xref: /qemu/hw/arm/virt.c (revision 73b49878)
1 /*
2  * ARM mach-virt emulation
3  *
4  * Copyright (c) 2013 Linaro Limited
5  *
6  * This program is free software; you can redistribute it and/or modify it
7  * under the terms and conditions of the GNU General Public License,
8  * version 2 or later, as published by the Free Software Foundation.
9  *
10  * This program is distributed in the hope it will be useful, but WITHOUT
11  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
12  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
13  * more details.
14  *
15  * You should have received a copy of the GNU General Public License along with
16  * this program.  If not, see <http://www.gnu.org/licenses/>.
17  *
18  * Emulate a virtual board which works by passing Linux all the information
19  * it needs about what devices are present via the device tree.
20  * There are some restrictions about what we can do here:
21  *  + we can only present devices whose Linux drivers will work based
22  *    purely on the device tree with no platform data at all
23  *  + we want to present a very stripped-down minimalist platform,
24  *    both because this reduces the security attack surface from the guest
25  *    and also because it reduces our exposure to being broken when
26  *    the kernel updates its device tree bindings and requires further
27  *    information in a device binding that we aren't providing.
28  * This is essentially the same approach kvmtool uses.
29  */
30 
31 #include "qemu/osdep.h"
32 #include "qemu/datadir.h"
33 #include "qemu/units.h"
34 #include "qemu/option.h"
35 #include "monitor/qdev.h"
36 #include "hw/sysbus.h"
37 #include "hw/arm/boot.h"
38 #include "hw/arm/primecell.h"
39 #include "hw/arm/virt.h"
40 #include "hw/block/flash.h"
41 #include "hw/vfio/vfio-calxeda-xgmac.h"
42 #include "hw/vfio/vfio-amd-xgbe.h"
43 #include "hw/display/ramfb.h"
44 #include "net/net.h"
45 #include "sysemu/device_tree.h"
46 #include "sysemu/numa.h"
47 #include "sysemu/runstate.h"
48 #include "sysemu/tpm.h"
49 #include "sysemu/tcg.h"
50 #include "sysemu/kvm.h"
51 #include "sysemu/hvf.h"
52 #include "sysemu/qtest.h"
53 #include "hw/loader.h"
54 #include "qapi/error.h"
55 #include "qemu/bitops.h"
56 #include "qemu/error-report.h"
57 #include "qemu/module.h"
58 #include "hw/pci-host/gpex.h"
59 #include "hw/virtio/virtio-pci.h"
60 #include "hw/core/sysbus-fdt.h"
61 #include "hw/platform-bus.h"
62 #include "hw/qdev-properties.h"
63 #include "hw/arm/fdt.h"
64 #include "hw/intc/arm_gic.h"
65 #include "hw/intc/arm_gicv3_common.h"
66 #include "hw/intc/arm_gicv3_its_common.h"
67 #include "hw/irq.h"
68 #include "kvm_arm.h"
69 #include "hw/firmware/smbios.h"
70 #include "qapi/visitor.h"
71 #include "qapi/qapi-visit-common.h"
72 #include "qapi/qmp/qlist.h"
73 #include "standard-headers/linux/input.h"
74 #include "hw/arm/smmuv3.h"
75 #include "hw/acpi/acpi.h"
76 #include "target/arm/cpu-qom.h"
77 #include "target/arm/internals.h"
78 #include "target/arm/multiprocessing.h"
79 #include "target/arm/gtimer.h"
80 #include "hw/mem/pc-dimm.h"
81 #include "hw/mem/nvdimm.h"
82 #include "hw/acpi/generic_event_device.h"
83 #include "hw/virtio/virtio-md-pci.h"
84 #include "hw/virtio/virtio-iommu.h"
85 #include "hw/char/pl011.h"
86 #include "qemu/guest-random.h"
87 
88 #define DEFINE_VIRT_MACHINE_LATEST(major, minor, latest) \
89     static void virt_##major##_##minor##_class_init(ObjectClass *oc, \
90                                                     void *data) \
91     { \
92         MachineClass *mc = MACHINE_CLASS(oc); \
93         virt_machine_##major##_##minor##_options(mc); \
94         mc->desc = "QEMU " # major "." # minor " ARM Virtual Machine"; \
95         if (latest) { \
96             mc->alias = "virt"; \
97         } \
98     } \
99     static const TypeInfo machvirt_##major##_##minor##_info = { \
100         .name = MACHINE_TYPE_NAME("virt-" # major "." # minor), \
101         .parent = TYPE_VIRT_MACHINE, \
102         .class_init = virt_##major##_##minor##_class_init, \
103     }; \
104     static void machvirt_machine_##major##_##minor##_init(void) \
105     { \
106         type_register_static(&machvirt_##major##_##minor##_info); \
107     } \
108     type_init(machvirt_machine_##major##_##minor##_init);
109 
110 #define DEFINE_VIRT_MACHINE_AS_LATEST(major, minor) \
111     DEFINE_VIRT_MACHINE_LATEST(major, minor, true)
112 #define DEFINE_VIRT_MACHINE(major, minor) \
113     DEFINE_VIRT_MACHINE_LATEST(major, minor, false)
114 
115 
116 /* Number of external interrupt lines to configure the GIC with */
117 #define NUM_IRQS 256
118 
119 #define PLATFORM_BUS_NUM_IRQS 64
120 
121 /* Legacy RAM limit in GB (< version 4.0) */
122 #define LEGACY_RAMLIMIT_GB 255
123 #define LEGACY_RAMLIMIT_BYTES (LEGACY_RAMLIMIT_GB * GiB)
124 
125 /* Addresses and sizes of our components.
126  * 0..128MB is space for a flash device so we can run bootrom code such as UEFI.
127  * 128MB..256MB is used for miscellaneous device I/O.
128  * 256MB..1GB is reserved for possible future PCI support (ie where the
129  * PCI memory window will go if we add a PCI host controller).
130  * 1GB and up is RAM (which may happily spill over into the
131  * high memory region beyond 4GB).
132  * This represents a compromise between how much RAM can be given to
133  * a 32 bit VM and leaving space for expansion and in particular for PCI.
134  * Note that devices should generally be placed at multiples of 0x10000,
135  * to accommodate guests using 64K pages.
136  */
137 static const MemMapEntry base_memmap[] = {
138     /* Space up to 0x8000000 is reserved for a boot ROM */
139     [VIRT_FLASH] =              {          0, 0x08000000 },
140     [VIRT_CPUPERIPHS] =         { 0x08000000, 0x00020000 },
141     /* GIC distributor and CPU interfaces sit inside the CPU peripheral space */
142     [VIRT_GIC_DIST] =           { 0x08000000, 0x00010000 },
143     [VIRT_GIC_CPU] =            { 0x08010000, 0x00010000 },
144     [VIRT_GIC_V2M] =            { 0x08020000, 0x00001000 },
145     [VIRT_GIC_HYP] =            { 0x08030000, 0x00010000 },
146     [VIRT_GIC_VCPU] =           { 0x08040000, 0x00010000 },
147     /* The space in between here is reserved for GICv3 CPU/vCPU/HYP */
148     [VIRT_GIC_ITS] =            { 0x08080000, 0x00020000 },
149     /* This redistributor space allows up to 2*64kB*123 CPUs */
150     [VIRT_GIC_REDIST] =         { 0x080A0000, 0x00F60000 },
151     [VIRT_UART] =               { 0x09000000, 0x00001000 },
152     [VIRT_RTC] =                { 0x09010000, 0x00001000 },
153     [VIRT_FW_CFG] =             { 0x09020000, 0x00000018 },
154     [VIRT_GPIO] =               { 0x09030000, 0x00001000 },
155     [VIRT_SECURE_UART] =        { 0x09040000, 0x00001000 },
156     [VIRT_SMMU] =               { 0x09050000, 0x00020000 },
157     [VIRT_PCDIMM_ACPI] =        { 0x09070000, MEMORY_HOTPLUG_IO_LEN },
158     [VIRT_ACPI_GED] =           { 0x09080000, ACPI_GED_EVT_SEL_LEN },
159     [VIRT_NVDIMM_ACPI] =        { 0x09090000, NVDIMM_ACPI_IO_LEN},
160     [VIRT_PVTIME] =             { 0x090a0000, 0x00010000 },
161     [VIRT_SECURE_GPIO] =        { 0x090b0000, 0x00001000 },
162     [VIRT_MMIO] =               { 0x0a000000, 0x00000200 },
163     /* ...repeating for a total of NUM_VIRTIO_TRANSPORTS, each of that size */
164     [VIRT_PLATFORM_BUS] =       { 0x0c000000, 0x02000000 },
165     [VIRT_SECURE_MEM] =         { 0x0e000000, 0x01000000 },
166     [VIRT_PCIE_MMIO] =          { 0x10000000, 0x2eff0000 },
167     [VIRT_PCIE_PIO] =           { 0x3eff0000, 0x00010000 },
168     [VIRT_PCIE_ECAM] =          { 0x3f000000, 0x01000000 },
169     /* Actual RAM size depends on initial RAM and device memory settings */
170     [VIRT_MEM] =                { GiB, LEGACY_RAMLIMIT_BYTES },
171 };
172 
173 /*
174  * Highmem IO Regions: This memory map is floating, located after the RAM.
175  * Each MemMapEntry base (GPA) will be dynamically computed, depending on the
176  * top of the RAM, so that its base get the same alignment as the size,
177  * ie. a 512GiB entry will be aligned on a 512GiB boundary. If there is
178  * less than 256GiB of RAM, the floating area starts at the 256GiB mark.
179  * Note the extended_memmap is sized so that it eventually also includes the
180  * base_memmap entries (VIRT_HIGH_GIC_REDIST2 index is greater than the last
181  * index of base_memmap).
182  *
183  * The memory map for these Highmem IO Regions can be in legacy or compact
184  * layout, depending on 'compact-highmem' property. With legacy layout, the
185  * PA space for one specific region is always reserved, even if the region
186  * has been disabled or doesn't fit into the PA space. However, the PA space
187  * for the region won't be reserved in these circumstances with compact layout.
188  */
189 static MemMapEntry extended_memmap[] = {
190     /* Additional 64 MB redist region (can contain up to 512 redistributors) */
191     [VIRT_HIGH_GIC_REDIST2] =   { 0x0, 64 * MiB },
192     [VIRT_HIGH_PCIE_ECAM] =     { 0x0, 256 * MiB },
193     /* Second PCIe window */
194     [VIRT_HIGH_PCIE_MMIO] =     { 0x0, 512 * GiB },
195 };
196 
197 static const int a15irqmap[] = {
198     [VIRT_UART] = 1,
199     [VIRT_RTC] = 2,
200     [VIRT_PCIE] = 3, /* ... to 6 */
201     [VIRT_GPIO] = 7,
202     [VIRT_SECURE_UART] = 8,
203     [VIRT_ACPI_GED] = 9,
204     [VIRT_MMIO] = 16, /* ...to 16 + NUM_VIRTIO_TRANSPORTS - 1 */
205     [VIRT_GIC_V2M] = 48, /* ...to 48 + NUM_GICV2M_SPIS - 1 */
206     [VIRT_SMMU] = 74,    /* ...to 74 + NUM_SMMU_IRQS - 1 */
207     [VIRT_PLATFORM_BUS] = 112, /* ...to 112 + PLATFORM_BUS_NUM_IRQS -1 */
208 };
209 
210 static void create_randomness(MachineState *ms, const char *node)
211 {
212     struct {
213         uint64_t kaslr;
214         uint8_t rng[32];
215     } seed;
216 
217     if (qemu_guest_getrandom(&seed, sizeof(seed), NULL)) {
218         return;
219     }
220     qemu_fdt_setprop_u64(ms->fdt, node, "kaslr-seed", seed.kaslr);
221     qemu_fdt_setprop(ms->fdt, node, "rng-seed", seed.rng, sizeof(seed.rng));
222 }
223 
224 static void create_fdt(VirtMachineState *vms)
225 {
226     MachineState *ms = MACHINE(vms);
227     int nb_numa_nodes = ms->numa_state->num_nodes;
228     void *fdt = create_device_tree(&vms->fdt_size);
229 
230     if (!fdt) {
231         error_report("create_device_tree() failed");
232         exit(1);
233     }
234 
235     ms->fdt = fdt;
236 
237     /* Header */
238     qemu_fdt_setprop_string(fdt, "/", "compatible", "linux,dummy-virt");
239     qemu_fdt_setprop_cell(fdt, "/", "#address-cells", 0x2);
240     qemu_fdt_setprop_cell(fdt, "/", "#size-cells", 0x2);
241     qemu_fdt_setprop_string(fdt, "/", "model", "linux,dummy-virt");
242 
243     /* /chosen must exist for load_dtb to fill in necessary properties later */
244     qemu_fdt_add_subnode(fdt, "/chosen");
245     if (vms->dtb_randomness) {
246         create_randomness(ms, "/chosen");
247     }
248 
249     if (vms->secure) {
250         qemu_fdt_add_subnode(fdt, "/secure-chosen");
251         if (vms->dtb_randomness) {
252             create_randomness(ms, "/secure-chosen");
253         }
254     }
255 
256     /* Clock node, for the benefit of the UART. The kernel device tree
257      * binding documentation claims the PL011 node clock properties are
258      * optional but in practice if you omit them the kernel refuses to
259      * probe for the device.
260      */
261     vms->clock_phandle = qemu_fdt_alloc_phandle(fdt);
262     qemu_fdt_add_subnode(fdt, "/apb-pclk");
263     qemu_fdt_setprop_string(fdt, "/apb-pclk", "compatible", "fixed-clock");
264     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "#clock-cells", 0x0);
265     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "clock-frequency", 24000000);
266     qemu_fdt_setprop_string(fdt, "/apb-pclk", "clock-output-names",
267                                 "clk24mhz");
268     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "phandle", vms->clock_phandle);
269 
270     if (nb_numa_nodes > 0 && ms->numa_state->have_numa_distance) {
271         int size = nb_numa_nodes * nb_numa_nodes * 3 * sizeof(uint32_t);
272         uint32_t *matrix = g_malloc0(size);
273         int idx, i, j;
274 
275         for (i = 0; i < nb_numa_nodes; i++) {
276             for (j = 0; j < nb_numa_nodes; j++) {
277                 idx = (i * nb_numa_nodes + j) * 3;
278                 matrix[idx + 0] = cpu_to_be32(i);
279                 matrix[idx + 1] = cpu_to_be32(j);
280                 matrix[idx + 2] =
281                     cpu_to_be32(ms->numa_state->nodes[i].distance[j]);
282             }
283         }
284 
285         qemu_fdt_add_subnode(fdt, "/distance-map");
286         qemu_fdt_setprop_string(fdt, "/distance-map", "compatible",
287                                 "numa-distance-map-v1");
288         qemu_fdt_setprop(fdt, "/distance-map", "distance-matrix",
289                          matrix, size);
290         g_free(matrix);
291     }
292 }
293 
294 static void fdt_add_timer_nodes(const VirtMachineState *vms)
295 {
296     /* On real hardware these interrupts are level-triggered.
297      * On KVM they were edge-triggered before host kernel version 4.4,
298      * and level-triggered afterwards.
299      * On emulated QEMU they are level-triggered.
300      *
301      * Getting the DTB info about them wrong is awkward for some
302      * guest kernels:
303      *  pre-4.8 ignore the DT and leave the interrupt configured
304      *   with whatever the GIC reset value (or the bootloader) left it at
305      *  4.8 before rc6 honour the incorrect data by programming it back
306      *   into the GIC, causing problems
307      *  4.8rc6 and later ignore the DT and always write "level triggered"
308      *   into the GIC
309      *
310      * For backwards-compatibility, virt-2.8 and earlier will continue
311      * to say these are edge-triggered, but later machines will report
312      * the correct information.
313      */
314     ARMCPU *armcpu;
315     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
316     uint32_t irqflags = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
317     MachineState *ms = MACHINE(vms);
318 
319     if (vmc->claim_edge_triggered_timers) {
320         irqflags = GIC_FDT_IRQ_FLAGS_EDGE_LO_HI;
321     }
322 
323     if (vms->gic_version == VIRT_GIC_VERSION_2) {
324         irqflags = deposit32(irqflags, GIC_FDT_IRQ_PPI_CPU_START,
325                              GIC_FDT_IRQ_PPI_CPU_WIDTH,
326                              (1 << MACHINE(vms)->smp.cpus) - 1);
327     }
328 
329     qemu_fdt_add_subnode(ms->fdt, "/timer");
330 
331     armcpu = ARM_CPU(qemu_get_cpu(0));
332     if (arm_feature(&armcpu->env, ARM_FEATURE_V8)) {
333         const char compat[] = "arm,armv8-timer\0arm,armv7-timer";
334         qemu_fdt_setprop(ms->fdt, "/timer", "compatible",
335                          compat, sizeof(compat));
336     } else {
337         qemu_fdt_setprop_string(ms->fdt, "/timer", "compatible",
338                                 "arm,armv7-timer");
339     }
340     qemu_fdt_setprop(ms->fdt, "/timer", "always-on", NULL, 0);
341     qemu_fdt_setprop_cells(ms->fdt, "/timer", "interrupts",
342                            GIC_FDT_IRQ_TYPE_PPI,
343                            INTID_TO_PPI(ARCH_TIMER_S_EL1_IRQ), irqflags,
344                            GIC_FDT_IRQ_TYPE_PPI,
345                            INTID_TO_PPI(ARCH_TIMER_NS_EL1_IRQ), irqflags,
346                            GIC_FDT_IRQ_TYPE_PPI,
347                            INTID_TO_PPI(ARCH_TIMER_VIRT_IRQ), irqflags,
348                            GIC_FDT_IRQ_TYPE_PPI,
349                            INTID_TO_PPI(ARCH_TIMER_NS_EL2_IRQ), irqflags);
350 }
351 
352 static void fdt_add_cpu_nodes(const VirtMachineState *vms)
353 {
354     int cpu;
355     int addr_cells = 1;
356     const MachineState *ms = MACHINE(vms);
357     const VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
358     int smp_cpus = ms->smp.cpus;
359 
360     /*
361      * See Linux Documentation/devicetree/bindings/arm/cpus.yaml
362      * On ARM v8 64-bit systems value should be set to 2,
363      * that corresponds to the MPIDR_EL1 register size.
364      * If MPIDR_EL1[63:32] value is equal to 0 on all CPUs
365      * in the system, #address-cells can be set to 1, since
366      * MPIDR_EL1[63:32] bits are not used for CPUs
367      * identification.
368      *
369      * Here we actually don't know whether our system is 32- or 64-bit one.
370      * The simplest way to go is to examine affinity IDs of all our CPUs. If
371      * at least one of them has Aff3 populated, we set #address-cells to 2.
372      */
373     for (cpu = 0; cpu < smp_cpus; cpu++) {
374         ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(cpu));
375 
376         if (arm_cpu_mp_affinity(armcpu) & ARM_AFF3_MASK) {
377             addr_cells = 2;
378             break;
379         }
380     }
381 
382     qemu_fdt_add_subnode(ms->fdt, "/cpus");
383     qemu_fdt_setprop_cell(ms->fdt, "/cpus", "#address-cells", addr_cells);
384     qemu_fdt_setprop_cell(ms->fdt, "/cpus", "#size-cells", 0x0);
385 
386     for (cpu = smp_cpus - 1; cpu >= 0; cpu--) {
387         char *nodename = g_strdup_printf("/cpus/cpu@%d", cpu);
388         ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(cpu));
389         CPUState *cs = CPU(armcpu);
390 
391         qemu_fdt_add_subnode(ms->fdt, nodename);
392         qemu_fdt_setprop_string(ms->fdt, nodename, "device_type", "cpu");
393         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
394                                     armcpu->dtb_compatible);
395 
396         if (vms->psci_conduit != QEMU_PSCI_CONDUIT_DISABLED && smp_cpus > 1) {
397             qemu_fdt_setprop_string(ms->fdt, nodename,
398                                         "enable-method", "psci");
399         }
400 
401         if (addr_cells == 2) {
402             qemu_fdt_setprop_u64(ms->fdt, nodename, "reg",
403                                  arm_cpu_mp_affinity(armcpu));
404         } else {
405             qemu_fdt_setprop_cell(ms->fdt, nodename, "reg",
406                                   arm_cpu_mp_affinity(armcpu));
407         }
408 
409         if (ms->possible_cpus->cpus[cs->cpu_index].props.has_node_id) {
410             qemu_fdt_setprop_cell(ms->fdt, nodename, "numa-node-id",
411                 ms->possible_cpus->cpus[cs->cpu_index].props.node_id);
412         }
413 
414         if (!vmc->no_cpu_topology) {
415             qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle",
416                                   qemu_fdt_alloc_phandle(ms->fdt));
417         }
418 
419         g_free(nodename);
420     }
421 
422     if (!vmc->no_cpu_topology) {
423         /*
424          * Add vCPU topology description through fdt node cpu-map.
425          *
426          * See Linux Documentation/devicetree/bindings/cpu/cpu-topology.txt
427          * In a SMP system, the hierarchy of CPUs can be defined through
428          * four entities that are used to describe the layout of CPUs in
429          * the system: socket/cluster/core/thread.
430          *
431          * A socket node represents the boundary of system physical package
432          * and its child nodes must be one or more cluster nodes. A system
433          * can contain several layers of clustering within a single physical
434          * package and cluster nodes can be contained in parent cluster nodes.
435          *
436          * Note: currently we only support one layer of clustering within
437          * each physical package.
438          */
439         qemu_fdt_add_subnode(ms->fdt, "/cpus/cpu-map");
440 
441         for (cpu = smp_cpus - 1; cpu >= 0; cpu--) {
442             char *cpu_path = g_strdup_printf("/cpus/cpu@%d", cpu);
443             char *map_path;
444 
445             if (ms->smp.threads > 1) {
446                 map_path = g_strdup_printf(
447                     "/cpus/cpu-map/socket%d/cluster%d/core%d/thread%d",
448                     cpu / (ms->smp.clusters * ms->smp.cores * ms->smp.threads),
449                     (cpu / (ms->smp.cores * ms->smp.threads)) % ms->smp.clusters,
450                     (cpu / ms->smp.threads) % ms->smp.cores,
451                     cpu % ms->smp.threads);
452             } else {
453                 map_path = g_strdup_printf(
454                     "/cpus/cpu-map/socket%d/cluster%d/core%d",
455                     cpu / (ms->smp.clusters * ms->smp.cores),
456                     (cpu / ms->smp.cores) % ms->smp.clusters,
457                     cpu % ms->smp.cores);
458             }
459             qemu_fdt_add_path(ms->fdt, map_path);
460             qemu_fdt_setprop_phandle(ms->fdt, map_path, "cpu", cpu_path);
461 
462             g_free(map_path);
463             g_free(cpu_path);
464         }
465     }
466 }
467 
468 static void fdt_add_its_gic_node(VirtMachineState *vms)
469 {
470     char *nodename;
471     MachineState *ms = MACHINE(vms);
472 
473     vms->msi_phandle = qemu_fdt_alloc_phandle(ms->fdt);
474     nodename = g_strdup_printf("/intc/its@%" PRIx64,
475                                vms->memmap[VIRT_GIC_ITS].base);
476     qemu_fdt_add_subnode(ms->fdt, nodename);
477     qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
478                             "arm,gic-v3-its");
479     qemu_fdt_setprop(ms->fdt, nodename, "msi-controller", NULL, 0);
480     qemu_fdt_setprop_cell(ms->fdt, nodename, "#msi-cells", 1);
481     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
482                                  2, vms->memmap[VIRT_GIC_ITS].base,
483                                  2, vms->memmap[VIRT_GIC_ITS].size);
484     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", vms->msi_phandle);
485     g_free(nodename);
486 }
487 
488 static void fdt_add_v2m_gic_node(VirtMachineState *vms)
489 {
490     MachineState *ms = MACHINE(vms);
491     char *nodename;
492 
493     nodename = g_strdup_printf("/intc/v2m@%" PRIx64,
494                                vms->memmap[VIRT_GIC_V2M].base);
495     vms->msi_phandle = qemu_fdt_alloc_phandle(ms->fdt);
496     qemu_fdt_add_subnode(ms->fdt, nodename);
497     qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
498                             "arm,gic-v2m-frame");
499     qemu_fdt_setprop(ms->fdt, nodename, "msi-controller", NULL, 0);
500     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
501                                  2, vms->memmap[VIRT_GIC_V2M].base,
502                                  2, vms->memmap[VIRT_GIC_V2M].size);
503     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", vms->msi_phandle);
504     g_free(nodename);
505 }
506 
507 static void fdt_add_gic_node(VirtMachineState *vms)
508 {
509     MachineState *ms = MACHINE(vms);
510     char *nodename;
511 
512     vms->gic_phandle = qemu_fdt_alloc_phandle(ms->fdt);
513     qemu_fdt_setprop_cell(ms->fdt, "/", "interrupt-parent", vms->gic_phandle);
514 
515     nodename = g_strdup_printf("/intc@%" PRIx64,
516                                vms->memmap[VIRT_GIC_DIST].base);
517     qemu_fdt_add_subnode(ms->fdt, nodename);
518     qemu_fdt_setprop_cell(ms->fdt, nodename, "#interrupt-cells", 3);
519     qemu_fdt_setprop(ms->fdt, nodename, "interrupt-controller", NULL, 0);
520     qemu_fdt_setprop_cell(ms->fdt, nodename, "#address-cells", 0x2);
521     qemu_fdt_setprop_cell(ms->fdt, nodename, "#size-cells", 0x2);
522     qemu_fdt_setprop(ms->fdt, nodename, "ranges", NULL, 0);
523     if (vms->gic_version != VIRT_GIC_VERSION_2) {
524         int nb_redist_regions = virt_gicv3_redist_region_count(vms);
525 
526         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
527                                 "arm,gic-v3");
528 
529         qemu_fdt_setprop_cell(ms->fdt, nodename,
530                               "#redistributor-regions", nb_redist_regions);
531 
532         if (nb_redist_regions == 1) {
533             qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
534                                          2, vms->memmap[VIRT_GIC_DIST].base,
535                                          2, vms->memmap[VIRT_GIC_DIST].size,
536                                          2, vms->memmap[VIRT_GIC_REDIST].base,
537                                          2, vms->memmap[VIRT_GIC_REDIST].size);
538         } else {
539             qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
540                                  2, vms->memmap[VIRT_GIC_DIST].base,
541                                  2, vms->memmap[VIRT_GIC_DIST].size,
542                                  2, vms->memmap[VIRT_GIC_REDIST].base,
543                                  2, vms->memmap[VIRT_GIC_REDIST].size,
544                                  2, vms->memmap[VIRT_HIGH_GIC_REDIST2].base,
545                                  2, vms->memmap[VIRT_HIGH_GIC_REDIST2].size);
546         }
547 
548         if (vms->virt) {
549             qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
550                                    GIC_FDT_IRQ_TYPE_PPI,
551                                    INTID_TO_PPI(ARCH_GIC_MAINT_IRQ),
552                                    GIC_FDT_IRQ_FLAGS_LEVEL_HI);
553         }
554     } else {
555         /* 'cortex-a15-gic' means 'GIC v2' */
556         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
557                                 "arm,cortex-a15-gic");
558         if (!vms->virt) {
559             qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
560                                          2, vms->memmap[VIRT_GIC_DIST].base,
561                                          2, vms->memmap[VIRT_GIC_DIST].size,
562                                          2, vms->memmap[VIRT_GIC_CPU].base,
563                                          2, vms->memmap[VIRT_GIC_CPU].size);
564         } else {
565             qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
566                                          2, vms->memmap[VIRT_GIC_DIST].base,
567                                          2, vms->memmap[VIRT_GIC_DIST].size,
568                                          2, vms->memmap[VIRT_GIC_CPU].base,
569                                          2, vms->memmap[VIRT_GIC_CPU].size,
570                                          2, vms->memmap[VIRT_GIC_HYP].base,
571                                          2, vms->memmap[VIRT_GIC_HYP].size,
572                                          2, vms->memmap[VIRT_GIC_VCPU].base,
573                                          2, vms->memmap[VIRT_GIC_VCPU].size);
574             qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
575                                    GIC_FDT_IRQ_TYPE_PPI,
576                                    INTID_TO_PPI(ARCH_GIC_MAINT_IRQ),
577                                    GIC_FDT_IRQ_FLAGS_LEVEL_HI);
578         }
579     }
580 
581     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", vms->gic_phandle);
582     g_free(nodename);
583 }
584 
585 static void fdt_add_pmu_nodes(const VirtMachineState *vms)
586 {
587     ARMCPU *armcpu = ARM_CPU(first_cpu);
588     uint32_t irqflags = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
589     MachineState *ms = MACHINE(vms);
590 
591     if (!arm_feature(&armcpu->env, ARM_FEATURE_PMU)) {
592         assert(!object_property_get_bool(OBJECT(armcpu), "pmu", NULL));
593         return;
594     }
595 
596     if (vms->gic_version == VIRT_GIC_VERSION_2) {
597         irqflags = deposit32(irqflags, GIC_FDT_IRQ_PPI_CPU_START,
598                              GIC_FDT_IRQ_PPI_CPU_WIDTH,
599                              (1 << MACHINE(vms)->smp.cpus) - 1);
600     }
601 
602     qemu_fdt_add_subnode(ms->fdt, "/pmu");
603     if (arm_feature(&armcpu->env, ARM_FEATURE_V8)) {
604         const char compat[] = "arm,armv8-pmuv3";
605         qemu_fdt_setprop(ms->fdt, "/pmu", "compatible",
606                          compat, sizeof(compat));
607         qemu_fdt_setprop_cells(ms->fdt, "/pmu", "interrupts",
608                                GIC_FDT_IRQ_TYPE_PPI,
609                                INTID_TO_PPI(VIRTUAL_PMU_IRQ), irqflags);
610     }
611 }
612 
613 static inline DeviceState *create_acpi_ged(VirtMachineState *vms)
614 {
615     DeviceState *dev;
616     MachineState *ms = MACHINE(vms);
617     int irq = vms->irqmap[VIRT_ACPI_GED];
618     uint32_t event = ACPI_GED_PWR_DOWN_EVT;
619 
620     if (ms->ram_slots) {
621         event |= ACPI_GED_MEM_HOTPLUG_EVT;
622     }
623 
624     if (ms->nvdimms_state->is_enabled) {
625         event |= ACPI_GED_NVDIMM_HOTPLUG_EVT;
626     }
627 
628     dev = qdev_new(TYPE_ACPI_GED);
629     qdev_prop_set_uint32(dev, "ged-event", event);
630     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
631 
632     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_ACPI_GED].base);
633     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 1, vms->memmap[VIRT_PCDIMM_ACPI].base);
634     sysbus_connect_irq(SYS_BUS_DEVICE(dev), 0, qdev_get_gpio_in(vms->gic, irq));
635 
636     return dev;
637 }
638 
639 static void create_its(VirtMachineState *vms)
640 {
641     const char *itsclass = its_class_name();
642     DeviceState *dev;
643 
644     if (!strcmp(itsclass, "arm-gicv3-its")) {
645         if (!vms->tcg_its) {
646             itsclass = NULL;
647         }
648     }
649 
650     if (!itsclass) {
651         /* Do nothing if not supported */
652         return;
653     }
654 
655     dev = qdev_new(itsclass);
656 
657     object_property_set_link(OBJECT(dev), "parent-gicv3", OBJECT(vms->gic),
658                              &error_abort);
659     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
660     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_GIC_ITS].base);
661 
662     fdt_add_its_gic_node(vms);
663     vms->msi_controller = VIRT_MSI_CTRL_ITS;
664 }
665 
666 static void create_v2m(VirtMachineState *vms)
667 {
668     int i;
669     int irq = vms->irqmap[VIRT_GIC_V2M];
670     DeviceState *dev;
671 
672     dev = qdev_new("arm-gicv2m");
673     qdev_prop_set_uint32(dev, "base-spi", irq);
674     qdev_prop_set_uint32(dev, "num-spi", NUM_GICV2M_SPIS);
675     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
676     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_GIC_V2M].base);
677 
678     for (i = 0; i < NUM_GICV2M_SPIS; i++) {
679         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
680                            qdev_get_gpio_in(vms->gic, irq + i));
681     }
682 
683     fdt_add_v2m_gic_node(vms);
684     vms->msi_controller = VIRT_MSI_CTRL_GICV2M;
685 }
686 
687 static void create_gic(VirtMachineState *vms, MemoryRegion *mem)
688 {
689     MachineState *ms = MACHINE(vms);
690     /* We create a standalone GIC */
691     SysBusDevice *gicbusdev;
692     const char *gictype;
693     int i;
694     unsigned int smp_cpus = ms->smp.cpus;
695     uint32_t nb_redist_regions = 0;
696     int revision;
697 
698     if (vms->gic_version == VIRT_GIC_VERSION_2) {
699         gictype = gic_class_name();
700     } else {
701         gictype = gicv3_class_name();
702     }
703 
704     switch (vms->gic_version) {
705     case VIRT_GIC_VERSION_2:
706         revision = 2;
707         break;
708     case VIRT_GIC_VERSION_3:
709         revision = 3;
710         break;
711     case VIRT_GIC_VERSION_4:
712         revision = 4;
713         break;
714     default:
715         g_assert_not_reached();
716     }
717     vms->gic = qdev_new(gictype);
718     qdev_prop_set_uint32(vms->gic, "revision", revision);
719     qdev_prop_set_uint32(vms->gic, "num-cpu", smp_cpus);
720     /* Note that the num-irq property counts both internal and external
721      * interrupts; there are always 32 of the former (mandated by GIC spec).
722      */
723     qdev_prop_set_uint32(vms->gic, "num-irq", NUM_IRQS + 32);
724     if (!kvm_irqchip_in_kernel()) {
725         qdev_prop_set_bit(vms->gic, "has-security-extensions", vms->secure);
726     }
727 
728     if (vms->gic_version != VIRT_GIC_VERSION_2) {
729         QList *redist_region_count;
730         uint32_t redist0_capacity = virt_redist_capacity(vms, VIRT_GIC_REDIST);
731         uint32_t redist0_count = MIN(smp_cpus, redist0_capacity);
732 
733         nb_redist_regions = virt_gicv3_redist_region_count(vms);
734 
735         redist_region_count = qlist_new();
736         qlist_append_int(redist_region_count, redist0_count);
737         if (nb_redist_regions == 2) {
738             uint32_t redist1_capacity =
739                 virt_redist_capacity(vms, VIRT_HIGH_GIC_REDIST2);
740 
741             qlist_append_int(redist_region_count,
742                 MIN(smp_cpus - redist0_count, redist1_capacity));
743         }
744         qdev_prop_set_array(vms->gic, "redist-region-count",
745                             redist_region_count);
746 
747         if (!kvm_irqchip_in_kernel()) {
748             if (vms->tcg_its) {
749                 object_property_set_link(OBJECT(vms->gic), "sysmem",
750                                          OBJECT(mem), &error_fatal);
751                 qdev_prop_set_bit(vms->gic, "has-lpi", true);
752             }
753         }
754     } else {
755         if (!kvm_irqchip_in_kernel()) {
756             qdev_prop_set_bit(vms->gic, "has-virtualization-extensions",
757                               vms->virt);
758         }
759     }
760     gicbusdev = SYS_BUS_DEVICE(vms->gic);
761     sysbus_realize_and_unref(gicbusdev, &error_fatal);
762     sysbus_mmio_map(gicbusdev, 0, vms->memmap[VIRT_GIC_DIST].base);
763     if (vms->gic_version != VIRT_GIC_VERSION_2) {
764         sysbus_mmio_map(gicbusdev, 1, vms->memmap[VIRT_GIC_REDIST].base);
765         if (nb_redist_regions == 2) {
766             sysbus_mmio_map(gicbusdev, 2,
767                             vms->memmap[VIRT_HIGH_GIC_REDIST2].base);
768         }
769     } else {
770         sysbus_mmio_map(gicbusdev, 1, vms->memmap[VIRT_GIC_CPU].base);
771         if (vms->virt) {
772             sysbus_mmio_map(gicbusdev, 2, vms->memmap[VIRT_GIC_HYP].base);
773             sysbus_mmio_map(gicbusdev, 3, vms->memmap[VIRT_GIC_VCPU].base);
774         }
775     }
776 
777     /* Wire the outputs from each CPU's generic timer and the GICv3
778      * maintenance interrupt signal to the appropriate GIC PPI inputs,
779      * and the GIC's IRQ/FIQ/VIRQ/VFIQ interrupt outputs to the CPU's inputs.
780      */
781     for (i = 0; i < smp_cpus; i++) {
782         DeviceState *cpudev = DEVICE(qemu_get_cpu(i));
783         int intidbase = NUM_IRQS + i * GIC_INTERNAL;
784         /* Mapping from the output timer irq lines from the CPU to the
785          * GIC PPI inputs we use for the virt board.
786          */
787         const int timer_irq[] = {
788             [GTIMER_PHYS] = ARCH_TIMER_NS_EL1_IRQ,
789             [GTIMER_VIRT] = ARCH_TIMER_VIRT_IRQ,
790             [GTIMER_HYP]  = ARCH_TIMER_NS_EL2_IRQ,
791             [GTIMER_SEC]  = ARCH_TIMER_S_EL1_IRQ,
792         };
793 
794         for (unsigned irq = 0; irq < ARRAY_SIZE(timer_irq); irq++) {
795             qdev_connect_gpio_out(cpudev, irq,
796                                   qdev_get_gpio_in(vms->gic,
797                                                    intidbase + timer_irq[irq]));
798         }
799 
800         if (vms->gic_version != VIRT_GIC_VERSION_2) {
801             qemu_irq irq = qdev_get_gpio_in(vms->gic,
802                                             intidbase + ARCH_GIC_MAINT_IRQ);
803             qdev_connect_gpio_out_named(cpudev, "gicv3-maintenance-interrupt",
804                                         0, irq);
805         } else if (vms->virt) {
806             qemu_irq irq = qdev_get_gpio_in(vms->gic,
807                                             intidbase + ARCH_GIC_MAINT_IRQ);
808             sysbus_connect_irq(gicbusdev, i + 4 * smp_cpus, irq);
809         }
810 
811         qdev_connect_gpio_out_named(cpudev, "pmu-interrupt", 0,
812                                     qdev_get_gpio_in(vms->gic, intidbase
813                                                      + VIRTUAL_PMU_IRQ));
814 
815         sysbus_connect_irq(gicbusdev, i, qdev_get_gpio_in(cpudev, ARM_CPU_IRQ));
816         sysbus_connect_irq(gicbusdev, i + smp_cpus,
817                            qdev_get_gpio_in(cpudev, ARM_CPU_FIQ));
818         sysbus_connect_irq(gicbusdev, i + 2 * smp_cpus,
819                            qdev_get_gpio_in(cpudev, ARM_CPU_VIRQ));
820         sysbus_connect_irq(gicbusdev, i + 3 * smp_cpus,
821                            qdev_get_gpio_in(cpudev, ARM_CPU_VFIQ));
822     }
823 
824     fdt_add_gic_node(vms);
825 
826     if (vms->gic_version != VIRT_GIC_VERSION_2 && vms->its) {
827         create_its(vms);
828     } else if (vms->gic_version == VIRT_GIC_VERSION_2) {
829         create_v2m(vms);
830     }
831 }
832 
833 static void create_uart(const VirtMachineState *vms, int uart,
834                         MemoryRegion *mem, Chardev *chr)
835 {
836     char *nodename;
837     hwaddr base = vms->memmap[uart].base;
838     hwaddr size = vms->memmap[uart].size;
839     int irq = vms->irqmap[uart];
840     const char compat[] = "arm,pl011\0arm,primecell";
841     const char clocknames[] = "uartclk\0apb_pclk";
842     DeviceState *dev = qdev_new(TYPE_PL011);
843     SysBusDevice *s = SYS_BUS_DEVICE(dev);
844     MachineState *ms = MACHINE(vms);
845 
846     qdev_prop_set_chr(dev, "chardev", chr);
847     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
848     memory_region_add_subregion(mem, base,
849                                 sysbus_mmio_get_region(s, 0));
850     sysbus_connect_irq(s, 0, qdev_get_gpio_in(vms->gic, irq));
851 
852     nodename = g_strdup_printf("/pl011@%" PRIx64, base);
853     qemu_fdt_add_subnode(ms->fdt, nodename);
854     /* Note that we can't use setprop_string because of the embedded NUL */
855     qemu_fdt_setprop(ms->fdt, nodename, "compatible",
856                          compat, sizeof(compat));
857     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
858                                      2, base, 2, size);
859     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
860                                GIC_FDT_IRQ_TYPE_SPI, irq,
861                                GIC_FDT_IRQ_FLAGS_LEVEL_HI);
862     qemu_fdt_setprop_cells(ms->fdt, nodename, "clocks",
863                                vms->clock_phandle, vms->clock_phandle);
864     qemu_fdt_setprop(ms->fdt, nodename, "clock-names",
865                          clocknames, sizeof(clocknames));
866 
867     if (uart == VIRT_UART) {
868         qemu_fdt_setprop_string(ms->fdt, "/chosen", "stdout-path", nodename);
869     } else {
870         /* Mark as not usable by the normal world */
871         qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
872         qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
873 
874         qemu_fdt_setprop_string(ms->fdt, "/secure-chosen", "stdout-path",
875                                 nodename);
876     }
877 
878     g_free(nodename);
879 }
880 
881 static void create_rtc(const VirtMachineState *vms)
882 {
883     char *nodename;
884     hwaddr base = vms->memmap[VIRT_RTC].base;
885     hwaddr size = vms->memmap[VIRT_RTC].size;
886     int irq = vms->irqmap[VIRT_RTC];
887     const char compat[] = "arm,pl031\0arm,primecell";
888     MachineState *ms = MACHINE(vms);
889 
890     sysbus_create_simple("pl031", base, qdev_get_gpio_in(vms->gic, irq));
891 
892     nodename = g_strdup_printf("/pl031@%" PRIx64, base);
893     qemu_fdt_add_subnode(ms->fdt, nodename);
894     qemu_fdt_setprop(ms->fdt, nodename, "compatible", compat, sizeof(compat));
895     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
896                                  2, base, 2, size);
897     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
898                            GIC_FDT_IRQ_TYPE_SPI, irq,
899                            GIC_FDT_IRQ_FLAGS_LEVEL_HI);
900     qemu_fdt_setprop_cell(ms->fdt, nodename, "clocks", vms->clock_phandle);
901     qemu_fdt_setprop_string(ms->fdt, nodename, "clock-names", "apb_pclk");
902     g_free(nodename);
903 }
904 
905 static DeviceState *gpio_key_dev;
906 static void virt_powerdown_req(Notifier *n, void *opaque)
907 {
908     VirtMachineState *s = container_of(n, VirtMachineState, powerdown_notifier);
909 
910     if (s->acpi_dev) {
911         acpi_send_event(s->acpi_dev, ACPI_POWER_DOWN_STATUS);
912     } else {
913         /* use gpio Pin 3 for power button event */
914         qemu_set_irq(qdev_get_gpio_in(gpio_key_dev, 0), 1);
915     }
916 }
917 
918 static void create_gpio_keys(char *fdt, DeviceState *pl061_dev,
919                              uint32_t phandle)
920 {
921     gpio_key_dev = sysbus_create_simple("gpio-key", -1,
922                                         qdev_get_gpio_in(pl061_dev, 3));
923 
924     qemu_fdt_add_subnode(fdt, "/gpio-keys");
925     qemu_fdt_setprop_string(fdt, "/gpio-keys", "compatible", "gpio-keys");
926 
927     qemu_fdt_add_subnode(fdt, "/gpio-keys/poweroff");
928     qemu_fdt_setprop_string(fdt, "/gpio-keys/poweroff",
929                             "label", "GPIO Key Poweroff");
930     qemu_fdt_setprop_cell(fdt, "/gpio-keys/poweroff", "linux,code",
931                           KEY_POWER);
932     qemu_fdt_setprop_cells(fdt, "/gpio-keys/poweroff",
933                            "gpios", phandle, 3, 0);
934 }
935 
936 #define SECURE_GPIO_POWEROFF 0
937 #define SECURE_GPIO_RESET    1
938 
939 static void create_secure_gpio_pwr(char *fdt, DeviceState *pl061_dev,
940                                    uint32_t phandle)
941 {
942     DeviceState *gpio_pwr_dev;
943 
944     /* gpio-pwr */
945     gpio_pwr_dev = sysbus_create_simple("gpio-pwr", -1, NULL);
946 
947     /* connect secure pl061 to gpio-pwr */
948     qdev_connect_gpio_out(pl061_dev, SECURE_GPIO_RESET,
949                           qdev_get_gpio_in_named(gpio_pwr_dev, "reset", 0));
950     qdev_connect_gpio_out(pl061_dev, SECURE_GPIO_POWEROFF,
951                           qdev_get_gpio_in_named(gpio_pwr_dev, "shutdown", 0));
952 
953     qemu_fdt_add_subnode(fdt, "/gpio-poweroff");
954     qemu_fdt_setprop_string(fdt, "/gpio-poweroff", "compatible",
955                             "gpio-poweroff");
956     qemu_fdt_setprop_cells(fdt, "/gpio-poweroff",
957                            "gpios", phandle, SECURE_GPIO_POWEROFF, 0);
958     qemu_fdt_setprop_string(fdt, "/gpio-poweroff", "status", "disabled");
959     qemu_fdt_setprop_string(fdt, "/gpio-poweroff", "secure-status",
960                             "okay");
961 
962     qemu_fdt_add_subnode(fdt, "/gpio-restart");
963     qemu_fdt_setprop_string(fdt, "/gpio-restart", "compatible",
964                             "gpio-restart");
965     qemu_fdt_setprop_cells(fdt, "/gpio-restart",
966                            "gpios", phandle, SECURE_GPIO_RESET, 0);
967     qemu_fdt_setprop_string(fdt, "/gpio-restart", "status", "disabled");
968     qemu_fdt_setprop_string(fdt, "/gpio-restart", "secure-status",
969                             "okay");
970 }
971 
972 static void create_gpio_devices(const VirtMachineState *vms, int gpio,
973                                 MemoryRegion *mem)
974 {
975     char *nodename;
976     DeviceState *pl061_dev;
977     hwaddr base = vms->memmap[gpio].base;
978     hwaddr size = vms->memmap[gpio].size;
979     int irq = vms->irqmap[gpio];
980     const char compat[] = "arm,pl061\0arm,primecell";
981     SysBusDevice *s;
982     MachineState *ms = MACHINE(vms);
983 
984     pl061_dev = qdev_new("pl061");
985     /* Pull lines down to 0 if not driven by the PL061 */
986     qdev_prop_set_uint32(pl061_dev, "pullups", 0);
987     qdev_prop_set_uint32(pl061_dev, "pulldowns", 0xff);
988     s = SYS_BUS_DEVICE(pl061_dev);
989     sysbus_realize_and_unref(s, &error_fatal);
990     memory_region_add_subregion(mem, base, sysbus_mmio_get_region(s, 0));
991     sysbus_connect_irq(s, 0, qdev_get_gpio_in(vms->gic, irq));
992 
993     uint32_t phandle = qemu_fdt_alloc_phandle(ms->fdt);
994     nodename = g_strdup_printf("/pl061@%" PRIx64, base);
995     qemu_fdt_add_subnode(ms->fdt, nodename);
996     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
997                                  2, base, 2, size);
998     qemu_fdt_setprop(ms->fdt, nodename, "compatible", compat, sizeof(compat));
999     qemu_fdt_setprop_cell(ms->fdt, nodename, "#gpio-cells", 2);
1000     qemu_fdt_setprop(ms->fdt, nodename, "gpio-controller", NULL, 0);
1001     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
1002                            GIC_FDT_IRQ_TYPE_SPI, irq,
1003                            GIC_FDT_IRQ_FLAGS_LEVEL_HI);
1004     qemu_fdt_setprop_cell(ms->fdt, nodename, "clocks", vms->clock_phandle);
1005     qemu_fdt_setprop_string(ms->fdt, nodename, "clock-names", "apb_pclk");
1006     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", phandle);
1007 
1008     if (gpio != VIRT_GPIO) {
1009         /* Mark as not usable by the normal world */
1010         qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
1011         qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
1012     }
1013     g_free(nodename);
1014 
1015     /* Child gpio devices */
1016     if (gpio == VIRT_GPIO) {
1017         create_gpio_keys(ms->fdt, pl061_dev, phandle);
1018     } else {
1019         create_secure_gpio_pwr(ms->fdt, pl061_dev, phandle);
1020     }
1021 }
1022 
1023 static void create_virtio_devices(const VirtMachineState *vms)
1024 {
1025     int i;
1026     hwaddr size = vms->memmap[VIRT_MMIO].size;
1027     MachineState *ms = MACHINE(vms);
1028 
1029     /* We create the transports in forwards order. Since qbus_realize()
1030      * prepends (not appends) new child buses, the incrementing loop below will
1031      * create a list of virtio-mmio buses with decreasing base addresses.
1032      *
1033      * When a -device option is processed from the command line,
1034      * qbus_find_recursive() picks the next free virtio-mmio bus in forwards
1035      * order. The upshot is that -device options in increasing command line
1036      * order are mapped to virtio-mmio buses with decreasing base addresses.
1037      *
1038      * When this code was originally written, that arrangement ensured that the
1039      * guest Linux kernel would give the lowest "name" (/dev/vda, eth0, etc) to
1040      * the first -device on the command line. (The end-to-end order is a
1041      * function of this loop, qbus_realize(), qbus_find_recursive(), and the
1042      * guest kernel's name-to-address assignment strategy.)
1043      *
1044      * Meanwhile, the kernel's traversal seems to have been reversed; see eg.
1045      * the message, if not necessarily the code, of commit 70161ff336.
1046      * Therefore the loop now establishes the inverse of the original intent.
1047      *
1048      * Unfortunately, we can't counteract the kernel change by reversing the
1049      * loop; it would break existing command lines.
1050      *
1051      * In any case, the kernel makes no guarantee about the stability of
1052      * enumeration order of virtio devices (as demonstrated by it changing
1053      * between kernel versions). For reliable and stable identification
1054      * of disks users must use UUIDs or similar mechanisms.
1055      */
1056     for (i = 0; i < NUM_VIRTIO_TRANSPORTS; i++) {
1057         int irq = vms->irqmap[VIRT_MMIO] + i;
1058         hwaddr base = vms->memmap[VIRT_MMIO].base + i * size;
1059 
1060         sysbus_create_simple("virtio-mmio", base,
1061                              qdev_get_gpio_in(vms->gic, irq));
1062     }
1063 
1064     /* We add dtb nodes in reverse order so that they appear in the finished
1065      * device tree lowest address first.
1066      *
1067      * Note that this mapping is independent of the loop above. The previous
1068      * loop influences virtio device to virtio transport assignment, whereas
1069      * this loop controls how virtio transports are laid out in the dtb.
1070      */
1071     for (i = NUM_VIRTIO_TRANSPORTS - 1; i >= 0; i--) {
1072         char *nodename;
1073         int irq = vms->irqmap[VIRT_MMIO] + i;
1074         hwaddr base = vms->memmap[VIRT_MMIO].base + i * size;
1075 
1076         nodename = g_strdup_printf("/virtio_mmio@%" PRIx64, base);
1077         qemu_fdt_add_subnode(ms->fdt, nodename);
1078         qemu_fdt_setprop_string(ms->fdt, nodename,
1079                                 "compatible", "virtio,mmio");
1080         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1081                                      2, base, 2, size);
1082         qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
1083                                GIC_FDT_IRQ_TYPE_SPI, irq,
1084                                GIC_FDT_IRQ_FLAGS_EDGE_LO_HI);
1085         qemu_fdt_setprop(ms->fdt, nodename, "dma-coherent", NULL, 0);
1086         g_free(nodename);
1087     }
1088 }
1089 
1090 #define VIRT_FLASH_SECTOR_SIZE (256 * KiB)
1091 
1092 static PFlashCFI01 *virt_flash_create1(VirtMachineState *vms,
1093                                         const char *name,
1094                                         const char *alias_prop_name)
1095 {
1096     /*
1097      * Create a single flash device.  We use the same parameters as
1098      * the flash devices on the Versatile Express board.
1099      */
1100     DeviceState *dev = qdev_new(TYPE_PFLASH_CFI01);
1101 
1102     qdev_prop_set_uint64(dev, "sector-length", VIRT_FLASH_SECTOR_SIZE);
1103     qdev_prop_set_uint8(dev, "width", 4);
1104     qdev_prop_set_uint8(dev, "device-width", 2);
1105     qdev_prop_set_bit(dev, "big-endian", false);
1106     qdev_prop_set_uint16(dev, "id0", 0x89);
1107     qdev_prop_set_uint16(dev, "id1", 0x18);
1108     qdev_prop_set_uint16(dev, "id2", 0x00);
1109     qdev_prop_set_uint16(dev, "id3", 0x00);
1110     qdev_prop_set_string(dev, "name", name);
1111     object_property_add_child(OBJECT(vms), name, OBJECT(dev));
1112     object_property_add_alias(OBJECT(vms), alias_prop_name,
1113                               OBJECT(dev), "drive");
1114     return PFLASH_CFI01(dev);
1115 }
1116 
1117 static void virt_flash_create(VirtMachineState *vms)
1118 {
1119     vms->flash[0] = virt_flash_create1(vms, "virt.flash0", "pflash0");
1120     vms->flash[1] = virt_flash_create1(vms, "virt.flash1", "pflash1");
1121 }
1122 
1123 static void virt_flash_map1(PFlashCFI01 *flash,
1124                             hwaddr base, hwaddr size,
1125                             MemoryRegion *sysmem)
1126 {
1127     DeviceState *dev = DEVICE(flash);
1128 
1129     assert(QEMU_IS_ALIGNED(size, VIRT_FLASH_SECTOR_SIZE));
1130     assert(size / VIRT_FLASH_SECTOR_SIZE <= UINT32_MAX);
1131     qdev_prop_set_uint32(dev, "num-blocks", size / VIRT_FLASH_SECTOR_SIZE);
1132     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1133 
1134     memory_region_add_subregion(sysmem, base,
1135                                 sysbus_mmio_get_region(SYS_BUS_DEVICE(dev),
1136                                                        0));
1137 }
1138 
1139 static void virt_flash_map(VirtMachineState *vms,
1140                            MemoryRegion *sysmem,
1141                            MemoryRegion *secure_sysmem)
1142 {
1143     /*
1144      * Map two flash devices to fill the VIRT_FLASH space in the memmap.
1145      * sysmem is the system memory space. secure_sysmem is the secure view
1146      * of the system, and the first flash device should be made visible only
1147      * there. The second flash device is visible to both secure and nonsecure.
1148      * If sysmem == secure_sysmem this means there is no separate Secure
1149      * address space and both flash devices are generally visible.
1150      */
1151     hwaddr flashsize = vms->memmap[VIRT_FLASH].size / 2;
1152     hwaddr flashbase = vms->memmap[VIRT_FLASH].base;
1153 
1154     virt_flash_map1(vms->flash[0], flashbase, flashsize,
1155                     secure_sysmem);
1156     virt_flash_map1(vms->flash[1], flashbase + flashsize, flashsize,
1157                     sysmem);
1158 }
1159 
1160 static void virt_flash_fdt(VirtMachineState *vms,
1161                            MemoryRegion *sysmem,
1162                            MemoryRegion *secure_sysmem)
1163 {
1164     hwaddr flashsize = vms->memmap[VIRT_FLASH].size / 2;
1165     hwaddr flashbase = vms->memmap[VIRT_FLASH].base;
1166     MachineState *ms = MACHINE(vms);
1167     char *nodename;
1168 
1169     if (sysmem == secure_sysmem) {
1170         /* Report both flash devices as a single node in the DT */
1171         nodename = g_strdup_printf("/flash@%" PRIx64, flashbase);
1172         qemu_fdt_add_subnode(ms->fdt, nodename);
1173         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible", "cfi-flash");
1174         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1175                                      2, flashbase, 2, flashsize,
1176                                      2, flashbase + flashsize, 2, flashsize);
1177         qemu_fdt_setprop_cell(ms->fdt, nodename, "bank-width", 4);
1178         g_free(nodename);
1179     } else {
1180         /*
1181          * Report the devices as separate nodes so we can mark one as
1182          * only visible to the secure world.
1183          */
1184         nodename = g_strdup_printf("/secflash@%" PRIx64, flashbase);
1185         qemu_fdt_add_subnode(ms->fdt, nodename);
1186         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible", "cfi-flash");
1187         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1188                                      2, flashbase, 2, flashsize);
1189         qemu_fdt_setprop_cell(ms->fdt, nodename, "bank-width", 4);
1190         qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
1191         qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
1192         g_free(nodename);
1193 
1194         nodename = g_strdup_printf("/flash@%" PRIx64, flashbase + flashsize);
1195         qemu_fdt_add_subnode(ms->fdt, nodename);
1196         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible", "cfi-flash");
1197         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1198                                      2, flashbase + flashsize, 2, flashsize);
1199         qemu_fdt_setprop_cell(ms->fdt, nodename, "bank-width", 4);
1200         g_free(nodename);
1201     }
1202 }
1203 
1204 static bool virt_firmware_init(VirtMachineState *vms,
1205                                MemoryRegion *sysmem,
1206                                MemoryRegion *secure_sysmem)
1207 {
1208     int i;
1209     const char *bios_name;
1210     BlockBackend *pflash_blk0;
1211 
1212     /* Map legacy -drive if=pflash to machine properties */
1213     for (i = 0; i < ARRAY_SIZE(vms->flash); i++) {
1214         pflash_cfi01_legacy_drive(vms->flash[i],
1215                                   drive_get(IF_PFLASH, 0, i));
1216     }
1217 
1218     virt_flash_map(vms, sysmem, secure_sysmem);
1219 
1220     pflash_blk0 = pflash_cfi01_get_blk(vms->flash[0]);
1221 
1222     bios_name = MACHINE(vms)->firmware;
1223     if (bios_name) {
1224         char *fname;
1225         MemoryRegion *mr;
1226         int image_size;
1227 
1228         if (pflash_blk0) {
1229             error_report("The contents of the first flash device may be "
1230                          "specified with -bios or with -drive if=pflash... "
1231                          "but you cannot use both options at once");
1232             exit(1);
1233         }
1234 
1235         /* Fall back to -bios */
1236 
1237         fname = qemu_find_file(QEMU_FILE_TYPE_BIOS, bios_name);
1238         if (!fname) {
1239             error_report("Could not find ROM image '%s'", bios_name);
1240             exit(1);
1241         }
1242         mr = sysbus_mmio_get_region(SYS_BUS_DEVICE(vms->flash[0]), 0);
1243         image_size = load_image_mr(fname, mr);
1244         g_free(fname);
1245         if (image_size < 0) {
1246             error_report("Could not load ROM image '%s'", bios_name);
1247             exit(1);
1248         }
1249     }
1250 
1251     return pflash_blk0 || bios_name;
1252 }
1253 
1254 static FWCfgState *create_fw_cfg(const VirtMachineState *vms, AddressSpace *as)
1255 {
1256     MachineState *ms = MACHINE(vms);
1257     hwaddr base = vms->memmap[VIRT_FW_CFG].base;
1258     hwaddr size = vms->memmap[VIRT_FW_CFG].size;
1259     FWCfgState *fw_cfg;
1260     char *nodename;
1261 
1262     fw_cfg = fw_cfg_init_mem_wide(base + 8, base, 8, base + 16, as);
1263     fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, (uint16_t)ms->smp.cpus);
1264 
1265     nodename = g_strdup_printf("/fw-cfg@%" PRIx64, base);
1266     qemu_fdt_add_subnode(ms->fdt, nodename);
1267     qemu_fdt_setprop_string(ms->fdt, nodename,
1268                             "compatible", "qemu,fw-cfg-mmio");
1269     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1270                                  2, base, 2, size);
1271     qemu_fdt_setprop(ms->fdt, nodename, "dma-coherent", NULL, 0);
1272     g_free(nodename);
1273     return fw_cfg;
1274 }
1275 
1276 static void create_pcie_irq_map(const MachineState *ms,
1277                                 uint32_t gic_phandle,
1278                                 int first_irq, const char *nodename)
1279 {
1280     int devfn, pin;
1281     uint32_t full_irq_map[4 * 4 * 10] = { 0 };
1282     uint32_t *irq_map = full_irq_map;
1283 
1284     for (devfn = 0; devfn <= 0x18; devfn += 0x8) {
1285         for (pin = 0; pin < 4; pin++) {
1286             int irq_type = GIC_FDT_IRQ_TYPE_SPI;
1287             int irq_nr = first_irq + ((pin + PCI_SLOT(devfn)) % PCI_NUM_PINS);
1288             int irq_level = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
1289             int i;
1290 
1291             uint32_t map[] = {
1292                 devfn << 8, 0, 0,                           /* devfn */
1293                 pin + 1,                                    /* PCI pin */
1294                 gic_phandle, 0, 0, irq_type, irq_nr, irq_level }; /* GIC irq */
1295 
1296             /* Convert map to big endian */
1297             for (i = 0; i < 10; i++) {
1298                 irq_map[i] = cpu_to_be32(map[i]);
1299             }
1300             irq_map += 10;
1301         }
1302     }
1303 
1304     qemu_fdt_setprop(ms->fdt, nodename, "interrupt-map",
1305                      full_irq_map, sizeof(full_irq_map));
1306 
1307     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupt-map-mask",
1308                            cpu_to_be16(PCI_DEVFN(3, 0)), /* Slot 3 */
1309                            0, 0,
1310                            0x7           /* PCI irq */);
1311 }
1312 
1313 static void create_smmu(const VirtMachineState *vms,
1314                         PCIBus *bus)
1315 {
1316     char *node;
1317     const char compat[] = "arm,smmu-v3";
1318     int irq =  vms->irqmap[VIRT_SMMU];
1319     int i;
1320     hwaddr base = vms->memmap[VIRT_SMMU].base;
1321     hwaddr size = vms->memmap[VIRT_SMMU].size;
1322     const char irq_names[] = "eventq\0priq\0cmdq-sync\0gerror";
1323     DeviceState *dev;
1324     MachineState *ms = MACHINE(vms);
1325 
1326     if (vms->iommu != VIRT_IOMMU_SMMUV3 || !vms->iommu_phandle) {
1327         return;
1328     }
1329 
1330     dev = qdev_new(TYPE_ARM_SMMUV3);
1331 
1332     object_property_set_link(OBJECT(dev), "primary-bus", OBJECT(bus),
1333                              &error_abort);
1334     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1335     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, base);
1336     for (i = 0; i < NUM_SMMU_IRQS; i++) {
1337         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
1338                            qdev_get_gpio_in(vms->gic, irq + i));
1339     }
1340 
1341     node = g_strdup_printf("/smmuv3@%" PRIx64, base);
1342     qemu_fdt_add_subnode(ms->fdt, node);
1343     qemu_fdt_setprop(ms->fdt, node, "compatible", compat, sizeof(compat));
1344     qemu_fdt_setprop_sized_cells(ms->fdt, node, "reg", 2, base, 2, size);
1345 
1346     qemu_fdt_setprop_cells(ms->fdt, node, "interrupts",
1347             GIC_FDT_IRQ_TYPE_SPI, irq    , GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1348             GIC_FDT_IRQ_TYPE_SPI, irq + 1, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1349             GIC_FDT_IRQ_TYPE_SPI, irq + 2, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1350             GIC_FDT_IRQ_TYPE_SPI, irq + 3, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI);
1351 
1352     qemu_fdt_setprop(ms->fdt, node, "interrupt-names", irq_names,
1353                      sizeof(irq_names));
1354 
1355     qemu_fdt_setprop(ms->fdt, node, "dma-coherent", NULL, 0);
1356 
1357     qemu_fdt_setprop_cell(ms->fdt, node, "#iommu-cells", 1);
1358 
1359     qemu_fdt_setprop_cell(ms->fdt, node, "phandle", vms->iommu_phandle);
1360     g_free(node);
1361 }
1362 
1363 static void create_virtio_iommu_dt_bindings(VirtMachineState *vms)
1364 {
1365     const char compat[] = "virtio,pci-iommu\0pci1af4,1057";
1366     uint16_t bdf = vms->virtio_iommu_bdf;
1367     MachineState *ms = MACHINE(vms);
1368     char *node;
1369 
1370     vms->iommu_phandle = qemu_fdt_alloc_phandle(ms->fdt);
1371 
1372     node = g_strdup_printf("%s/virtio_iommu@%x,%x", vms->pciehb_nodename,
1373                            PCI_SLOT(bdf), PCI_FUNC(bdf));
1374     qemu_fdt_add_subnode(ms->fdt, node);
1375     qemu_fdt_setprop(ms->fdt, node, "compatible", compat, sizeof(compat));
1376     qemu_fdt_setprop_sized_cells(ms->fdt, node, "reg",
1377                                  1, bdf << 8, 1, 0, 1, 0,
1378                                  1, 0, 1, 0);
1379 
1380     qemu_fdt_setprop_cell(ms->fdt, node, "#iommu-cells", 1);
1381     qemu_fdt_setprop_cell(ms->fdt, node, "phandle", vms->iommu_phandle);
1382     g_free(node);
1383 
1384     qemu_fdt_setprop_cells(ms->fdt, vms->pciehb_nodename, "iommu-map",
1385                            0x0, vms->iommu_phandle, 0x0, bdf,
1386                            bdf + 1, vms->iommu_phandle, bdf + 1, 0xffff - bdf);
1387 }
1388 
1389 static void create_pcie(VirtMachineState *vms)
1390 {
1391     hwaddr base_mmio = vms->memmap[VIRT_PCIE_MMIO].base;
1392     hwaddr size_mmio = vms->memmap[VIRT_PCIE_MMIO].size;
1393     hwaddr base_mmio_high = vms->memmap[VIRT_HIGH_PCIE_MMIO].base;
1394     hwaddr size_mmio_high = vms->memmap[VIRT_HIGH_PCIE_MMIO].size;
1395     hwaddr base_pio = vms->memmap[VIRT_PCIE_PIO].base;
1396     hwaddr size_pio = vms->memmap[VIRT_PCIE_PIO].size;
1397     hwaddr base_ecam, size_ecam;
1398     hwaddr base = base_mmio;
1399     int nr_pcie_buses;
1400     int irq = vms->irqmap[VIRT_PCIE];
1401     MemoryRegion *mmio_alias;
1402     MemoryRegion *mmio_reg;
1403     MemoryRegion *ecam_alias;
1404     MemoryRegion *ecam_reg;
1405     DeviceState *dev;
1406     char *nodename;
1407     int i, ecam_id;
1408     PCIHostState *pci;
1409     MachineState *ms = MACHINE(vms);
1410     MachineClass *mc = MACHINE_GET_CLASS(ms);
1411 
1412     dev = qdev_new(TYPE_GPEX_HOST);
1413     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1414 
1415     ecam_id = VIRT_ECAM_ID(vms->highmem_ecam);
1416     base_ecam = vms->memmap[ecam_id].base;
1417     size_ecam = vms->memmap[ecam_id].size;
1418     nr_pcie_buses = size_ecam / PCIE_MMCFG_SIZE_MIN;
1419     /* Map only the first size_ecam bytes of ECAM space */
1420     ecam_alias = g_new0(MemoryRegion, 1);
1421     ecam_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 0);
1422     memory_region_init_alias(ecam_alias, OBJECT(dev), "pcie-ecam",
1423                              ecam_reg, 0, size_ecam);
1424     memory_region_add_subregion(get_system_memory(), base_ecam, ecam_alias);
1425 
1426     /* Map the MMIO window into system address space so as to expose
1427      * the section of PCI MMIO space which starts at the same base address
1428      * (ie 1:1 mapping for that part of PCI MMIO space visible through
1429      * the window).
1430      */
1431     mmio_alias = g_new0(MemoryRegion, 1);
1432     mmio_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 1);
1433     memory_region_init_alias(mmio_alias, OBJECT(dev), "pcie-mmio",
1434                              mmio_reg, base_mmio, size_mmio);
1435     memory_region_add_subregion(get_system_memory(), base_mmio, mmio_alias);
1436 
1437     if (vms->highmem_mmio) {
1438         /* Map high MMIO space */
1439         MemoryRegion *high_mmio_alias = g_new0(MemoryRegion, 1);
1440 
1441         memory_region_init_alias(high_mmio_alias, OBJECT(dev), "pcie-mmio-high",
1442                                  mmio_reg, base_mmio_high, size_mmio_high);
1443         memory_region_add_subregion(get_system_memory(), base_mmio_high,
1444                                     high_mmio_alias);
1445     }
1446 
1447     /* Map IO port space */
1448     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 2, base_pio);
1449 
1450     for (i = 0; i < GPEX_NUM_IRQS; i++) {
1451         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
1452                            qdev_get_gpio_in(vms->gic, irq + i));
1453         gpex_set_irq_num(GPEX_HOST(dev), i, irq + i);
1454     }
1455 
1456     pci = PCI_HOST_BRIDGE(dev);
1457     pci->bypass_iommu = vms->default_bus_bypass_iommu;
1458     vms->bus = pci->bus;
1459     if (vms->bus) {
1460         for (i = 0; i < nb_nics; i++) {
1461             pci_nic_init_nofail(&nd_table[i], pci->bus, mc->default_nic, NULL);
1462         }
1463     }
1464 
1465     nodename = vms->pciehb_nodename = g_strdup_printf("/pcie@%" PRIx64, base);
1466     qemu_fdt_add_subnode(ms->fdt, nodename);
1467     qemu_fdt_setprop_string(ms->fdt, nodename,
1468                             "compatible", "pci-host-ecam-generic");
1469     qemu_fdt_setprop_string(ms->fdt, nodename, "device_type", "pci");
1470     qemu_fdt_setprop_cell(ms->fdt, nodename, "#address-cells", 3);
1471     qemu_fdt_setprop_cell(ms->fdt, nodename, "#size-cells", 2);
1472     qemu_fdt_setprop_cell(ms->fdt, nodename, "linux,pci-domain", 0);
1473     qemu_fdt_setprop_cells(ms->fdt, nodename, "bus-range", 0,
1474                            nr_pcie_buses - 1);
1475     qemu_fdt_setprop(ms->fdt, nodename, "dma-coherent", NULL, 0);
1476 
1477     if (vms->msi_phandle) {
1478         qemu_fdt_setprop_cells(ms->fdt, nodename, "msi-map",
1479                                0, vms->msi_phandle, 0, 0x10000);
1480     }
1481 
1482     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1483                                  2, base_ecam, 2, size_ecam);
1484 
1485     if (vms->highmem_mmio) {
1486         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "ranges",
1487                                      1, FDT_PCI_RANGE_IOPORT, 2, 0,
1488                                      2, base_pio, 2, size_pio,
1489                                      1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
1490                                      2, base_mmio, 2, size_mmio,
1491                                      1, FDT_PCI_RANGE_MMIO_64BIT,
1492                                      2, base_mmio_high,
1493                                      2, base_mmio_high, 2, size_mmio_high);
1494     } else {
1495         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "ranges",
1496                                      1, FDT_PCI_RANGE_IOPORT, 2, 0,
1497                                      2, base_pio, 2, size_pio,
1498                                      1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
1499                                      2, base_mmio, 2, size_mmio);
1500     }
1501 
1502     qemu_fdt_setprop_cell(ms->fdt, nodename, "#interrupt-cells", 1);
1503     create_pcie_irq_map(ms, vms->gic_phandle, irq, nodename);
1504 
1505     if (vms->iommu) {
1506         vms->iommu_phandle = qemu_fdt_alloc_phandle(ms->fdt);
1507 
1508         switch (vms->iommu) {
1509         case VIRT_IOMMU_SMMUV3:
1510             create_smmu(vms, vms->bus);
1511             qemu_fdt_setprop_cells(ms->fdt, nodename, "iommu-map",
1512                                    0x0, vms->iommu_phandle, 0x0, 0x10000);
1513             break;
1514         default:
1515             g_assert_not_reached();
1516         }
1517     }
1518 }
1519 
1520 static void create_platform_bus(VirtMachineState *vms)
1521 {
1522     DeviceState *dev;
1523     SysBusDevice *s;
1524     int i;
1525     MemoryRegion *sysmem = get_system_memory();
1526 
1527     dev = qdev_new(TYPE_PLATFORM_BUS_DEVICE);
1528     dev->id = g_strdup(TYPE_PLATFORM_BUS_DEVICE);
1529     qdev_prop_set_uint32(dev, "num_irqs", PLATFORM_BUS_NUM_IRQS);
1530     qdev_prop_set_uint32(dev, "mmio_size", vms->memmap[VIRT_PLATFORM_BUS].size);
1531     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1532     vms->platform_bus_dev = dev;
1533 
1534     s = SYS_BUS_DEVICE(dev);
1535     for (i = 0; i < PLATFORM_BUS_NUM_IRQS; i++) {
1536         int irq = vms->irqmap[VIRT_PLATFORM_BUS] + i;
1537         sysbus_connect_irq(s, i, qdev_get_gpio_in(vms->gic, irq));
1538     }
1539 
1540     memory_region_add_subregion(sysmem,
1541                                 vms->memmap[VIRT_PLATFORM_BUS].base,
1542                                 sysbus_mmio_get_region(s, 0));
1543 }
1544 
1545 static void create_tag_ram(MemoryRegion *tag_sysmem,
1546                            hwaddr base, hwaddr size,
1547                            const char *name)
1548 {
1549     MemoryRegion *tagram = g_new(MemoryRegion, 1);
1550 
1551     memory_region_init_ram(tagram, NULL, name, size / 32, &error_fatal);
1552     memory_region_add_subregion(tag_sysmem, base / 32, tagram);
1553 }
1554 
1555 static void create_secure_ram(VirtMachineState *vms,
1556                               MemoryRegion *secure_sysmem,
1557                               MemoryRegion *secure_tag_sysmem)
1558 {
1559     MemoryRegion *secram = g_new(MemoryRegion, 1);
1560     char *nodename;
1561     hwaddr base = vms->memmap[VIRT_SECURE_MEM].base;
1562     hwaddr size = vms->memmap[VIRT_SECURE_MEM].size;
1563     MachineState *ms = MACHINE(vms);
1564 
1565     memory_region_init_ram(secram, NULL, "virt.secure-ram", size,
1566                            &error_fatal);
1567     memory_region_add_subregion(secure_sysmem, base, secram);
1568 
1569     nodename = g_strdup_printf("/secram@%" PRIx64, base);
1570     qemu_fdt_add_subnode(ms->fdt, nodename);
1571     qemu_fdt_setprop_string(ms->fdt, nodename, "device_type", "memory");
1572     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg", 2, base, 2, size);
1573     qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
1574     qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
1575 
1576     if (secure_tag_sysmem) {
1577         create_tag_ram(secure_tag_sysmem, base, size, "mach-virt.secure-tag");
1578     }
1579 
1580     g_free(nodename);
1581 }
1582 
1583 static void *machvirt_dtb(const struct arm_boot_info *binfo, int *fdt_size)
1584 {
1585     const VirtMachineState *board = container_of(binfo, VirtMachineState,
1586                                                  bootinfo);
1587     MachineState *ms = MACHINE(board);
1588 
1589 
1590     *fdt_size = board->fdt_size;
1591     return ms->fdt;
1592 }
1593 
1594 static void virt_build_smbios(VirtMachineState *vms)
1595 {
1596     MachineClass *mc = MACHINE_GET_CLASS(vms);
1597     MachineState *ms = MACHINE(vms);
1598     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
1599     uint8_t *smbios_tables, *smbios_anchor;
1600     size_t smbios_tables_len, smbios_anchor_len;
1601     struct smbios_phys_mem_area mem_array;
1602     const char *product = "QEMU Virtual Machine";
1603 
1604     if (kvm_enabled()) {
1605         product = "KVM Virtual Machine";
1606     }
1607 
1608     smbios_set_defaults("QEMU", product,
1609                         vmc->smbios_old_sys_ver ? "1.0" : mc->name, false,
1610                         true, SMBIOS_ENTRY_POINT_TYPE_64);
1611 
1612     /* build the array of physical mem area from base_memmap */
1613     mem_array.address = vms->memmap[VIRT_MEM].base;
1614     mem_array.length = ms->ram_size;
1615 
1616     smbios_get_tables(ms, &mem_array, 1,
1617                       &smbios_tables, &smbios_tables_len,
1618                       &smbios_anchor, &smbios_anchor_len,
1619                       &error_fatal);
1620 
1621     if (smbios_anchor) {
1622         fw_cfg_add_file(vms->fw_cfg, "etc/smbios/smbios-tables",
1623                         smbios_tables, smbios_tables_len);
1624         fw_cfg_add_file(vms->fw_cfg, "etc/smbios/smbios-anchor",
1625                         smbios_anchor, smbios_anchor_len);
1626     }
1627 }
1628 
1629 static
1630 void virt_machine_done(Notifier *notifier, void *data)
1631 {
1632     VirtMachineState *vms = container_of(notifier, VirtMachineState,
1633                                          machine_done);
1634     MachineState *ms = MACHINE(vms);
1635     ARMCPU *cpu = ARM_CPU(first_cpu);
1636     struct arm_boot_info *info = &vms->bootinfo;
1637     AddressSpace *as = arm_boot_address_space(cpu, info);
1638 
1639     /*
1640      * If the user provided a dtb, we assume the dynamic sysbus nodes
1641      * already are integrated there. This corresponds to a use case where
1642      * the dynamic sysbus nodes are complex and their generation is not yet
1643      * supported. In that case the user can take charge of the guest dt
1644      * while qemu takes charge of the qom stuff.
1645      */
1646     if (info->dtb_filename == NULL) {
1647         platform_bus_add_all_fdt_nodes(ms->fdt, "/intc",
1648                                        vms->memmap[VIRT_PLATFORM_BUS].base,
1649                                        vms->memmap[VIRT_PLATFORM_BUS].size,
1650                                        vms->irqmap[VIRT_PLATFORM_BUS]);
1651     }
1652     if (arm_load_dtb(info->dtb_start, info, info->dtb_limit, as, ms) < 0) {
1653         exit(1);
1654     }
1655 
1656     fw_cfg_add_extra_pci_roots(vms->bus, vms->fw_cfg);
1657 
1658     virt_acpi_setup(vms);
1659     virt_build_smbios(vms);
1660 }
1661 
1662 static uint64_t virt_cpu_mp_affinity(VirtMachineState *vms, int idx)
1663 {
1664     uint8_t clustersz = ARM_DEFAULT_CPUS_PER_CLUSTER;
1665     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
1666 
1667     if (!vmc->disallow_affinity_adjustment) {
1668         /* Adjust MPIDR like 64-bit KVM hosts, which incorporate the
1669          * GIC's target-list limitations. 32-bit KVM hosts currently
1670          * always create clusters of 4 CPUs, but that is expected to
1671          * change when they gain support for gicv3. When KVM is enabled
1672          * it will override the changes we make here, therefore our
1673          * purposes are to make TCG consistent (with 64-bit KVM hosts)
1674          * and to improve SGI efficiency.
1675          */
1676         if (vms->gic_version == VIRT_GIC_VERSION_2) {
1677             clustersz = GIC_TARGETLIST_BITS;
1678         } else {
1679             clustersz = GICV3_TARGETLIST_BITS;
1680         }
1681     }
1682     return arm_build_mp_affinity(idx, clustersz);
1683 }
1684 
1685 static inline bool *virt_get_high_memmap_enabled(VirtMachineState *vms,
1686                                                  int index)
1687 {
1688     bool *enabled_array[] = {
1689         &vms->highmem_redists,
1690         &vms->highmem_ecam,
1691         &vms->highmem_mmio,
1692     };
1693 
1694     assert(ARRAY_SIZE(extended_memmap) - VIRT_LOWMEMMAP_LAST ==
1695            ARRAY_SIZE(enabled_array));
1696     assert(index - VIRT_LOWMEMMAP_LAST < ARRAY_SIZE(enabled_array));
1697 
1698     return enabled_array[index - VIRT_LOWMEMMAP_LAST];
1699 }
1700 
1701 static void virt_set_high_memmap(VirtMachineState *vms,
1702                                  hwaddr base, int pa_bits)
1703 {
1704     hwaddr region_base, region_size;
1705     bool *region_enabled, fits;
1706     int i;
1707 
1708     for (i = VIRT_LOWMEMMAP_LAST; i < ARRAY_SIZE(extended_memmap); i++) {
1709         region_enabled = virt_get_high_memmap_enabled(vms, i);
1710         region_base = ROUND_UP(base, extended_memmap[i].size);
1711         region_size = extended_memmap[i].size;
1712 
1713         vms->memmap[i].base = region_base;
1714         vms->memmap[i].size = region_size;
1715 
1716         /*
1717          * Check each device to see if it fits in the PA space,
1718          * moving highest_gpa as we go. For compatibility, move
1719          * highest_gpa for disabled fitting devices as well, if
1720          * the compact layout has been disabled.
1721          *
1722          * For each device that doesn't fit, disable it.
1723          */
1724         fits = (region_base + region_size) <= BIT_ULL(pa_bits);
1725         *region_enabled &= fits;
1726         if (vms->highmem_compact && !*region_enabled) {
1727             continue;
1728         }
1729 
1730         base = region_base + region_size;
1731         if (fits) {
1732             vms->highest_gpa = base - 1;
1733         }
1734     }
1735 }
1736 
1737 static void virt_set_memmap(VirtMachineState *vms, int pa_bits)
1738 {
1739     MachineState *ms = MACHINE(vms);
1740     hwaddr base, device_memory_base, device_memory_size, memtop;
1741     int i;
1742 
1743     vms->memmap = extended_memmap;
1744 
1745     for (i = 0; i < ARRAY_SIZE(base_memmap); i++) {
1746         vms->memmap[i] = base_memmap[i];
1747     }
1748 
1749     if (ms->ram_slots > ACPI_MAX_RAM_SLOTS) {
1750         error_report("unsupported number of memory slots: %"PRIu64,
1751                      ms->ram_slots);
1752         exit(EXIT_FAILURE);
1753     }
1754 
1755     /*
1756      * !highmem is exactly the same as limiting the PA space to 32bit,
1757      * irrespective of the underlying capabilities of the HW.
1758      */
1759     if (!vms->highmem) {
1760         pa_bits = 32;
1761     }
1762 
1763     /*
1764      * We compute the base of the high IO region depending on the
1765      * amount of initial and device memory. The device memory start/size
1766      * is aligned on 1GiB. We never put the high IO region below 256GiB
1767      * so that if maxram_size is < 255GiB we keep the legacy memory map.
1768      * The device region size assumes 1GiB page max alignment per slot.
1769      */
1770     device_memory_base =
1771         ROUND_UP(vms->memmap[VIRT_MEM].base + ms->ram_size, GiB);
1772     device_memory_size = ms->maxram_size - ms->ram_size + ms->ram_slots * GiB;
1773 
1774     /* Base address of the high IO region */
1775     memtop = base = device_memory_base + ROUND_UP(device_memory_size, GiB);
1776     if (memtop > BIT_ULL(pa_bits)) {
1777         error_report("Addressing limited to %d bits, but memory exceeds it by %llu bytes",
1778                      pa_bits, memtop - BIT_ULL(pa_bits));
1779         exit(EXIT_FAILURE);
1780     }
1781     if (base < device_memory_base) {
1782         error_report("maxmem/slots too huge");
1783         exit(EXIT_FAILURE);
1784     }
1785     if (base < vms->memmap[VIRT_MEM].base + LEGACY_RAMLIMIT_BYTES) {
1786         base = vms->memmap[VIRT_MEM].base + LEGACY_RAMLIMIT_BYTES;
1787     }
1788 
1789     /* We know for sure that at least the memory fits in the PA space */
1790     vms->highest_gpa = memtop - 1;
1791 
1792     virt_set_high_memmap(vms, base, pa_bits);
1793 
1794     if (device_memory_size > 0) {
1795         machine_memory_devices_init(ms, device_memory_base, device_memory_size);
1796     }
1797 }
1798 
1799 static VirtGICType finalize_gic_version_do(const char *accel_name,
1800                                            VirtGICType gic_version,
1801                                            int gics_supported,
1802                                            unsigned int max_cpus)
1803 {
1804     /* Convert host/max/nosel to GIC version number */
1805     switch (gic_version) {
1806     case VIRT_GIC_VERSION_HOST:
1807         if (!kvm_enabled()) {
1808             error_report("gic-version=host requires KVM");
1809             exit(1);
1810         }
1811 
1812         /* For KVM, gic-version=host means gic-version=max */
1813         return finalize_gic_version_do(accel_name, VIRT_GIC_VERSION_MAX,
1814                                        gics_supported, max_cpus);
1815     case VIRT_GIC_VERSION_MAX:
1816         if (gics_supported & VIRT_GIC_VERSION_4_MASK) {
1817             gic_version = VIRT_GIC_VERSION_4;
1818         } else if (gics_supported & VIRT_GIC_VERSION_3_MASK) {
1819             gic_version = VIRT_GIC_VERSION_3;
1820         } else {
1821             gic_version = VIRT_GIC_VERSION_2;
1822         }
1823         break;
1824     case VIRT_GIC_VERSION_NOSEL:
1825         if ((gics_supported & VIRT_GIC_VERSION_2_MASK) &&
1826             max_cpus <= GIC_NCPU) {
1827             gic_version = VIRT_GIC_VERSION_2;
1828         } else if (gics_supported & VIRT_GIC_VERSION_3_MASK) {
1829             /*
1830              * in case the host does not support v2 emulation or
1831              * the end-user requested more than 8 VCPUs we now default
1832              * to v3. In any case defaulting to v2 would be broken.
1833              */
1834             gic_version = VIRT_GIC_VERSION_3;
1835         } else if (max_cpus > GIC_NCPU) {
1836             error_report("%s only supports GICv2 emulation but more than 8 "
1837                          "vcpus are requested", accel_name);
1838             exit(1);
1839         }
1840         break;
1841     case VIRT_GIC_VERSION_2:
1842     case VIRT_GIC_VERSION_3:
1843     case VIRT_GIC_VERSION_4:
1844         break;
1845     }
1846 
1847     /* Check chosen version is effectively supported */
1848     switch (gic_version) {
1849     case VIRT_GIC_VERSION_2:
1850         if (!(gics_supported & VIRT_GIC_VERSION_2_MASK)) {
1851             error_report("%s does not support GICv2 emulation", accel_name);
1852             exit(1);
1853         }
1854         break;
1855     case VIRT_GIC_VERSION_3:
1856         if (!(gics_supported & VIRT_GIC_VERSION_3_MASK)) {
1857             error_report("%s does not support GICv3 emulation", accel_name);
1858             exit(1);
1859         }
1860         break;
1861     case VIRT_GIC_VERSION_4:
1862         if (!(gics_supported & VIRT_GIC_VERSION_4_MASK)) {
1863             error_report("%s does not support GICv4 emulation, is virtualization=on?",
1864                          accel_name);
1865             exit(1);
1866         }
1867         break;
1868     default:
1869         error_report("logic error in finalize_gic_version");
1870         exit(1);
1871         break;
1872     }
1873 
1874     return gic_version;
1875 }
1876 
1877 /*
1878  * finalize_gic_version - Determines the final gic_version
1879  * according to the gic-version property
1880  *
1881  * Default GIC type is v2
1882  */
1883 static void finalize_gic_version(VirtMachineState *vms)
1884 {
1885     const char *accel_name = current_accel_name();
1886     unsigned int max_cpus = MACHINE(vms)->smp.max_cpus;
1887     int gics_supported = 0;
1888 
1889     /* Determine which GIC versions the current environment supports */
1890     if (kvm_enabled() && kvm_irqchip_in_kernel()) {
1891         int probe_bitmap = kvm_arm_vgic_probe();
1892 
1893         if (!probe_bitmap) {
1894             error_report("Unable to determine GIC version supported by host");
1895             exit(1);
1896         }
1897 
1898         if (probe_bitmap & KVM_ARM_VGIC_V2) {
1899             gics_supported |= VIRT_GIC_VERSION_2_MASK;
1900         }
1901         if (probe_bitmap & KVM_ARM_VGIC_V3) {
1902             gics_supported |= VIRT_GIC_VERSION_3_MASK;
1903         }
1904     } else if (kvm_enabled() && !kvm_irqchip_in_kernel()) {
1905         /* KVM w/o kernel irqchip can only deal with GICv2 */
1906         gics_supported |= VIRT_GIC_VERSION_2_MASK;
1907         accel_name = "KVM with kernel-irqchip=off";
1908     } else if (tcg_enabled() || hvf_enabled() || qtest_enabled())  {
1909         gics_supported |= VIRT_GIC_VERSION_2_MASK;
1910         if (module_object_class_by_name("arm-gicv3")) {
1911             gics_supported |= VIRT_GIC_VERSION_3_MASK;
1912             if (vms->virt) {
1913                 /* GICv4 only makes sense if CPU has EL2 */
1914                 gics_supported |= VIRT_GIC_VERSION_4_MASK;
1915             }
1916         }
1917     } else {
1918         error_report("Unsupported accelerator, can not determine GIC support");
1919         exit(1);
1920     }
1921 
1922     /*
1923      * Then convert helpers like host/max to concrete GIC versions and ensure
1924      * the desired version is supported
1925      */
1926     vms->gic_version = finalize_gic_version_do(accel_name, vms->gic_version,
1927                                                gics_supported, max_cpus);
1928 }
1929 
1930 /*
1931  * virt_cpu_post_init() must be called after the CPUs have
1932  * been realized and the GIC has been created.
1933  */
1934 static void virt_cpu_post_init(VirtMachineState *vms, MemoryRegion *sysmem)
1935 {
1936     int max_cpus = MACHINE(vms)->smp.max_cpus;
1937     bool aarch64, pmu, steal_time;
1938     CPUState *cpu;
1939 
1940     aarch64 = object_property_get_bool(OBJECT(first_cpu), "aarch64", NULL);
1941     pmu = object_property_get_bool(OBJECT(first_cpu), "pmu", NULL);
1942     steal_time = object_property_get_bool(OBJECT(first_cpu),
1943                                           "kvm-steal-time", NULL);
1944 
1945     if (kvm_enabled()) {
1946         hwaddr pvtime_reg_base = vms->memmap[VIRT_PVTIME].base;
1947         hwaddr pvtime_reg_size = vms->memmap[VIRT_PVTIME].size;
1948 
1949         if (steal_time) {
1950             MemoryRegion *pvtime = g_new(MemoryRegion, 1);
1951             hwaddr pvtime_size = max_cpus * PVTIME_SIZE_PER_CPU;
1952 
1953             /* The memory region size must be a multiple of host page size. */
1954             pvtime_size = REAL_HOST_PAGE_ALIGN(pvtime_size);
1955 
1956             if (pvtime_size > pvtime_reg_size) {
1957                 error_report("pvtime requires a %" HWADDR_PRId
1958                              " byte memory region for %d CPUs,"
1959                              " but only %" HWADDR_PRId " has been reserved",
1960                              pvtime_size, max_cpus, pvtime_reg_size);
1961                 exit(1);
1962             }
1963 
1964             memory_region_init_ram(pvtime, NULL, "pvtime", pvtime_size, NULL);
1965             memory_region_add_subregion(sysmem, pvtime_reg_base, pvtime);
1966         }
1967 
1968         CPU_FOREACH(cpu) {
1969             if (pmu) {
1970                 assert(arm_feature(&ARM_CPU(cpu)->env, ARM_FEATURE_PMU));
1971                 if (kvm_irqchip_in_kernel()) {
1972                     kvm_arm_pmu_set_irq(ARM_CPU(cpu), VIRTUAL_PMU_IRQ);
1973                 }
1974                 kvm_arm_pmu_init(ARM_CPU(cpu));
1975             }
1976             if (steal_time) {
1977                 kvm_arm_pvtime_init(ARM_CPU(cpu), pvtime_reg_base
1978                                                   + cpu->cpu_index
1979                                                     * PVTIME_SIZE_PER_CPU);
1980             }
1981         }
1982     } else {
1983         if (aarch64 && vms->highmem) {
1984             int requested_pa_size = 64 - clz64(vms->highest_gpa);
1985             int pamax = arm_pamax(ARM_CPU(first_cpu));
1986 
1987             if (pamax < requested_pa_size) {
1988                 error_report("VCPU supports less PA bits (%d) than "
1989                              "requested by the memory map (%d)",
1990                              pamax, requested_pa_size);
1991                 exit(1);
1992             }
1993         }
1994     }
1995 }
1996 
1997 static void machvirt_init(MachineState *machine)
1998 {
1999     VirtMachineState *vms = VIRT_MACHINE(machine);
2000     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(machine);
2001     MachineClass *mc = MACHINE_GET_CLASS(machine);
2002     const CPUArchIdList *possible_cpus;
2003     MemoryRegion *sysmem = get_system_memory();
2004     MemoryRegion *secure_sysmem = NULL;
2005     MemoryRegion *tag_sysmem = NULL;
2006     MemoryRegion *secure_tag_sysmem = NULL;
2007     int n, virt_max_cpus;
2008     bool firmware_loaded;
2009     bool aarch64 = true;
2010     bool has_ged = !vmc->no_ged;
2011     unsigned int smp_cpus = machine->smp.cpus;
2012     unsigned int max_cpus = machine->smp.max_cpus;
2013 
2014     possible_cpus = mc->possible_cpu_arch_ids(machine);
2015 
2016     /*
2017      * In accelerated mode, the memory map is computed earlier in kvm_type()
2018      * to create a VM with the right number of IPA bits.
2019      */
2020     if (!vms->memmap) {
2021         Object *cpuobj;
2022         ARMCPU *armcpu;
2023         int pa_bits;
2024 
2025         /*
2026          * Instantiate a temporary CPU object to find out about what
2027          * we are about to deal with. Once this is done, get rid of
2028          * the object.
2029          */
2030         cpuobj = object_new(possible_cpus->cpus[0].type);
2031         armcpu = ARM_CPU(cpuobj);
2032 
2033         pa_bits = arm_pamax(armcpu);
2034 
2035         object_unref(cpuobj);
2036 
2037         virt_set_memmap(vms, pa_bits);
2038     }
2039 
2040     /* We can probe only here because during property set
2041      * KVM is not available yet
2042      */
2043     finalize_gic_version(vms);
2044 
2045     if (vms->secure) {
2046         /*
2047          * The Secure view of the world is the same as the NonSecure,
2048          * but with a few extra devices. Create it as a container region
2049          * containing the system memory at low priority; any secure-only
2050          * devices go in at higher priority and take precedence.
2051          */
2052         secure_sysmem = g_new(MemoryRegion, 1);
2053         memory_region_init(secure_sysmem, OBJECT(machine), "secure-memory",
2054                            UINT64_MAX);
2055         memory_region_add_subregion_overlap(secure_sysmem, 0, sysmem, -1);
2056     }
2057 
2058     firmware_loaded = virt_firmware_init(vms, sysmem,
2059                                          secure_sysmem ?: sysmem);
2060 
2061     /* If we have an EL3 boot ROM then the assumption is that it will
2062      * implement PSCI itself, so disable QEMU's internal implementation
2063      * so it doesn't get in the way. Instead of starting secondary
2064      * CPUs in PSCI powerdown state we will start them all running and
2065      * let the boot ROM sort them out.
2066      * The usual case is that we do use QEMU's PSCI implementation;
2067      * if the guest has EL2 then we will use SMC as the conduit,
2068      * and otherwise we will use HVC (for backwards compatibility and
2069      * because if we're using KVM then we must use HVC).
2070      */
2071     if (vms->secure && firmware_loaded) {
2072         vms->psci_conduit = QEMU_PSCI_CONDUIT_DISABLED;
2073     } else if (vms->virt) {
2074         vms->psci_conduit = QEMU_PSCI_CONDUIT_SMC;
2075     } else {
2076         vms->psci_conduit = QEMU_PSCI_CONDUIT_HVC;
2077     }
2078 
2079     /*
2080      * The maximum number of CPUs depends on the GIC version, or on how
2081      * many redistributors we can fit into the memory map (which in turn
2082      * depends on whether this is a GICv3 or v4).
2083      */
2084     if (vms->gic_version == VIRT_GIC_VERSION_2) {
2085         virt_max_cpus = GIC_NCPU;
2086     } else {
2087         virt_max_cpus = virt_redist_capacity(vms, VIRT_GIC_REDIST);
2088         if (vms->highmem_redists) {
2089             virt_max_cpus += virt_redist_capacity(vms, VIRT_HIGH_GIC_REDIST2);
2090         }
2091     }
2092 
2093     if (max_cpus > virt_max_cpus) {
2094         error_report("Number of SMP CPUs requested (%d) exceeds max CPUs "
2095                      "supported by machine 'mach-virt' (%d)",
2096                      max_cpus, virt_max_cpus);
2097         if (vms->gic_version != VIRT_GIC_VERSION_2 && !vms->highmem_redists) {
2098             error_printf("Try 'highmem-redists=on' for more CPUs\n");
2099         }
2100 
2101         exit(1);
2102     }
2103 
2104     if (vms->secure && (kvm_enabled() || hvf_enabled())) {
2105         error_report("mach-virt: %s does not support providing "
2106                      "Security extensions (TrustZone) to the guest CPU",
2107                      current_accel_name());
2108         exit(1);
2109     }
2110 
2111     if (vms->virt && (kvm_enabled() || hvf_enabled())) {
2112         error_report("mach-virt: %s does not support providing "
2113                      "Virtualization extensions to the guest CPU",
2114                      current_accel_name());
2115         exit(1);
2116     }
2117 
2118     if (vms->mte && (kvm_enabled() || hvf_enabled())) {
2119         error_report("mach-virt: %s does not support providing "
2120                      "MTE to the guest CPU",
2121                      current_accel_name());
2122         exit(1);
2123     }
2124 
2125     create_fdt(vms);
2126 
2127     assert(possible_cpus->len == max_cpus);
2128     for (n = 0; n < possible_cpus->len; n++) {
2129         Object *cpuobj;
2130         CPUState *cs;
2131 
2132         if (n >= smp_cpus) {
2133             break;
2134         }
2135 
2136         cpuobj = object_new(possible_cpus->cpus[n].type);
2137         object_property_set_int(cpuobj, "mp-affinity",
2138                                 possible_cpus->cpus[n].arch_id, NULL);
2139 
2140         cs = CPU(cpuobj);
2141         cs->cpu_index = n;
2142 
2143         numa_cpu_pre_plug(&possible_cpus->cpus[cs->cpu_index], DEVICE(cpuobj),
2144                           &error_fatal);
2145 
2146         aarch64 &= object_property_get_bool(cpuobj, "aarch64", NULL);
2147 
2148         if (!vms->secure) {
2149             object_property_set_bool(cpuobj, "has_el3", false, NULL);
2150         }
2151 
2152         if (!vms->virt && object_property_find(cpuobj, "has_el2")) {
2153             object_property_set_bool(cpuobj, "has_el2", false, NULL);
2154         }
2155 
2156         if (vmc->kvm_no_adjvtime &&
2157             object_property_find(cpuobj, "kvm-no-adjvtime")) {
2158             object_property_set_bool(cpuobj, "kvm-no-adjvtime", true, NULL);
2159         }
2160 
2161         if (vmc->no_kvm_steal_time &&
2162             object_property_find(cpuobj, "kvm-steal-time")) {
2163             object_property_set_bool(cpuobj, "kvm-steal-time", false, NULL);
2164         }
2165 
2166         if (vmc->no_pmu && object_property_find(cpuobj, "pmu")) {
2167             object_property_set_bool(cpuobj, "pmu", false, NULL);
2168         }
2169 
2170         if (vmc->no_tcg_lpa2 && object_property_find(cpuobj, "lpa2")) {
2171             object_property_set_bool(cpuobj, "lpa2", false, NULL);
2172         }
2173 
2174         if (object_property_find(cpuobj, "reset-cbar")) {
2175             object_property_set_int(cpuobj, "reset-cbar",
2176                                     vms->memmap[VIRT_CPUPERIPHS].base,
2177                                     &error_abort);
2178         }
2179 
2180         object_property_set_link(cpuobj, "memory", OBJECT(sysmem),
2181                                  &error_abort);
2182         if (vms->secure) {
2183             object_property_set_link(cpuobj, "secure-memory",
2184                                      OBJECT(secure_sysmem), &error_abort);
2185         }
2186 
2187         if (vms->mte) {
2188             /* Create the memory region only once, but link to all cpus. */
2189             if (!tag_sysmem) {
2190                 /*
2191                  * The property exists only if MemTag is supported.
2192                  * If it is, we must allocate the ram to back that up.
2193                  */
2194                 if (!object_property_find(cpuobj, "tag-memory")) {
2195                     error_report("MTE requested, but not supported "
2196                                  "by the guest CPU");
2197                     exit(1);
2198                 }
2199 
2200                 tag_sysmem = g_new(MemoryRegion, 1);
2201                 memory_region_init(tag_sysmem, OBJECT(machine),
2202                                    "tag-memory", UINT64_MAX / 32);
2203 
2204                 if (vms->secure) {
2205                     secure_tag_sysmem = g_new(MemoryRegion, 1);
2206                     memory_region_init(secure_tag_sysmem, OBJECT(machine),
2207                                        "secure-tag-memory", UINT64_MAX / 32);
2208 
2209                     /* As with ram, secure-tag takes precedence over tag.  */
2210                     memory_region_add_subregion_overlap(secure_tag_sysmem, 0,
2211                                                         tag_sysmem, -1);
2212                 }
2213             }
2214 
2215             object_property_set_link(cpuobj, "tag-memory", OBJECT(tag_sysmem),
2216                                      &error_abort);
2217             if (vms->secure) {
2218                 object_property_set_link(cpuobj, "secure-tag-memory",
2219                                          OBJECT(secure_tag_sysmem),
2220                                          &error_abort);
2221             }
2222         }
2223 
2224         qdev_realize(DEVICE(cpuobj), NULL, &error_fatal);
2225         object_unref(cpuobj);
2226     }
2227     fdt_add_timer_nodes(vms);
2228     fdt_add_cpu_nodes(vms);
2229 
2230     memory_region_add_subregion(sysmem, vms->memmap[VIRT_MEM].base,
2231                                 machine->ram);
2232 
2233     virt_flash_fdt(vms, sysmem, secure_sysmem ?: sysmem);
2234 
2235     create_gic(vms, sysmem);
2236 
2237     virt_cpu_post_init(vms, sysmem);
2238 
2239     fdt_add_pmu_nodes(vms);
2240 
2241     create_uart(vms, VIRT_UART, sysmem, serial_hd(0));
2242 
2243     if (vms->secure) {
2244         create_secure_ram(vms, secure_sysmem, secure_tag_sysmem);
2245         create_uart(vms, VIRT_SECURE_UART, secure_sysmem, serial_hd(1));
2246     }
2247 
2248     if (tag_sysmem) {
2249         create_tag_ram(tag_sysmem, vms->memmap[VIRT_MEM].base,
2250                        machine->ram_size, "mach-virt.tag");
2251     }
2252 
2253     vms->highmem_ecam &= (!firmware_loaded || aarch64);
2254 
2255     create_rtc(vms);
2256 
2257     create_pcie(vms);
2258 
2259     if (has_ged && aarch64 && firmware_loaded && virt_is_acpi_enabled(vms)) {
2260         vms->acpi_dev = create_acpi_ged(vms);
2261     } else {
2262         create_gpio_devices(vms, VIRT_GPIO, sysmem);
2263     }
2264 
2265     if (vms->secure && !vmc->no_secure_gpio) {
2266         create_gpio_devices(vms, VIRT_SECURE_GPIO, secure_sysmem);
2267     }
2268 
2269      /* connect powerdown request */
2270      vms->powerdown_notifier.notify = virt_powerdown_req;
2271      qemu_register_powerdown_notifier(&vms->powerdown_notifier);
2272 
2273     /* Create mmio transports, so the user can create virtio backends
2274      * (which will be automatically plugged in to the transports). If
2275      * no backend is created the transport will just sit harmlessly idle.
2276      */
2277     create_virtio_devices(vms);
2278 
2279     vms->fw_cfg = create_fw_cfg(vms, &address_space_memory);
2280     rom_set_fw(vms->fw_cfg);
2281 
2282     create_platform_bus(vms);
2283 
2284     if (machine->nvdimms_state->is_enabled) {
2285         const struct AcpiGenericAddress arm_virt_nvdimm_acpi_dsmio = {
2286             .space_id = AML_AS_SYSTEM_MEMORY,
2287             .address = vms->memmap[VIRT_NVDIMM_ACPI].base,
2288             .bit_width = NVDIMM_ACPI_IO_LEN << 3
2289         };
2290 
2291         nvdimm_init_acpi_state(machine->nvdimms_state, sysmem,
2292                                arm_virt_nvdimm_acpi_dsmio,
2293                                vms->fw_cfg, OBJECT(vms));
2294     }
2295 
2296     vms->bootinfo.ram_size = machine->ram_size;
2297     vms->bootinfo.board_id = -1;
2298     vms->bootinfo.loader_start = vms->memmap[VIRT_MEM].base;
2299     vms->bootinfo.get_dtb = machvirt_dtb;
2300     vms->bootinfo.skip_dtb_autoload = true;
2301     vms->bootinfo.firmware_loaded = firmware_loaded;
2302     vms->bootinfo.psci_conduit = vms->psci_conduit;
2303     arm_load_kernel(ARM_CPU(first_cpu), machine, &vms->bootinfo);
2304 
2305     vms->machine_done.notify = virt_machine_done;
2306     qemu_add_machine_init_done_notifier(&vms->machine_done);
2307 }
2308 
2309 static bool virt_get_secure(Object *obj, Error **errp)
2310 {
2311     VirtMachineState *vms = VIRT_MACHINE(obj);
2312 
2313     return vms->secure;
2314 }
2315 
2316 static void virt_set_secure(Object *obj, bool value, Error **errp)
2317 {
2318     VirtMachineState *vms = VIRT_MACHINE(obj);
2319 
2320     vms->secure = value;
2321 }
2322 
2323 static bool virt_get_virt(Object *obj, Error **errp)
2324 {
2325     VirtMachineState *vms = VIRT_MACHINE(obj);
2326 
2327     return vms->virt;
2328 }
2329 
2330 static void virt_set_virt(Object *obj, bool value, Error **errp)
2331 {
2332     VirtMachineState *vms = VIRT_MACHINE(obj);
2333 
2334     vms->virt = value;
2335 }
2336 
2337 static bool virt_get_highmem(Object *obj, Error **errp)
2338 {
2339     VirtMachineState *vms = VIRT_MACHINE(obj);
2340 
2341     return vms->highmem;
2342 }
2343 
2344 static void virt_set_highmem(Object *obj, bool value, Error **errp)
2345 {
2346     VirtMachineState *vms = VIRT_MACHINE(obj);
2347 
2348     vms->highmem = value;
2349 }
2350 
2351 static bool virt_get_compact_highmem(Object *obj, Error **errp)
2352 {
2353     VirtMachineState *vms = VIRT_MACHINE(obj);
2354 
2355     return vms->highmem_compact;
2356 }
2357 
2358 static void virt_set_compact_highmem(Object *obj, bool value, Error **errp)
2359 {
2360     VirtMachineState *vms = VIRT_MACHINE(obj);
2361 
2362     vms->highmem_compact = value;
2363 }
2364 
2365 static bool virt_get_highmem_redists(Object *obj, Error **errp)
2366 {
2367     VirtMachineState *vms = VIRT_MACHINE(obj);
2368 
2369     return vms->highmem_redists;
2370 }
2371 
2372 static void virt_set_highmem_redists(Object *obj, bool value, Error **errp)
2373 {
2374     VirtMachineState *vms = VIRT_MACHINE(obj);
2375 
2376     vms->highmem_redists = value;
2377 }
2378 
2379 static bool virt_get_highmem_ecam(Object *obj, Error **errp)
2380 {
2381     VirtMachineState *vms = VIRT_MACHINE(obj);
2382 
2383     return vms->highmem_ecam;
2384 }
2385 
2386 static void virt_set_highmem_ecam(Object *obj, bool value, Error **errp)
2387 {
2388     VirtMachineState *vms = VIRT_MACHINE(obj);
2389 
2390     vms->highmem_ecam = value;
2391 }
2392 
2393 static bool virt_get_highmem_mmio(Object *obj, Error **errp)
2394 {
2395     VirtMachineState *vms = VIRT_MACHINE(obj);
2396 
2397     return vms->highmem_mmio;
2398 }
2399 
2400 static void virt_set_highmem_mmio(Object *obj, bool value, Error **errp)
2401 {
2402     VirtMachineState *vms = VIRT_MACHINE(obj);
2403 
2404     vms->highmem_mmio = value;
2405 }
2406 
2407 
2408 static bool virt_get_its(Object *obj, Error **errp)
2409 {
2410     VirtMachineState *vms = VIRT_MACHINE(obj);
2411 
2412     return vms->its;
2413 }
2414 
2415 static void virt_set_its(Object *obj, bool value, Error **errp)
2416 {
2417     VirtMachineState *vms = VIRT_MACHINE(obj);
2418 
2419     vms->its = value;
2420 }
2421 
2422 static bool virt_get_dtb_randomness(Object *obj, Error **errp)
2423 {
2424     VirtMachineState *vms = VIRT_MACHINE(obj);
2425 
2426     return vms->dtb_randomness;
2427 }
2428 
2429 static void virt_set_dtb_randomness(Object *obj, bool value, Error **errp)
2430 {
2431     VirtMachineState *vms = VIRT_MACHINE(obj);
2432 
2433     vms->dtb_randomness = value;
2434 }
2435 
2436 static char *virt_get_oem_id(Object *obj, Error **errp)
2437 {
2438     VirtMachineState *vms = VIRT_MACHINE(obj);
2439 
2440     return g_strdup(vms->oem_id);
2441 }
2442 
2443 static void virt_set_oem_id(Object *obj, const char *value, Error **errp)
2444 {
2445     VirtMachineState *vms = VIRT_MACHINE(obj);
2446     size_t len = strlen(value);
2447 
2448     if (len > 6) {
2449         error_setg(errp,
2450                    "User specified oem-id value is bigger than 6 bytes in size");
2451         return;
2452     }
2453 
2454     strncpy(vms->oem_id, value, 6);
2455 }
2456 
2457 static char *virt_get_oem_table_id(Object *obj, Error **errp)
2458 {
2459     VirtMachineState *vms = VIRT_MACHINE(obj);
2460 
2461     return g_strdup(vms->oem_table_id);
2462 }
2463 
2464 static void virt_set_oem_table_id(Object *obj, const char *value,
2465                                   Error **errp)
2466 {
2467     VirtMachineState *vms = VIRT_MACHINE(obj);
2468     size_t len = strlen(value);
2469 
2470     if (len > 8) {
2471         error_setg(errp,
2472                    "User specified oem-table-id value is bigger than 8 bytes in size");
2473         return;
2474     }
2475     strncpy(vms->oem_table_id, value, 8);
2476 }
2477 
2478 
2479 bool virt_is_acpi_enabled(VirtMachineState *vms)
2480 {
2481     if (vms->acpi == ON_OFF_AUTO_OFF) {
2482         return false;
2483     }
2484     return true;
2485 }
2486 
2487 static void virt_get_acpi(Object *obj, Visitor *v, const char *name,
2488                           void *opaque, Error **errp)
2489 {
2490     VirtMachineState *vms = VIRT_MACHINE(obj);
2491     OnOffAuto acpi = vms->acpi;
2492 
2493     visit_type_OnOffAuto(v, name, &acpi, errp);
2494 }
2495 
2496 static void virt_set_acpi(Object *obj, Visitor *v, const char *name,
2497                           void *opaque, Error **errp)
2498 {
2499     VirtMachineState *vms = VIRT_MACHINE(obj);
2500 
2501     visit_type_OnOffAuto(v, name, &vms->acpi, errp);
2502 }
2503 
2504 static bool virt_get_ras(Object *obj, Error **errp)
2505 {
2506     VirtMachineState *vms = VIRT_MACHINE(obj);
2507 
2508     return vms->ras;
2509 }
2510 
2511 static void virt_set_ras(Object *obj, bool value, Error **errp)
2512 {
2513     VirtMachineState *vms = VIRT_MACHINE(obj);
2514 
2515     vms->ras = value;
2516 }
2517 
2518 static bool virt_get_mte(Object *obj, Error **errp)
2519 {
2520     VirtMachineState *vms = VIRT_MACHINE(obj);
2521 
2522     return vms->mte;
2523 }
2524 
2525 static void virt_set_mte(Object *obj, bool value, Error **errp)
2526 {
2527     VirtMachineState *vms = VIRT_MACHINE(obj);
2528 
2529     vms->mte = value;
2530 }
2531 
2532 static char *virt_get_gic_version(Object *obj, Error **errp)
2533 {
2534     VirtMachineState *vms = VIRT_MACHINE(obj);
2535     const char *val;
2536 
2537     switch (vms->gic_version) {
2538     case VIRT_GIC_VERSION_4:
2539         val = "4";
2540         break;
2541     case VIRT_GIC_VERSION_3:
2542         val = "3";
2543         break;
2544     default:
2545         val = "2";
2546         break;
2547     }
2548     return g_strdup(val);
2549 }
2550 
2551 static void virt_set_gic_version(Object *obj, const char *value, Error **errp)
2552 {
2553     VirtMachineState *vms = VIRT_MACHINE(obj);
2554 
2555     if (!strcmp(value, "4")) {
2556         vms->gic_version = VIRT_GIC_VERSION_4;
2557     } else if (!strcmp(value, "3")) {
2558         vms->gic_version = VIRT_GIC_VERSION_3;
2559     } else if (!strcmp(value, "2")) {
2560         vms->gic_version = VIRT_GIC_VERSION_2;
2561     } else if (!strcmp(value, "host")) {
2562         vms->gic_version = VIRT_GIC_VERSION_HOST; /* Will probe later */
2563     } else if (!strcmp(value, "max")) {
2564         vms->gic_version = VIRT_GIC_VERSION_MAX; /* Will probe later */
2565     } else {
2566         error_setg(errp, "Invalid gic-version value");
2567         error_append_hint(errp, "Valid values are 3, 2, host, max.\n");
2568     }
2569 }
2570 
2571 static char *virt_get_iommu(Object *obj, Error **errp)
2572 {
2573     VirtMachineState *vms = VIRT_MACHINE(obj);
2574 
2575     switch (vms->iommu) {
2576     case VIRT_IOMMU_NONE:
2577         return g_strdup("none");
2578     case VIRT_IOMMU_SMMUV3:
2579         return g_strdup("smmuv3");
2580     default:
2581         g_assert_not_reached();
2582     }
2583 }
2584 
2585 static void virt_set_iommu(Object *obj, const char *value, Error **errp)
2586 {
2587     VirtMachineState *vms = VIRT_MACHINE(obj);
2588 
2589     if (!strcmp(value, "smmuv3")) {
2590         vms->iommu = VIRT_IOMMU_SMMUV3;
2591     } else if (!strcmp(value, "none")) {
2592         vms->iommu = VIRT_IOMMU_NONE;
2593     } else {
2594         error_setg(errp, "Invalid iommu value");
2595         error_append_hint(errp, "Valid values are none, smmuv3.\n");
2596     }
2597 }
2598 
2599 static bool virt_get_default_bus_bypass_iommu(Object *obj, Error **errp)
2600 {
2601     VirtMachineState *vms = VIRT_MACHINE(obj);
2602 
2603     return vms->default_bus_bypass_iommu;
2604 }
2605 
2606 static void virt_set_default_bus_bypass_iommu(Object *obj, bool value,
2607                                               Error **errp)
2608 {
2609     VirtMachineState *vms = VIRT_MACHINE(obj);
2610 
2611     vms->default_bus_bypass_iommu = value;
2612 }
2613 
2614 static CpuInstanceProperties
2615 virt_cpu_index_to_props(MachineState *ms, unsigned cpu_index)
2616 {
2617     MachineClass *mc = MACHINE_GET_CLASS(ms);
2618     const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(ms);
2619 
2620     assert(cpu_index < possible_cpus->len);
2621     return possible_cpus->cpus[cpu_index].props;
2622 }
2623 
2624 static int64_t virt_get_default_cpu_node_id(const MachineState *ms, int idx)
2625 {
2626     int64_t socket_id = ms->possible_cpus->cpus[idx].props.socket_id;
2627 
2628     return socket_id % ms->numa_state->num_nodes;
2629 }
2630 
2631 static const CPUArchIdList *virt_possible_cpu_arch_ids(MachineState *ms)
2632 {
2633     int n;
2634     unsigned int max_cpus = ms->smp.max_cpus;
2635     VirtMachineState *vms = VIRT_MACHINE(ms);
2636     MachineClass *mc = MACHINE_GET_CLASS(vms);
2637 
2638     if (ms->possible_cpus) {
2639         assert(ms->possible_cpus->len == max_cpus);
2640         return ms->possible_cpus;
2641     }
2642 
2643     ms->possible_cpus = g_malloc0(sizeof(CPUArchIdList) +
2644                                   sizeof(CPUArchId) * max_cpus);
2645     ms->possible_cpus->len = max_cpus;
2646     for (n = 0; n < ms->possible_cpus->len; n++) {
2647         ms->possible_cpus->cpus[n].type = ms->cpu_type;
2648         ms->possible_cpus->cpus[n].arch_id =
2649             virt_cpu_mp_affinity(vms, n);
2650 
2651         assert(!mc->smp_props.dies_supported);
2652         ms->possible_cpus->cpus[n].props.has_socket_id = true;
2653         ms->possible_cpus->cpus[n].props.socket_id =
2654             n / (ms->smp.clusters * ms->smp.cores * ms->smp.threads);
2655         ms->possible_cpus->cpus[n].props.has_cluster_id = true;
2656         ms->possible_cpus->cpus[n].props.cluster_id =
2657             (n / (ms->smp.cores * ms->smp.threads)) % ms->smp.clusters;
2658         ms->possible_cpus->cpus[n].props.has_core_id = true;
2659         ms->possible_cpus->cpus[n].props.core_id =
2660             (n / ms->smp.threads) % ms->smp.cores;
2661         ms->possible_cpus->cpus[n].props.has_thread_id = true;
2662         ms->possible_cpus->cpus[n].props.thread_id =
2663             n % ms->smp.threads;
2664     }
2665     return ms->possible_cpus;
2666 }
2667 
2668 static void virt_memory_pre_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
2669                                  Error **errp)
2670 {
2671     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2672     const MachineState *ms = MACHINE(hotplug_dev);
2673     const bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2674 
2675     if (!vms->acpi_dev) {
2676         error_setg(errp,
2677                    "memory hotplug is not enabled: missing acpi-ged device");
2678         return;
2679     }
2680 
2681     if (vms->mte) {
2682         error_setg(errp, "memory hotplug is not enabled: MTE is enabled");
2683         return;
2684     }
2685 
2686     if (is_nvdimm && !ms->nvdimms_state->is_enabled) {
2687         error_setg(errp, "nvdimm is not enabled: add 'nvdimm=on' to '-M'");
2688         return;
2689     }
2690 
2691     pc_dimm_pre_plug(PC_DIMM(dev), MACHINE(hotplug_dev), NULL, errp);
2692 }
2693 
2694 static void virt_memory_plug(HotplugHandler *hotplug_dev,
2695                              DeviceState *dev, Error **errp)
2696 {
2697     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2698     MachineState *ms = MACHINE(hotplug_dev);
2699     bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2700 
2701     pc_dimm_plug(PC_DIMM(dev), MACHINE(vms));
2702 
2703     if (is_nvdimm) {
2704         nvdimm_plug(ms->nvdimms_state);
2705     }
2706 
2707     hotplug_handler_plug(HOTPLUG_HANDLER(vms->acpi_dev),
2708                          dev, &error_abort);
2709 }
2710 
2711 static void virt_machine_device_pre_plug_cb(HotplugHandler *hotplug_dev,
2712                                             DeviceState *dev, Error **errp)
2713 {
2714     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2715 
2716     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2717         virt_memory_pre_plug(hotplug_dev, dev, errp);
2718     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI)) {
2719         virtio_md_pci_pre_plug(VIRTIO_MD_PCI(dev), MACHINE(hotplug_dev), errp);
2720     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2721         hwaddr db_start = 0, db_end = 0;
2722         QList *reserved_regions;
2723         char *resv_prop_str;
2724 
2725         if (vms->iommu != VIRT_IOMMU_NONE) {
2726             error_setg(errp, "virt machine does not support multiple IOMMUs");
2727             return;
2728         }
2729 
2730         switch (vms->msi_controller) {
2731         case VIRT_MSI_CTRL_NONE:
2732             return;
2733         case VIRT_MSI_CTRL_ITS:
2734             /* GITS_TRANSLATER page */
2735             db_start = base_memmap[VIRT_GIC_ITS].base + 0x10000;
2736             db_end = base_memmap[VIRT_GIC_ITS].base +
2737                      base_memmap[VIRT_GIC_ITS].size - 1;
2738             break;
2739         case VIRT_MSI_CTRL_GICV2M:
2740             /* MSI_SETSPI_NS page */
2741             db_start = base_memmap[VIRT_GIC_V2M].base;
2742             db_end = db_start + base_memmap[VIRT_GIC_V2M].size - 1;
2743             break;
2744         }
2745         resv_prop_str = g_strdup_printf("0x%"PRIx64":0x%"PRIx64":%u",
2746                                         db_start, db_end,
2747                                         VIRTIO_IOMMU_RESV_MEM_T_MSI);
2748 
2749         reserved_regions = qlist_new();
2750         qlist_append_str(reserved_regions, resv_prop_str);
2751         qdev_prop_set_array(dev, "reserved-regions", reserved_regions);
2752         g_free(resv_prop_str);
2753     }
2754 }
2755 
2756 static void virt_machine_device_plug_cb(HotplugHandler *hotplug_dev,
2757                                         DeviceState *dev, Error **errp)
2758 {
2759     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2760 
2761     if (vms->platform_bus_dev) {
2762         MachineClass *mc = MACHINE_GET_CLASS(vms);
2763 
2764         if (device_is_dynamic_sysbus(mc, dev)) {
2765             platform_bus_link_device(PLATFORM_BUS_DEVICE(vms->platform_bus_dev),
2766                                      SYS_BUS_DEVICE(dev));
2767         }
2768     }
2769 
2770     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2771         virt_memory_plug(hotplug_dev, dev, errp);
2772     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI)) {
2773         virtio_md_pci_plug(VIRTIO_MD_PCI(dev), MACHINE(hotplug_dev), errp);
2774     }
2775 
2776     if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2777         PCIDevice *pdev = PCI_DEVICE(dev);
2778 
2779         vms->iommu = VIRT_IOMMU_VIRTIO;
2780         vms->virtio_iommu_bdf = pci_get_bdf(pdev);
2781         create_virtio_iommu_dt_bindings(vms);
2782     }
2783 }
2784 
2785 static void virt_dimm_unplug_request(HotplugHandler *hotplug_dev,
2786                                      DeviceState *dev, Error **errp)
2787 {
2788     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2789 
2790     if (!vms->acpi_dev) {
2791         error_setg(errp,
2792                    "memory hotplug is not enabled: missing acpi-ged device");
2793         return;
2794     }
2795 
2796     if (object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM)) {
2797         error_setg(errp, "nvdimm device hot unplug is not supported yet.");
2798         return;
2799     }
2800 
2801     hotplug_handler_unplug_request(HOTPLUG_HANDLER(vms->acpi_dev), dev,
2802                                    errp);
2803 }
2804 
2805 static void virt_dimm_unplug(HotplugHandler *hotplug_dev,
2806                              DeviceState *dev, Error **errp)
2807 {
2808     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2809     Error *local_err = NULL;
2810 
2811     hotplug_handler_unplug(HOTPLUG_HANDLER(vms->acpi_dev), dev, &local_err);
2812     if (local_err) {
2813         goto out;
2814     }
2815 
2816     pc_dimm_unplug(PC_DIMM(dev), MACHINE(vms));
2817     qdev_unrealize(dev);
2818 
2819 out:
2820     error_propagate(errp, local_err);
2821 }
2822 
2823 static void virt_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
2824                                           DeviceState *dev, Error **errp)
2825 {
2826     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2827         virt_dimm_unplug_request(hotplug_dev, dev, errp);
2828     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI)) {
2829         virtio_md_pci_unplug_request(VIRTIO_MD_PCI(dev), MACHINE(hotplug_dev),
2830                                      errp);
2831     } else {
2832         error_setg(errp, "device unplug request for unsupported device"
2833                    " type: %s", object_get_typename(OBJECT(dev)));
2834     }
2835 }
2836 
2837 static void virt_machine_device_unplug_cb(HotplugHandler *hotplug_dev,
2838                                           DeviceState *dev, Error **errp)
2839 {
2840     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2841         virt_dimm_unplug(hotplug_dev, dev, errp);
2842     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI)) {
2843         virtio_md_pci_unplug(VIRTIO_MD_PCI(dev), MACHINE(hotplug_dev), errp);
2844     } else {
2845         error_setg(errp, "virt: device unplug for unsupported device"
2846                    " type: %s", object_get_typename(OBJECT(dev)));
2847     }
2848 }
2849 
2850 static HotplugHandler *virt_machine_get_hotplug_handler(MachineState *machine,
2851                                                         DeviceState *dev)
2852 {
2853     MachineClass *mc = MACHINE_GET_CLASS(machine);
2854 
2855     if (device_is_dynamic_sysbus(mc, dev) ||
2856         object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) ||
2857         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI) ||
2858         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2859         return HOTPLUG_HANDLER(machine);
2860     }
2861     return NULL;
2862 }
2863 
2864 /*
2865  * for arm64 kvm_type [7-0] encodes the requested number of bits
2866  * in the IPA address space
2867  */
2868 static int virt_kvm_type(MachineState *ms, const char *type_str)
2869 {
2870     VirtMachineState *vms = VIRT_MACHINE(ms);
2871     int max_vm_pa_size, requested_pa_size;
2872     bool fixed_ipa;
2873 
2874     max_vm_pa_size = kvm_arm_get_max_vm_ipa_size(ms, &fixed_ipa);
2875 
2876     /* we freeze the memory map to compute the highest gpa */
2877     virt_set_memmap(vms, max_vm_pa_size);
2878 
2879     requested_pa_size = 64 - clz64(vms->highest_gpa);
2880 
2881     /*
2882      * KVM requires the IPA size to be at least 32 bits.
2883      */
2884     if (requested_pa_size < 32) {
2885         requested_pa_size = 32;
2886     }
2887 
2888     if (requested_pa_size > max_vm_pa_size) {
2889         error_report("-m and ,maxmem option values "
2890                      "require an IPA range (%d bits) larger than "
2891                      "the one supported by the host (%d bits)",
2892                      requested_pa_size, max_vm_pa_size);
2893         return -1;
2894     }
2895     /*
2896      * We return the requested PA log size, unless KVM only supports
2897      * the implicit legacy 40b IPA setting, in which case the kvm_type
2898      * must be 0.
2899      */
2900     return fixed_ipa ? 0 : requested_pa_size;
2901 }
2902 
2903 static void virt_machine_class_init(ObjectClass *oc, void *data)
2904 {
2905     MachineClass *mc = MACHINE_CLASS(oc);
2906     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
2907     static const char * const valid_cpu_types[] = {
2908 #ifdef CONFIG_TCG
2909         ARM_CPU_TYPE_NAME("cortex-a7"),
2910         ARM_CPU_TYPE_NAME("cortex-a15"),
2911 #ifdef TARGET_AARCH64
2912         ARM_CPU_TYPE_NAME("cortex-a35"),
2913         ARM_CPU_TYPE_NAME("cortex-a55"),
2914         ARM_CPU_TYPE_NAME("cortex-a72"),
2915         ARM_CPU_TYPE_NAME("cortex-a76"),
2916         ARM_CPU_TYPE_NAME("cortex-a710"),
2917         ARM_CPU_TYPE_NAME("a64fx"),
2918         ARM_CPU_TYPE_NAME("neoverse-n1"),
2919         ARM_CPU_TYPE_NAME("neoverse-v1"),
2920         ARM_CPU_TYPE_NAME("neoverse-n2"),
2921 #endif /* TARGET_AARCH64 */
2922 #endif /* CONFIG_TCG */
2923 #ifdef TARGET_AARCH64
2924         ARM_CPU_TYPE_NAME("cortex-a53"),
2925         ARM_CPU_TYPE_NAME("cortex-a57"),
2926 #if defined(CONFIG_KVM) || defined(CONFIG_HVF)
2927         ARM_CPU_TYPE_NAME("host"),
2928 #endif /* CONFIG_KVM || CONFIG_HVF */
2929 #endif /* TARGET_AARCH64 */
2930         ARM_CPU_TYPE_NAME("max"),
2931         NULL
2932     };
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 #ifdef CONFIG_TCG
2955     mc->default_cpu_type = ARM_CPU_TYPE_NAME("cortex-a15");
2956 #else
2957     mc->default_cpu_type = ARM_CPU_TYPE_NAME("max");
2958 #endif
2959     mc->valid_cpu_types = valid_cpu_types;
2960     mc->get_default_cpu_node_id = virt_get_default_cpu_node_id;
2961     mc->kvm_type = virt_kvm_type;
2962     assert(!mc->get_hotplug_handler);
2963     mc->get_hotplug_handler = virt_machine_get_hotplug_handler;
2964     hc->pre_plug = virt_machine_device_pre_plug_cb;
2965     hc->plug = virt_machine_device_plug_cb;
2966     hc->unplug_request = virt_machine_device_unplug_request_cb;
2967     hc->unplug = virt_machine_device_unplug_cb;
2968     mc->nvdimm_supported = true;
2969     mc->smp_props.clusters_supported = true;
2970     mc->auto_enable_numa_with_memhp = true;
2971     mc->auto_enable_numa_with_memdev = true;
2972     /* platform instead of architectural choice */
2973     mc->cpu_cluster_has_numa_boundary = true;
2974     mc->default_ram_id = "mach-virt.ram";
2975     mc->default_nic = "virtio-net-pci";
2976 
2977     object_class_property_add(oc, "acpi", "OnOffAuto",
2978         virt_get_acpi, virt_set_acpi,
2979         NULL, NULL);
2980     object_class_property_set_description(oc, "acpi",
2981         "Enable ACPI");
2982     object_class_property_add_bool(oc, "secure", virt_get_secure,
2983                                    virt_set_secure);
2984     object_class_property_set_description(oc, "secure",
2985                                                 "Set on/off to enable/disable the ARM "
2986                                                 "Security Extensions (TrustZone)");
2987 
2988     object_class_property_add_bool(oc, "virtualization", virt_get_virt,
2989                                    virt_set_virt);
2990     object_class_property_set_description(oc, "virtualization",
2991                                           "Set on/off to enable/disable emulating a "
2992                                           "guest CPU which implements the ARM "
2993                                           "Virtualization Extensions");
2994 
2995     object_class_property_add_bool(oc, "highmem", virt_get_highmem,
2996                                    virt_set_highmem);
2997     object_class_property_set_description(oc, "highmem",
2998                                           "Set on/off to enable/disable using "
2999                                           "physical address space above 32 bits");
3000 
3001     object_class_property_add_bool(oc, "compact-highmem",
3002                                    virt_get_compact_highmem,
3003                                    virt_set_compact_highmem);
3004     object_class_property_set_description(oc, "compact-highmem",
3005                                           "Set on/off to enable/disable compact "
3006                                           "layout for high memory regions");
3007 
3008     object_class_property_add_bool(oc, "highmem-redists",
3009                                    virt_get_highmem_redists,
3010                                    virt_set_highmem_redists);
3011     object_class_property_set_description(oc, "highmem-redists",
3012                                           "Set on/off to enable/disable high "
3013                                           "memory region for GICv3 or GICv4 "
3014                                           "redistributor");
3015 
3016     object_class_property_add_bool(oc, "highmem-ecam",
3017                                    virt_get_highmem_ecam,
3018                                    virt_set_highmem_ecam);
3019     object_class_property_set_description(oc, "highmem-ecam",
3020                                           "Set on/off to enable/disable high "
3021                                           "memory region for PCI ECAM");
3022 
3023     object_class_property_add_bool(oc, "highmem-mmio",
3024                                    virt_get_highmem_mmio,
3025                                    virt_set_highmem_mmio);
3026     object_class_property_set_description(oc, "highmem-mmio",
3027                                           "Set on/off to enable/disable high "
3028                                           "memory region for PCI MMIO");
3029 
3030     object_class_property_add_str(oc, "gic-version", virt_get_gic_version,
3031                                   virt_set_gic_version);
3032     object_class_property_set_description(oc, "gic-version",
3033                                           "Set GIC version. "
3034                                           "Valid values are 2, 3, 4, host and max");
3035 
3036     object_class_property_add_str(oc, "iommu", virt_get_iommu, virt_set_iommu);
3037     object_class_property_set_description(oc, "iommu",
3038                                           "Set the IOMMU type. "
3039                                           "Valid values are none and smmuv3");
3040 
3041     object_class_property_add_bool(oc, "default-bus-bypass-iommu",
3042                                    virt_get_default_bus_bypass_iommu,
3043                                    virt_set_default_bus_bypass_iommu);
3044     object_class_property_set_description(oc, "default-bus-bypass-iommu",
3045                                           "Set on/off to enable/disable "
3046                                           "bypass_iommu for default root bus");
3047 
3048     object_class_property_add_bool(oc, "ras", virt_get_ras,
3049                                    virt_set_ras);
3050     object_class_property_set_description(oc, "ras",
3051                                           "Set on/off to enable/disable reporting host memory errors "
3052                                           "to a KVM guest using ACPI and guest external abort exceptions");
3053 
3054     object_class_property_add_bool(oc, "mte", virt_get_mte, virt_set_mte);
3055     object_class_property_set_description(oc, "mte",
3056                                           "Set on/off to enable/disable emulating a "
3057                                           "guest CPU which implements the ARM "
3058                                           "Memory Tagging Extension");
3059 
3060     object_class_property_add_bool(oc, "its", virt_get_its,
3061                                    virt_set_its);
3062     object_class_property_set_description(oc, "its",
3063                                           "Set on/off to enable/disable "
3064                                           "ITS instantiation");
3065 
3066     object_class_property_add_bool(oc, "dtb-randomness",
3067                                    virt_get_dtb_randomness,
3068                                    virt_set_dtb_randomness);
3069     object_class_property_set_description(oc, "dtb-randomness",
3070                                           "Set off to disable passing random or "
3071                                           "non-deterministic dtb nodes to guest");
3072 
3073     object_class_property_add_bool(oc, "dtb-kaslr-seed",
3074                                    virt_get_dtb_randomness,
3075                                    virt_set_dtb_randomness);
3076     object_class_property_set_description(oc, "dtb-kaslr-seed",
3077                                           "Deprecated synonym of dtb-randomness");
3078 
3079     object_class_property_add_str(oc, "x-oem-id",
3080                                   virt_get_oem_id,
3081                                   virt_set_oem_id);
3082     object_class_property_set_description(oc, "x-oem-id",
3083                                           "Override the default value of field OEMID "
3084                                           "in ACPI table header."
3085                                           "The string may be up to 6 bytes in size");
3086 
3087 
3088     object_class_property_add_str(oc, "x-oem-table-id",
3089                                   virt_get_oem_table_id,
3090                                   virt_set_oem_table_id);
3091     object_class_property_set_description(oc, "x-oem-table-id",
3092                                           "Override the default value of field OEM Table ID "
3093                                           "in ACPI table header."
3094                                           "The string may be up to 8 bytes in size");
3095 
3096 }
3097 
3098 static void virt_instance_init(Object *obj)
3099 {
3100     VirtMachineState *vms = VIRT_MACHINE(obj);
3101     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
3102 
3103     /* EL3 is disabled by default on virt: this makes us consistent
3104      * between KVM and TCG for this board, and it also allows us to
3105      * boot UEFI blobs which assume no TrustZone support.
3106      */
3107     vms->secure = false;
3108 
3109     /* EL2 is also disabled by default, for similar reasons */
3110     vms->virt = false;
3111 
3112     /* High memory is enabled by default */
3113     vms->highmem = true;
3114     vms->highmem_compact = !vmc->no_highmem_compact;
3115     vms->gic_version = VIRT_GIC_VERSION_NOSEL;
3116 
3117     vms->highmem_ecam = !vmc->no_highmem_ecam;
3118     vms->highmem_mmio = true;
3119     vms->highmem_redists = true;
3120 
3121     if (vmc->no_its) {
3122         vms->its = false;
3123     } else {
3124         /* Default allows ITS instantiation */
3125         vms->its = true;
3126 
3127         if (vmc->no_tcg_its) {
3128             vms->tcg_its = false;
3129         } else {
3130             vms->tcg_its = true;
3131         }
3132     }
3133 
3134     /* Default disallows iommu instantiation */
3135     vms->iommu = VIRT_IOMMU_NONE;
3136 
3137     /* The default root bus is attached to iommu by default */
3138     vms->default_bus_bypass_iommu = false;
3139 
3140     /* Default disallows RAS instantiation */
3141     vms->ras = false;
3142 
3143     /* MTE is disabled by default.  */
3144     vms->mte = false;
3145 
3146     /* Supply kaslr-seed and rng-seed by default */
3147     vms->dtb_randomness = true;
3148 
3149     vms->irqmap = a15irqmap;
3150 
3151     virt_flash_create(vms);
3152 
3153     vms->oem_id = g_strndup(ACPI_BUILD_APPNAME6, 6);
3154     vms->oem_table_id = g_strndup(ACPI_BUILD_APPNAME8, 8);
3155 }
3156 
3157 static const TypeInfo virt_machine_info = {
3158     .name          = TYPE_VIRT_MACHINE,
3159     .parent        = TYPE_MACHINE,
3160     .abstract      = true,
3161     .instance_size = sizeof(VirtMachineState),
3162     .class_size    = sizeof(VirtMachineClass),
3163     .class_init    = virt_machine_class_init,
3164     .instance_init = virt_instance_init,
3165     .interfaces = (InterfaceInfo[]) {
3166          { TYPE_HOTPLUG_HANDLER },
3167          { }
3168     },
3169 };
3170 
3171 static void machvirt_machine_init(void)
3172 {
3173     type_register_static(&virt_machine_info);
3174 }
3175 type_init(machvirt_machine_init);
3176 
3177 static void virt_machine_9_0_options(MachineClass *mc)
3178 {
3179 }
3180 DEFINE_VIRT_MACHINE_AS_LATEST(9, 0)
3181 
3182 static void virt_machine_8_2_options(MachineClass *mc)
3183 {
3184     virt_machine_9_0_options(mc);
3185     compat_props_add(mc->compat_props, hw_compat_8_2, hw_compat_8_2_len);
3186 }
3187 DEFINE_VIRT_MACHINE(8, 2)
3188 
3189 static void virt_machine_8_1_options(MachineClass *mc)
3190 {
3191     virt_machine_8_2_options(mc);
3192     compat_props_add(mc->compat_props, hw_compat_8_1, hw_compat_8_1_len);
3193 }
3194 DEFINE_VIRT_MACHINE(8, 1)
3195 
3196 static void virt_machine_8_0_options(MachineClass *mc)
3197 {
3198     virt_machine_8_1_options(mc);
3199     compat_props_add(mc->compat_props, hw_compat_8_0, hw_compat_8_0_len);
3200 }
3201 DEFINE_VIRT_MACHINE(8, 0)
3202 
3203 static void virt_machine_7_2_options(MachineClass *mc)
3204 {
3205     virt_machine_8_0_options(mc);
3206     compat_props_add(mc->compat_props, hw_compat_7_2, hw_compat_7_2_len);
3207 }
3208 DEFINE_VIRT_MACHINE(7, 2)
3209 
3210 static void virt_machine_7_1_options(MachineClass *mc)
3211 {
3212     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3213 
3214     virt_machine_7_2_options(mc);
3215     compat_props_add(mc->compat_props, hw_compat_7_1, hw_compat_7_1_len);
3216     /* Compact layout for high memory regions was introduced with 7.2 */
3217     vmc->no_highmem_compact = true;
3218 }
3219 DEFINE_VIRT_MACHINE(7, 1)
3220 
3221 static void virt_machine_7_0_options(MachineClass *mc)
3222 {
3223     virt_machine_7_1_options(mc);
3224     compat_props_add(mc->compat_props, hw_compat_7_0, hw_compat_7_0_len);
3225 }
3226 DEFINE_VIRT_MACHINE(7, 0)
3227 
3228 static void virt_machine_6_2_options(MachineClass *mc)
3229 {
3230     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3231 
3232     virt_machine_7_0_options(mc);
3233     compat_props_add(mc->compat_props, hw_compat_6_2, hw_compat_6_2_len);
3234     vmc->no_tcg_lpa2 = true;
3235 }
3236 DEFINE_VIRT_MACHINE(6, 2)
3237 
3238 static void virt_machine_6_1_options(MachineClass *mc)
3239 {
3240     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3241 
3242     virt_machine_6_2_options(mc);
3243     compat_props_add(mc->compat_props, hw_compat_6_1, hw_compat_6_1_len);
3244     mc->smp_props.prefer_sockets = true;
3245     vmc->no_cpu_topology = true;
3246 
3247     /* qemu ITS was introduced with 6.2 */
3248     vmc->no_tcg_its = true;
3249 }
3250 DEFINE_VIRT_MACHINE(6, 1)
3251 
3252 static void virt_machine_6_0_options(MachineClass *mc)
3253 {
3254     virt_machine_6_1_options(mc);
3255     compat_props_add(mc->compat_props, hw_compat_6_0, hw_compat_6_0_len);
3256 }
3257 DEFINE_VIRT_MACHINE(6, 0)
3258 
3259 static void virt_machine_5_2_options(MachineClass *mc)
3260 {
3261     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3262 
3263     virt_machine_6_0_options(mc);
3264     compat_props_add(mc->compat_props, hw_compat_5_2, hw_compat_5_2_len);
3265     vmc->no_secure_gpio = true;
3266 }
3267 DEFINE_VIRT_MACHINE(5, 2)
3268 
3269 static void virt_machine_5_1_options(MachineClass *mc)
3270 {
3271     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3272 
3273     virt_machine_5_2_options(mc);
3274     compat_props_add(mc->compat_props, hw_compat_5_1, hw_compat_5_1_len);
3275     vmc->no_kvm_steal_time = true;
3276 }
3277 DEFINE_VIRT_MACHINE(5, 1)
3278 
3279 static void virt_machine_5_0_options(MachineClass *mc)
3280 {
3281     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3282 
3283     virt_machine_5_1_options(mc);
3284     compat_props_add(mc->compat_props, hw_compat_5_0, hw_compat_5_0_len);
3285     mc->numa_mem_supported = true;
3286     vmc->acpi_expose_flash = true;
3287     mc->auto_enable_numa_with_memdev = false;
3288 }
3289 DEFINE_VIRT_MACHINE(5, 0)
3290 
3291 static void virt_machine_4_2_options(MachineClass *mc)
3292 {
3293     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3294 
3295     virt_machine_5_0_options(mc);
3296     compat_props_add(mc->compat_props, hw_compat_4_2, hw_compat_4_2_len);
3297     vmc->kvm_no_adjvtime = true;
3298 }
3299 DEFINE_VIRT_MACHINE(4, 2)
3300 
3301 static void virt_machine_4_1_options(MachineClass *mc)
3302 {
3303     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3304 
3305     virt_machine_4_2_options(mc);
3306     compat_props_add(mc->compat_props, hw_compat_4_1, hw_compat_4_1_len);
3307     vmc->no_ged = true;
3308     mc->auto_enable_numa_with_memhp = false;
3309 }
3310 DEFINE_VIRT_MACHINE(4, 1)
3311 
3312 static void virt_machine_4_0_options(MachineClass *mc)
3313 {
3314     virt_machine_4_1_options(mc);
3315     compat_props_add(mc->compat_props, hw_compat_4_0, hw_compat_4_0_len);
3316 }
3317 DEFINE_VIRT_MACHINE(4, 0)
3318 
3319 static void virt_machine_3_1_options(MachineClass *mc)
3320 {
3321     virt_machine_4_0_options(mc);
3322     compat_props_add(mc->compat_props, hw_compat_3_1, hw_compat_3_1_len);
3323 }
3324 DEFINE_VIRT_MACHINE(3, 1)
3325 
3326 static void virt_machine_3_0_options(MachineClass *mc)
3327 {
3328     virt_machine_3_1_options(mc);
3329     compat_props_add(mc->compat_props, hw_compat_3_0, hw_compat_3_0_len);
3330 }
3331 DEFINE_VIRT_MACHINE(3, 0)
3332 
3333 static void virt_machine_2_12_options(MachineClass *mc)
3334 {
3335     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3336 
3337     virt_machine_3_0_options(mc);
3338     compat_props_add(mc->compat_props, hw_compat_2_12, hw_compat_2_12_len);
3339     vmc->no_highmem_ecam = true;
3340     mc->max_cpus = 255;
3341 }
3342 DEFINE_VIRT_MACHINE(2, 12)
3343 
3344 static void virt_machine_2_11_options(MachineClass *mc)
3345 {
3346     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3347 
3348     virt_machine_2_12_options(mc);
3349     compat_props_add(mc->compat_props, hw_compat_2_11, hw_compat_2_11_len);
3350     vmc->smbios_old_sys_ver = true;
3351 }
3352 DEFINE_VIRT_MACHINE(2, 11)
3353 
3354 static void virt_machine_2_10_options(MachineClass *mc)
3355 {
3356     virt_machine_2_11_options(mc);
3357     compat_props_add(mc->compat_props, hw_compat_2_10, hw_compat_2_10_len);
3358     /* before 2.11 we never faulted accesses to bad addresses */
3359     mc->ignore_memory_transaction_failures = true;
3360 }
3361 DEFINE_VIRT_MACHINE(2, 10)
3362 
3363 static void virt_machine_2_9_options(MachineClass *mc)
3364 {
3365     virt_machine_2_10_options(mc);
3366     compat_props_add(mc->compat_props, hw_compat_2_9, hw_compat_2_9_len);
3367 }
3368 DEFINE_VIRT_MACHINE(2, 9)
3369 
3370 static void virt_machine_2_8_options(MachineClass *mc)
3371 {
3372     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3373 
3374     virt_machine_2_9_options(mc);
3375     compat_props_add(mc->compat_props, hw_compat_2_8, hw_compat_2_8_len);
3376     /* For 2.8 and earlier we falsely claimed in the DT that
3377      * our timers were edge-triggered, not level-triggered.
3378      */
3379     vmc->claim_edge_triggered_timers = true;
3380 }
3381 DEFINE_VIRT_MACHINE(2, 8)
3382 
3383 static void virt_machine_2_7_options(MachineClass *mc)
3384 {
3385     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3386 
3387     virt_machine_2_8_options(mc);
3388     compat_props_add(mc->compat_props, hw_compat_2_7, hw_compat_2_7_len);
3389     /* ITS was introduced with 2.8 */
3390     vmc->no_its = true;
3391     /* Stick with 1K pages for migration compatibility */
3392     mc->minimum_page_bits = 0;
3393 }
3394 DEFINE_VIRT_MACHINE(2, 7)
3395 
3396 static void virt_machine_2_6_options(MachineClass *mc)
3397 {
3398     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3399 
3400     virt_machine_2_7_options(mc);
3401     compat_props_add(mc->compat_props, hw_compat_2_6, hw_compat_2_6_len);
3402     vmc->disallow_affinity_adjustment = true;
3403     /* Disable PMU for 2.6 as PMU support was first introduced in 2.7 */
3404     vmc->no_pmu = true;
3405 }
3406 DEFINE_VIRT_MACHINE(2, 6)
3407