xref: /openbmc/qemu/hw/arm/virt.c (revision f40408a9fe5d1db70a75a33d2b26c8af8a5d57b0)
1 /*
2  * ARM mach-virt emulation
3  *
4  * Copyright (c) 2013 Linaro Limited
5  *
6  * This program is free software; you can redistribute it and/or modify it
7  * under the terms and conditions of the GNU General Public License,
8  * version 2 or later, as published by the Free Software Foundation.
9  *
10  * This program is distributed in the hope it will be useful, but WITHOUT
11  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
12  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
13  * more details.
14  *
15  * You should have received a copy of the GNU General Public License along with
16  * this program.  If not, see <http://www.gnu.org/licenses/>.
17  *
18  * Emulate a virtual board which works by passing Linux all the information
19  * it needs about what devices are present via the device tree.
20  * There are some restrictions about what we can do here:
21  *  + we can only present devices whose Linux drivers will work based
22  *    purely on the device tree with no platform data at all
23  *  + we want to present a very stripped-down minimalist platform,
24  *    both because this reduces the security attack surface from the guest
25  *    and also because it reduces our exposure to being broken when
26  *    the kernel updates its device tree bindings and requires further
27  *    information in a device binding that we aren't providing.
28  * This is essentially the same approach kvmtool uses.
29  */
30 
31 #include "qemu/osdep.h"
32 #include "qemu/datadir.h"
33 #include "qemu/units.h"
34 #include "qemu/option.h"
35 #include "monitor/qdev.h"
36 #include "qapi/error.h"
37 #include "hw/sysbus.h"
38 #include "hw/arm/boot.h"
39 #include "hw/arm/primecell.h"
40 #include "hw/arm/virt.h"
41 #include "hw/block/flash.h"
42 #include "hw/vfio/vfio-calxeda-xgmac.h"
43 #include "hw/vfio/vfio-amd-xgbe.h"
44 #include "hw/display/ramfb.h"
45 #include "net/net.h"
46 #include "sysemu/device_tree.h"
47 #include "sysemu/numa.h"
48 #include "sysemu/runstate.h"
49 #include "sysemu/tpm.h"
50 #include "sysemu/kvm.h"
51 #include "sysemu/hvf.h"
52 #include "hw/loader.h"
53 #include "qapi/error.h"
54 #include "qemu/bitops.h"
55 #include "qemu/error-report.h"
56 #include "qemu/module.h"
57 #include "hw/pci-host/gpex.h"
58 #include "hw/virtio/virtio-pci.h"
59 #include "hw/core/sysbus-fdt.h"
60 #include "hw/platform-bus.h"
61 #include "hw/qdev-properties.h"
62 #include "hw/arm/fdt.h"
63 #include "hw/intc/arm_gic.h"
64 #include "hw/intc/arm_gicv3_common.h"
65 #include "hw/irq.h"
66 #include "kvm_arm.h"
67 #include "hw/firmware/smbios.h"
68 #include "qapi/visitor.h"
69 #include "qapi/qapi-visit-common.h"
70 #include "standard-headers/linux/input.h"
71 #include "hw/arm/smmuv3.h"
72 #include "hw/acpi/acpi.h"
73 #include "target/arm/internals.h"
74 #include "hw/mem/memory-device.h"
75 #include "hw/mem/pc-dimm.h"
76 #include "hw/mem/nvdimm.h"
77 #include "hw/acpi/generic_event_device.h"
78 #include "hw/virtio/virtio-mem-pci.h"
79 #include "hw/virtio/virtio-iommu.h"
80 #include "hw/char/pl011.h"
81 #include "qemu/guest-random.h"
82 
83 #define DEFINE_VIRT_MACHINE_LATEST(major, minor, latest) \
84     static void virt_##major##_##minor##_class_init(ObjectClass *oc, \
85                                                     void *data) \
86     { \
87         MachineClass *mc = MACHINE_CLASS(oc); \
88         virt_machine_##major##_##minor##_options(mc); \
89         mc->desc = "QEMU " # major "." # minor " ARM Virtual Machine"; \
90         if (latest) { \
91             mc->alias = "virt"; \
92         } \
93     } \
94     static const TypeInfo machvirt_##major##_##minor##_info = { \
95         .name = MACHINE_TYPE_NAME("virt-" # major "." # minor), \
96         .parent = TYPE_VIRT_MACHINE, \
97         .class_init = virt_##major##_##minor##_class_init, \
98     }; \
99     static void machvirt_machine_##major##_##minor##_init(void) \
100     { \
101         type_register_static(&machvirt_##major##_##minor##_info); \
102     } \
103     type_init(machvirt_machine_##major##_##minor##_init);
104 
105 #define DEFINE_VIRT_MACHINE_AS_LATEST(major, minor) \
106     DEFINE_VIRT_MACHINE_LATEST(major, minor, true)
107 #define DEFINE_VIRT_MACHINE(major, minor) \
108     DEFINE_VIRT_MACHINE_LATEST(major, minor, false)
109 
110 
111 /* Number of external interrupt lines to configure the GIC with */
112 #define NUM_IRQS 256
113 
114 #define PLATFORM_BUS_NUM_IRQS 64
115 
116 /* Legacy RAM limit in GB (< version 4.0) */
117 #define LEGACY_RAMLIMIT_GB 255
118 #define LEGACY_RAMLIMIT_BYTES (LEGACY_RAMLIMIT_GB * GiB)
119 
120 /* Addresses and sizes of our components.
121  * 0..128MB is space for a flash device so we can run bootrom code such as UEFI.
122  * 128MB..256MB is used for miscellaneous device I/O.
123  * 256MB..1GB is reserved for possible future PCI support (ie where the
124  * PCI memory window will go if we add a PCI host controller).
125  * 1GB and up is RAM (which may happily spill over into the
126  * high memory region beyond 4GB).
127  * This represents a compromise between how much RAM can be given to
128  * a 32 bit VM and leaving space for expansion and in particular for PCI.
129  * Note that devices should generally be placed at multiples of 0x10000,
130  * to accommodate guests using 64K pages.
131  */
132 static const MemMapEntry base_memmap[] = {
133     /* Space up to 0x8000000 is reserved for a boot ROM */
134     [VIRT_FLASH] =              {          0, 0x08000000 },
135     [VIRT_CPUPERIPHS] =         { 0x08000000, 0x00020000 },
136     /* GIC distributor and CPU interfaces sit inside the CPU peripheral space */
137     [VIRT_GIC_DIST] =           { 0x08000000, 0x00010000 },
138     [VIRT_GIC_CPU] =            { 0x08010000, 0x00010000 },
139     [VIRT_GIC_V2M] =            { 0x08020000, 0x00001000 },
140     [VIRT_GIC_HYP] =            { 0x08030000, 0x00010000 },
141     [VIRT_GIC_VCPU] =           { 0x08040000, 0x00010000 },
142     /* The space in between here is reserved for GICv3 CPU/vCPU/HYP */
143     [VIRT_GIC_ITS] =            { 0x08080000, 0x00020000 },
144     /* This redistributor space allows up to 2*64kB*123 CPUs */
145     [VIRT_GIC_REDIST] =         { 0x080A0000, 0x00F60000 },
146     [VIRT_UART] =               { 0x09000000, 0x00001000 },
147     [VIRT_RTC] =                { 0x09010000, 0x00001000 },
148     [VIRT_FW_CFG] =             { 0x09020000, 0x00000018 },
149     [VIRT_GPIO] =               { 0x09030000, 0x00001000 },
150     [VIRT_SECURE_UART] =        { 0x09040000, 0x00001000 },
151     [VIRT_SMMU] =               { 0x09050000, 0x00020000 },
152     [VIRT_PCDIMM_ACPI] =        { 0x09070000, MEMORY_HOTPLUG_IO_LEN },
153     [VIRT_ACPI_GED] =           { 0x09080000, ACPI_GED_EVT_SEL_LEN },
154     [VIRT_NVDIMM_ACPI] =        { 0x09090000, NVDIMM_ACPI_IO_LEN},
155     [VIRT_PVTIME] =             { 0x090a0000, 0x00010000 },
156     [VIRT_SECURE_GPIO] =        { 0x090b0000, 0x00001000 },
157     [VIRT_MMIO] =               { 0x0a000000, 0x00000200 },
158     /* ...repeating for a total of NUM_VIRTIO_TRANSPORTS, each of that size */
159     [VIRT_PLATFORM_BUS] =       { 0x0c000000, 0x02000000 },
160     [VIRT_SECURE_MEM] =         { 0x0e000000, 0x01000000 },
161     [VIRT_PCIE_MMIO] =          { 0x10000000, 0x2eff0000 },
162     [VIRT_PCIE_PIO] =           { 0x3eff0000, 0x00010000 },
163     [VIRT_PCIE_ECAM] =          { 0x3f000000, 0x01000000 },
164     /* Actual RAM size depends on initial RAM and device memory settings */
165     [VIRT_MEM] =                { GiB, LEGACY_RAMLIMIT_BYTES },
166 };
167 
168 /*
169  * Highmem IO Regions: This memory map is floating, located after the RAM.
170  * Each MemMapEntry base (GPA) will be dynamically computed, depending on the
171  * top of the RAM, so that its base get the same alignment as the size,
172  * ie. a 512GiB entry will be aligned on a 512GiB boundary. If there is
173  * less than 256GiB of RAM, the floating area starts at the 256GiB mark.
174  * Note the extended_memmap is sized so that it eventually also includes the
175  * base_memmap entries (VIRT_HIGH_GIC_REDIST2 index is greater than the last
176  * index of base_memmap).
177  *
178  * The memory map for these Highmem IO Regions can be in legacy or compact
179  * layout, depending on 'compact-highmem' property. With legacy layout, the
180  * PA space for one specific region is always reserved, even if the region
181  * has been disabled or doesn't fit into the PA space. However, the PA space
182  * for the region won't be reserved in these circumstances with compact layout.
183  */
184 static MemMapEntry extended_memmap[] = {
185     /* Additional 64 MB redist region (can contain up to 512 redistributors) */
186     [VIRT_HIGH_GIC_REDIST2] =   { 0x0, 64 * MiB },
187     [VIRT_HIGH_PCIE_ECAM] =     { 0x0, 256 * MiB },
188     /* Second PCIe window */
189     [VIRT_HIGH_PCIE_MMIO] =     { 0x0, 512 * GiB },
190 };
191 
192 static const int a15irqmap[] = {
193     [VIRT_UART] = 1,
194     [VIRT_RTC] = 2,
195     [VIRT_PCIE] = 3, /* ... to 6 */
196     [VIRT_GPIO] = 7,
197     [VIRT_SECURE_UART] = 8,
198     [VIRT_ACPI_GED] = 9,
199     [VIRT_MMIO] = 16, /* ...to 16 + NUM_VIRTIO_TRANSPORTS - 1 */
200     [VIRT_GIC_V2M] = 48, /* ...to 48 + NUM_GICV2M_SPIS - 1 */
201     [VIRT_SMMU] = 74,    /* ...to 74 + NUM_SMMU_IRQS - 1 */
202     [VIRT_PLATFORM_BUS] = 112, /* ...to 112 + PLATFORM_BUS_NUM_IRQS -1 */
203 };
204 
205 static const char *valid_cpus[] = {
206     ARM_CPU_TYPE_NAME("cortex-a7"),
207     ARM_CPU_TYPE_NAME("cortex-a15"),
208     ARM_CPU_TYPE_NAME("cortex-a35"),
209     ARM_CPU_TYPE_NAME("cortex-a53"),
210     ARM_CPU_TYPE_NAME("cortex-a57"),
211     ARM_CPU_TYPE_NAME("cortex-a72"),
212     ARM_CPU_TYPE_NAME("cortex-a76"),
213     ARM_CPU_TYPE_NAME("a64fx"),
214     ARM_CPU_TYPE_NAME("neoverse-n1"),
215     ARM_CPU_TYPE_NAME("host"),
216     ARM_CPU_TYPE_NAME("max"),
217 };
218 
219 static bool cpu_type_valid(const char *cpu)
220 {
221     int i;
222 
223     for (i = 0; i < ARRAY_SIZE(valid_cpus); i++) {
224         if (strcmp(cpu, valid_cpus[i]) == 0) {
225             return true;
226         }
227     }
228     return false;
229 }
230 
231 static void create_randomness(MachineState *ms, const char *node)
232 {
233     struct {
234         uint64_t kaslr;
235         uint8_t rng[32];
236     } seed;
237 
238     if (qemu_guest_getrandom(&seed, sizeof(seed), NULL)) {
239         return;
240     }
241     qemu_fdt_setprop_u64(ms->fdt, node, "kaslr-seed", seed.kaslr);
242     qemu_fdt_setprop(ms->fdt, node, "rng-seed", seed.rng, sizeof(seed.rng));
243 }
244 
245 static void create_fdt(VirtMachineState *vms)
246 {
247     MachineState *ms = MACHINE(vms);
248     int nb_numa_nodes = ms->numa_state->num_nodes;
249     void *fdt = create_device_tree(&vms->fdt_size);
250 
251     if (!fdt) {
252         error_report("create_device_tree() failed");
253         exit(1);
254     }
255 
256     ms->fdt = fdt;
257 
258     /* Header */
259     qemu_fdt_setprop_string(fdt, "/", "compatible", "linux,dummy-virt");
260     qemu_fdt_setprop_cell(fdt, "/", "#address-cells", 0x2);
261     qemu_fdt_setprop_cell(fdt, "/", "#size-cells", 0x2);
262     qemu_fdt_setprop_string(fdt, "/", "model", "linux,dummy-virt");
263 
264     /* /chosen must exist for load_dtb to fill in necessary properties later */
265     qemu_fdt_add_subnode(fdt, "/chosen");
266     if (vms->dtb_randomness) {
267         create_randomness(ms, "/chosen");
268     }
269 
270     if (vms->secure) {
271         qemu_fdt_add_subnode(fdt, "/secure-chosen");
272         if (vms->dtb_randomness) {
273             create_randomness(ms, "/secure-chosen");
274         }
275     }
276 
277     /* Clock node, for the benefit of the UART. The kernel device tree
278      * binding documentation claims the PL011 node clock properties are
279      * optional but in practice if you omit them the kernel refuses to
280      * probe for the device.
281      */
282     vms->clock_phandle = qemu_fdt_alloc_phandle(fdt);
283     qemu_fdt_add_subnode(fdt, "/apb-pclk");
284     qemu_fdt_setprop_string(fdt, "/apb-pclk", "compatible", "fixed-clock");
285     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "#clock-cells", 0x0);
286     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "clock-frequency", 24000000);
287     qemu_fdt_setprop_string(fdt, "/apb-pclk", "clock-output-names",
288                                 "clk24mhz");
289     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "phandle", vms->clock_phandle);
290 
291     if (nb_numa_nodes > 0 && ms->numa_state->have_numa_distance) {
292         int size = nb_numa_nodes * nb_numa_nodes * 3 * sizeof(uint32_t);
293         uint32_t *matrix = g_malloc0(size);
294         int idx, i, j;
295 
296         for (i = 0; i < nb_numa_nodes; i++) {
297             for (j = 0; j < nb_numa_nodes; j++) {
298                 idx = (i * nb_numa_nodes + j) * 3;
299                 matrix[idx + 0] = cpu_to_be32(i);
300                 matrix[idx + 1] = cpu_to_be32(j);
301                 matrix[idx + 2] =
302                     cpu_to_be32(ms->numa_state->nodes[i].distance[j]);
303             }
304         }
305 
306         qemu_fdt_add_subnode(fdt, "/distance-map");
307         qemu_fdt_setprop_string(fdt, "/distance-map", "compatible",
308                                 "numa-distance-map-v1");
309         qemu_fdt_setprop(fdt, "/distance-map", "distance-matrix",
310                          matrix, size);
311         g_free(matrix);
312     }
313 }
314 
315 static void fdt_add_timer_nodes(const VirtMachineState *vms)
316 {
317     /* On real hardware these interrupts are level-triggered.
318      * On KVM they were edge-triggered before host kernel version 4.4,
319      * and level-triggered afterwards.
320      * On emulated QEMU they are level-triggered.
321      *
322      * Getting the DTB info about them wrong is awkward for some
323      * guest kernels:
324      *  pre-4.8 ignore the DT and leave the interrupt configured
325      *   with whatever the GIC reset value (or the bootloader) left it at
326      *  4.8 before rc6 honour the incorrect data by programming it back
327      *   into the GIC, causing problems
328      *  4.8rc6 and later ignore the DT and always write "level triggered"
329      *   into the GIC
330      *
331      * For backwards-compatibility, virt-2.8 and earlier will continue
332      * to say these are edge-triggered, but later machines will report
333      * the correct information.
334      */
335     ARMCPU *armcpu;
336     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
337     uint32_t irqflags = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
338     MachineState *ms = MACHINE(vms);
339 
340     if (vmc->claim_edge_triggered_timers) {
341         irqflags = GIC_FDT_IRQ_FLAGS_EDGE_LO_HI;
342     }
343 
344     if (vms->gic_version == VIRT_GIC_VERSION_2) {
345         irqflags = deposit32(irqflags, GIC_FDT_IRQ_PPI_CPU_START,
346                              GIC_FDT_IRQ_PPI_CPU_WIDTH,
347                              (1 << MACHINE(vms)->smp.cpus) - 1);
348     }
349 
350     qemu_fdt_add_subnode(ms->fdt, "/timer");
351 
352     armcpu = ARM_CPU(qemu_get_cpu(0));
353     if (arm_feature(&armcpu->env, ARM_FEATURE_V8)) {
354         const char compat[] = "arm,armv8-timer\0arm,armv7-timer";
355         qemu_fdt_setprop(ms->fdt, "/timer", "compatible",
356                          compat, sizeof(compat));
357     } else {
358         qemu_fdt_setprop_string(ms->fdt, "/timer", "compatible",
359                                 "arm,armv7-timer");
360     }
361     qemu_fdt_setprop(ms->fdt, "/timer", "always-on", NULL, 0);
362     qemu_fdt_setprop_cells(ms->fdt, "/timer", "interrupts",
363                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_S_EL1_IRQ, irqflags,
364                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_NS_EL1_IRQ, irqflags,
365                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_VIRT_IRQ, irqflags,
366                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_NS_EL2_IRQ, irqflags);
367 }
368 
369 static void fdt_add_cpu_nodes(const VirtMachineState *vms)
370 {
371     int cpu;
372     int addr_cells = 1;
373     const MachineState *ms = MACHINE(vms);
374     const VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
375     int smp_cpus = ms->smp.cpus;
376 
377     /*
378      * See Linux Documentation/devicetree/bindings/arm/cpus.yaml
379      * On ARM v8 64-bit systems value should be set to 2,
380      * that corresponds to the MPIDR_EL1 register size.
381      * If MPIDR_EL1[63:32] value is equal to 0 on all CPUs
382      * in the system, #address-cells can be set to 1, since
383      * MPIDR_EL1[63:32] bits are not used for CPUs
384      * identification.
385      *
386      * Here we actually don't know whether our system is 32- or 64-bit one.
387      * The simplest way to go is to examine affinity IDs of all our CPUs. If
388      * at least one of them has Aff3 populated, we set #address-cells to 2.
389      */
390     for (cpu = 0; cpu < smp_cpus; cpu++) {
391         ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(cpu));
392 
393         if (armcpu->mp_affinity & ARM_AFF3_MASK) {
394             addr_cells = 2;
395             break;
396         }
397     }
398 
399     qemu_fdt_add_subnode(ms->fdt, "/cpus");
400     qemu_fdt_setprop_cell(ms->fdt, "/cpus", "#address-cells", addr_cells);
401     qemu_fdt_setprop_cell(ms->fdt, "/cpus", "#size-cells", 0x0);
402 
403     for (cpu = smp_cpus - 1; cpu >= 0; cpu--) {
404         char *nodename = g_strdup_printf("/cpus/cpu@%d", cpu);
405         ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(cpu));
406         CPUState *cs = CPU(armcpu);
407 
408         qemu_fdt_add_subnode(ms->fdt, nodename);
409         qemu_fdt_setprop_string(ms->fdt, nodename, "device_type", "cpu");
410         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
411                                     armcpu->dtb_compatible);
412 
413         if (vms->psci_conduit != QEMU_PSCI_CONDUIT_DISABLED && smp_cpus > 1) {
414             qemu_fdt_setprop_string(ms->fdt, nodename,
415                                         "enable-method", "psci");
416         }
417 
418         if (addr_cells == 2) {
419             qemu_fdt_setprop_u64(ms->fdt, nodename, "reg",
420                                  armcpu->mp_affinity);
421         } else {
422             qemu_fdt_setprop_cell(ms->fdt, nodename, "reg",
423                                   armcpu->mp_affinity);
424         }
425 
426         if (ms->possible_cpus->cpus[cs->cpu_index].props.has_node_id) {
427             qemu_fdt_setprop_cell(ms->fdt, nodename, "numa-node-id",
428                 ms->possible_cpus->cpus[cs->cpu_index].props.node_id);
429         }
430 
431         if (!vmc->no_cpu_topology) {
432             qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle",
433                                   qemu_fdt_alloc_phandle(ms->fdt));
434         }
435 
436         g_free(nodename);
437     }
438 
439     if (!vmc->no_cpu_topology) {
440         /*
441          * Add vCPU topology description through fdt node cpu-map.
442          *
443          * See Linux Documentation/devicetree/bindings/cpu/cpu-topology.txt
444          * In a SMP system, the hierarchy of CPUs can be defined through
445          * four entities that are used to describe the layout of CPUs in
446          * the system: socket/cluster/core/thread.
447          *
448          * A socket node represents the boundary of system physical package
449          * and its child nodes must be one or more cluster nodes. A system
450          * can contain several layers of clustering within a single physical
451          * package and cluster nodes can be contained in parent cluster nodes.
452          *
453          * Note: currently we only support one layer of clustering within
454          * each physical package.
455          */
456         qemu_fdt_add_subnode(ms->fdt, "/cpus/cpu-map");
457 
458         for (cpu = smp_cpus - 1; cpu >= 0; cpu--) {
459             char *cpu_path = g_strdup_printf("/cpus/cpu@%d", cpu);
460             char *map_path;
461 
462             if (ms->smp.threads > 1) {
463                 map_path = g_strdup_printf(
464                     "/cpus/cpu-map/socket%d/cluster%d/core%d/thread%d",
465                     cpu / (ms->smp.clusters * ms->smp.cores * ms->smp.threads),
466                     (cpu / (ms->smp.cores * ms->smp.threads)) % ms->smp.clusters,
467                     (cpu / ms->smp.threads) % ms->smp.cores,
468                     cpu % ms->smp.threads);
469             } else {
470                 map_path = g_strdup_printf(
471                     "/cpus/cpu-map/socket%d/cluster%d/core%d",
472                     cpu / (ms->smp.clusters * ms->smp.cores),
473                     (cpu / ms->smp.cores) % ms->smp.clusters,
474                     cpu % ms->smp.cores);
475             }
476             qemu_fdt_add_path(ms->fdt, map_path);
477             qemu_fdt_setprop_phandle(ms->fdt, map_path, "cpu", cpu_path);
478 
479             g_free(map_path);
480             g_free(cpu_path);
481         }
482     }
483 }
484 
485 static void fdt_add_its_gic_node(VirtMachineState *vms)
486 {
487     char *nodename;
488     MachineState *ms = MACHINE(vms);
489 
490     vms->msi_phandle = qemu_fdt_alloc_phandle(ms->fdt);
491     nodename = g_strdup_printf("/intc/its@%" PRIx64,
492                                vms->memmap[VIRT_GIC_ITS].base);
493     qemu_fdt_add_subnode(ms->fdt, nodename);
494     qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
495                             "arm,gic-v3-its");
496     qemu_fdt_setprop(ms->fdt, nodename, "msi-controller", NULL, 0);
497     qemu_fdt_setprop_cell(ms->fdt, nodename, "#msi-cells", 1);
498     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
499                                  2, vms->memmap[VIRT_GIC_ITS].base,
500                                  2, vms->memmap[VIRT_GIC_ITS].size);
501     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", vms->msi_phandle);
502     g_free(nodename);
503 }
504 
505 static void fdt_add_v2m_gic_node(VirtMachineState *vms)
506 {
507     MachineState *ms = MACHINE(vms);
508     char *nodename;
509 
510     nodename = g_strdup_printf("/intc/v2m@%" PRIx64,
511                                vms->memmap[VIRT_GIC_V2M].base);
512     vms->msi_phandle = qemu_fdt_alloc_phandle(ms->fdt);
513     qemu_fdt_add_subnode(ms->fdt, nodename);
514     qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
515                             "arm,gic-v2m-frame");
516     qemu_fdt_setprop(ms->fdt, nodename, "msi-controller", NULL, 0);
517     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
518                                  2, vms->memmap[VIRT_GIC_V2M].base,
519                                  2, vms->memmap[VIRT_GIC_V2M].size);
520     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", vms->msi_phandle);
521     g_free(nodename);
522 }
523 
524 static void fdt_add_gic_node(VirtMachineState *vms)
525 {
526     MachineState *ms = MACHINE(vms);
527     char *nodename;
528 
529     vms->gic_phandle = qemu_fdt_alloc_phandle(ms->fdt);
530     qemu_fdt_setprop_cell(ms->fdt, "/", "interrupt-parent", vms->gic_phandle);
531 
532     nodename = g_strdup_printf("/intc@%" PRIx64,
533                                vms->memmap[VIRT_GIC_DIST].base);
534     qemu_fdt_add_subnode(ms->fdt, nodename);
535     qemu_fdt_setprop_cell(ms->fdt, nodename, "#interrupt-cells", 3);
536     qemu_fdt_setprop(ms->fdt, nodename, "interrupt-controller", NULL, 0);
537     qemu_fdt_setprop_cell(ms->fdt, nodename, "#address-cells", 0x2);
538     qemu_fdt_setprop_cell(ms->fdt, nodename, "#size-cells", 0x2);
539     qemu_fdt_setprop(ms->fdt, nodename, "ranges", NULL, 0);
540     if (vms->gic_version != VIRT_GIC_VERSION_2) {
541         int nb_redist_regions = virt_gicv3_redist_region_count(vms);
542 
543         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
544                                 "arm,gic-v3");
545 
546         qemu_fdt_setprop_cell(ms->fdt, nodename,
547                               "#redistributor-regions", nb_redist_regions);
548 
549         if (nb_redist_regions == 1) {
550             qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
551                                          2, vms->memmap[VIRT_GIC_DIST].base,
552                                          2, vms->memmap[VIRT_GIC_DIST].size,
553                                          2, vms->memmap[VIRT_GIC_REDIST].base,
554                                          2, vms->memmap[VIRT_GIC_REDIST].size);
555         } else {
556             qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
557                                  2, vms->memmap[VIRT_GIC_DIST].base,
558                                  2, vms->memmap[VIRT_GIC_DIST].size,
559                                  2, vms->memmap[VIRT_GIC_REDIST].base,
560                                  2, vms->memmap[VIRT_GIC_REDIST].size,
561                                  2, vms->memmap[VIRT_HIGH_GIC_REDIST2].base,
562                                  2, vms->memmap[VIRT_HIGH_GIC_REDIST2].size);
563         }
564 
565         if (vms->virt) {
566             qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
567                                    GIC_FDT_IRQ_TYPE_PPI, ARCH_GIC_MAINT_IRQ,
568                                    GIC_FDT_IRQ_FLAGS_LEVEL_HI);
569         }
570     } else {
571         /* 'cortex-a15-gic' means 'GIC v2' */
572         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
573                                 "arm,cortex-a15-gic");
574         if (!vms->virt) {
575             qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
576                                          2, vms->memmap[VIRT_GIC_DIST].base,
577                                          2, vms->memmap[VIRT_GIC_DIST].size,
578                                          2, vms->memmap[VIRT_GIC_CPU].base,
579                                          2, vms->memmap[VIRT_GIC_CPU].size);
580         } else {
581             qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
582                                          2, vms->memmap[VIRT_GIC_DIST].base,
583                                          2, vms->memmap[VIRT_GIC_DIST].size,
584                                          2, vms->memmap[VIRT_GIC_CPU].base,
585                                          2, vms->memmap[VIRT_GIC_CPU].size,
586                                          2, vms->memmap[VIRT_GIC_HYP].base,
587                                          2, vms->memmap[VIRT_GIC_HYP].size,
588                                          2, vms->memmap[VIRT_GIC_VCPU].base,
589                                          2, vms->memmap[VIRT_GIC_VCPU].size);
590             qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
591                                    GIC_FDT_IRQ_TYPE_PPI, ARCH_GIC_MAINT_IRQ,
592                                    GIC_FDT_IRQ_FLAGS_LEVEL_HI);
593         }
594     }
595 
596     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", vms->gic_phandle);
597     g_free(nodename);
598 }
599 
600 static void fdt_add_pmu_nodes(const VirtMachineState *vms)
601 {
602     ARMCPU *armcpu = ARM_CPU(first_cpu);
603     uint32_t irqflags = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
604     MachineState *ms = MACHINE(vms);
605 
606     if (!arm_feature(&armcpu->env, ARM_FEATURE_PMU)) {
607         assert(!object_property_get_bool(OBJECT(armcpu), "pmu", NULL));
608         return;
609     }
610 
611     if (vms->gic_version == VIRT_GIC_VERSION_2) {
612         irqflags = deposit32(irqflags, GIC_FDT_IRQ_PPI_CPU_START,
613                              GIC_FDT_IRQ_PPI_CPU_WIDTH,
614                              (1 << MACHINE(vms)->smp.cpus) - 1);
615     }
616 
617     qemu_fdt_add_subnode(ms->fdt, "/pmu");
618     if (arm_feature(&armcpu->env, ARM_FEATURE_V8)) {
619         const char compat[] = "arm,armv8-pmuv3";
620         qemu_fdt_setprop(ms->fdt, "/pmu", "compatible",
621                          compat, sizeof(compat));
622         qemu_fdt_setprop_cells(ms->fdt, "/pmu", "interrupts",
623                                GIC_FDT_IRQ_TYPE_PPI, VIRTUAL_PMU_IRQ, irqflags);
624     }
625 }
626 
627 static inline DeviceState *create_acpi_ged(VirtMachineState *vms)
628 {
629     DeviceState *dev;
630     MachineState *ms = MACHINE(vms);
631     int irq = vms->irqmap[VIRT_ACPI_GED];
632     uint32_t event = ACPI_GED_PWR_DOWN_EVT;
633 
634     if (ms->ram_slots) {
635         event |= ACPI_GED_MEM_HOTPLUG_EVT;
636     }
637 
638     if (ms->nvdimms_state->is_enabled) {
639         event |= ACPI_GED_NVDIMM_HOTPLUG_EVT;
640     }
641 
642     dev = qdev_new(TYPE_ACPI_GED);
643     qdev_prop_set_uint32(dev, "ged-event", event);
644 
645     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_ACPI_GED].base);
646     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 1, vms->memmap[VIRT_PCDIMM_ACPI].base);
647     sysbus_connect_irq(SYS_BUS_DEVICE(dev), 0, qdev_get_gpio_in(vms->gic, irq));
648 
649     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
650 
651     return dev;
652 }
653 
654 static void create_its(VirtMachineState *vms)
655 {
656     const char *itsclass = its_class_name();
657     DeviceState *dev;
658 
659     if (!strcmp(itsclass, "arm-gicv3-its")) {
660         if (!vms->tcg_its) {
661             itsclass = NULL;
662         }
663     }
664 
665     if (!itsclass) {
666         /* Do nothing if not supported */
667         return;
668     }
669 
670     dev = qdev_new(itsclass);
671 
672     object_property_set_link(OBJECT(dev), "parent-gicv3", OBJECT(vms->gic),
673                              &error_abort);
674     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
675     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_GIC_ITS].base);
676 
677     fdt_add_its_gic_node(vms);
678     vms->msi_controller = VIRT_MSI_CTRL_ITS;
679 }
680 
681 static void create_v2m(VirtMachineState *vms)
682 {
683     int i;
684     int irq = vms->irqmap[VIRT_GIC_V2M];
685     DeviceState *dev;
686 
687     dev = qdev_new("arm-gicv2m");
688     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_GIC_V2M].base);
689     qdev_prop_set_uint32(dev, "base-spi", irq);
690     qdev_prop_set_uint32(dev, "num-spi", NUM_GICV2M_SPIS);
691     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
692 
693     for (i = 0; i < NUM_GICV2M_SPIS; i++) {
694         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
695                            qdev_get_gpio_in(vms->gic, irq + i));
696     }
697 
698     fdt_add_v2m_gic_node(vms);
699     vms->msi_controller = VIRT_MSI_CTRL_GICV2M;
700 }
701 
702 static void create_gic(VirtMachineState *vms, MemoryRegion *mem)
703 {
704     MachineState *ms = MACHINE(vms);
705     /* We create a standalone GIC */
706     SysBusDevice *gicbusdev;
707     const char *gictype;
708     int i;
709     unsigned int smp_cpus = ms->smp.cpus;
710     uint32_t nb_redist_regions = 0;
711     int revision;
712 
713     if (vms->gic_version == VIRT_GIC_VERSION_2) {
714         gictype = gic_class_name();
715     } else {
716         gictype = gicv3_class_name();
717     }
718 
719     switch (vms->gic_version) {
720     case VIRT_GIC_VERSION_2:
721         revision = 2;
722         break;
723     case VIRT_GIC_VERSION_3:
724         revision = 3;
725         break;
726     case VIRT_GIC_VERSION_4:
727         revision = 4;
728         break;
729     default:
730         g_assert_not_reached();
731     }
732     vms->gic = qdev_new(gictype);
733     qdev_prop_set_uint32(vms->gic, "revision", revision);
734     qdev_prop_set_uint32(vms->gic, "num-cpu", smp_cpus);
735     /* Note that the num-irq property counts both internal and external
736      * interrupts; there are always 32 of the former (mandated by GIC spec).
737      */
738     qdev_prop_set_uint32(vms->gic, "num-irq", NUM_IRQS + 32);
739     if (!kvm_irqchip_in_kernel()) {
740         qdev_prop_set_bit(vms->gic, "has-security-extensions", vms->secure);
741     }
742 
743     if (vms->gic_version != VIRT_GIC_VERSION_2) {
744         uint32_t redist0_capacity = virt_redist_capacity(vms, VIRT_GIC_REDIST);
745         uint32_t redist0_count = MIN(smp_cpus, redist0_capacity);
746 
747         nb_redist_regions = virt_gicv3_redist_region_count(vms);
748 
749         qdev_prop_set_uint32(vms->gic, "len-redist-region-count",
750                              nb_redist_regions);
751         qdev_prop_set_uint32(vms->gic, "redist-region-count[0]", redist0_count);
752 
753         if (!kvm_irqchip_in_kernel()) {
754             if (vms->tcg_its) {
755                 object_property_set_link(OBJECT(vms->gic), "sysmem",
756                                          OBJECT(mem), &error_fatal);
757                 qdev_prop_set_bit(vms->gic, "has-lpi", true);
758             }
759         }
760 
761         if (nb_redist_regions == 2) {
762             uint32_t redist1_capacity =
763                 virt_redist_capacity(vms, VIRT_HIGH_GIC_REDIST2);
764 
765             qdev_prop_set_uint32(vms->gic, "redist-region-count[1]",
766                 MIN(smp_cpus - redist0_count, redist1_capacity));
767         }
768     } else {
769         if (!kvm_irqchip_in_kernel()) {
770             qdev_prop_set_bit(vms->gic, "has-virtualization-extensions",
771                               vms->virt);
772         }
773     }
774     gicbusdev = SYS_BUS_DEVICE(vms->gic);
775     sysbus_realize_and_unref(gicbusdev, &error_fatal);
776     sysbus_mmio_map(gicbusdev, 0, vms->memmap[VIRT_GIC_DIST].base);
777     if (vms->gic_version != VIRT_GIC_VERSION_2) {
778         sysbus_mmio_map(gicbusdev, 1, vms->memmap[VIRT_GIC_REDIST].base);
779         if (nb_redist_regions == 2) {
780             sysbus_mmio_map(gicbusdev, 2,
781                             vms->memmap[VIRT_HIGH_GIC_REDIST2].base);
782         }
783     } else {
784         sysbus_mmio_map(gicbusdev, 1, vms->memmap[VIRT_GIC_CPU].base);
785         if (vms->virt) {
786             sysbus_mmio_map(gicbusdev, 2, vms->memmap[VIRT_GIC_HYP].base);
787             sysbus_mmio_map(gicbusdev, 3, vms->memmap[VIRT_GIC_VCPU].base);
788         }
789     }
790 
791     /* Wire the outputs from each CPU's generic timer and the GICv3
792      * maintenance interrupt signal to the appropriate GIC PPI inputs,
793      * and the GIC's IRQ/FIQ/VIRQ/VFIQ interrupt outputs to the CPU's inputs.
794      */
795     for (i = 0; i < smp_cpus; i++) {
796         DeviceState *cpudev = DEVICE(qemu_get_cpu(i));
797         int ppibase = NUM_IRQS + i * GIC_INTERNAL + GIC_NR_SGIS;
798         int irq;
799         /* Mapping from the output timer irq lines from the CPU to the
800          * GIC PPI inputs we use for the virt board.
801          */
802         const int timer_irq[] = {
803             [GTIMER_PHYS] = ARCH_TIMER_NS_EL1_IRQ,
804             [GTIMER_VIRT] = ARCH_TIMER_VIRT_IRQ,
805             [GTIMER_HYP]  = ARCH_TIMER_NS_EL2_IRQ,
806             [GTIMER_SEC]  = ARCH_TIMER_S_EL1_IRQ,
807         };
808 
809         for (irq = 0; irq < ARRAY_SIZE(timer_irq); irq++) {
810             qdev_connect_gpio_out(cpudev, irq,
811                                   qdev_get_gpio_in(vms->gic,
812                                                    ppibase + timer_irq[irq]));
813         }
814 
815         if (vms->gic_version != VIRT_GIC_VERSION_2) {
816             qemu_irq irq = qdev_get_gpio_in(vms->gic,
817                                             ppibase + ARCH_GIC_MAINT_IRQ);
818             qdev_connect_gpio_out_named(cpudev, "gicv3-maintenance-interrupt",
819                                         0, irq);
820         } else if (vms->virt) {
821             qemu_irq irq = qdev_get_gpio_in(vms->gic,
822                                             ppibase + ARCH_GIC_MAINT_IRQ);
823             sysbus_connect_irq(gicbusdev, i + 4 * smp_cpus, irq);
824         }
825 
826         qdev_connect_gpio_out_named(cpudev, "pmu-interrupt", 0,
827                                     qdev_get_gpio_in(vms->gic, ppibase
828                                                      + VIRTUAL_PMU_IRQ));
829 
830         sysbus_connect_irq(gicbusdev, i, qdev_get_gpio_in(cpudev, ARM_CPU_IRQ));
831         sysbus_connect_irq(gicbusdev, i + smp_cpus,
832                            qdev_get_gpio_in(cpudev, ARM_CPU_FIQ));
833         sysbus_connect_irq(gicbusdev, i + 2 * smp_cpus,
834                            qdev_get_gpio_in(cpudev, ARM_CPU_VIRQ));
835         sysbus_connect_irq(gicbusdev, i + 3 * smp_cpus,
836                            qdev_get_gpio_in(cpudev, ARM_CPU_VFIQ));
837     }
838 
839     fdt_add_gic_node(vms);
840 
841     if (vms->gic_version != VIRT_GIC_VERSION_2 && vms->its) {
842         create_its(vms);
843     } else if (vms->gic_version == VIRT_GIC_VERSION_2) {
844         create_v2m(vms);
845     }
846 }
847 
848 static void create_uart(const VirtMachineState *vms, int uart,
849                         MemoryRegion *mem, Chardev *chr)
850 {
851     char *nodename;
852     hwaddr base = vms->memmap[uart].base;
853     hwaddr size = vms->memmap[uart].size;
854     int irq = vms->irqmap[uart];
855     const char compat[] = "arm,pl011\0arm,primecell";
856     const char clocknames[] = "uartclk\0apb_pclk";
857     DeviceState *dev = qdev_new(TYPE_PL011);
858     SysBusDevice *s = SYS_BUS_DEVICE(dev);
859     MachineState *ms = MACHINE(vms);
860 
861     qdev_prop_set_chr(dev, "chardev", chr);
862     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
863     memory_region_add_subregion(mem, base,
864                                 sysbus_mmio_get_region(s, 0));
865     sysbus_connect_irq(s, 0, qdev_get_gpio_in(vms->gic, irq));
866 
867     nodename = g_strdup_printf("/pl011@%" PRIx64, base);
868     qemu_fdt_add_subnode(ms->fdt, nodename);
869     /* Note that we can't use setprop_string because of the embedded NUL */
870     qemu_fdt_setprop(ms->fdt, nodename, "compatible",
871                          compat, sizeof(compat));
872     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
873                                      2, base, 2, size);
874     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
875                                GIC_FDT_IRQ_TYPE_SPI, irq,
876                                GIC_FDT_IRQ_FLAGS_LEVEL_HI);
877     qemu_fdt_setprop_cells(ms->fdt, nodename, "clocks",
878                                vms->clock_phandle, vms->clock_phandle);
879     qemu_fdt_setprop(ms->fdt, nodename, "clock-names",
880                          clocknames, sizeof(clocknames));
881 
882     if (uart == VIRT_UART) {
883         qemu_fdt_setprop_string(ms->fdt, "/chosen", "stdout-path", nodename);
884     } else {
885         /* Mark as not usable by the normal world */
886         qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
887         qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
888 
889         qemu_fdt_setprop_string(ms->fdt, "/secure-chosen", "stdout-path",
890                                 nodename);
891     }
892 
893     g_free(nodename);
894 }
895 
896 static void create_rtc(const VirtMachineState *vms)
897 {
898     char *nodename;
899     hwaddr base = vms->memmap[VIRT_RTC].base;
900     hwaddr size = vms->memmap[VIRT_RTC].size;
901     int irq = vms->irqmap[VIRT_RTC];
902     const char compat[] = "arm,pl031\0arm,primecell";
903     MachineState *ms = MACHINE(vms);
904 
905     sysbus_create_simple("pl031", base, qdev_get_gpio_in(vms->gic, irq));
906 
907     nodename = g_strdup_printf("/pl031@%" PRIx64, base);
908     qemu_fdt_add_subnode(ms->fdt, nodename);
909     qemu_fdt_setprop(ms->fdt, nodename, "compatible", compat, sizeof(compat));
910     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
911                                  2, base, 2, size);
912     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
913                            GIC_FDT_IRQ_TYPE_SPI, irq,
914                            GIC_FDT_IRQ_FLAGS_LEVEL_HI);
915     qemu_fdt_setprop_cell(ms->fdt, nodename, "clocks", vms->clock_phandle);
916     qemu_fdt_setprop_string(ms->fdt, nodename, "clock-names", "apb_pclk");
917     g_free(nodename);
918 }
919 
920 static DeviceState *gpio_key_dev;
921 static void virt_powerdown_req(Notifier *n, void *opaque)
922 {
923     VirtMachineState *s = container_of(n, VirtMachineState, powerdown_notifier);
924 
925     if (s->acpi_dev) {
926         acpi_send_event(s->acpi_dev, ACPI_POWER_DOWN_STATUS);
927     } else {
928         /* use gpio Pin 3 for power button event */
929         qemu_set_irq(qdev_get_gpio_in(gpio_key_dev, 0), 1);
930     }
931 }
932 
933 static void create_gpio_keys(char *fdt, DeviceState *pl061_dev,
934                              uint32_t phandle)
935 {
936     gpio_key_dev = sysbus_create_simple("gpio-key", -1,
937                                         qdev_get_gpio_in(pl061_dev, 3));
938 
939     qemu_fdt_add_subnode(fdt, "/gpio-keys");
940     qemu_fdt_setprop_string(fdt, "/gpio-keys", "compatible", "gpio-keys");
941 
942     qemu_fdt_add_subnode(fdt, "/gpio-keys/poweroff");
943     qemu_fdt_setprop_string(fdt, "/gpio-keys/poweroff",
944                             "label", "GPIO Key Poweroff");
945     qemu_fdt_setprop_cell(fdt, "/gpio-keys/poweroff", "linux,code",
946                           KEY_POWER);
947     qemu_fdt_setprop_cells(fdt, "/gpio-keys/poweroff",
948                            "gpios", phandle, 3, 0);
949 }
950 
951 #define SECURE_GPIO_POWEROFF 0
952 #define SECURE_GPIO_RESET    1
953 
954 static void create_secure_gpio_pwr(char *fdt, DeviceState *pl061_dev,
955                                    uint32_t phandle)
956 {
957     DeviceState *gpio_pwr_dev;
958 
959     /* gpio-pwr */
960     gpio_pwr_dev = sysbus_create_simple("gpio-pwr", -1, NULL);
961 
962     /* connect secure pl061 to gpio-pwr */
963     qdev_connect_gpio_out(pl061_dev, SECURE_GPIO_RESET,
964                           qdev_get_gpio_in_named(gpio_pwr_dev, "reset", 0));
965     qdev_connect_gpio_out(pl061_dev, SECURE_GPIO_POWEROFF,
966                           qdev_get_gpio_in_named(gpio_pwr_dev, "shutdown", 0));
967 
968     qemu_fdt_add_subnode(fdt, "/gpio-poweroff");
969     qemu_fdt_setprop_string(fdt, "/gpio-poweroff", "compatible",
970                             "gpio-poweroff");
971     qemu_fdt_setprop_cells(fdt, "/gpio-poweroff",
972                            "gpios", phandle, SECURE_GPIO_POWEROFF, 0);
973     qemu_fdt_setprop_string(fdt, "/gpio-poweroff", "status", "disabled");
974     qemu_fdt_setprop_string(fdt, "/gpio-poweroff", "secure-status",
975                             "okay");
976 
977     qemu_fdt_add_subnode(fdt, "/gpio-restart");
978     qemu_fdt_setprop_string(fdt, "/gpio-restart", "compatible",
979                             "gpio-restart");
980     qemu_fdt_setprop_cells(fdt, "/gpio-restart",
981                            "gpios", phandle, SECURE_GPIO_RESET, 0);
982     qemu_fdt_setprop_string(fdt, "/gpio-restart", "status", "disabled");
983     qemu_fdt_setprop_string(fdt, "/gpio-restart", "secure-status",
984                             "okay");
985 }
986 
987 static void create_gpio_devices(const VirtMachineState *vms, int gpio,
988                                 MemoryRegion *mem)
989 {
990     char *nodename;
991     DeviceState *pl061_dev;
992     hwaddr base = vms->memmap[gpio].base;
993     hwaddr size = vms->memmap[gpio].size;
994     int irq = vms->irqmap[gpio];
995     const char compat[] = "arm,pl061\0arm,primecell";
996     SysBusDevice *s;
997     MachineState *ms = MACHINE(vms);
998 
999     pl061_dev = qdev_new("pl061");
1000     /* Pull lines down to 0 if not driven by the PL061 */
1001     qdev_prop_set_uint32(pl061_dev, "pullups", 0);
1002     qdev_prop_set_uint32(pl061_dev, "pulldowns", 0xff);
1003     s = SYS_BUS_DEVICE(pl061_dev);
1004     sysbus_realize_and_unref(s, &error_fatal);
1005     memory_region_add_subregion(mem, base, sysbus_mmio_get_region(s, 0));
1006     sysbus_connect_irq(s, 0, qdev_get_gpio_in(vms->gic, irq));
1007 
1008     uint32_t phandle = qemu_fdt_alloc_phandle(ms->fdt);
1009     nodename = g_strdup_printf("/pl061@%" PRIx64, base);
1010     qemu_fdt_add_subnode(ms->fdt, nodename);
1011     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1012                                  2, base, 2, size);
1013     qemu_fdt_setprop(ms->fdt, nodename, "compatible", compat, sizeof(compat));
1014     qemu_fdt_setprop_cell(ms->fdt, nodename, "#gpio-cells", 2);
1015     qemu_fdt_setprop(ms->fdt, nodename, "gpio-controller", NULL, 0);
1016     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
1017                            GIC_FDT_IRQ_TYPE_SPI, irq,
1018                            GIC_FDT_IRQ_FLAGS_LEVEL_HI);
1019     qemu_fdt_setprop_cell(ms->fdt, nodename, "clocks", vms->clock_phandle);
1020     qemu_fdt_setprop_string(ms->fdt, nodename, "clock-names", "apb_pclk");
1021     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", phandle);
1022 
1023     if (gpio != VIRT_GPIO) {
1024         /* Mark as not usable by the normal world */
1025         qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
1026         qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
1027     }
1028     g_free(nodename);
1029 
1030     /* Child gpio devices */
1031     if (gpio == VIRT_GPIO) {
1032         create_gpio_keys(ms->fdt, pl061_dev, phandle);
1033     } else {
1034         create_secure_gpio_pwr(ms->fdt, pl061_dev, phandle);
1035     }
1036 }
1037 
1038 static void create_virtio_devices(const VirtMachineState *vms)
1039 {
1040     int i;
1041     hwaddr size = vms->memmap[VIRT_MMIO].size;
1042     MachineState *ms = MACHINE(vms);
1043 
1044     /* We create the transports in forwards order. Since qbus_realize()
1045      * prepends (not appends) new child buses, the incrementing loop below will
1046      * create a list of virtio-mmio buses with decreasing base addresses.
1047      *
1048      * When a -device option is processed from the command line,
1049      * qbus_find_recursive() picks the next free virtio-mmio bus in forwards
1050      * order. The upshot is that -device options in increasing command line
1051      * order are mapped to virtio-mmio buses with decreasing base addresses.
1052      *
1053      * When this code was originally written, that arrangement ensured that the
1054      * guest Linux kernel would give the lowest "name" (/dev/vda, eth0, etc) to
1055      * the first -device on the command line. (The end-to-end order is a
1056      * function of this loop, qbus_realize(), qbus_find_recursive(), and the
1057      * guest kernel's name-to-address assignment strategy.)
1058      *
1059      * Meanwhile, the kernel's traversal seems to have been reversed; see eg.
1060      * the message, if not necessarily the code, of commit 70161ff336.
1061      * Therefore the loop now establishes the inverse of the original intent.
1062      *
1063      * Unfortunately, we can't counteract the kernel change by reversing the
1064      * loop; it would break existing command lines.
1065      *
1066      * In any case, the kernel makes no guarantee about the stability of
1067      * enumeration order of virtio devices (as demonstrated by it changing
1068      * between kernel versions). For reliable and stable identification
1069      * of disks users must use UUIDs or similar mechanisms.
1070      */
1071     for (i = 0; i < NUM_VIRTIO_TRANSPORTS; i++) {
1072         int irq = vms->irqmap[VIRT_MMIO] + i;
1073         hwaddr base = vms->memmap[VIRT_MMIO].base + i * size;
1074 
1075         sysbus_create_simple("virtio-mmio", base,
1076                              qdev_get_gpio_in(vms->gic, irq));
1077     }
1078 
1079     /* We add dtb nodes in reverse order so that they appear in the finished
1080      * device tree lowest address first.
1081      *
1082      * Note that this mapping is independent of the loop above. The previous
1083      * loop influences virtio device to virtio transport assignment, whereas
1084      * this loop controls how virtio transports are laid out in the dtb.
1085      */
1086     for (i = NUM_VIRTIO_TRANSPORTS - 1; i >= 0; i--) {
1087         char *nodename;
1088         int irq = vms->irqmap[VIRT_MMIO] + i;
1089         hwaddr base = vms->memmap[VIRT_MMIO].base + i * size;
1090 
1091         nodename = g_strdup_printf("/virtio_mmio@%" PRIx64, base);
1092         qemu_fdt_add_subnode(ms->fdt, nodename);
1093         qemu_fdt_setprop_string(ms->fdt, nodename,
1094                                 "compatible", "virtio,mmio");
1095         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1096                                      2, base, 2, size);
1097         qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
1098                                GIC_FDT_IRQ_TYPE_SPI, irq,
1099                                GIC_FDT_IRQ_FLAGS_EDGE_LO_HI);
1100         qemu_fdt_setprop(ms->fdt, nodename, "dma-coherent", NULL, 0);
1101         g_free(nodename);
1102     }
1103 }
1104 
1105 #define VIRT_FLASH_SECTOR_SIZE (256 * KiB)
1106 
1107 static PFlashCFI01 *virt_flash_create1(VirtMachineState *vms,
1108                                         const char *name,
1109                                         const char *alias_prop_name)
1110 {
1111     /*
1112      * Create a single flash device.  We use the same parameters as
1113      * the flash devices on the Versatile Express board.
1114      */
1115     DeviceState *dev = qdev_new(TYPE_PFLASH_CFI01);
1116 
1117     qdev_prop_set_uint64(dev, "sector-length", VIRT_FLASH_SECTOR_SIZE);
1118     qdev_prop_set_uint8(dev, "width", 4);
1119     qdev_prop_set_uint8(dev, "device-width", 2);
1120     qdev_prop_set_bit(dev, "big-endian", false);
1121     qdev_prop_set_uint16(dev, "id0", 0x89);
1122     qdev_prop_set_uint16(dev, "id1", 0x18);
1123     qdev_prop_set_uint16(dev, "id2", 0x00);
1124     qdev_prop_set_uint16(dev, "id3", 0x00);
1125     qdev_prop_set_string(dev, "name", name);
1126     object_property_add_child(OBJECT(vms), name, OBJECT(dev));
1127     object_property_add_alias(OBJECT(vms), alias_prop_name,
1128                               OBJECT(dev), "drive");
1129     return PFLASH_CFI01(dev);
1130 }
1131 
1132 static void virt_flash_create(VirtMachineState *vms)
1133 {
1134     vms->flash[0] = virt_flash_create1(vms, "virt.flash0", "pflash0");
1135     vms->flash[1] = virt_flash_create1(vms, "virt.flash1", "pflash1");
1136 }
1137 
1138 static void virt_flash_map1(PFlashCFI01 *flash,
1139                             hwaddr base, hwaddr size,
1140                             MemoryRegion *sysmem)
1141 {
1142     DeviceState *dev = DEVICE(flash);
1143 
1144     assert(QEMU_IS_ALIGNED(size, VIRT_FLASH_SECTOR_SIZE));
1145     assert(size / VIRT_FLASH_SECTOR_SIZE <= UINT32_MAX);
1146     qdev_prop_set_uint32(dev, "num-blocks", size / VIRT_FLASH_SECTOR_SIZE);
1147     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1148 
1149     memory_region_add_subregion(sysmem, base,
1150                                 sysbus_mmio_get_region(SYS_BUS_DEVICE(dev),
1151                                                        0));
1152 }
1153 
1154 static void virt_flash_map(VirtMachineState *vms,
1155                            MemoryRegion *sysmem,
1156                            MemoryRegion *secure_sysmem)
1157 {
1158     /*
1159      * Map two flash devices to fill the VIRT_FLASH space in the memmap.
1160      * sysmem is the system memory space. secure_sysmem is the secure view
1161      * of the system, and the first flash device should be made visible only
1162      * there. The second flash device is visible to both secure and nonsecure.
1163      * If sysmem == secure_sysmem this means there is no separate Secure
1164      * address space and both flash devices are generally visible.
1165      */
1166     hwaddr flashsize = vms->memmap[VIRT_FLASH].size / 2;
1167     hwaddr flashbase = vms->memmap[VIRT_FLASH].base;
1168 
1169     virt_flash_map1(vms->flash[0], flashbase, flashsize,
1170                     secure_sysmem);
1171     virt_flash_map1(vms->flash[1], flashbase + flashsize, flashsize,
1172                     sysmem);
1173 }
1174 
1175 static void virt_flash_fdt(VirtMachineState *vms,
1176                            MemoryRegion *sysmem,
1177                            MemoryRegion *secure_sysmem)
1178 {
1179     hwaddr flashsize = vms->memmap[VIRT_FLASH].size / 2;
1180     hwaddr flashbase = vms->memmap[VIRT_FLASH].base;
1181     MachineState *ms = MACHINE(vms);
1182     char *nodename;
1183 
1184     if (sysmem == secure_sysmem) {
1185         /* Report both flash devices as a single node in the DT */
1186         nodename = g_strdup_printf("/flash@%" PRIx64, flashbase);
1187         qemu_fdt_add_subnode(ms->fdt, nodename);
1188         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible", "cfi-flash");
1189         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1190                                      2, flashbase, 2, flashsize,
1191                                      2, flashbase + flashsize, 2, flashsize);
1192         qemu_fdt_setprop_cell(ms->fdt, nodename, "bank-width", 4);
1193         g_free(nodename);
1194     } else {
1195         /*
1196          * Report the devices as separate nodes so we can mark one as
1197          * only visible to the secure world.
1198          */
1199         nodename = g_strdup_printf("/secflash@%" PRIx64, flashbase);
1200         qemu_fdt_add_subnode(ms->fdt, nodename);
1201         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible", "cfi-flash");
1202         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1203                                      2, flashbase, 2, flashsize);
1204         qemu_fdt_setprop_cell(ms->fdt, nodename, "bank-width", 4);
1205         qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
1206         qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
1207         g_free(nodename);
1208 
1209         nodename = g_strdup_printf("/flash@%" PRIx64, flashbase + flashsize);
1210         qemu_fdt_add_subnode(ms->fdt, nodename);
1211         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible", "cfi-flash");
1212         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1213                                      2, flashbase + flashsize, 2, flashsize);
1214         qemu_fdt_setprop_cell(ms->fdt, nodename, "bank-width", 4);
1215         g_free(nodename);
1216     }
1217 }
1218 
1219 static bool virt_firmware_init(VirtMachineState *vms,
1220                                MemoryRegion *sysmem,
1221                                MemoryRegion *secure_sysmem)
1222 {
1223     int i;
1224     const char *bios_name;
1225     BlockBackend *pflash_blk0;
1226 
1227     /* Map legacy -drive if=pflash to machine properties */
1228     for (i = 0; i < ARRAY_SIZE(vms->flash); i++) {
1229         pflash_cfi01_legacy_drive(vms->flash[i],
1230                                   drive_get(IF_PFLASH, 0, i));
1231     }
1232 
1233     virt_flash_map(vms, sysmem, secure_sysmem);
1234 
1235     pflash_blk0 = pflash_cfi01_get_blk(vms->flash[0]);
1236 
1237     bios_name = MACHINE(vms)->firmware;
1238     if (bios_name) {
1239         char *fname;
1240         MemoryRegion *mr;
1241         int image_size;
1242 
1243         if (pflash_blk0) {
1244             error_report("The contents of the first flash device may be "
1245                          "specified with -bios or with -drive if=pflash... "
1246                          "but you cannot use both options at once");
1247             exit(1);
1248         }
1249 
1250         /* Fall back to -bios */
1251 
1252         fname = qemu_find_file(QEMU_FILE_TYPE_BIOS, bios_name);
1253         if (!fname) {
1254             error_report("Could not find ROM image '%s'", bios_name);
1255             exit(1);
1256         }
1257         mr = sysbus_mmio_get_region(SYS_BUS_DEVICE(vms->flash[0]), 0);
1258         image_size = load_image_mr(fname, mr);
1259         g_free(fname);
1260         if (image_size < 0) {
1261             error_report("Could not load ROM image '%s'", bios_name);
1262             exit(1);
1263         }
1264     }
1265 
1266     return pflash_blk0 || bios_name;
1267 }
1268 
1269 static FWCfgState *create_fw_cfg(const VirtMachineState *vms, AddressSpace *as)
1270 {
1271     MachineState *ms = MACHINE(vms);
1272     hwaddr base = vms->memmap[VIRT_FW_CFG].base;
1273     hwaddr size = vms->memmap[VIRT_FW_CFG].size;
1274     FWCfgState *fw_cfg;
1275     char *nodename;
1276 
1277     fw_cfg = fw_cfg_init_mem_wide(base + 8, base, 8, base + 16, as);
1278     fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, (uint16_t)ms->smp.cpus);
1279 
1280     nodename = g_strdup_printf("/fw-cfg@%" PRIx64, base);
1281     qemu_fdt_add_subnode(ms->fdt, nodename);
1282     qemu_fdt_setprop_string(ms->fdt, nodename,
1283                             "compatible", "qemu,fw-cfg-mmio");
1284     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1285                                  2, base, 2, size);
1286     qemu_fdt_setprop(ms->fdt, nodename, "dma-coherent", NULL, 0);
1287     g_free(nodename);
1288     return fw_cfg;
1289 }
1290 
1291 static void create_pcie_irq_map(const MachineState *ms,
1292                                 uint32_t gic_phandle,
1293                                 int first_irq, const char *nodename)
1294 {
1295     int devfn, pin;
1296     uint32_t full_irq_map[4 * 4 * 10] = { 0 };
1297     uint32_t *irq_map = full_irq_map;
1298 
1299     for (devfn = 0; devfn <= 0x18; devfn += 0x8) {
1300         for (pin = 0; pin < 4; pin++) {
1301             int irq_type = GIC_FDT_IRQ_TYPE_SPI;
1302             int irq_nr = first_irq + ((pin + PCI_SLOT(devfn)) % PCI_NUM_PINS);
1303             int irq_level = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
1304             int i;
1305 
1306             uint32_t map[] = {
1307                 devfn << 8, 0, 0,                           /* devfn */
1308                 pin + 1,                                    /* PCI pin */
1309                 gic_phandle, 0, 0, irq_type, irq_nr, irq_level }; /* GIC irq */
1310 
1311             /* Convert map to big endian */
1312             for (i = 0; i < 10; i++) {
1313                 irq_map[i] = cpu_to_be32(map[i]);
1314             }
1315             irq_map += 10;
1316         }
1317     }
1318 
1319     qemu_fdt_setprop(ms->fdt, nodename, "interrupt-map",
1320                      full_irq_map, sizeof(full_irq_map));
1321 
1322     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupt-map-mask",
1323                            cpu_to_be16(PCI_DEVFN(3, 0)), /* Slot 3 */
1324                            0, 0,
1325                            0x7           /* PCI irq */);
1326 }
1327 
1328 static void create_smmu(const VirtMachineState *vms,
1329                         PCIBus *bus)
1330 {
1331     char *node;
1332     const char compat[] = "arm,smmu-v3";
1333     int irq =  vms->irqmap[VIRT_SMMU];
1334     int i;
1335     hwaddr base = vms->memmap[VIRT_SMMU].base;
1336     hwaddr size = vms->memmap[VIRT_SMMU].size;
1337     const char irq_names[] = "eventq\0priq\0cmdq-sync\0gerror";
1338     DeviceState *dev;
1339     MachineState *ms = MACHINE(vms);
1340 
1341     if (vms->iommu != VIRT_IOMMU_SMMUV3 || !vms->iommu_phandle) {
1342         return;
1343     }
1344 
1345     dev = qdev_new("arm-smmuv3");
1346 
1347     object_property_set_link(OBJECT(dev), "primary-bus", OBJECT(bus),
1348                              &error_abort);
1349     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1350     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, base);
1351     for (i = 0; i < NUM_SMMU_IRQS; i++) {
1352         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
1353                            qdev_get_gpio_in(vms->gic, irq + i));
1354     }
1355 
1356     node = g_strdup_printf("/smmuv3@%" PRIx64, base);
1357     qemu_fdt_add_subnode(ms->fdt, node);
1358     qemu_fdt_setprop(ms->fdt, node, "compatible", compat, sizeof(compat));
1359     qemu_fdt_setprop_sized_cells(ms->fdt, node, "reg", 2, base, 2, size);
1360 
1361     qemu_fdt_setprop_cells(ms->fdt, node, "interrupts",
1362             GIC_FDT_IRQ_TYPE_SPI, irq    , GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1363             GIC_FDT_IRQ_TYPE_SPI, irq + 1, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1364             GIC_FDT_IRQ_TYPE_SPI, irq + 2, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1365             GIC_FDT_IRQ_TYPE_SPI, irq + 3, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI);
1366 
1367     qemu_fdt_setprop(ms->fdt, node, "interrupt-names", irq_names,
1368                      sizeof(irq_names));
1369 
1370     qemu_fdt_setprop(ms->fdt, node, "dma-coherent", NULL, 0);
1371 
1372     qemu_fdt_setprop_cell(ms->fdt, node, "#iommu-cells", 1);
1373 
1374     qemu_fdt_setprop_cell(ms->fdt, node, "phandle", vms->iommu_phandle);
1375     g_free(node);
1376 }
1377 
1378 static void create_virtio_iommu_dt_bindings(VirtMachineState *vms)
1379 {
1380     const char compat[] = "virtio,pci-iommu\0pci1af4,1057";
1381     uint16_t bdf = vms->virtio_iommu_bdf;
1382     MachineState *ms = MACHINE(vms);
1383     char *node;
1384 
1385     vms->iommu_phandle = qemu_fdt_alloc_phandle(ms->fdt);
1386 
1387     node = g_strdup_printf("%s/virtio_iommu@%x,%x", vms->pciehb_nodename,
1388                            PCI_SLOT(bdf), PCI_FUNC(bdf));
1389     qemu_fdt_add_subnode(ms->fdt, node);
1390     qemu_fdt_setprop(ms->fdt, node, "compatible", compat, sizeof(compat));
1391     qemu_fdt_setprop_sized_cells(ms->fdt, node, "reg",
1392                                  1, bdf << 8, 1, 0, 1, 0,
1393                                  1, 0, 1, 0);
1394 
1395     qemu_fdt_setprop_cell(ms->fdt, node, "#iommu-cells", 1);
1396     qemu_fdt_setprop_cell(ms->fdt, node, "phandle", vms->iommu_phandle);
1397     g_free(node);
1398 
1399     qemu_fdt_setprop_cells(ms->fdt, vms->pciehb_nodename, "iommu-map",
1400                            0x0, vms->iommu_phandle, 0x0, bdf,
1401                            bdf + 1, vms->iommu_phandle, bdf + 1, 0xffff - bdf);
1402 }
1403 
1404 static void create_pcie(VirtMachineState *vms)
1405 {
1406     hwaddr base_mmio = vms->memmap[VIRT_PCIE_MMIO].base;
1407     hwaddr size_mmio = vms->memmap[VIRT_PCIE_MMIO].size;
1408     hwaddr base_mmio_high = vms->memmap[VIRT_HIGH_PCIE_MMIO].base;
1409     hwaddr size_mmio_high = vms->memmap[VIRT_HIGH_PCIE_MMIO].size;
1410     hwaddr base_pio = vms->memmap[VIRT_PCIE_PIO].base;
1411     hwaddr size_pio = vms->memmap[VIRT_PCIE_PIO].size;
1412     hwaddr base_ecam, size_ecam;
1413     hwaddr base = base_mmio;
1414     int nr_pcie_buses;
1415     int irq = vms->irqmap[VIRT_PCIE];
1416     MemoryRegion *mmio_alias;
1417     MemoryRegion *mmio_reg;
1418     MemoryRegion *ecam_alias;
1419     MemoryRegion *ecam_reg;
1420     DeviceState *dev;
1421     char *nodename;
1422     int i, ecam_id;
1423     PCIHostState *pci;
1424     MachineState *ms = MACHINE(vms);
1425 
1426     dev = qdev_new(TYPE_GPEX_HOST);
1427     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1428 
1429     ecam_id = VIRT_ECAM_ID(vms->highmem_ecam);
1430     base_ecam = vms->memmap[ecam_id].base;
1431     size_ecam = vms->memmap[ecam_id].size;
1432     nr_pcie_buses = size_ecam / PCIE_MMCFG_SIZE_MIN;
1433     /* Map only the first size_ecam bytes of ECAM space */
1434     ecam_alias = g_new0(MemoryRegion, 1);
1435     ecam_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 0);
1436     memory_region_init_alias(ecam_alias, OBJECT(dev), "pcie-ecam",
1437                              ecam_reg, 0, size_ecam);
1438     memory_region_add_subregion(get_system_memory(), base_ecam, ecam_alias);
1439 
1440     /* Map the MMIO window into system address space so as to expose
1441      * the section of PCI MMIO space which starts at the same base address
1442      * (ie 1:1 mapping for that part of PCI MMIO space visible through
1443      * the window).
1444      */
1445     mmio_alias = g_new0(MemoryRegion, 1);
1446     mmio_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 1);
1447     memory_region_init_alias(mmio_alias, OBJECT(dev), "pcie-mmio",
1448                              mmio_reg, base_mmio, size_mmio);
1449     memory_region_add_subregion(get_system_memory(), base_mmio, mmio_alias);
1450 
1451     if (vms->highmem_mmio) {
1452         /* Map high MMIO space */
1453         MemoryRegion *high_mmio_alias = g_new0(MemoryRegion, 1);
1454 
1455         memory_region_init_alias(high_mmio_alias, OBJECT(dev), "pcie-mmio-high",
1456                                  mmio_reg, base_mmio_high, size_mmio_high);
1457         memory_region_add_subregion(get_system_memory(), base_mmio_high,
1458                                     high_mmio_alias);
1459     }
1460 
1461     /* Map IO port space */
1462     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 2, base_pio);
1463 
1464     for (i = 0; i < GPEX_NUM_IRQS; i++) {
1465         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
1466                            qdev_get_gpio_in(vms->gic, irq + i));
1467         gpex_set_irq_num(GPEX_HOST(dev), i, irq + i);
1468     }
1469 
1470     pci = PCI_HOST_BRIDGE(dev);
1471     pci->bypass_iommu = vms->default_bus_bypass_iommu;
1472     vms->bus = pci->bus;
1473     if (vms->bus) {
1474         for (i = 0; i < nb_nics; i++) {
1475             NICInfo *nd = &nd_table[i];
1476 
1477             if (!nd->model) {
1478                 nd->model = g_strdup("virtio");
1479             }
1480 
1481             pci_nic_init_nofail(nd, pci->bus, nd->model, NULL);
1482         }
1483     }
1484 
1485     nodename = vms->pciehb_nodename = g_strdup_printf("/pcie@%" PRIx64, base);
1486     qemu_fdt_add_subnode(ms->fdt, nodename);
1487     qemu_fdt_setprop_string(ms->fdt, nodename,
1488                             "compatible", "pci-host-ecam-generic");
1489     qemu_fdt_setprop_string(ms->fdt, nodename, "device_type", "pci");
1490     qemu_fdt_setprop_cell(ms->fdt, nodename, "#address-cells", 3);
1491     qemu_fdt_setprop_cell(ms->fdt, nodename, "#size-cells", 2);
1492     qemu_fdt_setprop_cell(ms->fdt, nodename, "linux,pci-domain", 0);
1493     qemu_fdt_setprop_cells(ms->fdt, nodename, "bus-range", 0,
1494                            nr_pcie_buses - 1);
1495     qemu_fdt_setprop(ms->fdt, nodename, "dma-coherent", NULL, 0);
1496 
1497     if (vms->msi_phandle) {
1498         qemu_fdt_setprop_cells(ms->fdt, nodename, "msi-map",
1499                                0, vms->msi_phandle, 0, 0x10000);
1500     }
1501 
1502     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1503                                  2, base_ecam, 2, size_ecam);
1504 
1505     if (vms->highmem_mmio) {
1506         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "ranges",
1507                                      1, FDT_PCI_RANGE_IOPORT, 2, 0,
1508                                      2, base_pio, 2, size_pio,
1509                                      1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
1510                                      2, base_mmio, 2, size_mmio,
1511                                      1, FDT_PCI_RANGE_MMIO_64BIT,
1512                                      2, base_mmio_high,
1513                                      2, base_mmio_high, 2, size_mmio_high);
1514     } else {
1515         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "ranges",
1516                                      1, FDT_PCI_RANGE_IOPORT, 2, 0,
1517                                      2, base_pio, 2, size_pio,
1518                                      1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
1519                                      2, base_mmio, 2, size_mmio);
1520     }
1521 
1522     qemu_fdt_setprop_cell(ms->fdt, nodename, "#interrupt-cells", 1);
1523     create_pcie_irq_map(ms, vms->gic_phandle, irq, nodename);
1524 
1525     if (vms->iommu) {
1526         vms->iommu_phandle = qemu_fdt_alloc_phandle(ms->fdt);
1527 
1528         switch (vms->iommu) {
1529         case VIRT_IOMMU_SMMUV3:
1530             create_smmu(vms, vms->bus);
1531             qemu_fdt_setprop_cells(ms->fdt, nodename, "iommu-map",
1532                                    0x0, vms->iommu_phandle, 0x0, 0x10000);
1533             break;
1534         default:
1535             g_assert_not_reached();
1536         }
1537     }
1538 }
1539 
1540 static void create_platform_bus(VirtMachineState *vms)
1541 {
1542     DeviceState *dev;
1543     SysBusDevice *s;
1544     int i;
1545     MemoryRegion *sysmem = get_system_memory();
1546 
1547     dev = qdev_new(TYPE_PLATFORM_BUS_DEVICE);
1548     dev->id = g_strdup(TYPE_PLATFORM_BUS_DEVICE);
1549     qdev_prop_set_uint32(dev, "num_irqs", PLATFORM_BUS_NUM_IRQS);
1550     qdev_prop_set_uint32(dev, "mmio_size", vms->memmap[VIRT_PLATFORM_BUS].size);
1551     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1552     vms->platform_bus_dev = dev;
1553 
1554     s = SYS_BUS_DEVICE(dev);
1555     for (i = 0; i < PLATFORM_BUS_NUM_IRQS; i++) {
1556         int irq = vms->irqmap[VIRT_PLATFORM_BUS] + i;
1557         sysbus_connect_irq(s, i, qdev_get_gpio_in(vms->gic, irq));
1558     }
1559 
1560     memory_region_add_subregion(sysmem,
1561                                 vms->memmap[VIRT_PLATFORM_BUS].base,
1562                                 sysbus_mmio_get_region(s, 0));
1563 }
1564 
1565 static void create_tag_ram(MemoryRegion *tag_sysmem,
1566                            hwaddr base, hwaddr size,
1567                            const char *name)
1568 {
1569     MemoryRegion *tagram = g_new(MemoryRegion, 1);
1570 
1571     memory_region_init_ram(tagram, NULL, name, size / 32, &error_fatal);
1572     memory_region_add_subregion(tag_sysmem, base / 32, tagram);
1573 }
1574 
1575 static void create_secure_ram(VirtMachineState *vms,
1576                               MemoryRegion *secure_sysmem,
1577                               MemoryRegion *secure_tag_sysmem)
1578 {
1579     MemoryRegion *secram = g_new(MemoryRegion, 1);
1580     char *nodename;
1581     hwaddr base = vms->memmap[VIRT_SECURE_MEM].base;
1582     hwaddr size = vms->memmap[VIRT_SECURE_MEM].size;
1583     MachineState *ms = MACHINE(vms);
1584 
1585     memory_region_init_ram(secram, NULL, "virt.secure-ram", size,
1586                            &error_fatal);
1587     memory_region_add_subregion(secure_sysmem, base, secram);
1588 
1589     nodename = g_strdup_printf("/secram@%" PRIx64, base);
1590     qemu_fdt_add_subnode(ms->fdt, nodename);
1591     qemu_fdt_setprop_string(ms->fdt, nodename, "device_type", "memory");
1592     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg", 2, base, 2, size);
1593     qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
1594     qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
1595 
1596     if (secure_tag_sysmem) {
1597         create_tag_ram(secure_tag_sysmem, base, size, "mach-virt.secure-tag");
1598     }
1599 
1600     g_free(nodename);
1601 }
1602 
1603 static void *machvirt_dtb(const struct arm_boot_info *binfo, int *fdt_size)
1604 {
1605     const VirtMachineState *board = container_of(binfo, VirtMachineState,
1606                                                  bootinfo);
1607     MachineState *ms = MACHINE(board);
1608 
1609 
1610     *fdt_size = board->fdt_size;
1611     return ms->fdt;
1612 }
1613 
1614 static void virt_build_smbios(VirtMachineState *vms)
1615 {
1616     MachineClass *mc = MACHINE_GET_CLASS(vms);
1617     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
1618     uint8_t *smbios_tables, *smbios_anchor;
1619     size_t smbios_tables_len, smbios_anchor_len;
1620     const char *product = "QEMU Virtual Machine";
1621 
1622     if (kvm_enabled()) {
1623         product = "KVM Virtual Machine";
1624     }
1625 
1626     smbios_set_defaults("QEMU", product,
1627                         vmc->smbios_old_sys_ver ? "1.0" : mc->name, false,
1628                         true, SMBIOS_ENTRY_POINT_TYPE_64);
1629 
1630     smbios_get_tables(MACHINE(vms), NULL, 0,
1631                       &smbios_tables, &smbios_tables_len,
1632                       &smbios_anchor, &smbios_anchor_len,
1633                       &error_fatal);
1634 
1635     if (smbios_anchor) {
1636         fw_cfg_add_file(vms->fw_cfg, "etc/smbios/smbios-tables",
1637                         smbios_tables, smbios_tables_len);
1638         fw_cfg_add_file(vms->fw_cfg, "etc/smbios/smbios-anchor",
1639                         smbios_anchor, smbios_anchor_len);
1640     }
1641 }
1642 
1643 static
1644 void virt_machine_done(Notifier *notifier, void *data)
1645 {
1646     VirtMachineState *vms = container_of(notifier, VirtMachineState,
1647                                          machine_done);
1648     MachineState *ms = MACHINE(vms);
1649     ARMCPU *cpu = ARM_CPU(first_cpu);
1650     struct arm_boot_info *info = &vms->bootinfo;
1651     AddressSpace *as = arm_boot_address_space(cpu, info);
1652 
1653     /*
1654      * If the user provided a dtb, we assume the dynamic sysbus nodes
1655      * already are integrated there. This corresponds to a use case where
1656      * the dynamic sysbus nodes are complex and their generation is not yet
1657      * supported. In that case the user can take charge of the guest dt
1658      * while qemu takes charge of the qom stuff.
1659      */
1660     if (info->dtb_filename == NULL) {
1661         platform_bus_add_all_fdt_nodes(ms->fdt, "/intc",
1662                                        vms->memmap[VIRT_PLATFORM_BUS].base,
1663                                        vms->memmap[VIRT_PLATFORM_BUS].size,
1664                                        vms->irqmap[VIRT_PLATFORM_BUS]);
1665     }
1666     if (arm_load_dtb(info->dtb_start, info, info->dtb_limit, as, ms) < 0) {
1667         exit(1);
1668     }
1669 
1670     fw_cfg_add_extra_pci_roots(vms->bus, vms->fw_cfg);
1671 
1672     virt_acpi_setup(vms);
1673     virt_build_smbios(vms);
1674 }
1675 
1676 static uint64_t virt_cpu_mp_affinity(VirtMachineState *vms, int idx)
1677 {
1678     uint8_t clustersz = ARM_DEFAULT_CPUS_PER_CLUSTER;
1679     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
1680 
1681     if (!vmc->disallow_affinity_adjustment) {
1682         /* Adjust MPIDR like 64-bit KVM hosts, which incorporate the
1683          * GIC's target-list limitations. 32-bit KVM hosts currently
1684          * always create clusters of 4 CPUs, but that is expected to
1685          * change when they gain support for gicv3. When KVM is enabled
1686          * it will override the changes we make here, therefore our
1687          * purposes are to make TCG consistent (with 64-bit KVM hosts)
1688          * and to improve SGI efficiency.
1689          */
1690         if (vms->gic_version == VIRT_GIC_VERSION_2) {
1691             clustersz = GIC_TARGETLIST_BITS;
1692         } else {
1693             clustersz = GICV3_TARGETLIST_BITS;
1694         }
1695     }
1696     return arm_cpu_mp_affinity(idx, clustersz);
1697 }
1698 
1699 static inline bool *virt_get_high_memmap_enabled(VirtMachineState *vms,
1700                                                  int index)
1701 {
1702     bool *enabled_array[] = {
1703         &vms->highmem_redists,
1704         &vms->highmem_ecam,
1705         &vms->highmem_mmio,
1706     };
1707 
1708     assert(ARRAY_SIZE(extended_memmap) - VIRT_LOWMEMMAP_LAST ==
1709            ARRAY_SIZE(enabled_array));
1710     assert(index - VIRT_LOWMEMMAP_LAST < ARRAY_SIZE(enabled_array));
1711 
1712     return enabled_array[index - VIRT_LOWMEMMAP_LAST];
1713 }
1714 
1715 static void virt_set_high_memmap(VirtMachineState *vms,
1716                                  hwaddr base, int pa_bits)
1717 {
1718     hwaddr region_base, region_size;
1719     bool *region_enabled, fits;
1720     int i;
1721 
1722     for (i = VIRT_LOWMEMMAP_LAST; i < ARRAY_SIZE(extended_memmap); i++) {
1723         region_enabled = virt_get_high_memmap_enabled(vms, i);
1724         region_base = ROUND_UP(base, extended_memmap[i].size);
1725         region_size = extended_memmap[i].size;
1726 
1727         vms->memmap[i].base = region_base;
1728         vms->memmap[i].size = region_size;
1729 
1730         /*
1731          * Check each device to see if it fits in the PA space,
1732          * moving highest_gpa as we go. For compatibility, move
1733          * highest_gpa for disabled fitting devices as well, if
1734          * the compact layout has been disabled.
1735          *
1736          * For each device that doesn't fit, disable it.
1737          */
1738         fits = (region_base + region_size) <= BIT_ULL(pa_bits);
1739         *region_enabled &= fits;
1740         if (vms->highmem_compact && !*region_enabled) {
1741             continue;
1742         }
1743 
1744         base = region_base + region_size;
1745         if (fits) {
1746             vms->highest_gpa = base - 1;
1747         }
1748     }
1749 }
1750 
1751 static void virt_set_memmap(VirtMachineState *vms, int pa_bits)
1752 {
1753     MachineState *ms = MACHINE(vms);
1754     hwaddr base, device_memory_base, device_memory_size, memtop;
1755     int i;
1756 
1757     vms->memmap = extended_memmap;
1758 
1759     for (i = 0; i < ARRAY_SIZE(base_memmap); i++) {
1760         vms->memmap[i] = base_memmap[i];
1761     }
1762 
1763     if (ms->ram_slots > ACPI_MAX_RAM_SLOTS) {
1764         error_report("unsupported number of memory slots: %"PRIu64,
1765                      ms->ram_slots);
1766         exit(EXIT_FAILURE);
1767     }
1768 
1769     /*
1770      * !highmem is exactly the same as limiting the PA space to 32bit,
1771      * irrespective of the underlying capabilities of the HW.
1772      */
1773     if (!vms->highmem) {
1774         pa_bits = 32;
1775     }
1776 
1777     /*
1778      * We compute the base of the high IO region depending on the
1779      * amount of initial and device memory. The device memory start/size
1780      * is aligned on 1GiB. We never put the high IO region below 256GiB
1781      * so that if maxram_size is < 255GiB we keep the legacy memory map.
1782      * The device region size assumes 1GiB page max alignment per slot.
1783      */
1784     device_memory_base =
1785         ROUND_UP(vms->memmap[VIRT_MEM].base + ms->ram_size, GiB);
1786     device_memory_size = ms->maxram_size - ms->ram_size + ms->ram_slots * GiB;
1787 
1788     /* Base address of the high IO region */
1789     memtop = base = device_memory_base + ROUND_UP(device_memory_size, GiB);
1790     if (memtop > BIT_ULL(pa_bits)) {
1791 	    error_report("Addressing limited to %d bits, but memory exceeds it by %llu bytes\n",
1792 			 pa_bits, memtop - BIT_ULL(pa_bits));
1793         exit(EXIT_FAILURE);
1794     }
1795     if (base < device_memory_base) {
1796         error_report("maxmem/slots too huge");
1797         exit(EXIT_FAILURE);
1798     }
1799     if (base < vms->memmap[VIRT_MEM].base + LEGACY_RAMLIMIT_BYTES) {
1800         base = vms->memmap[VIRT_MEM].base + LEGACY_RAMLIMIT_BYTES;
1801     }
1802 
1803     /* We know for sure that at least the memory fits in the PA space */
1804     vms->highest_gpa = memtop - 1;
1805 
1806     virt_set_high_memmap(vms, base, pa_bits);
1807 
1808     if (device_memory_size > 0) {
1809         ms->device_memory = g_malloc0(sizeof(*ms->device_memory));
1810         ms->device_memory->base = device_memory_base;
1811         memory_region_init(&ms->device_memory->mr, OBJECT(vms),
1812                            "device-memory", device_memory_size);
1813     }
1814 }
1815 
1816 /*
1817  * finalize_gic_version - Determines the final gic_version
1818  * according to the gic-version property
1819  *
1820  * Default GIC type is v2
1821  */
1822 static void finalize_gic_version(VirtMachineState *vms)
1823 {
1824     unsigned int max_cpus = MACHINE(vms)->smp.max_cpus;
1825 
1826     if (kvm_enabled()) {
1827         int probe_bitmap;
1828 
1829         if (!kvm_irqchip_in_kernel()) {
1830             switch (vms->gic_version) {
1831             case VIRT_GIC_VERSION_HOST:
1832                 warn_report(
1833                     "gic-version=host not relevant with kernel-irqchip=off "
1834                      "as only userspace GICv2 is supported. Using v2 ...");
1835                 return;
1836             case VIRT_GIC_VERSION_MAX:
1837             case VIRT_GIC_VERSION_NOSEL:
1838                 vms->gic_version = VIRT_GIC_VERSION_2;
1839                 return;
1840             case VIRT_GIC_VERSION_2:
1841                 return;
1842             case VIRT_GIC_VERSION_3:
1843                 error_report(
1844                     "gic-version=3 is not supported with kernel-irqchip=off");
1845                 exit(1);
1846             case VIRT_GIC_VERSION_4:
1847                 error_report(
1848                     "gic-version=4 is not supported with kernel-irqchip=off");
1849                 exit(1);
1850             }
1851         }
1852 
1853         probe_bitmap = kvm_arm_vgic_probe();
1854         if (!probe_bitmap) {
1855             error_report("Unable to determine GIC version supported by host");
1856             exit(1);
1857         }
1858 
1859         switch (vms->gic_version) {
1860         case VIRT_GIC_VERSION_HOST:
1861         case VIRT_GIC_VERSION_MAX:
1862             if (probe_bitmap & KVM_ARM_VGIC_V3) {
1863                 vms->gic_version = VIRT_GIC_VERSION_3;
1864             } else {
1865                 vms->gic_version = VIRT_GIC_VERSION_2;
1866             }
1867             return;
1868         case VIRT_GIC_VERSION_NOSEL:
1869             if ((probe_bitmap & KVM_ARM_VGIC_V2) && max_cpus <= GIC_NCPU) {
1870                 vms->gic_version = VIRT_GIC_VERSION_2;
1871             } else if (probe_bitmap & KVM_ARM_VGIC_V3) {
1872                 /*
1873                  * in case the host does not support v2 in-kernel emulation or
1874                  * the end-user requested more than 8 VCPUs we now default
1875                  * to v3. In any case defaulting to v2 would be broken.
1876                  */
1877                 vms->gic_version = VIRT_GIC_VERSION_3;
1878             } else if (max_cpus > GIC_NCPU) {
1879                 error_report("host only supports in-kernel GICv2 emulation "
1880                              "but more than 8 vcpus are requested");
1881                 exit(1);
1882             }
1883             break;
1884         case VIRT_GIC_VERSION_2:
1885         case VIRT_GIC_VERSION_3:
1886             break;
1887         case VIRT_GIC_VERSION_4:
1888             error_report("gic-version=4 is not supported with KVM");
1889             exit(1);
1890         }
1891 
1892         /* Check chosen version is effectively supported by the host */
1893         if (vms->gic_version == VIRT_GIC_VERSION_2 &&
1894             !(probe_bitmap & KVM_ARM_VGIC_V2)) {
1895             error_report("host does not support in-kernel GICv2 emulation");
1896             exit(1);
1897         } else if (vms->gic_version == VIRT_GIC_VERSION_3 &&
1898                    !(probe_bitmap & KVM_ARM_VGIC_V3)) {
1899             error_report("host does not support in-kernel GICv3 emulation");
1900             exit(1);
1901         }
1902         return;
1903     }
1904 
1905     /* TCG mode */
1906     switch (vms->gic_version) {
1907     case VIRT_GIC_VERSION_NOSEL:
1908         vms->gic_version = VIRT_GIC_VERSION_2;
1909         break;
1910     case VIRT_GIC_VERSION_MAX:
1911         if (module_object_class_by_name("arm-gicv3")) {
1912             /* CONFIG_ARM_GICV3_TCG was set */
1913             if (vms->virt) {
1914                 /* GICv4 only makes sense if CPU has EL2 */
1915                 vms->gic_version = VIRT_GIC_VERSION_4;
1916             } else {
1917                 vms->gic_version = VIRT_GIC_VERSION_3;
1918             }
1919         } else {
1920             vms->gic_version = VIRT_GIC_VERSION_2;
1921         }
1922         break;
1923     case VIRT_GIC_VERSION_HOST:
1924         error_report("gic-version=host requires KVM");
1925         exit(1);
1926     case VIRT_GIC_VERSION_4:
1927         if (!vms->virt) {
1928             error_report("gic-version=4 requires virtualization enabled");
1929             exit(1);
1930         }
1931         break;
1932     case VIRT_GIC_VERSION_2:
1933     case VIRT_GIC_VERSION_3:
1934         break;
1935     }
1936 }
1937 
1938 /*
1939  * virt_cpu_post_init() must be called after the CPUs have
1940  * been realized and the GIC has been created.
1941  */
1942 static void virt_cpu_post_init(VirtMachineState *vms, MemoryRegion *sysmem)
1943 {
1944     int max_cpus = MACHINE(vms)->smp.max_cpus;
1945     bool aarch64, pmu, steal_time;
1946     CPUState *cpu;
1947 
1948     aarch64 = object_property_get_bool(OBJECT(first_cpu), "aarch64", NULL);
1949     pmu = object_property_get_bool(OBJECT(first_cpu), "pmu", NULL);
1950     steal_time = object_property_get_bool(OBJECT(first_cpu),
1951                                           "kvm-steal-time", NULL);
1952 
1953     if (kvm_enabled()) {
1954         hwaddr pvtime_reg_base = vms->memmap[VIRT_PVTIME].base;
1955         hwaddr pvtime_reg_size = vms->memmap[VIRT_PVTIME].size;
1956 
1957         if (steal_time) {
1958             MemoryRegion *pvtime = g_new(MemoryRegion, 1);
1959             hwaddr pvtime_size = max_cpus * PVTIME_SIZE_PER_CPU;
1960 
1961             /* The memory region size must be a multiple of host page size. */
1962             pvtime_size = REAL_HOST_PAGE_ALIGN(pvtime_size);
1963 
1964             if (pvtime_size > pvtime_reg_size) {
1965                 error_report("pvtime requires a %" HWADDR_PRId
1966                              " byte memory region for %d CPUs,"
1967                              " but only %" HWADDR_PRId " has been reserved",
1968                              pvtime_size, max_cpus, pvtime_reg_size);
1969                 exit(1);
1970             }
1971 
1972             memory_region_init_ram(pvtime, NULL, "pvtime", pvtime_size, NULL);
1973             memory_region_add_subregion(sysmem, pvtime_reg_base, pvtime);
1974         }
1975 
1976         CPU_FOREACH(cpu) {
1977             if (pmu) {
1978                 assert(arm_feature(&ARM_CPU(cpu)->env, ARM_FEATURE_PMU));
1979                 if (kvm_irqchip_in_kernel()) {
1980                     kvm_arm_pmu_set_irq(cpu, PPI(VIRTUAL_PMU_IRQ));
1981                 }
1982                 kvm_arm_pmu_init(cpu);
1983             }
1984             if (steal_time) {
1985                 kvm_arm_pvtime_init(cpu, pvtime_reg_base +
1986                                          cpu->cpu_index * PVTIME_SIZE_PER_CPU);
1987             }
1988         }
1989     } else {
1990         if (aarch64 && vms->highmem) {
1991             int requested_pa_size = 64 - clz64(vms->highest_gpa);
1992             int pamax = arm_pamax(ARM_CPU(first_cpu));
1993 
1994             if (pamax < requested_pa_size) {
1995                 error_report("VCPU supports less PA bits (%d) than "
1996                              "requested by the memory map (%d)",
1997                              pamax, requested_pa_size);
1998                 exit(1);
1999             }
2000         }
2001     }
2002 }
2003 
2004 static void machvirt_init(MachineState *machine)
2005 {
2006     VirtMachineState *vms = VIRT_MACHINE(machine);
2007     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(machine);
2008     MachineClass *mc = MACHINE_GET_CLASS(machine);
2009     const CPUArchIdList *possible_cpus;
2010     MemoryRegion *sysmem = get_system_memory();
2011     MemoryRegion *secure_sysmem = NULL;
2012     MemoryRegion *tag_sysmem = NULL;
2013     MemoryRegion *secure_tag_sysmem = NULL;
2014     int n, virt_max_cpus;
2015     bool firmware_loaded;
2016     bool aarch64 = true;
2017     bool has_ged = !vmc->no_ged;
2018     unsigned int smp_cpus = machine->smp.cpus;
2019     unsigned int max_cpus = machine->smp.max_cpus;
2020 
2021     if (!cpu_type_valid(machine->cpu_type)) {
2022         error_report("mach-virt: CPU type %s not supported", machine->cpu_type);
2023         exit(1);
2024     }
2025 
2026     possible_cpus = mc->possible_cpu_arch_ids(machine);
2027 
2028     /*
2029      * In accelerated mode, the memory map is computed earlier in kvm_type()
2030      * to create a VM with the right number of IPA bits.
2031      */
2032     if (!vms->memmap) {
2033         Object *cpuobj;
2034         ARMCPU *armcpu;
2035         int pa_bits;
2036 
2037         /*
2038          * Instanciate a temporary CPU object to find out about what
2039          * we are about to deal with. Once this is done, get rid of
2040          * the object.
2041          */
2042         cpuobj = object_new(possible_cpus->cpus[0].type);
2043         armcpu = ARM_CPU(cpuobj);
2044 
2045         pa_bits = arm_pamax(armcpu);
2046 
2047         object_unref(cpuobj);
2048 
2049         virt_set_memmap(vms, pa_bits);
2050     }
2051 
2052     /* We can probe only here because during property set
2053      * KVM is not available yet
2054      */
2055     finalize_gic_version(vms);
2056 
2057     if (vms->secure) {
2058         /*
2059          * The Secure view of the world is the same as the NonSecure,
2060          * but with a few extra devices. Create it as a container region
2061          * containing the system memory at low priority; any secure-only
2062          * devices go in at higher priority and take precedence.
2063          */
2064         secure_sysmem = g_new(MemoryRegion, 1);
2065         memory_region_init(secure_sysmem, OBJECT(machine), "secure-memory",
2066                            UINT64_MAX);
2067         memory_region_add_subregion_overlap(secure_sysmem, 0, sysmem, -1);
2068     }
2069 
2070     firmware_loaded = virt_firmware_init(vms, sysmem,
2071                                          secure_sysmem ?: sysmem);
2072 
2073     /* If we have an EL3 boot ROM then the assumption is that it will
2074      * implement PSCI itself, so disable QEMU's internal implementation
2075      * so it doesn't get in the way. Instead of starting secondary
2076      * CPUs in PSCI powerdown state we will start them all running and
2077      * let the boot ROM sort them out.
2078      * The usual case is that we do use QEMU's PSCI implementation;
2079      * if the guest has EL2 then we will use SMC as the conduit,
2080      * and otherwise we will use HVC (for backwards compatibility and
2081      * because if we're using KVM then we must use HVC).
2082      */
2083     if (vms->secure && firmware_loaded) {
2084         vms->psci_conduit = QEMU_PSCI_CONDUIT_DISABLED;
2085     } else if (vms->virt) {
2086         vms->psci_conduit = QEMU_PSCI_CONDUIT_SMC;
2087     } else {
2088         vms->psci_conduit = QEMU_PSCI_CONDUIT_HVC;
2089     }
2090 
2091     /*
2092      * The maximum number of CPUs depends on the GIC version, or on how
2093      * many redistributors we can fit into the memory map (which in turn
2094      * depends on whether this is a GICv3 or v4).
2095      */
2096     if (vms->gic_version == VIRT_GIC_VERSION_2) {
2097         virt_max_cpus = GIC_NCPU;
2098     } else {
2099         virt_max_cpus = virt_redist_capacity(vms, VIRT_GIC_REDIST) +
2100             virt_redist_capacity(vms, VIRT_HIGH_GIC_REDIST2);
2101     }
2102 
2103     if (max_cpus > virt_max_cpus) {
2104         error_report("Number of SMP CPUs requested (%d) exceeds max CPUs "
2105                      "supported by machine 'mach-virt' (%d)",
2106                      max_cpus, virt_max_cpus);
2107         exit(1);
2108     }
2109 
2110     if (vms->secure && (kvm_enabled() || hvf_enabled())) {
2111         error_report("mach-virt: %s does not support providing "
2112                      "Security extensions (TrustZone) to the guest CPU",
2113                      kvm_enabled() ? "KVM" : "HVF");
2114         exit(1);
2115     }
2116 
2117     if (vms->virt && (kvm_enabled() || hvf_enabled())) {
2118         error_report("mach-virt: %s does not support providing "
2119                      "Virtualization extensions to the guest CPU",
2120                      kvm_enabled() ? "KVM" : "HVF");
2121         exit(1);
2122     }
2123 
2124     if (vms->mte && (kvm_enabled() || hvf_enabled())) {
2125         error_report("mach-virt: %s does not support providing "
2126                      "MTE to the guest CPU",
2127                      kvm_enabled() ? "KVM" : "HVF");
2128         exit(1);
2129     }
2130 
2131     create_fdt(vms);
2132 
2133     assert(possible_cpus->len == max_cpus);
2134     for (n = 0; n < possible_cpus->len; n++) {
2135         Object *cpuobj;
2136         CPUState *cs;
2137 
2138         if (n >= smp_cpus) {
2139             break;
2140         }
2141 
2142         cpuobj = object_new(possible_cpus->cpus[n].type);
2143         object_property_set_int(cpuobj, "mp-affinity",
2144                                 possible_cpus->cpus[n].arch_id, NULL);
2145 
2146         cs = CPU(cpuobj);
2147         cs->cpu_index = n;
2148 
2149         numa_cpu_pre_plug(&possible_cpus->cpus[cs->cpu_index], DEVICE(cpuobj),
2150                           &error_fatal);
2151 
2152         aarch64 &= object_property_get_bool(cpuobj, "aarch64", NULL);
2153 
2154         if (!vms->secure) {
2155             object_property_set_bool(cpuobj, "has_el3", false, NULL);
2156         }
2157 
2158         if (!vms->virt && object_property_find(cpuobj, "has_el2")) {
2159             object_property_set_bool(cpuobj, "has_el2", false, NULL);
2160         }
2161 
2162         if (vmc->kvm_no_adjvtime &&
2163             object_property_find(cpuobj, "kvm-no-adjvtime")) {
2164             object_property_set_bool(cpuobj, "kvm-no-adjvtime", true, NULL);
2165         }
2166 
2167         if (vmc->no_kvm_steal_time &&
2168             object_property_find(cpuobj, "kvm-steal-time")) {
2169             object_property_set_bool(cpuobj, "kvm-steal-time", false, NULL);
2170         }
2171 
2172         if (vmc->no_pmu && object_property_find(cpuobj, "pmu")) {
2173             object_property_set_bool(cpuobj, "pmu", false, NULL);
2174         }
2175 
2176         if (vmc->no_tcg_lpa2 && object_property_find(cpuobj, "lpa2")) {
2177             object_property_set_bool(cpuobj, "lpa2", false, NULL);
2178         }
2179 
2180         if (object_property_find(cpuobj, "reset-cbar")) {
2181             object_property_set_int(cpuobj, "reset-cbar",
2182                                     vms->memmap[VIRT_CPUPERIPHS].base,
2183                                     &error_abort);
2184         }
2185 
2186         object_property_set_link(cpuobj, "memory", OBJECT(sysmem),
2187                                  &error_abort);
2188         if (vms->secure) {
2189             object_property_set_link(cpuobj, "secure-memory",
2190                                      OBJECT(secure_sysmem), &error_abort);
2191         }
2192 
2193         if (vms->mte) {
2194             /* Create the memory region only once, but link to all cpus. */
2195             if (!tag_sysmem) {
2196                 /*
2197                  * The property exists only if MemTag is supported.
2198                  * If it is, we must allocate the ram to back that up.
2199                  */
2200                 if (!object_property_find(cpuobj, "tag-memory")) {
2201                     error_report("MTE requested, but not supported "
2202                                  "by the guest CPU");
2203                     exit(1);
2204                 }
2205 
2206                 tag_sysmem = g_new(MemoryRegion, 1);
2207                 memory_region_init(tag_sysmem, OBJECT(machine),
2208                                    "tag-memory", UINT64_MAX / 32);
2209 
2210                 if (vms->secure) {
2211                     secure_tag_sysmem = g_new(MemoryRegion, 1);
2212                     memory_region_init(secure_tag_sysmem, OBJECT(machine),
2213                                        "secure-tag-memory", UINT64_MAX / 32);
2214 
2215                     /* As with ram, secure-tag takes precedence over tag.  */
2216                     memory_region_add_subregion_overlap(secure_tag_sysmem, 0,
2217                                                         tag_sysmem, -1);
2218                 }
2219             }
2220 
2221             object_property_set_link(cpuobj, "tag-memory", OBJECT(tag_sysmem),
2222                                      &error_abort);
2223             if (vms->secure) {
2224                 object_property_set_link(cpuobj, "secure-tag-memory",
2225                                          OBJECT(secure_tag_sysmem),
2226                                          &error_abort);
2227             }
2228         }
2229 
2230         qdev_realize(DEVICE(cpuobj), NULL, &error_fatal);
2231         object_unref(cpuobj);
2232     }
2233     fdt_add_timer_nodes(vms);
2234     fdt_add_cpu_nodes(vms);
2235 
2236     memory_region_add_subregion(sysmem, vms->memmap[VIRT_MEM].base,
2237                                 machine->ram);
2238     if (machine->device_memory) {
2239         memory_region_add_subregion(sysmem, machine->device_memory->base,
2240                                     &machine->device_memory->mr);
2241     }
2242 
2243     virt_flash_fdt(vms, sysmem, secure_sysmem ?: sysmem);
2244 
2245     create_gic(vms, sysmem);
2246 
2247     virt_cpu_post_init(vms, sysmem);
2248 
2249     fdt_add_pmu_nodes(vms);
2250 
2251     create_uart(vms, VIRT_UART, sysmem, serial_hd(0));
2252 
2253     if (vms->secure) {
2254         create_secure_ram(vms, secure_sysmem, secure_tag_sysmem);
2255         create_uart(vms, VIRT_SECURE_UART, secure_sysmem, serial_hd(1));
2256     }
2257 
2258     if (tag_sysmem) {
2259         create_tag_ram(tag_sysmem, vms->memmap[VIRT_MEM].base,
2260                        machine->ram_size, "mach-virt.tag");
2261     }
2262 
2263     vms->highmem_ecam &= (!firmware_loaded || aarch64);
2264 
2265     create_rtc(vms);
2266 
2267     create_pcie(vms);
2268 
2269     if (has_ged && aarch64 && firmware_loaded && virt_is_acpi_enabled(vms)) {
2270         vms->acpi_dev = create_acpi_ged(vms);
2271     } else {
2272         create_gpio_devices(vms, VIRT_GPIO, sysmem);
2273     }
2274 
2275     if (vms->secure && !vmc->no_secure_gpio) {
2276         create_gpio_devices(vms, VIRT_SECURE_GPIO, secure_sysmem);
2277     }
2278 
2279      /* connect powerdown request */
2280      vms->powerdown_notifier.notify = virt_powerdown_req;
2281      qemu_register_powerdown_notifier(&vms->powerdown_notifier);
2282 
2283     /* Create mmio transports, so the user can create virtio backends
2284      * (which will be automatically plugged in to the transports). If
2285      * no backend is created the transport will just sit harmlessly idle.
2286      */
2287     create_virtio_devices(vms);
2288 
2289     vms->fw_cfg = create_fw_cfg(vms, &address_space_memory);
2290     rom_set_fw(vms->fw_cfg);
2291 
2292     create_platform_bus(vms);
2293 
2294     if (machine->nvdimms_state->is_enabled) {
2295         const struct AcpiGenericAddress arm_virt_nvdimm_acpi_dsmio = {
2296             .space_id = AML_AS_SYSTEM_MEMORY,
2297             .address = vms->memmap[VIRT_NVDIMM_ACPI].base,
2298             .bit_width = NVDIMM_ACPI_IO_LEN << 3
2299         };
2300 
2301         nvdimm_init_acpi_state(machine->nvdimms_state, sysmem,
2302                                arm_virt_nvdimm_acpi_dsmio,
2303                                vms->fw_cfg, OBJECT(vms));
2304     }
2305 
2306     vms->bootinfo.ram_size = machine->ram_size;
2307     vms->bootinfo.board_id = -1;
2308     vms->bootinfo.loader_start = vms->memmap[VIRT_MEM].base;
2309     vms->bootinfo.get_dtb = machvirt_dtb;
2310     vms->bootinfo.skip_dtb_autoload = true;
2311     vms->bootinfo.firmware_loaded = firmware_loaded;
2312     vms->bootinfo.psci_conduit = vms->psci_conduit;
2313     arm_load_kernel(ARM_CPU(first_cpu), machine, &vms->bootinfo);
2314 
2315     vms->machine_done.notify = virt_machine_done;
2316     qemu_add_machine_init_done_notifier(&vms->machine_done);
2317 }
2318 
2319 static bool virt_get_secure(Object *obj, Error **errp)
2320 {
2321     VirtMachineState *vms = VIRT_MACHINE(obj);
2322 
2323     return vms->secure;
2324 }
2325 
2326 static void virt_set_secure(Object *obj, bool value, Error **errp)
2327 {
2328     VirtMachineState *vms = VIRT_MACHINE(obj);
2329 
2330     vms->secure = value;
2331 }
2332 
2333 static bool virt_get_virt(Object *obj, Error **errp)
2334 {
2335     VirtMachineState *vms = VIRT_MACHINE(obj);
2336 
2337     return vms->virt;
2338 }
2339 
2340 static void virt_set_virt(Object *obj, bool value, Error **errp)
2341 {
2342     VirtMachineState *vms = VIRT_MACHINE(obj);
2343 
2344     vms->virt = value;
2345 }
2346 
2347 static bool virt_get_highmem(Object *obj, Error **errp)
2348 {
2349     VirtMachineState *vms = VIRT_MACHINE(obj);
2350 
2351     return vms->highmem;
2352 }
2353 
2354 static void virt_set_highmem(Object *obj, bool value, Error **errp)
2355 {
2356     VirtMachineState *vms = VIRT_MACHINE(obj);
2357 
2358     vms->highmem = value;
2359 }
2360 
2361 static bool virt_get_compact_highmem(Object *obj, Error **errp)
2362 {
2363     VirtMachineState *vms = VIRT_MACHINE(obj);
2364 
2365     return vms->highmem_compact;
2366 }
2367 
2368 static void virt_set_compact_highmem(Object *obj, bool value, Error **errp)
2369 {
2370     VirtMachineState *vms = VIRT_MACHINE(obj);
2371 
2372     vms->highmem_compact = value;
2373 }
2374 
2375 static bool virt_get_its(Object *obj, Error **errp)
2376 {
2377     VirtMachineState *vms = VIRT_MACHINE(obj);
2378 
2379     return vms->its;
2380 }
2381 
2382 static void virt_set_its(Object *obj, bool value, Error **errp)
2383 {
2384     VirtMachineState *vms = VIRT_MACHINE(obj);
2385 
2386     vms->its = value;
2387 }
2388 
2389 static bool virt_get_dtb_randomness(Object *obj, Error **errp)
2390 {
2391     VirtMachineState *vms = VIRT_MACHINE(obj);
2392 
2393     return vms->dtb_randomness;
2394 }
2395 
2396 static void virt_set_dtb_randomness(Object *obj, bool value, Error **errp)
2397 {
2398     VirtMachineState *vms = VIRT_MACHINE(obj);
2399 
2400     vms->dtb_randomness = value;
2401 }
2402 
2403 static char *virt_get_oem_id(Object *obj, Error **errp)
2404 {
2405     VirtMachineState *vms = VIRT_MACHINE(obj);
2406 
2407     return g_strdup(vms->oem_id);
2408 }
2409 
2410 static void virt_set_oem_id(Object *obj, const char *value, Error **errp)
2411 {
2412     VirtMachineState *vms = VIRT_MACHINE(obj);
2413     size_t len = strlen(value);
2414 
2415     if (len > 6) {
2416         error_setg(errp,
2417                    "User specified oem-id value is bigger than 6 bytes in size");
2418         return;
2419     }
2420 
2421     strncpy(vms->oem_id, value, 6);
2422 }
2423 
2424 static char *virt_get_oem_table_id(Object *obj, Error **errp)
2425 {
2426     VirtMachineState *vms = VIRT_MACHINE(obj);
2427 
2428     return g_strdup(vms->oem_table_id);
2429 }
2430 
2431 static void virt_set_oem_table_id(Object *obj, const char *value,
2432                                   Error **errp)
2433 {
2434     VirtMachineState *vms = VIRT_MACHINE(obj);
2435     size_t len = strlen(value);
2436 
2437     if (len > 8) {
2438         error_setg(errp,
2439                    "User specified oem-table-id value is bigger than 8 bytes in size");
2440         return;
2441     }
2442     strncpy(vms->oem_table_id, value, 8);
2443 }
2444 
2445 
2446 bool virt_is_acpi_enabled(VirtMachineState *vms)
2447 {
2448     if (vms->acpi == ON_OFF_AUTO_OFF) {
2449         return false;
2450     }
2451     return true;
2452 }
2453 
2454 static void virt_get_acpi(Object *obj, Visitor *v, const char *name,
2455                           void *opaque, Error **errp)
2456 {
2457     VirtMachineState *vms = VIRT_MACHINE(obj);
2458     OnOffAuto acpi = vms->acpi;
2459 
2460     visit_type_OnOffAuto(v, name, &acpi, errp);
2461 }
2462 
2463 static void virt_set_acpi(Object *obj, Visitor *v, const char *name,
2464                           void *opaque, Error **errp)
2465 {
2466     VirtMachineState *vms = VIRT_MACHINE(obj);
2467 
2468     visit_type_OnOffAuto(v, name, &vms->acpi, errp);
2469 }
2470 
2471 static bool virt_get_ras(Object *obj, Error **errp)
2472 {
2473     VirtMachineState *vms = VIRT_MACHINE(obj);
2474 
2475     return vms->ras;
2476 }
2477 
2478 static void virt_set_ras(Object *obj, bool value, Error **errp)
2479 {
2480     VirtMachineState *vms = VIRT_MACHINE(obj);
2481 
2482     vms->ras = value;
2483 }
2484 
2485 static bool virt_get_mte(Object *obj, Error **errp)
2486 {
2487     VirtMachineState *vms = VIRT_MACHINE(obj);
2488 
2489     return vms->mte;
2490 }
2491 
2492 static void virt_set_mte(Object *obj, bool value, Error **errp)
2493 {
2494     VirtMachineState *vms = VIRT_MACHINE(obj);
2495 
2496     vms->mte = value;
2497 }
2498 
2499 static char *virt_get_gic_version(Object *obj, Error **errp)
2500 {
2501     VirtMachineState *vms = VIRT_MACHINE(obj);
2502     const char *val;
2503 
2504     switch (vms->gic_version) {
2505     case VIRT_GIC_VERSION_4:
2506         val = "4";
2507         break;
2508     case VIRT_GIC_VERSION_3:
2509         val = "3";
2510         break;
2511     default:
2512         val = "2";
2513         break;
2514     }
2515     return g_strdup(val);
2516 }
2517 
2518 static void virt_set_gic_version(Object *obj, const char *value, Error **errp)
2519 {
2520     VirtMachineState *vms = VIRT_MACHINE(obj);
2521 
2522     if (!strcmp(value, "4")) {
2523         vms->gic_version = VIRT_GIC_VERSION_4;
2524     } else if (!strcmp(value, "3")) {
2525         vms->gic_version = VIRT_GIC_VERSION_3;
2526     } else if (!strcmp(value, "2")) {
2527         vms->gic_version = VIRT_GIC_VERSION_2;
2528     } else if (!strcmp(value, "host")) {
2529         vms->gic_version = VIRT_GIC_VERSION_HOST; /* Will probe later */
2530     } else if (!strcmp(value, "max")) {
2531         vms->gic_version = VIRT_GIC_VERSION_MAX; /* Will probe later */
2532     } else {
2533         error_setg(errp, "Invalid gic-version value");
2534         error_append_hint(errp, "Valid values are 3, 2, host, max.\n");
2535     }
2536 }
2537 
2538 static char *virt_get_iommu(Object *obj, Error **errp)
2539 {
2540     VirtMachineState *vms = VIRT_MACHINE(obj);
2541 
2542     switch (vms->iommu) {
2543     case VIRT_IOMMU_NONE:
2544         return g_strdup("none");
2545     case VIRT_IOMMU_SMMUV3:
2546         return g_strdup("smmuv3");
2547     default:
2548         g_assert_not_reached();
2549     }
2550 }
2551 
2552 static void virt_set_iommu(Object *obj, const char *value, Error **errp)
2553 {
2554     VirtMachineState *vms = VIRT_MACHINE(obj);
2555 
2556     if (!strcmp(value, "smmuv3")) {
2557         vms->iommu = VIRT_IOMMU_SMMUV3;
2558     } else if (!strcmp(value, "none")) {
2559         vms->iommu = VIRT_IOMMU_NONE;
2560     } else {
2561         error_setg(errp, "Invalid iommu value");
2562         error_append_hint(errp, "Valid values are none, smmuv3.\n");
2563     }
2564 }
2565 
2566 static bool virt_get_default_bus_bypass_iommu(Object *obj, Error **errp)
2567 {
2568     VirtMachineState *vms = VIRT_MACHINE(obj);
2569 
2570     return vms->default_bus_bypass_iommu;
2571 }
2572 
2573 static void virt_set_default_bus_bypass_iommu(Object *obj, bool value,
2574                                               Error **errp)
2575 {
2576     VirtMachineState *vms = VIRT_MACHINE(obj);
2577 
2578     vms->default_bus_bypass_iommu = value;
2579 }
2580 
2581 static CpuInstanceProperties
2582 virt_cpu_index_to_props(MachineState *ms, unsigned cpu_index)
2583 {
2584     MachineClass *mc = MACHINE_GET_CLASS(ms);
2585     const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(ms);
2586 
2587     assert(cpu_index < possible_cpus->len);
2588     return possible_cpus->cpus[cpu_index].props;
2589 }
2590 
2591 static int64_t virt_get_default_cpu_node_id(const MachineState *ms, int idx)
2592 {
2593     int64_t socket_id = ms->possible_cpus->cpus[idx].props.socket_id;
2594 
2595     return socket_id % ms->numa_state->num_nodes;
2596 }
2597 
2598 static const CPUArchIdList *virt_possible_cpu_arch_ids(MachineState *ms)
2599 {
2600     int n;
2601     unsigned int max_cpus = ms->smp.max_cpus;
2602     VirtMachineState *vms = VIRT_MACHINE(ms);
2603     MachineClass *mc = MACHINE_GET_CLASS(vms);
2604 
2605     if (ms->possible_cpus) {
2606         assert(ms->possible_cpus->len == max_cpus);
2607         return ms->possible_cpus;
2608     }
2609 
2610     ms->possible_cpus = g_malloc0(sizeof(CPUArchIdList) +
2611                                   sizeof(CPUArchId) * max_cpus);
2612     ms->possible_cpus->len = max_cpus;
2613     for (n = 0; n < ms->possible_cpus->len; n++) {
2614         ms->possible_cpus->cpus[n].type = ms->cpu_type;
2615         ms->possible_cpus->cpus[n].arch_id =
2616             virt_cpu_mp_affinity(vms, n);
2617 
2618         assert(!mc->smp_props.dies_supported);
2619         ms->possible_cpus->cpus[n].props.has_socket_id = true;
2620         ms->possible_cpus->cpus[n].props.socket_id =
2621             n / (ms->smp.clusters * ms->smp.cores * ms->smp.threads);
2622         ms->possible_cpus->cpus[n].props.has_cluster_id = true;
2623         ms->possible_cpus->cpus[n].props.cluster_id =
2624             (n / (ms->smp.cores * ms->smp.threads)) % ms->smp.clusters;
2625         ms->possible_cpus->cpus[n].props.has_core_id = true;
2626         ms->possible_cpus->cpus[n].props.core_id =
2627             (n / ms->smp.threads) % ms->smp.cores;
2628         ms->possible_cpus->cpus[n].props.has_thread_id = true;
2629         ms->possible_cpus->cpus[n].props.thread_id =
2630             n % ms->smp.threads;
2631     }
2632     return ms->possible_cpus;
2633 }
2634 
2635 static void virt_memory_pre_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
2636                                  Error **errp)
2637 {
2638     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2639     const MachineState *ms = MACHINE(hotplug_dev);
2640     const bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2641 
2642     if (!vms->acpi_dev) {
2643         error_setg(errp,
2644                    "memory hotplug is not enabled: missing acpi-ged device");
2645         return;
2646     }
2647 
2648     if (vms->mte) {
2649         error_setg(errp, "memory hotplug is not enabled: MTE is enabled");
2650         return;
2651     }
2652 
2653     if (is_nvdimm && !ms->nvdimms_state->is_enabled) {
2654         error_setg(errp, "nvdimm is not enabled: add 'nvdimm=on' to '-M'");
2655         return;
2656     }
2657 
2658     pc_dimm_pre_plug(PC_DIMM(dev), MACHINE(hotplug_dev), NULL, errp);
2659 }
2660 
2661 static void virt_memory_plug(HotplugHandler *hotplug_dev,
2662                              DeviceState *dev, Error **errp)
2663 {
2664     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2665     MachineState *ms = MACHINE(hotplug_dev);
2666     bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2667 
2668     pc_dimm_plug(PC_DIMM(dev), MACHINE(vms));
2669 
2670     if (is_nvdimm) {
2671         nvdimm_plug(ms->nvdimms_state);
2672     }
2673 
2674     hotplug_handler_plug(HOTPLUG_HANDLER(vms->acpi_dev),
2675                          dev, &error_abort);
2676 }
2677 
2678 static void virt_virtio_md_pci_pre_plug(HotplugHandler *hotplug_dev,
2679                                         DeviceState *dev, Error **errp)
2680 {
2681     HotplugHandler *hotplug_dev2 = qdev_get_bus_hotplug_handler(dev);
2682     Error *local_err = NULL;
2683 
2684     if (!hotplug_dev2 && dev->hotplugged) {
2685         /*
2686          * Without a bus hotplug handler, we cannot control the plug/unplug
2687          * order. We should never reach this point when hotplugging on ARM.
2688          * However, it's nice to add a safety net, similar to what we have
2689          * on x86.
2690          */
2691         error_setg(errp, "hotplug of virtio based memory devices not supported"
2692                    " on this bus.");
2693         return;
2694     }
2695     /*
2696      * First, see if we can plug this memory device at all. If that
2697      * succeeds, branch of to the actual hotplug handler.
2698      */
2699     memory_device_pre_plug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev), NULL,
2700                            &local_err);
2701     if (!local_err && hotplug_dev2) {
2702         hotplug_handler_pre_plug(hotplug_dev2, dev, &local_err);
2703     }
2704     error_propagate(errp, local_err);
2705 }
2706 
2707 static void virt_virtio_md_pci_plug(HotplugHandler *hotplug_dev,
2708                                     DeviceState *dev, Error **errp)
2709 {
2710     HotplugHandler *hotplug_dev2 = qdev_get_bus_hotplug_handler(dev);
2711     Error *local_err = NULL;
2712 
2713     /*
2714      * Plug the memory device first and then branch off to the actual
2715      * hotplug handler. If that one fails, we can easily undo the memory
2716      * device bits.
2717      */
2718     memory_device_plug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev));
2719     if (hotplug_dev2) {
2720         hotplug_handler_plug(hotplug_dev2, dev, &local_err);
2721         if (local_err) {
2722             memory_device_unplug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev));
2723         }
2724     }
2725     error_propagate(errp, local_err);
2726 }
2727 
2728 static void virt_virtio_md_pci_unplug_request(HotplugHandler *hotplug_dev,
2729                                               DeviceState *dev, Error **errp)
2730 {
2731     /* We don't support hot unplug of virtio based memory devices */
2732     error_setg(errp, "virtio based memory devices cannot be unplugged.");
2733 }
2734 
2735 
2736 static void virt_machine_device_pre_plug_cb(HotplugHandler *hotplug_dev,
2737                                             DeviceState *dev, Error **errp)
2738 {
2739     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2740 
2741     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2742         virt_memory_pre_plug(hotplug_dev, dev, errp);
2743     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI)) {
2744         virt_virtio_md_pci_pre_plug(hotplug_dev, dev, errp);
2745     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2746         hwaddr db_start = 0, db_end = 0;
2747         char *resv_prop_str;
2748 
2749         if (vms->iommu != VIRT_IOMMU_NONE) {
2750             error_setg(errp, "virt machine does not support multiple IOMMUs");
2751             return;
2752         }
2753 
2754         switch (vms->msi_controller) {
2755         case VIRT_MSI_CTRL_NONE:
2756             return;
2757         case VIRT_MSI_CTRL_ITS:
2758             /* GITS_TRANSLATER page */
2759             db_start = base_memmap[VIRT_GIC_ITS].base + 0x10000;
2760             db_end = base_memmap[VIRT_GIC_ITS].base +
2761                      base_memmap[VIRT_GIC_ITS].size - 1;
2762             break;
2763         case VIRT_MSI_CTRL_GICV2M:
2764             /* MSI_SETSPI_NS page */
2765             db_start = base_memmap[VIRT_GIC_V2M].base;
2766             db_end = db_start + base_memmap[VIRT_GIC_V2M].size - 1;
2767             break;
2768         }
2769         resv_prop_str = g_strdup_printf("0x%"PRIx64":0x%"PRIx64":%u",
2770                                         db_start, db_end,
2771                                         VIRTIO_IOMMU_RESV_MEM_T_MSI);
2772 
2773         object_property_set_uint(OBJECT(dev), "len-reserved-regions", 1, errp);
2774         object_property_set_str(OBJECT(dev), "reserved-regions[0]",
2775                                 resv_prop_str, errp);
2776         g_free(resv_prop_str);
2777     }
2778 }
2779 
2780 static void virt_machine_device_plug_cb(HotplugHandler *hotplug_dev,
2781                                         DeviceState *dev, Error **errp)
2782 {
2783     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2784 
2785     if (vms->platform_bus_dev) {
2786         MachineClass *mc = MACHINE_GET_CLASS(vms);
2787 
2788         if (device_is_dynamic_sysbus(mc, dev)) {
2789             platform_bus_link_device(PLATFORM_BUS_DEVICE(vms->platform_bus_dev),
2790                                      SYS_BUS_DEVICE(dev));
2791         }
2792     }
2793     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2794         virt_memory_plug(hotplug_dev, dev, errp);
2795     }
2796 
2797     if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI)) {
2798         virt_virtio_md_pci_plug(hotplug_dev, dev, errp);
2799     }
2800 
2801     if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2802         PCIDevice *pdev = PCI_DEVICE(dev);
2803 
2804         vms->iommu = VIRT_IOMMU_VIRTIO;
2805         vms->virtio_iommu_bdf = pci_get_bdf(pdev);
2806         create_virtio_iommu_dt_bindings(vms);
2807     }
2808 }
2809 
2810 static void virt_dimm_unplug_request(HotplugHandler *hotplug_dev,
2811                                      DeviceState *dev, Error **errp)
2812 {
2813     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2814     Error *local_err = NULL;
2815 
2816     if (!vms->acpi_dev) {
2817         error_setg(&local_err,
2818                    "memory hotplug is not enabled: missing acpi-ged device");
2819         goto out;
2820     }
2821 
2822     if (object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM)) {
2823         error_setg(&local_err,
2824                    "nvdimm device hot unplug is not supported yet.");
2825         goto out;
2826     }
2827 
2828     hotplug_handler_unplug_request(HOTPLUG_HANDLER(vms->acpi_dev), dev,
2829                                    &local_err);
2830 out:
2831     error_propagate(errp, local_err);
2832 }
2833 
2834 static void virt_dimm_unplug(HotplugHandler *hotplug_dev,
2835                              DeviceState *dev, Error **errp)
2836 {
2837     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2838     Error *local_err = NULL;
2839 
2840     hotplug_handler_unplug(HOTPLUG_HANDLER(vms->acpi_dev), dev, &local_err);
2841     if (local_err) {
2842         goto out;
2843     }
2844 
2845     pc_dimm_unplug(PC_DIMM(dev), MACHINE(vms));
2846     qdev_unrealize(dev);
2847 
2848 out:
2849     error_propagate(errp, local_err);
2850 }
2851 
2852 static void virt_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
2853                                           DeviceState *dev, Error **errp)
2854 {
2855     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2856         virt_dimm_unplug_request(hotplug_dev, dev, errp);
2857     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI)) {
2858         virt_virtio_md_pci_unplug_request(hotplug_dev, dev, errp);
2859     } else {
2860         error_setg(errp, "device unplug request for unsupported device"
2861                    " type: %s", object_get_typename(OBJECT(dev)));
2862     }
2863 }
2864 
2865 static void virt_machine_device_unplug_cb(HotplugHandler *hotplug_dev,
2866                                           DeviceState *dev, Error **errp)
2867 {
2868     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2869         virt_dimm_unplug(hotplug_dev, dev, errp);
2870     } else {
2871         error_setg(errp, "virt: device unplug for unsupported device"
2872                    " type: %s", object_get_typename(OBJECT(dev)));
2873     }
2874 }
2875 
2876 static HotplugHandler *virt_machine_get_hotplug_handler(MachineState *machine,
2877                                                         DeviceState *dev)
2878 {
2879     MachineClass *mc = MACHINE_GET_CLASS(machine);
2880 
2881     if (device_is_dynamic_sysbus(mc, dev) ||
2882         object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) ||
2883         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI) ||
2884         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2885         return HOTPLUG_HANDLER(machine);
2886     }
2887     return NULL;
2888 }
2889 
2890 /*
2891  * for arm64 kvm_type [7-0] encodes the requested number of bits
2892  * in the IPA address space
2893  */
2894 static int virt_kvm_type(MachineState *ms, const char *type_str)
2895 {
2896     VirtMachineState *vms = VIRT_MACHINE(ms);
2897     int max_vm_pa_size, requested_pa_size;
2898     bool fixed_ipa;
2899 
2900     max_vm_pa_size = kvm_arm_get_max_vm_ipa_size(ms, &fixed_ipa);
2901 
2902     /* we freeze the memory map to compute the highest gpa */
2903     virt_set_memmap(vms, max_vm_pa_size);
2904 
2905     requested_pa_size = 64 - clz64(vms->highest_gpa);
2906 
2907     /*
2908      * KVM requires the IPA size to be at least 32 bits.
2909      */
2910     if (requested_pa_size < 32) {
2911         requested_pa_size = 32;
2912     }
2913 
2914     if (requested_pa_size > max_vm_pa_size) {
2915         error_report("-m and ,maxmem option values "
2916                      "require an IPA range (%d bits) larger than "
2917                      "the one supported by the host (%d bits)",
2918                      requested_pa_size, max_vm_pa_size);
2919         exit(1);
2920     }
2921     /*
2922      * We return the requested PA log size, unless KVM only supports
2923      * the implicit legacy 40b IPA setting, in which case the kvm_type
2924      * must be 0.
2925      */
2926     return fixed_ipa ? 0 : requested_pa_size;
2927 }
2928 
2929 static void virt_machine_class_init(ObjectClass *oc, void *data)
2930 {
2931     MachineClass *mc = MACHINE_CLASS(oc);
2932     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
2933 
2934     mc->init = machvirt_init;
2935     /* Start with max_cpus set to 512, which is the maximum supported by KVM.
2936      * The value may be reduced later when we have more information about the
2937      * configuration of the particular instance.
2938      */
2939     mc->max_cpus = 512;
2940     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_CALXEDA_XGMAC);
2941     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_AMD_XGBE);
2942     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_RAMFB_DEVICE);
2943     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_PLATFORM);
2944 #ifdef CONFIG_TPM
2945     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_TPM_TIS_SYSBUS);
2946 #endif
2947     mc->block_default_type = IF_VIRTIO;
2948     mc->no_cdrom = 1;
2949     mc->pci_allow_0_address = true;
2950     /* We know we will never create a pre-ARMv7 CPU which needs 1K pages */
2951     mc->minimum_page_bits = 12;
2952     mc->possible_cpu_arch_ids = virt_possible_cpu_arch_ids;
2953     mc->cpu_index_to_instance_props = virt_cpu_index_to_props;
2954     mc->default_cpu_type = ARM_CPU_TYPE_NAME("cortex-a15");
2955     mc->get_default_cpu_node_id = virt_get_default_cpu_node_id;
2956     mc->kvm_type = virt_kvm_type;
2957     assert(!mc->get_hotplug_handler);
2958     mc->get_hotplug_handler = virt_machine_get_hotplug_handler;
2959     hc->pre_plug = virt_machine_device_pre_plug_cb;
2960     hc->plug = virt_machine_device_plug_cb;
2961     hc->unplug_request = virt_machine_device_unplug_request_cb;
2962     hc->unplug = virt_machine_device_unplug_cb;
2963     mc->nvdimm_supported = true;
2964     mc->smp_props.clusters_supported = true;
2965     mc->auto_enable_numa_with_memhp = true;
2966     mc->auto_enable_numa_with_memdev = true;
2967     mc->default_ram_id = "mach-virt.ram";
2968 
2969     object_class_property_add(oc, "acpi", "OnOffAuto",
2970         virt_get_acpi, virt_set_acpi,
2971         NULL, NULL);
2972     object_class_property_set_description(oc, "acpi",
2973         "Enable ACPI");
2974     object_class_property_add_bool(oc, "secure", virt_get_secure,
2975                                    virt_set_secure);
2976     object_class_property_set_description(oc, "secure",
2977                                                 "Set on/off to enable/disable the ARM "
2978                                                 "Security Extensions (TrustZone)");
2979 
2980     object_class_property_add_bool(oc, "virtualization", virt_get_virt,
2981                                    virt_set_virt);
2982     object_class_property_set_description(oc, "virtualization",
2983                                           "Set on/off to enable/disable emulating a "
2984                                           "guest CPU which implements the ARM "
2985                                           "Virtualization Extensions");
2986 
2987     object_class_property_add_bool(oc, "highmem", virt_get_highmem,
2988                                    virt_set_highmem);
2989     object_class_property_set_description(oc, "highmem",
2990                                           "Set on/off to enable/disable using "
2991                                           "physical address space above 32 bits");
2992 
2993     object_class_property_add_bool(oc, "compact-highmem",
2994                                    virt_get_compact_highmem,
2995                                    virt_set_compact_highmem);
2996     object_class_property_set_description(oc, "compact-highmem",
2997                                           "Set on/off to enable/disable compact "
2998                                           "layout for high memory regions");
2999 
3000     object_class_property_add_str(oc, "gic-version", virt_get_gic_version,
3001                                   virt_set_gic_version);
3002     object_class_property_set_description(oc, "gic-version",
3003                                           "Set GIC version. "
3004                                           "Valid values are 2, 3, 4, host and max");
3005 
3006     object_class_property_add_str(oc, "iommu", virt_get_iommu, virt_set_iommu);
3007     object_class_property_set_description(oc, "iommu",
3008                                           "Set the IOMMU type. "
3009                                           "Valid values are none and smmuv3");
3010 
3011     object_class_property_add_bool(oc, "default-bus-bypass-iommu",
3012                                    virt_get_default_bus_bypass_iommu,
3013                                    virt_set_default_bus_bypass_iommu);
3014     object_class_property_set_description(oc, "default-bus-bypass-iommu",
3015                                           "Set on/off to enable/disable "
3016                                           "bypass_iommu for default root bus");
3017 
3018     object_class_property_add_bool(oc, "ras", virt_get_ras,
3019                                    virt_set_ras);
3020     object_class_property_set_description(oc, "ras",
3021                                           "Set on/off to enable/disable reporting host memory errors "
3022                                           "to a KVM guest using ACPI and guest external abort exceptions");
3023 
3024     object_class_property_add_bool(oc, "mte", virt_get_mte, virt_set_mte);
3025     object_class_property_set_description(oc, "mte",
3026                                           "Set on/off to enable/disable emulating a "
3027                                           "guest CPU which implements the ARM "
3028                                           "Memory Tagging Extension");
3029 
3030     object_class_property_add_bool(oc, "its", virt_get_its,
3031                                    virt_set_its);
3032     object_class_property_set_description(oc, "its",
3033                                           "Set on/off to enable/disable "
3034                                           "ITS instantiation");
3035 
3036     object_class_property_add_bool(oc, "dtb-randomness",
3037                                    virt_get_dtb_randomness,
3038                                    virt_set_dtb_randomness);
3039     object_class_property_set_description(oc, "dtb-randomness",
3040                                           "Set off to disable passing random or "
3041                                           "non-deterministic dtb nodes to guest");
3042 
3043     object_class_property_add_bool(oc, "dtb-kaslr-seed",
3044                                    virt_get_dtb_randomness,
3045                                    virt_set_dtb_randomness);
3046     object_class_property_set_description(oc, "dtb-kaslr-seed",
3047                                           "Deprecated synonym of dtb-randomness");
3048 
3049     object_class_property_add_str(oc, "x-oem-id",
3050                                   virt_get_oem_id,
3051                                   virt_set_oem_id);
3052     object_class_property_set_description(oc, "x-oem-id",
3053                                           "Override the default value of field OEMID "
3054                                           "in ACPI table header."
3055                                           "The string may be up to 6 bytes in size");
3056 
3057 
3058     object_class_property_add_str(oc, "x-oem-table-id",
3059                                   virt_get_oem_table_id,
3060                                   virt_set_oem_table_id);
3061     object_class_property_set_description(oc, "x-oem-table-id",
3062                                           "Override the default value of field OEM Table ID "
3063                                           "in ACPI table header."
3064                                           "The string may be up to 8 bytes in size");
3065 
3066 }
3067 
3068 static void virt_instance_init(Object *obj)
3069 {
3070     VirtMachineState *vms = VIRT_MACHINE(obj);
3071     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
3072 
3073     /* EL3 is disabled by default on virt: this makes us consistent
3074      * between KVM and TCG for this board, and it also allows us to
3075      * boot UEFI blobs which assume no TrustZone support.
3076      */
3077     vms->secure = false;
3078 
3079     /* EL2 is also disabled by default, for similar reasons */
3080     vms->virt = false;
3081 
3082     /* High memory is enabled by default */
3083     vms->highmem = true;
3084     vms->highmem_compact = !vmc->no_highmem_compact;
3085     vms->gic_version = VIRT_GIC_VERSION_NOSEL;
3086 
3087     vms->highmem_ecam = !vmc->no_highmem_ecam;
3088     vms->highmem_mmio = true;
3089     vms->highmem_redists = true;
3090 
3091     if (vmc->no_its) {
3092         vms->its = false;
3093     } else {
3094         /* Default allows ITS instantiation */
3095         vms->its = true;
3096 
3097         if (vmc->no_tcg_its) {
3098             vms->tcg_its = false;
3099         } else {
3100             vms->tcg_its = true;
3101         }
3102     }
3103 
3104     /* Default disallows iommu instantiation */
3105     vms->iommu = VIRT_IOMMU_NONE;
3106 
3107     /* The default root bus is attached to iommu by default */
3108     vms->default_bus_bypass_iommu = false;
3109 
3110     /* Default disallows RAS instantiation */
3111     vms->ras = false;
3112 
3113     /* MTE is disabled by default.  */
3114     vms->mte = false;
3115 
3116     /* Supply kaslr-seed and rng-seed by default */
3117     vms->dtb_randomness = true;
3118 
3119     vms->irqmap = a15irqmap;
3120 
3121     virt_flash_create(vms);
3122 
3123     vms->oem_id = g_strndup(ACPI_BUILD_APPNAME6, 6);
3124     vms->oem_table_id = g_strndup(ACPI_BUILD_APPNAME8, 8);
3125 }
3126 
3127 static const TypeInfo virt_machine_info = {
3128     .name          = TYPE_VIRT_MACHINE,
3129     .parent        = TYPE_MACHINE,
3130     .abstract      = true,
3131     .instance_size = sizeof(VirtMachineState),
3132     .class_size    = sizeof(VirtMachineClass),
3133     .class_init    = virt_machine_class_init,
3134     .instance_init = virt_instance_init,
3135     .interfaces = (InterfaceInfo[]) {
3136          { TYPE_HOTPLUG_HANDLER },
3137          { }
3138     },
3139 };
3140 
3141 static void machvirt_machine_init(void)
3142 {
3143     type_register_static(&virt_machine_info);
3144 }
3145 type_init(machvirt_machine_init);
3146 
3147 static void virt_machine_7_2_options(MachineClass *mc)
3148 {
3149 }
3150 DEFINE_VIRT_MACHINE_AS_LATEST(7, 2)
3151 
3152 static void virt_machine_7_1_options(MachineClass *mc)
3153 {
3154     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3155 
3156     virt_machine_7_2_options(mc);
3157     compat_props_add(mc->compat_props, hw_compat_7_1, hw_compat_7_1_len);
3158     /* Compact layout for high memory regions was introduced with 7.2 */
3159     vmc->no_highmem_compact = true;
3160 }
3161 DEFINE_VIRT_MACHINE(7, 1)
3162 
3163 static void virt_machine_7_0_options(MachineClass *mc)
3164 {
3165     virt_machine_7_1_options(mc);
3166     compat_props_add(mc->compat_props, hw_compat_7_0, hw_compat_7_0_len);
3167 }
3168 DEFINE_VIRT_MACHINE(7, 0)
3169 
3170 static void virt_machine_6_2_options(MachineClass *mc)
3171 {
3172     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3173 
3174     virt_machine_7_0_options(mc);
3175     compat_props_add(mc->compat_props, hw_compat_6_2, hw_compat_6_2_len);
3176     vmc->no_tcg_lpa2 = true;
3177 }
3178 DEFINE_VIRT_MACHINE(6, 2)
3179 
3180 static void virt_machine_6_1_options(MachineClass *mc)
3181 {
3182     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3183 
3184     virt_machine_6_2_options(mc);
3185     compat_props_add(mc->compat_props, hw_compat_6_1, hw_compat_6_1_len);
3186     mc->smp_props.prefer_sockets = true;
3187     vmc->no_cpu_topology = true;
3188 
3189     /* qemu ITS was introduced with 6.2 */
3190     vmc->no_tcg_its = true;
3191 }
3192 DEFINE_VIRT_MACHINE(6, 1)
3193 
3194 static void virt_machine_6_0_options(MachineClass *mc)
3195 {
3196     virt_machine_6_1_options(mc);
3197     compat_props_add(mc->compat_props, hw_compat_6_0, hw_compat_6_0_len);
3198 }
3199 DEFINE_VIRT_MACHINE(6, 0)
3200 
3201 static void virt_machine_5_2_options(MachineClass *mc)
3202 {
3203     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3204 
3205     virt_machine_6_0_options(mc);
3206     compat_props_add(mc->compat_props, hw_compat_5_2, hw_compat_5_2_len);
3207     vmc->no_secure_gpio = true;
3208 }
3209 DEFINE_VIRT_MACHINE(5, 2)
3210 
3211 static void virt_machine_5_1_options(MachineClass *mc)
3212 {
3213     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3214 
3215     virt_machine_5_2_options(mc);
3216     compat_props_add(mc->compat_props, hw_compat_5_1, hw_compat_5_1_len);
3217     vmc->no_kvm_steal_time = true;
3218 }
3219 DEFINE_VIRT_MACHINE(5, 1)
3220 
3221 static void virt_machine_5_0_options(MachineClass *mc)
3222 {
3223     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3224 
3225     virt_machine_5_1_options(mc);
3226     compat_props_add(mc->compat_props, hw_compat_5_0, hw_compat_5_0_len);
3227     mc->numa_mem_supported = true;
3228     vmc->acpi_expose_flash = true;
3229     mc->auto_enable_numa_with_memdev = false;
3230 }
3231 DEFINE_VIRT_MACHINE(5, 0)
3232 
3233 static void virt_machine_4_2_options(MachineClass *mc)
3234 {
3235     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3236 
3237     virt_machine_5_0_options(mc);
3238     compat_props_add(mc->compat_props, hw_compat_4_2, hw_compat_4_2_len);
3239     vmc->kvm_no_adjvtime = true;
3240 }
3241 DEFINE_VIRT_MACHINE(4, 2)
3242 
3243 static void virt_machine_4_1_options(MachineClass *mc)
3244 {
3245     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3246 
3247     virt_machine_4_2_options(mc);
3248     compat_props_add(mc->compat_props, hw_compat_4_1, hw_compat_4_1_len);
3249     vmc->no_ged = true;
3250     mc->auto_enable_numa_with_memhp = false;
3251 }
3252 DEFINE_VIRT_MACHINE(4, 1)
3253 
3254 static void virt_machine_4_0_options(MachineClass *mc)
3255 {
3256     virt_machine_4_1_options(mc);
3257     compat_props_add(mc->compat_props, hw_compat_4_0, hw_compat_4_0_len);
3258 }
3259 DEFINE_VIRT_MACHINE(4, 0)
3260 
3261 static void virt_machine_3_1_options(MachineClass *mc)
3262 {
3263     virt_machine_4_0_options(mc);
3264     compat_props_add(mc->compat_props, hw_compat_3_1, hw_compat_3_1_len);
3265 }
3266 DEFINE_VIRT_MACHINE(3, 1)
3267 
3268 static void virt_machine_3_0_options(MachineClass *mc)
3269 {
3270     virt_machine_3_1_options(mc);
3271     compat_props_add(mc->compat_props, hw_compat_3_0, hw_compat_3_0_len);
3272 }
3273 DEFINE_VIRT_MACHINE(3, 0)
3274 
3275 static void virt_machine_2_12_options(MachineClass *mc)
3276 {
3277     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3278 
3279     virt_machine_3_0_options(mc);
3280     compat_props_add(mc->compat_props, hw_compat_2_12, hw_compat_2_12_len);
3281     vmc->no_highmem_ecam = true;
3282     mc->max_cpus = 255;
3283 }
3284 DEFINE_VIRT_MACHINE(2, 12)
3285 
3286 static void virt_machine_2_11_options(MachineClass *mc)
3287 {
3288     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3289 
3290     virt_machine_2_12_options(mc);
3291     compat_props_add(mc->compat_props, hw_compat_2_11, hw_compat_2_11_len);
3292     vmc->smbios_old_sys_ver = true;
3293 }
3294 DEFINE_VIRT_MACHINE(2, 11)
3295 
3296 static void virt_machine_2_10_options(MachineClass *mc)
3297 {
3298     virt_machine_2_11_options(mc);
3299     compat_props_add(mc->compat_props, hw_compat_2_10, hw_compat_2_10_len);
3300     /* before 2.11 we never faulted accesses to bad addresses */
3301     mc->ignore_memory_transaction_failures = true;
3302 }
3303 DEFINE_VIRT_MACHINE(2, 10)
3304 
3305 static void virt_machine_2_9_options(MachineClass *mc)
3306 {
3307     virt_machine_2_10_options(mc);
3308     compat_props_add(mc->compat_props, hw_compat_2_9, hw_compat_2_9_len);
3309 }
3310 DEFINE_VIRT_MACHINE(2, 9)
3311 
3312 static void virt_machine_2_8_options(MachineClass *mc)
3313 {
3314     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3315 
3316     virt_machine_2_9_options(mc);
3317     compat_props_add(mc->compat_props, hw_compat_2_8, hw_compat_2_8_len);
3318     /* For 2.8 and earlier we falsely claimed in the DT that
3319      * our timers were edge-triggered, not level-triggered.
3320      */
3321     vmc->claim_edge_triggered_timers = true;
3322 }
3323 DEFINE_VIRT_MACHINE(2, 8)
3324 
3325 static void virt_machine_2_7_options(MachineClass *mc)
3326 {
3327     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3328 
3329     virt_machine_2_8_options(mc);
3330     compat_props_add(mc->compat_props, hw_compat_2_7, hw_compat_2_7_len);
3331     /* ITS was introduced with 2.8 */
3332     vmc->no_its = true;
3333     /* Stick with 1K pages for migration compatibility */
3334     mc->minimum_page_bits = 0;
3335 }
3336 DEFINE_VIRT_MACHINE(2, 7)
3337 
3338 static void virt_machine_2_6_options(MachineClass *mc)
3339 {
3340     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3341 
3342     virt_machine_2_7_options(mc);
3343     compat_props_add(mc->compat_props, hw_compat_2_6, hw_compat_2_6_len);
3344     vmc->disallow_affinity_adjustment = true;
3345     /* Disable PMU for 2.6 as PMU support was first introduced in 2.7 */
3346     vmc->no_pmu = true;
3347 }
3348 DEFINE_VIRT_MACHINE(2, 6)
3349