
3342     virt_machine_4_2_options(mc);
3343     compat_props_add(mc->compat_props, hw_compat_4_1, hw_compat_4_1_len);
3344     vmc->no_ged = true;
3345     mc->auto_enable_numa_with_memhp = false;
3346 }
3347 DEFINE_VIRT_MACHINE(4, 1)
3348 
3349 static void virt_machine_4_0_options(MachineClass *mc)
3350 {
3351     virt_machine_4_1_options(mc);
3352     compat_props_add(mc->compat_props, hw_compat_4_0, hw_compat_4_0_len);
3353 }
3354 DEFINE_VIRT_MACHINE(4, 0)
3355 
3356 static void virt_machine_3_1_options(MachineClass *mc)
3357 {
3358     virt_machine_4_0_options(mc);
3359     compat_props_add(mc->compat_props, hw_compat_3_1, hw_compat_3_1_len);
3360 }
3361 DEFINE_VIRT_MACHINE(3, 1)
3362 
3363 static void virt_machine_3_0_options(MachineClass *mc)
3364 {
3365     virt_machine_3_1_options(mc);
3366     compat_props_add(mc->compat_props, hw_compat_3_0, hw_compat_3_0_len);
3367 }
3368 DEFINE_VIRT_MACHINE(3, 0)
3369 
3370 static void virt_machine_2_12_options(MachineClass *mc)
3371 {
3372     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3373 
3374     virt_machine_3_0_options(mc);
3375     compat_props_add(mc->compat_props, hw_compat_2_12, hw_compat_2_12_len);
3376     vmc->no_highmem_ecam = true;
3377     mc->max_cpus = 255;
3378 }
3379 DEFINE_VIRT_MACHINE(2, 12)
3380 
3381 static void virt_machine_2_11_options(MachineClass *mc)
3382 {
3383     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3384 
3385     virt_machine_2_12_options(mc);
3386     compat_props_add(mc->compat_props, hw_compat_2_11, hw_compat_2_11_len);
3387     vmc->smbios_old_sys_ver = true;
3388 }
3389 DEFINE_VIRT_MACHINE(2, 11)
3390 
3391 static void virt_machine_2_10_options(MachineClass *mc)
3392 {
3393     virt_machine_2_11_options(mc);
3394     compat_props_add(mc->compat_props, hw_compat_2_10, hw_compat_2_10_len);
3395     /* before 2.11 we never faulted accesses to bad addresses */
3396     mc->ignore_memory_transaction_failures = true;
3397 }
3398 DEFINE_VIRT_MACHINE(2, 10)
3399 
3400 static void virt_machine_2_9_options(MachineClass *mc)
3401 {
3402     virt_machine_2_10_options(mc);
3403     compat_props_add(mc->compat_props, hw_compat_2_9, hw_compat_2_9_len);
3404 }
3405 DEFINE_VIRT_MACHINE(2, 9)
3406 
3407 static void virt_machine_2_8_options(MachineClass *mc)
3408 {
3409     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3410 
3411     virt_machine_2_9_options(mc);
3412     compat_props_add(mc->compat_props, hw_compat_2_8, hw_compat_2_8_len);
3413     /* For 2.8 and earlier we falsely claimed in the DT that
3414      * our timers were edge-triggered, not level-triggered.
3415      */
3416     vmc->claim_edge_triggered_timers = true;
3417 }
3418 DEFINE_VIRT_MACHINE(2, 8)
3419 
3420 static void virt_machine_2_7_options(MachineClass *mc)
3421 {
3422     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3423 
3424     virt_machine_2_8_options(mc);
3425     compat_props_add(mc->compat_props, hw_compat_2_7, hw_compat_2_7_len);
3426     /* ITS was introduced with 2.8 */
3427     vmc->no_its = true;
3428     /* Stick with 1K pages for migration compatibility */
3429     mc->minimum_page_bits = 0;
3430 }
3431 DEFINE_VIRT_MACHINE(2, 7)
3432 
3433 static void virt_machine_2_6_options(MachineClass *mc)
3434 {
3435     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3436 
3437     virt_machine_2_7_options(mc);
3438     compat_props_add(mc->compat_props, hw_compat_2_6, hw_compat_2_6_len);
3439     vmc->disallow_affinity_adjustment = true;
3440     /* Disable PMU for 2.6 as PMU support was first introduced in 2.7 */
3441     vmc->no_pmu = true;
3442 }
3443 DEFINE_VIRT_MACHINE(2, 6)
3444