xref: /qemu/hw/arm/virt.c (revision 2bf40d08)
1 /*
2  * ARM mach-virt emulation
3  *
4  * Copyright (c) 2013 Linaro Limited
5  *
6  * This program is free software; you can redistribute it and/or modify it
7  * under the terms and conditions of the GNU General Public License,
8  * version 2 or later, as published by the Free Software Foundation.
9  *
10  * This program is distributed in the hope it will be useful, but WITHOUT
11  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
12  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
13  * more details.
14  *
15  * You should have received a copy of the GNU General Public License along with
16  * this program.  If not, see <http://www.gnu.org/licenses/>.
17  *
18  * Emulate a virtual board which works by passing Linux all the information
19  * it needs about what devices are present via the device tree.
20  * There are some restrictions about what we can do here:
21  *  + we can only present devices whose Linux drivers will work based
22  *    purely on the device tree with no platform data at all
23  *  + we want to present a very stripped-down minimalist platform,
24  *    both because this reduces the security attack surface from the guest
25  *    and also because it reduces our exposure to being broken when
26  *    the kernel updates its device tree bindings and requires further
27  *    information in a device binding that we aren't providing.
28  * This is essentially the same approach kvmtool uses.
29  */
30 
31 #include "qemu/osdep.h"
32 #include "qemu/datadir.h"
33 #include "qemu/units.h"
34 #include "qemu/option.h"
35 #include "monitor/qdev.h"
36 #include "qapi/error.h"
37 #include "hw/sysbus.h"
38 #include "hw/arm/boot.h"
39 #include "hw/arm/primecell.h"
40 #include "hw/arm/virt.h"
41 #include "hw/block/flash.h"
42 #include "hw/vfio/vfio-calxeda-xgmac.h"
43 #include "hw/vfio/vfio-amd-xgbe.h"
44 #include "hw/display/ramfb.h"
45 #include "net/net.h"
46 #include "sysemu/device_tree.h"
47 #include "sysemu/numa.h"
48 #include "sysemu/runstate.h"
49 #include "sysemu/tpm.h"
50 #include "sysemu/kvm.h"
51 #include "sysemu/hvf.h"
52 #include "hw/loader.h"
53 #include "qapi/error.h"
54 #include "qemu/bitops.h"
55 #include "qemu/error-report.h"
56 #include "qemu/module.h"
57 #include "hw/pci-host/gpex.h"
58 #include "hw/virtio/virtio-pci.h"
59 #include "hw/arm/sysbus-fdt.h"
60 #include "hw/platform-bus.h"
61 #include "hw/qdev-properties.h"
62 #include "hw/arm/fdt.h"
63 #include "hw/intc/arm_gic.h"
64 #include "hw/intc/arm_gicv3_common.h"
65 #include "hw/irq.h"
66 #include "kvm_arm.h"
67 #include "hw/firmware/smbios.h"
68 #include "qapi/visitor.h"
69 #include "qapi/qapi-visit-common.h"
70 #include "standard-headers/linux/input.h"
71 #include "hw/arm/smmuv3.h"
72 #include "hw/acpi/acpi.h"
73 #include "target/arm/internals.h"
74 #include "hw/mem/pc-dimm.h"
75 #include "hw/mem/nvdimm.h"
76 #include "hw/acpi/generic_event_device.h"
77 #include "hw/virtio/virtio-iommu.h"
78 #include "hw/char/pl011.h"
79 #include "qemu/guest-random.h"
80 
81 #define DEFINE_VIRT_MACHINE_LATEST(major, minor, latest) \
82     static void virt_##major##_##minor##_class_init(ObjectClass *oc, \
83                                                     void *data) \
84     { \
85         MachineClass *mc = MACHINE_CLASS(oc); \
86         virt_machine_##major##_##minor##_options(mc); \
87         mc->desc = "QEMU " # major "." # minor " ARM Virtual Machine"; \
88         if (latest) { \
89             mc->alias = "virt"; \
90         } \
91     } \
92     static const TypeInfo machvirt_##major##_##minor##_info = { \
93         .name = MACHINE_TYPE_NAME("virt-" # major "." # minor), \
94         .parent = TYPE_VIRT_MACHINE, \
95         .class_init = virt_##major##_##minor##_class_init, \
96     }; \
97     static void machvirt_machine_##major##_##minor##_init(void) \
98     { \
99         type_register_static(&machvirt_##major##_##minor##_info); \
100     } \
101     type_init(machvirt_machine_##major##_##minor##_init);
102 
103 #define DEFINE_VIRT_MACHINE_AS_LATEST(major, minor) \
104     DEFINE_VIRT_MACHINE_LATEST(major, minor, true)
105 #define DEFINE_VIRT_MACHINE(major, minor) \
106     DEFINE_VIRT_MACHINE_LATEST(major, minor, false)
107 
108 
109 /* Number of external interrupt lines to configure the GIC with */
110 #define NUM_IRQS 256
111 
112 #define PLATFORM_BUS_NUM_IRQS 64
113 
114 /* Legacy RAM limit in GB (< version 4.0) */
115 #define LEGACY_RAMLIMIT_GB 255
116 #define LEGACY_RAMLIMIT_BYTES (LEGACY_RAMLIMIT_GB * GiB)
117 
118 /* Addresses and sizes of our components.
119  * 0..128MB is space for a flash device so we can run bootrom code such as UEFI.
120  * 128MB..256MB is used for miscellaneous device I/O.
121  * 256MB..1GB is reserved for possible future PCI support (ie where the
122  * PCI memory window will go if we add a PCI host controller).
123  * 1GB and up is RAM (which may happily spill over into the
124  * high memory region beyond 4GB).
125  * This represents a compromise between how much RAM can be given to
126  * a 32 bit VM and leaving space for expansion and in particular for PCI.
127  * Note that devices should generally be placed at multiples of 0x10000,
128  * to accommodate guests using 64K pages.
129  */
130 static const MemMapEntry base_memmap[] = {
131     /* Space up to 0x8000000 is reserved for a boot ROM */
132     [VIRT_FLASH] =              {          0, 0x08000000 },
133     [VIRT_CPUPERIPHS] =         { 0x08000000, 0x00020000 },
134     /* GIC distributor and CPU interfaces sit inside the CPU peripheral space */
135     [VIRT_GIC_DIST] =           { 0x08000000, 0x00010000 },
136     [VIRT_GIC_CPU] =            { 0x08010000, 0x00010000 },
137     [VIRT_GIC_V2M] =            { 0x08020000, 0x00001000 },
138     [VIRT_GIC_HYP] =            { 0x08030000, 0x00010000 },
139     [VIRT_GIC_VCPU] =           { 0x08040000, 0x00010000 },
140     /* The space in between here is reserved for GICv3 CPU/vCPU/HYP */
141     [VIRT_GIC_ITS] =            { 0x08080000, 0x00020000 },
142     /* This redistributor space allows up to 2*64kB*123 CPUs */
143     [VIRT_GIC_REDIST] =         { 0x080A0000, 0x00F60000 },
144     [VIRT_UART] =               { 0x09000000, 0x00001000 },
145     [VIRT_RTC] =                { 0x09010000, 0x00001000 },
146     [VIRT_FW_CFG] =             { 0x09020000, 0x00000018 },
147     [VIRT_GPIO] =               { 0x09030000, 0x00001000 },
148     [VIRT_SECURE_UART] =        { 0x09040000, 0x00001000 },
149     [VIRT_SMMU] =               { 0x09050000, 0x00020000 },
150     [VIRT_PCDIMM_ACPI] =        { 0x09070000, MEMORY_HOTPLUG_IO_LEN },
151     [VIRT_ACPI_GED] =           { 0x09080000, ACPI_GED_EVT_SEL_LEN },
152     [VIRT_NVDIMM_ACPI] =        { 0x09090000, NVDIMM_ACPI_IO_LEN},
153     [VIRT_PVTIME] =             { 0x090a0000, 0x00010000 },
154     [VIRT_SECURE_GPIO] =        { 0x090b0000, 0x00001000 },
155     [VIRT_MMIO] =               { 0x0a000000, 0x00000200 },
156     /* ...repeating for a total of NUM_VIRTIO_TRANSPORTS, each of that size */
157     [VIRT_PLATFORM_BUS] =       { 0x0c000000, 0x02000000 },
158     [VIRT_SECURE_MEM] =         { 0x0e000000, 0x01000000 },
159     [VIRT_PCIE_MMIO] =          { 0x10000000, 0x2eff0000 },
160     [VIRT_PCIE_PIO] =           { 0x3eff0000, 0x00010000 },
161     [VIRT_PCIE_ECAM] =          { 0x3f000000, 0x01000000 },
162     /* Actual RAM size depends on initial RAM and device memory settings */
163     [VIRT_MEM] =                { GiB, LEGACY_RAMLIMIT_BYTES },
164 };
165 
166 /*
167  * Highmem IO Regions: This memory map is floating, located after the RAM.
168  * Each MemMapEntry base (GPA) will be dynamically computed, depending on the
169  * top of the RAM, so that its base get the same alignment as the size,
170  * ie. a 512GiB entry will be aligned on a 512GiB boundary. If there is
171  * less than 256GiB of RAM, the floating area starts at the 256GiB mark.
172  * Note the extended_memmap is sized so that it eventually also includes the
173  * base_memmap entries (VIRT_HIGH_GIC_REDIST2 index is greater than the last
174  * index of base_memmap).
175  */
176 static MemMapEntry extended_memmap[] = {
177     /* Additional 64 MB redist region (can contain up to 512 redistributors) */
178     [VIRT_HIGH_GIC_REDIST2] =   { 0x0, 64 * MiB },
179     [VIRT_HIGH_PCIE_ECAM] =     { 0x0, 256 * MiB },
180     /* Second PCIe window */
181     [VIRT_HIGH_PCIE_MMIO] =     { 0x0, 512 * GiB },
182 };
183 
184 static const int a15irqmap[] = {
185     [VIRT_UART] = 1,
186     [VIRT_RTC] = 2,
187     [VIRT_PCIE] = 3, /* ... to 6 */
188     [VIRT_GPIO] = 7,
189     [VIRT_SECURE_UART] = 8,
190     [VIRT_ACPI_GED] = 9,
191     [VIRT_MMIO] = 16, /* ...to 16 + NUM_VIRTIO_TRANSPORTS - 1 */
192     [VIRT_GIC_V2M] = 48, /* ...to 48 + NUM_GICV2M_SPIS - 1 */
193     [VIRT_SMMU] = 74,    /* ...to 74 + NUM_SMMU_IRQS - 1 */
194     [VIRT_PLATFORM_BUS] = 112, /* ...to 112 + PLATFORM_BUS_NUM_IRQS -1 */
195 };
196 
197 static const char *valid_cpus[] = {
198     ARM_CPU_TYPE_NAME("cortex-a7"),
199     ARM_CPU_TYPE_NAME("cortex-a15"),
200     ARM_CPU_TYPE_NAME("cortex-a53"),
201     ARM_CPU_TYPE_NAME("cortex-a57"),
202     ARM_CPU_TYPE_NAME("cortex-a72"),
203     ARM_CPU_TYPE_NAME("a64fx"),
204     ARM_CPU_TYPE_NAME("host"),
205     ARM_CPU_TYPE_NAME("max"),
206 };
207 
208 static bool cpu_type_valid(const char *cpu)
209 {
210     int i;
211 
212     for (i = 0; i < ARRAY_SIZE(valid_cpus); i++) {
213         if (strcmp(cpu, valid_cpus[i]) == 0) {
214             return true;
215         }
216     }
217     return false;
218 }
219 
220 static void create_kaslr_seed(MachineState *ms, const char *node)
221 {
222     uint64_t seed;
223 
224     if (qemu_guest_getrandom(&seed, sizeof(seed), NULL)) {
225         return;
226     }
227     qemu_fdt_setprop_u64(ms->fdt, node, "kaslr-seed", seed);
228 }
229 
230 static void create_fdt(VirtMachineState *vms)
231 {
232     MachineState *ms = MACHINE(vms);
233     int nb_numa_nodes = ms->numa_state->num_nodes;
234     void *fdt = create_device_tree(&vms->fdt_size);
235 
236     if (!fdt) {
237         error_report("create_device_tree() failed");
238         exit(1);
239     }
240 
241     ms->fdt = fdt;
242 
243     /* Header */
244     qemu_fdt_setprop_string(fdt, "/", "compatible", "linux,dummy-virt");
245     qemu_fdt_setprop_cell(fdt, "/", "#address-cells", 0x2);
246     qemu_fdt_setprop_cell(fdt, "/", "#size-cells", 0x2);
247 
248     /* /chosen must exist for load_dtb to fill in necessary properties later */
249     qemu_fdt_add_subnode(fdt, "/chosen");
250     create_kaslr_seed(ms, "/chosen");
251 
252     if (vms->secure) {
253         qemu_fdt_add_subnode(fdt, "/secure-chosen");
254         create_kaslr_seed(ms, "/secure-chosen");
255     }
256 
257     /* Clock node, for the benefit of the UART. The kernel device tree
258      * binding documentation claims the PL011 node clock properties are
259      * optional but in practice if you omit them the kernel refuses to
260      * probe for the device.
261      */
262     vms->clock_phandle = qemu_fdt_alloc_phandle(fdt);
263     qemu_fdt_add_subnode(fdt, "/apb-pclk");
264     qemu_fdt_setprop_string(fdt, "/apb-pclk", "compatible", "fixed-clock");
265     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "#clock-cells", 0x0);
266     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "clock-frequency", 24000000);
267     qemu_fdt_setprop_string(fdt, "/apb-pclk", "clock-output-names",
268                                 "clk24mhz");
269     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "phandle", vms->clock_phandle);
270 
271     if (nb_numa_nodes > 0 && ms->numa_state->have_numa_distance) {
272         int size = nb_numa_nodes * nb_numa_nodes * 3 * sizeof(uint32_t);
273         uint32_t *matrix = g_malloc0(size);
274         int idx, i, j;
275 
276         for (i = 0; i < nb_numa_nodes; i++) {
277             for (j = 0; j < nb_numa_nodes; j++) {
278                 idx = (i * nb_numa_nodes + j) * 3;
279                 matrix[idx + 0] = cpu_to_be32(i);
280                 matrix[idx + 1] = cpu_to_be32(j);
281                 matrix[idx + 2] =
282                     cpu_to_be32(ms->numa_state->nodes[i].distance[j]);
283             }
284         }
285 
286         qemu_fdt_add_subnode(fdt, "/distance-map");
287         qemu_fdt_setprop_string(fdt, "/distance-map", "compatible",
288                                 "numa-distance-map-v1");
289         qemu_fdt_setprop(fdt, "/distance-map", "distance-matrix",
290                          matrix, size);
291         g_free(matrix);
292     }
293 }
294 
295 static void fdt_add_timer_nodes(const VirtMachineState *vms)
296 {
297     /* On real hardware these interrupts are level-triggered.
298      * On KVM they were edge-triggered before host kernel version 4.4,
299      * and level-triggered afterwards.
300      * On emulated QEMU they are level-triggered.
301      *
302      * Getting the DTB info about them wrong is awkward for some
303      * guest kernels:
304      *  pre-4.8 ignore the DT and leave the interrupt configured
305      *   with whatever the GIC reset value (or the bootloader) left it at
306      *  4.8 before rc6 honour the incorrect data by programming it back
307      *   into the GIC, causing problems
308      *  4.8rc6 and later ignore the DT and always write "level triggered"
309      *   into the GIC
310      *
311      * For backwards-compatibility, virt-2.8 and earlier will continue
312      * to say these are edge-triggered, but later machines will report
313      * the correct information.
314      */
315     ARMCPU *armcpu;
316     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
317     uint32_t irqflags = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
318     MachineState *ms = MACHINE(vms);
319 
320     if (vmc->claim_edge_triggered_timers) {
321         irqflags = GIC_FDT_IRQ_FLAGS_EDGE_LO_HI;
322     }
323 
324     if (vms->gic_version == VIRT_GIC_VERSION_2) {
325         irqflags = deposit32(irqflags, GIC_FDT_IRQ_PPI_CPU_START,
326                              GIC_FDT_IRQ_PPI_CPU_WIDTH,
327                              (1 << MACHINE(vms)->smp.cpus) - 1);
328     }
329 
330     qemu_fdt_add_subnode(ms->fdt, "/timer");
331 
332     armcpu = ARM_CPU(qemu_get_cpu(0));
333     if (arm_feature(&armcpu->env, ARM_FEATURE_V8)) {
334         const char compat[] = "arm,armv8-timer\0arm,armv7-timer";
335         qemu_fdt_setprop(ms->fdt, "/timer", "compatible",
336                          compat, sizeof(compat));
337     } else {
338         qemu_fdt_setprop_string(ms->fdt, "/timer", "compatible",
339                                 "arm,armv7-timer");
340     }
341     qemu_fdt_setprop(ms->fdt, "/timer", "always-on", NULL, 0);
342     qemu_fdt_setprop_cells(ms->fdt, "/timer", "interrupts",
343                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_S_EL1_IRQ, irqflags,
344                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_NS_EL1_IRQ, irqflags,
345                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_VIRT_IRQ, irqflags,
346                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_NS_EL2_IRQ, irqflags);
347 }
348 
349 static void fdt_add_cpu_nodes(const VirtMachineState *vms)
350 {
351     int cpu;
352     int addr_cells = 1;
353     const MachineState *ms = MACHINE(vms);
354     const VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
355     int smp_cpus = ms->smp.cpus;
356 
357     /*
358      * See Linux Documentation/devicetree/bindings/arm/cpus.yaml
359      * On ARM v8 64-bit systems value should be set to 2,
360      * that corresponds to the MPIDR_EL1 register size.
361      * If MPIDR_EL1[63:32] value is equal to 0 on all CPUs
362      * in the system, #address-cells can be set to 1, since
363      * MPIDR_EL1[63:32] bits are not used for CPUs
364      * identification.
365      *
366      * Here we actually don't know whether our system is 32- or 64-bit one.
367      * The simplest way to go is to examine affinity IDs of all our CPUs. If
368      * at least one of them has Aff3 populated, we set #address-cells to 2.
369      */
370     for (cpu = 0; cpu < smp_cpus; cpu++) {
371         ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(cpu));
372 
373         if (armcpu->mp_affinity & ARM_AFF3_MASK) {
374             addr_cells = 2;
375             break;
376         }
377     }
378 
379     qemu_fdt_add_subnode(ms->fdt, "/cpus");
380     qemu_fdt_setprop_cell(ms->fdt, "/cpus", "#address-cells", addr_cells);
381     qemu_fdt_setprop_cell(ms->fdt, "/cpus", "#size-cells", 0x0);
382 
383     for (cpu = smp_cpus - 1; cpu >= 0; cpu--) {
384         char *nodename = g_strdup_printf("/cpus/cpu@%d", cpu);
385         ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(cpu));
386         CPUState *cs = CPU(armcpu);
387 
388         qemu_fdt_add_subnode(ms->fdt, nodename);
389         qemu_fdt_setprop_string(ms->fdt, nodename, "device_type", "cpu");
390         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
391                                     armcpu->dtb_compatible);
392 
393         if (vms->psci_conduit != QEMU_PSCI_CONDUIT_DISABLED && smp_cpus > 1) {
394             qemu_fdt_setprop_string(ms->fdt, nodename,
395                                         "enable-method", "psci");
396         }
397 
398         if (addr_cells == 2) {
399             qemu_fdt_setprop_u64(ms->fdt, nodename, "reg",
400                                  armcpu->mp_affinity);
401         } else {
402             qemu_fdt_setprop_cell(ms->fdt, nodename, "reg",
403                                   armcpu->mp_affinity);
404         }
405 
406         if (ms->possible_cpus->cpus[cs->cpu_index].props.has_node_id) {
407             qemu_fdt_setprop_cell(ms->fdt, nodename, "numa-node-id",
408                 ms->possible_cpus->cpus[cs->cpu_index].props.node_id);
409         }
410 
411         if (!vmc->no_cpu_topology) {
412             qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle",
413                                   qemu_fdt_alloc_phandle(ms->fdt));
414         }
415 
416         g_free(nodename);
417     }
418 
419     if (!vmc->no_cpu_topology) {
420         /*
421          * Add vCPU topology description through fdt node cpu-map.
422          *
423          * See Linux Documentation/devicetree/bindings/cpu/cpu-topology.txt
424          * In a SMP system, the hierarchy of CPUs can be defined through
425          * four entities that are used to describe the layout of CPUs in
426          * the system: socket/cluster/core/thread.
427          *
428          * A socket node represents the boundary of system physical package
429          * and its child nodes must be one or more cluster nodes. A system
430          * can contain several layers of clustering within a single physical
431          * package and cluster nodes can be contained in parent cluster nodes.
432          *
433          * Given that cluster is not yet supported in the vCPU topology,
434          * we currently generate one cluster node within each socket node
435          * by default.
436          */
437         qemu_fdt_add_subnode(ms->fdt, "/cpus/cpu-map");
438 
439         for (cpu = smp_cpus - 1; cpu >= 0; cpu--) {
440             char *cpu_path = g_strdup_printf("/cpus/cpu@%d", cpu);
441             char *map_path;
442 
443             if (ms->smp.threads > 1) {
444                 map_path = g_strdup_printf(
445                     "/cpus/cpu-map/socket%d/cluster0/core%d/thread%d",
446                     cpu / (ms->smp.cores * ms->smp.threads),
447                     (cpu / ms->smp.threads) % ms->smp.cores,
448                     cpu % ms->smp.threads);
449             } else {
450                 map_path = g_strdup_printf(
451                     "/cpus/cpu-map/socket%d/cluster0/core%d",
452                     cpu / ms->smp.cores,
453                     cpu % ms->smp.cores);
454             }
455             qemu_fdt_add_path(ms->fdt, map_path);
456             qemu_fdt_setprop_phandle(ms->fdt, map_path, "cpu", cpu_path);
457 
458             g_free(map_path);
459             g_free(cpu_path);
460         }
461     }
462 }
463 
464 static void fdt_add_its_gic_node(VirtMachineState *vms)
465 {
466     char *nodename;
467     MachineState *ms = MACHINE(vms);
468 
469     vms->msi_phandle = qemu_fdt_alloc_phandle(ms->fdt);
470     nodename = g_strdup_printf("/intc/its@%" PRIx64,
471                                vms->memmap[VIRT_GIC_ITS].base);
472     qemu_fdt_add_subnode(ms->fdt, nodename);
473     qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
474                             "arm,gic-v3-its");
475     qemu_fdt_setprop(ms->fdt, nodename, "msi-controller", NULL, 0);
476     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
477                                  2, vms->memmap[VIRT_GIC_ITS].base,
478                                  2, vms->memmap[VIRT_GIC_ITS].size);
479     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", vms->msi_phandle);
480     g_free(nodename);
481 }
482 
483 static void fdt_add_v2m_gic_node(VirtMachineState *vms)
484 {
485     MachineState *ms = MACHINE(vms);
486     char *nodename;
487 
488     nodename = g_strdup_printf("/intc/v2m@%" PRIx64,
489                                vms->memmap[VIRT_GIC_V2M].base);
490     vms->msi_phandle = qemu_fdt_alloc_phandle(ms->fdt);
491     qemu_fdt_add_subnode(ms->fdt, nodename);
492     qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
493                             "arm,gic-v2m-frame");
494     qemu_fdt_setprop(ms->fdt, nodename, "msi-controller", NULL, 0);
495     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
496                                  2, vms->memmap[VIRT_GIC_V2M].base,
497                                  2, vms->memmap[VIRT_GIC_V2M].size);
498     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", vms->msi_phandle);
499     g_free(nodename);
500 }
501 
502 static void fdt_add_gic_node(VirtMachineState *vms)
503 {
504     MachineState *ms = MACHINE(vms);
505     char *nodename;
506 
507     vms->gic_phandle = qemu_fdt_alloc_phandle(ms->fdt);
508     qemu_fdt_setprop_cell(ms->fdt, "/", "interrupt-parent", vms->gic_phandle);
509 
510     nodename = g_strdup_printf("/intc@%" PRIx64,
511                                vms->memmap[VIRT_GIC_DIST].base);
512     qemu_fdt_add_subnode(ms->fdt, nodename);
513     qemu_fdt_setprop_cell(ms->fdt, nodename, "#interrupt-cells", 3);
514     qemu_fdt_setprop(ms->fdt, nodename, "interrupt-controller", NULL, 0);
515     qemu_fdt_setprop_cell(ms->fdt, nodename, "#address-cells", 0x2);
516     qemu_fdt_setprop_cell(ms->fdt, nodename, "#size-cells", 0x2);
517     qemu_fdt_setprop(ms->fdt, nodename, "ranges", NULL, 0);
518     if (vms->gic_version == VIRT_GIC_VERSION_3) {
519         int nb_redist_regions = virt_gicv3_redist_region_count(vms);
520 
521         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
522                                 "arm,gic-v3");
523 
524         qemu_fdt_setprop_cell(ms->fdt, nodename,
525                               "#redistributor-regions", nb_redist_regions);
526 
527         if (nb_redist_regions == 1) {
528             qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
529                                          2, vms->memmap[VIRT_GIC_DIST].base,
530                                          2, vms->memmap[VIRT_GIC_DIST].size,
531                                          2, vms->memmap[VIRT_GIC_REDIST].base,
532                                          2, vms->memmap[VIRT_GIC_REDIST].size);
533         } else {
534             qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
535                                  2, vms->memmap[VIRT_GIC_DIST].base,
536                                  2, vms->memmap[VIRT_GIC_DIST].size,
537                                  2, vms->memmap[VIRT_GIC_REDIST].base,
538                                  2, vms->memmap[VIRT_GIC_REDIST].size,
539                                  2, vms->memmap[VIRT_HIGH_GIC_REDIST2].base,
540                                  2, vms->memmap[VIRT_HIGH_GIC_REDIST2].size);
541         }
542 
543         if (vms->virt) {
544             qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
545                                    GIC_FDT_IRQ_TYPE_PPI, ARCH_GIC_MAINT_IRQ,
546                                    GIC_FDT_IRQ_FLAGS_LEVEL_HI);
547         }
548     } else {
549         /* 'cortex-a15-gic' means 'GIC v2' */
550         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
551                                 "arm,cortex-a15-gic");
552         if (!vms->virt) {
553             qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
554                                          2, vms->memmap[VIRT_GIC_DIST].base,
555                                          2, vms->memmap[VIRT_GIC_DIST].size,
556                                          2, vms->memmap[VIRT_GIC_CPU].base,
557                                          2, vms->memmap[VIRT_GIC_CPU].size);
558         } else {
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                                          2, vms->memmap[VIRT_GIC_HYP].base,
565                                          2, vms->memmap[VIRT_GIC_HYP].size,
566                                          2, vms->memmap[VIRT_GIC_VCPU].base,
567                                          2, vms->memmap[VIRT_GIC_VCPU].size);
568             qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
569                                    GIC_FDT_IRQ_TYPE_PPI, ARCH_GIC_MAINT_IRQ,
570                                    GIC_FDT_IRQ_FLAGS_LEVEL_HI);
571         }
572     }
573 
574     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", vms->gic_phandle);
575     g_free(nodename);
576 }
577 
578 static void fdt_add_pmu_nodes(const VirtMachineState *vms)
579 {
580     ARMCPU *armcpu = ARM_CPU(first_cpu);
581     uint32_t irqflags = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
582     MachineState *ms = MACHINE(vms);
583 
584     if (!arm_feature(&armcpu->env, ARM_FEATURE_PMU)) {
585         assert(!object_property_get_bool(OBJECT(armcpu), "pmu", NULL));
586         return;
587     }
588 
589     if (vms->gic_version == VIRT_GIC_VERSION_2) {
590         irqflags = deposit32(irqflags, GIC_FDT_IRQ_PPI_CPU_START,
591                              GIC_FDT_IRQ_PPI_CPU_WIDTH,
592                              (1 << MACHINE(vms)->smp.cpus) - 1);
593     }
594 
595     qemu_fdt_add_subnode(ms->fdt, "/pmu");
596     if (arm_feature(&armcpu->env, ARM_FEATURE_V8)) {
597         const char compat[] = "arm,armv8-pmuv3";
598         qemu_fdt_setprop(ms->fdt, "/pmu", "compatible",
599                          compat, sizeof(compat));
600         qemu_fdt_setprop_cells(ms->fdt, "/pmu", "interrupts",
601                                GIC_FDT_IRQ_TYPE_PPI, VIRTUAL_PMU_IRQ, irqflags);
602     }
603 }
604 
605 static inline DeviceState *create_acpi_ged(VirtMachineState *vms)
606 {
607     DeviceState *dev;
608     MachineState *ms = MACHINE(vms);
609     int irq = vms->irqmap[VIRT_ACPI_GED];
610     uint32_t event = ACPI_GED_PWR_DOWN_EVT;
611 
612     if (ms->ram_slots) {
613         event |= ACPI_GED_MEM_HOTPLUG_EVT;
614     }
615 
616     if (ms->nvdimms_state->is_enabled) {
617         event |= ACPI_GED_NVDIMM_HOTPLUG_EVT;
618     }
619 
620     dev = qdev_new(TYPE_ACPI_GED);
621     qdev_prop_set_uint32(dev, "ged-event", event);
622 
623     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_ACPI_GED].base);
624     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 1, vms->memmap[VIRT_PCDIMM_ACPI].base);
625     sysbus_connect_irq(SYS_BUS_DEVICE(dev), 0, qdev_get_gpio_in(vms->gic, irq));
626 
627     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
628 
629     return dev;
630 }
631 
632 static void create_its(VirtMachineState *vms)
633 {
634     const char *itsclass = its_class_name();
635     DeviceState *dev;
636 
637     if (!strcmp(itsclass, "arm-gicv3-its")) {
638         if (!vms->tcg_its) {
639             itsclass = NULL;
640         }
641     }
642 
643     if (!itsclass) {
644         /* Do nothing if not supported */
645         return;
646     }
647 
648     dev = qdev_new(itsclass);
649 
650     object_property_set_link(OBJECT(dev), "parent-gicv3", OBJECT(vms->gic),
651                              &error_abort);
652     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
653     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_GIC_ITS].base);
654 
655     fdt_add_its_gic_node(vms);
656     vms->msi_controller = VIRT_MSI_CTRL_ITS;
657 }
658 
659 static void create_v2m(VirtMachineState *vms)
660 {
661     int i;
662     int irq = vms->irqmap[VIRT_GIC_V2M];
663     DeviceState *dev;
664 
665     dev = qdev_new("arm-gicv2m");
666     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_GIC_V2M].base);
667     qdev_prop_set_uint32(dev, "base-spi", irq);
668     qdev_prop_set_uint32(dev, "num-spi", NUM_GICV2M_SPIS);
669     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
670 
671     for (i = 0; i < NUM_GICV2M_SPIS; i++) {
672         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
673                            qdev_get_gpio_in(vms->gic, irq + i));
674     }
675 
676     fdt_add_v2m_gic_node(vms);
677     vms->msi_controller = VIRT_MSI_CTRL_GICV2M;
678 }
679 
680 static void create_gic(VirtMachineState *vms, MemoryRegion *mem)
681 {
682     MachineState *ms = MACHINE(vms);
683     /* We create a standalone GIC */
684     SysBusDevice *gicbusdev;
685     const char *gictype;
686     int type = vms->gic_version, i;
687     unsigned int smp_cpus = ms->smp.cpus;
688     uint32_t nb_redist_regions = 0;
689 
690     gictype = (type == 3) ? gicv3_class_name() : gic_class_name();
691 
692     vms->gic = qdev_new(gictype);
693     qdev_prop_set_uint32(vms->gic, "revision", type);
694     qdev_prop_set_uint32(vms->gic, "num-cpu", smp_cpus);
695     /* Note that the num-irq property counts both internal and external
696      * interrupts; there are always 32 of the former (mandated by GIC spec).
697      */
698     qdev_prop_set_uint32(vms->gic, "num-irq", NUM_IRQS + 32);
699     if (!kvm_irqchip_in_kernel()) {
700         qdev_prop_set_bit(vms->gic, "has-security-extensions", vms->secure);
701     }
702 
703     if (type == 3) {
704         uint32_t redist0_capacity =
705                     vms->memmap[VIRT_GIC_REDIST].size / GICV3_REDIST_SIZE;
706         uint32_t redist0_count = MIN(smp_cpus, redist0_capacity);
707 
708         nb_redist_regions = virt_gicv3_redist_region_count(vms);
709 
710         qdev_prop_set_uint32(vms->gic, "len-redist-region-count",
711                              nb_redist_regions);
712         qdev_prop_set_uint32(vms->gic, "redist-region-count[0]", redist0_count);
713 
714         if (!kvm_irqchip_in_kernel()) {
715             if (vms->tcg_its) {
716                 object_property_set_link(OBJECT(vms->gic), "sysmem",
717                                          OBJECT(mem), &error_fatal);
718                 qdev_prop_set_bit(vms->gic, "has-lpi", true);
719             }
720         }
721 
722         if (nb_redist_regions == 2) {
723             uint32_t redist1_capacity =
724                     vms->memmap[VIRT_HIGH_GIC_REDIST2].size / GICV3_REDIST_SIZE;
725 
726             qdev_prop_set_uint32(vms->gic, "redist-region-count[1]",
727                 MIN(smp_cpus - redist0_count, redist1_capacity));
728         }
729     } else {
730         if (!kvm_irqchip_in_kernel()) {
731             qdev_prop_set_bit(vms->gic, "has-virtualization-extensions",
732                               vms->virt);
733         }
734     }
735     gicbusdev = SYS_BUS_DEVICE(vms->gic);
736     sysbus_realize_and_unref(gicbusdev, &error_fatal);
737     sysbus_mmio_map(gicbusdev, 0, vms->memmap[VIRT_GIC_DIST].base);
738     if (type == 3) {
739         sysbus_mmio_map(gicbusdev, 1, vms->memmap[VIRT_GIC_REDIST].base);
740         if (nb_redist_regions == 2) {
741             sysbus_mmio_map(gicbusdev, 2,
742                             vms->memmap[VIRT_HIGH_GIC_REDIST2].base);
743         }
744     } else {
745         sysbus_mmio_map(gicbusdev, 1, vms->memmap[VIRT_GIC_CPU].base);
746         if (vms->virt) {
747             sysbus_mmio_map(gicbusdev, 2, vms->memmap[VIRT_GIC_HYP].base);
748             sysbus_mmio_map(gicbusdev, 3, vms->memmap[VIRT_GIC_VCPU].base);
749         }
750     }
751 
752     /* Wire the outputs from each CPU's generic timer and the GICv3
753      * maintenance interrupt signal to the appropriate GIC PPI inputs,
754      * and the GIC's IRQ/FIQ/VIRQ/VFIQ interrupt outputs to the CPU's inputs.
755      */
756     for (i = 0; i < smp_cpus; i++) {
757         DeviceState *cpudev = DEVICE(qemu_get_cpu(i));
758         int ppibase = NUM_IRQS + i * GIC_INTERNAL + GIC_NR_SGIS;
759         int irq;
760         /* Mapping from the output timer irq lines from the CPU to the
761          * GIC PPI inputs we use for the virt board.
762          */
763         const int timer_irq[] = {
764             [GTIMER_PHYS] = ARCH_TIMER_NS_EL1_IRQ,
765             [GTIMER_VIRT] = ARCH_TIMER_VIRT_IRQ,
766             [GTIMER_HYP]  = ARCH_TIMER_NS_EL2_IRQ,
767             [GTIMER_SEC]  = ARCH_TIMER_S_EL1_IRQ,
768         };
769 
770         for (irq = 0; irq < ARRAY_SIZE(timer_irq); irq++) {
771             qdev_connect_gpio_out(cpudev, irq,
772                                   qdev_get_gpio_in(vms->gic,
773                                                    ppibase + timer_irq[irq]));
774         }
775 
776         if (type == 3) {
777             qemu_irq irq = qdev_get_gpio_in(vms->gic,
778                                             ppibase + ARCH_GIC_MAINT_IRQ);
779             qdev_connect_gpio_out_named(cpudev, "gicv3-maintenance-interrupt",
780                                         0, irq);
781         } else if (vms->virt) {
782             qemu_irq irq = qdev_get_gpio_in(vms->gic,
783                                             ppibase + ARCH_GIC_MAINT_IRQ);
784             sysbus_connect_irq(gicbusdev, i + 4 * smp_cpus, irq);
785         }
786 
787         qdev_connect_gpio_out_named(cpudev, "pmu-interrupt", 0,
788                                     qdev_get_gpio_in(vms->gic, ppibase
789                                                      + VIRTUAL_PMU_IRQ));
790 
791         sysbus_connect_irq(gicbusdev, i, qdev_get_gpio_in(cpudev, ARM_CPU_IRQ));
792         sysbus_connect_irq(gicbusdev, i + smp_cpus,
793                            qdev_get_gpio_in(cpudev, ARM_CPU_FIQ));
794         sysbus_connect_irq(gicbusdev, i + 2 * smp_cpus,
795                            qdev_get_gpio_in(cpudev, ARM_CPU_VIRQ));
796         sysbus_connect_irq(gicbusdev, i + 3 * smp_cpus,
797                            qdev_get_gpio_in(cpudev, ARM_CPU_VFIQ));
798     }
799 
800     fdt_add_gic_node(vms);
801 
802     if (type == 3 && vms->its) {
803         create_its(vms);
804     } else if (type == 2) {
805         create_v2m(vms);
806     }
807 }
808 
809 static void create_uart(const VirtMachineState *vms, int uart,
810                         MemoryRegion *mem, Chardev *chr)
811 {
812     char *nodename;
813     hwaddr base = vms->memmap[uart].base;
814     hwaddr size = vms->memmap[uart].size;
815     int irq = vms->irqmap[uart];
816     const char compat[] = "arm,pl011\0arm,primecell";
817     const char clocknames[] = "uartclk\0apb_pclk";
818     DeviceState *dev = qdev_new(TYPE_PL011);
819     SysBusDevice *s = SYS_BUS_DEVICE(dev);
820     MachineState *ms = MACHINE(vms);
821 
822     qdev_prop_set_chr(dev, "chardev", chr);
823     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
824     memory_region_add_subregion(mem, base,
825                                 sysbus_mmio_get_region(s, 0));
826     sysbus_connect_irq(s, 0, qdev_get_gpio_in(vms->gic, irq));
827 
828     nodename = g_strdup_printf("/pl011@%" PRIx64, base);
829     qemu_fdt_add_subnode(ms->fdt, nodename);
830     /* Note that we can't use setprop_string because of the embedded NUL */
831     qemu_fdt_setprop(ms->fdt, nodename, "compatible",
832                          compat, sizeof(compat));
833     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
834                                      2, base, 2, size);
835     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
836                                GIC_FDT_IRQ_TYPE_SPI, irq,
837                                GIC_FDT_IRQ_FLAGS_LEVEL_HI);
838     qemu_fdt_setprop_cells(ms->fdt, nodename, "clocks",
839                                vms->clock_phandle, vms->clock_phandle);
840     qemu_fdt_setprop(ms->fdt, nodename, "clock-names",
841                          clocknames, sizeof(clocknames));
842 
843     if (uart == VIRT_UART) {
844         qemu_fdt_setprop_string(ms->fdt, "/chosen", "stdout-path", nodename);
845     } else {
846         /* Mark as not usable by the normal world */
847         qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
848         qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
849 
850         qemu_fdt_setprop_string(ms->fdt, "/secure-chosen", "stdout-path",
851                                 nodename);
852     }
853 
854     g_free(nodename);
855 }
856 
857 static void create_rtc(const VirtMachineState *vms)
858 {
859     char *nodename;
860     hwaddr base = vms->memmap[VIRT_RTC].base;
861     hwaddr size = vms->memmap[VIRT_RTC].size;
862     int irq = vms->irqmap[VIRT_RTC];
863     const char compat[] = "arm,pl031\0arm,primecell";
864     MachineState *ms = MACHINE(vms);
865 
866     sysbus_create_simple("pl031", base, qdev_get_gpio_in(vms->gic, irq));
867 
868     nodename = g_strdup_printf("/pl031@%" PRIx64, base);
869     qemu_fdt_add_subnode(ms->fdt, nodename);
870     qemu_fdt_setprop(ms->fdt, nodename, "compatible", compat, sizeof(compat));
871     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
872                                  2, base, 2, size);
873     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
874                            GIC_FDT_IRQ_TYPE_SPI, irq,
875                            GIC_FDT_IRQ_FLAGS_LEVEL_HI);
876     qemu_fdt_setprop_cell(ms->fdt, nodename, "clocks", vms->clock_phandle);
877     qemu_fdt_setprop_string(ms->fdt, nodename, "clock-names", "apb_pclk");
878     g_free(nodename);
879 }
880 
881 static DeviceState *gpio_key_dev;
882 static void virt_powerdown_req(Notifier *n, void *opaque)
883 {
884     VirtMachineState *s = container_of(n, VirtMachineState, powerdown_notifier);
885 
886     if (s->acpi_dev) {
887         acpi_send_event(s->acpi_dev, ACPI_POWER_DOWN_STATUS);
888     } else {
889         /* use gpio Pin 3 for power button event */
890         qemu_set_irq(qdev_get_gpio_in(gpio_key_dev, 0), 1);
891     }
892 }
893 
894 static void create_gpio_keys(char *fdt, DeviceState *pl061_dev,
895                              uint32_t phandle)
896 {
897     gpio_key_dev = sysbus_create_simple("gpio-key", -1,
898                                         qdev_get_gpio_in(pl061_dev, 3));
899 
900     qemu_fdt_add_subnode(fdt, "/gpio-keys");
901     qemu_fdt_setprop_string(fdt, "/gpio-keys", "compatible", "gpio-keys");
902     qemu_fdt_setprop_cell(fdt, "/gpio-keys", "#size-cells", 0);
903     qemu_fdt_setprop_cell(fdt, "/gpio-keys", "#address-cells", 1);
904 
905     qemu_fdt_add_subnode(fdt, "/gpio-keys/poweroff");
906     qemu_fdt_setprop_string(fdt, "/gpio-keys/poweroff",
907                             "label", "GPIO Key Poweroff");
908     qemu_fdt_setprop_cell(fdt, "/gpio-keys/poweroff", "linux,code",
909                           KEY_POWER);
910     qemu_fdt_setprop_cells(fdt, "/gpio-keys/poweroff",
911                            "gpios", phandle, 3, 0);
912 }
913 
914 #define SECURE_GPIO_POWEROFF 0
915 #define SECURE_GPIO_RESET    1
916 
917 static void create_secure_gpio_pwr(char *fdt, DeviceState *pl061_dev,
918                                    uint32_t phandle)
919 {
920     DeviceState *gpio_pwr_dev;
921 
922     /* gpio-pwr */
923     gpio_pwr_dev = sysbus_create_simple("gpio-pwr", -1, NULL);
924 
925     /* connect secure pl061 to gpio-pwr */
926     qdev_connect_gpio_out(pl061_dev, SECURE_GPIO_RESET,
927                           qdev_get_gpio_in_named(gpio_pwr_dev, "reset", 0));
928     qdev_connect_gpio_out(pl061_dev, SECURE_GPIO_POWEROFF,
929                           qdev_get_gpio_in_named(gpio_pwr_dev, "shutdown", 0));
930 
931     qemu_fdt_add_subnode(fdt, "/gpio-poweroff");
932     qemu_fdt_setprop_string(fdt, "/gpio-poweroff", "compatible",
933                             "gpio-poweroff");
934     qemu_fdt_setprop_cells(fdt, "/gpio-poweroff",
935                            "gpios", phandle, SECURE_GPIO_POWEROFF, 0);
936     qemu_fdt_setprop_string(fdt, "/gpio-poweroff", "status", "disabled");
937     qemu_fdt_setprop_string(fdt, "/gpio-poweroff", "secure-status",
938                             "okay");
939 
940     qemu_fdt_add_subnode(fdt, "/gpio-restart");
941     qemu_fdt_setprop_string(fdt, "/gpio-restart", "compatible",
942                             "gpio-restart");
943     qemu_fdt_setprop_cells(fdt, "/gpio-restart",
944                            "gpios", phandle, SECURE_GPIO_RESET, 0);
945     qemu_fdt_setprop_string(fdt, "/gpio-restart", "status", "disabled");
946     qemu_fdt_setprop_string(fdt, "/gpio-restart", "secure-status",
947                             "okay");
948 }
949 
950 static void create_gpio_devices(const VirtMachineState *vms, int gpio,
951                                 MemoryRegion *mem)
952 {
953     char *nodename;
954     DeviceState *pl061_dev;
955     hwaddr base = vms->memmap[gpio].base;
956     hwaddr size = vms->memmap[gpio].size;
957     int irq = vms->irqmap[gpio];
958     const char compat[] = "arm,pl061\0arm,primecell";
959     SysBusDevice *s;
960     MachineState *ms = MACHINE(vms);
961 
962     pl061_dev = qdev_new("pl061");
963     /* Pull lines down to 0 if not driven by the PL061 */
964     qdev_prop_set_uint32(pl061_dev, "pullups", 0);
965     qdev_prop_set_uint32(pl061_dev, "pulldowns", 0xff);
966     s = SYS_BUS_DEVICE(pl061_dev);
967     sysbus_realize_and_unref(s, &error_fatal);
968     memory_region_add_subregion(mem, base, sysbus_mmio_get_region(s, 0));
969     sysbus_connect_irq(s, 0, qdev_get_gpio_in(vms->gic, irq));
970 
971     uint32_t phandle = qemu_fdt_alloc_phandle(ms->fdt);
972     nodename = g_strdup_printf("/pl061@%" PRIx64, base);
973     qemu_fdt_add_subnode(ms->fdt, nodename);
974     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
975                                  2, base, 2, size);
976     qemu_fdt_setprop(ms->fdt, nodename, "compatible", compat, sizeof(compat));
977     qemu_fdt_setprop_cell(ms->fdt, nodename, "#gpio-cells", 2);
978     qemu_fdt_setprop(ms->fdt, nodename, "gpio-controller", NULL, 0);
979     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
980                            GIC_FDT_IRQ_TYPE_SPI, irq,
981                            GIC_FDT_IRQ_FLAGS_LEVEL_HI);
982     qemu_fdt_setprop_cell(ms->fdt, nodename, "clocks", vms->clock_phandle);
983     qemu_fdt_setprop_string(ms->fdt, nodename, "clock-names", "apb_pclk");
984     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", phandle);
985 
986     if (gpio != VIRT_GPIO) {
987         /* Mark as not usable by the normal world */
988         qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
989         qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
990     }
991     g_free(nodename);
992 
993     /* Child gpio devices */
994     if (gpio == VIRT_GPIO) {
995         create_gpio_keys(ms->fdt, pl061_dev, phandle);
996     } else {
997         create_secure_gpio_pwr(ms->fdt, pl061_dev, phandle);
998     }
999 }
1000 
1001 static void create_virtio_devices(const VirtMachineState *vms)
1002 {
1003     int i;
1004     hwaddr size = vms->memmap[VIRT_MMIO].size;
1005     MachineState *ms = MACHINE(vms);
1006 
1007     /* We create the transports in forwards order. Since qbus_realize()
1008      * prepends (not appends) new child buses, the incrementing loop below will
1009      * create a list of virtio-mmio buses with decreasing base addresses.
1010      *
1011      * When a -device option is processed from the command line,
1012      * qbus_find_recursive() picks the next free virtio-mmio bus in forwards
1013      * order. The upshot is that -device options in increasing command line
1014      * order are mapped to virtio-mmio buses with decreasing base addresses.
1015      *
1016      * When this code was originally written, that arrangement ensured that the
1017      * guest Linux kernel would give the lowest "name" (/dev/vda, eth0, etc) to
1018      * the first -device on the command line. (The end-to-end order is a
1019      * function of this loop, qbus_realize(), qbus_find_recursive(), and the
1020      * guest kernel's name-to-address assignment strategy.)
1021      *
1022      * Meanwhile, the kernel's traversal seems to have been reversed; see eg.
1023      * the message, if not necessarily the code, of commit 70161ff336.
1024      * Therefore the loop now establishes the inverse of the original intent.
1025      *
1026      * Unfortunately, we can't counteract the kernel change by reversing the
1027      * loop; it would break existing command lines.
1028      *
1029      * In any case, the kernel makes no guarantee about the stability of
1030      * enumeration order of virtio devices (as demonstrated by it changing
1031      * between kernel versions). For reliable and stable identification
1032      * of disks users must use UUIDs or similar mechanisms.
1033      */
1034     for (i = 0; i < NUM_VIRTIO_TRANSPORTS; i++) {
1035         int irq = vms->irqmap[VIRT_MMIO] + i;
1036         hwaddr base = vms->memmap[VIRT_MMIO].base + i * size;
1037 
1038         sysbus_create_simple("virtio-mmio", base,
1039                              qdev_get_gpio_in(vms->gic, irq));
1040     }
1041 
1042     /* We add dtb nodes in reverse order so that they appear in the finished
1043      * device tree lowest address first.
1044      *
1045      * Note that this mapping is independent of the loop above. The previous
1046      * loop influences virtio device to virtio transport assignment, whereas
1047      * this loop controls how virtio transports are laid out in the dtb.
1048      */
1049     for (i = NUM_VIRTIO_TRANSPORTS - 1; i >= 0; i--) {
1050         char *nodename;
1051         int irq = vms->irqmap[VIRT_MMIO] + i;
1052         hwaddr base = vms->memmap[VIRT_MMIO].base + i * size;
1053 
1054         nodename = g_strdup_printf("/virtio_mmio@%" PRIx64, base);
1055         qemu_fdt_add_subnode(ms->fdt, nodename);
1056         qemu_fdt_setprop_string(ms->fdt, nodename,
1057                                 "compatible", "virtio,mmio");
1058         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1059                                      2, base, 2, size);
1060         qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
1061                                GIC_FDT_IRQ_TYPE_SPI, irq,
1062                                GIC_FDT_IRQ_FLAGS_EDGE_LO_HI);
1063         qemu_fdt_setprop(ms->fdt, nodename, "dma-coherent", NULL, 0);
1064         g_free(nodename);
1065     }
1066 }
1067 
1068 #define VIRT_FLASH_SECTOR_SIZE (256 * KiB)
1069 
1070 static PFlashCFI01 *virt_flash_create1(VirtMachineState *vms,
1071                                         const char *name,
1072                                         const char *alias_prop_name)
1073 {
1074     /*
1075      * Create a single flash device.  We use the same parameters as
1076      * the flash devices on the Versatile Express board.
1077      */
1078     DeviceState *dev = qdev_new(TYPE_PFLASH_CFI01);
1079 
1080     qdev_prop_set_uint64(dev, "sector-length", VIRT_FLASH_SECTOR_SIZE);
1081     qdev_prop_set_uint8(dev, "width", 4);
1082     qdev_prop_set_uint8(dev, "device-width", 2);
1083     qdev_prop_set_bit(dev, "big-endian", false);
1084     qdev_prop_set_uint16(dev, "id0", 0x89);
1085     qdev_prop_set_uint16(dev, "id1", 0x18);
1086     qdev_prop_set_uint16(dev, "id2", 0x00);
1087     qdev_prop_set_uint16(dev, "id3", 0x00);
1088     qdev_prop_set_string(dev, "name", name);
1089     object_property_add_child(OBJECT(vms), name, OBJECT(dev));
1090     object_property_add_alias(OBJECT(vms), alias_prop_name,
1091                               OBJECT(dev), "drive");
1092     return PFLASH_CFI01(dev);
1093 }
1094 
1095 static void virt_flash_create(VirtMachineState *vms)
1096 {
1097     vms->flash[0] = virt_flash_create1(vms, "virt.flash0", "pflash0");
1098     vms->flash[1] = virt_flash_create1(vms, "virt.flash1", "pflash1");
1099 }
1100 
1101 static void virt_flash_map1(PFlashCFI01 *flash,
1102                             hwaddr base, hwaddr size,
1103                             MemoryRegion *sysmem)
1104 {
1105     DeviceState *dev = DEVICE(flash);
1106 
1107     assert(QEMU_IS_ALIGNED(size, VIRT_FLASH_SECTOR_SIZE));
1108     assert(size / VIRT_FLASH_SECTOR_SIZE <= UINT32_MAX);
1109     qdev_prop_set_uint32(dev, "num-blocks", size / VIRT_FLASH_SECTOR_SIZE);
1110     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1111 
1112     memory_region_add_subregion(sysmem, base,
1113                                 sysbus_mmio_get_region(SYS_BUS_DEVICE(dev),
1114                                                        0));
1115 }
1116 
1117 static void virt_flash_map(VirtMachineState *vms,
1118                            MemoryRegion *sysmem,
1119                            MemoryRegion *secure_sysmem)
1120 {
1121     /*
1122      * Map two flash devices to fill the VIRT_FLASH space in the memmap.
1123      * sysmem is the system memory space. secure_sysmem is the secure view
1124      * of the system, and the first flash device should be made visible only
1125      * there. The second flash device is visible to both secure and nonsecure.
1126      * If sysmem == secure_sysmem this means there is no separate Secure
1127      * address space and both flash devices are generally visible.
1128      */
1129     hwaddr flashsize = vms->memmap[VIRT_FLASH].size / 2;
1130     hwaddr flashbase = vms->memmap[VIRT_FLASH].base;
1131 
1132     virt_flash_map1(vms->flash[0], flashbase, flashsize,
1133                     secure_sysmem);
1134     virt_flash_map1(vms->flash[1], flashbase + flashsize, flashsize,
1135                     sysmem);
1136 }
1137 
1138 static void virt_flash_fdt(VirtMachineState *vms,
1139                            MemoryRegion *sysmem,
1140                            MemoryRegion *secure_sysmem)
1141 {
1142     hwaddr flashsize = vms->memmap[VIRT_FLASH].size / 2;
1143     hwaddr flashbase = vms->memmap[VIRT_FLASH].base;
1144     MachineState *ms = MACHINE(vms);
1145     char *nodename;
1146 
1147     if (sysmem == secure_sysmem) {
1148         /* Report both flash devices as a single node in the DT */
1149         nodename = g_strdup_printf("/flash@%" PRIx64, flashbase);
1150         qemu_fdt_add_subnode(ms->fdt, nodename);
1151         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible", "cfi-flash");
1152         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1153                                      2, flashbase, 2, flashsize,
1154                                      2, flashbase + flashsize, 2, flashsize);
1155         qemu_fdt_setprop_cell(ms->fdt, nodename, "bank-width", 4);
1156         g_free(nodename);
1157     } else {
1158         /*
1159          * Report the devices as separate nodes so we can mark one as
1160          * only visible to the secure world.
1161          */
1162         nodename = g_strdup_printf("/secflash@%" PRIx64, flashbase);
1163         qemu_fdt_add_subnode(ms->fdt, nodename);
1164         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible", "cfi-flash");
1165         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1166                                      2, flashbase, 2, flashsize);
1167         qemu_fdt_setprop_cell(ms->fdt, nodename, "bank-width", 4);
1168         qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
1169         qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
1170         g_free(nodename);
1171 
1172         nodename = g_strdup_printf("/flash@%" PRIx64, flashbase);
1173         qemu_fdt_add_subnode(ms->fdt, nodename);
1174         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible", "cfi-flash");
1175         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1176                                      2, flashbase + flashsize, 2, flashsize);
1177         qemu_fdt_setprop_cell(ms->fdt, nodename, "bank-width", 4);
1178         g_free(nodename);
1179     }
1180 }
1181 
1182 static bool virt_firmware_init(VirtMachineState *vms,
1183                                MemoryRegion *sysmem,
1184                                MemoryRegion *secure_sysmem)
1185 {
1186     int i;
1187     const char *bios_name;
1188     BlockBackend *pflash_blk0;
1189 
1190     /* Map legacy -drive if=pflash to machine properties */
1191     for (i = 0; i < ARRAY_SIZE(vms->flash); i++) {
1192         pflash_cfi01_legacy_drive(vms->flash[i],
1193                                   drive_get(IF_PFLASH, 0, i));
1194     }
1195 
1196     virt_flash_map(vms, sysmem, secure_sysmem);
1197 
1198     pflash_blk0 = pflash_cfi01_get_blk(vms->flash[0]);
1199 
1200     bios_name = MACHINE(vms)->firmware;
1201     if (bios_name) {
1202         char *fname;
1203         MemoryRegion *mr;
1204         int image_size;
1205 
1206         if (pflash_blk0) {
1207             error_report("The contents of the first flash device may be "
1208                          "specified with -bios or with -drive if=pflash... "
1209                          "but you cannot use both options at once");
1210             exit(1);
1211         }
1212 
1213         /* Fall back to -bios */
1214 
1215         fname = qemu_find_file(QEMU_FILE_TYPE_BIOS, bios_name);
1216         if (!fname) {
1217             error_report("Could not find ROM image '%s'", bios_name);
1218             exit(1);
1219         }
1220         mr = sysbus_mmio_get_region(SYS_BUS_DEVICE(vms->flash[0]), 0);
1221         image_size = load_image_mr(fname, mr);
1222         g_free(fname);
1223         if (image_size < 0) {
1224             error_report("Could not load ROM image '%s'", bios_name);
1225             exit(1);
1226         }
1227     }
1228 
1229     return pflash_blk0 || bios_name;
1230 }
1231 
1232 static FWCfgState *create_fw_cfg(const VirtMachineState *vms, AddressSpace *as)
1233 {
1234     MachineState *ms = MACHINE(vms);
1235     hwaddr base = vms->memmap[VIRT_FW_CFG].base;
1236     hwaddr size = vms->memmap[VIRT_FW_CFG].size;
1237     FWCfgState *fw_cfg;
1238     char *nodename;
1239 
1240     fw_cfg = fw_cfg_init_mem_wide(base + 8, base, 8, base + 16, as);
1241     fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, (uint16_t)ms->smp.cpus);
1242 
1243     nodename = g_strdup_printf("/fw-cfg@%" PRIx64, base);
1244     qemu_fdt_add_subnode(ms->fdt, nodename);
1245     qemu_fdt_setprop_string(ms->fdt, nodename,
1246                             "compatible", "qemu,fw-cfg-mmio");
1247     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1248                                  2, base, 2, size);
1249     qemu_fdt_setprop(ms->fdt, nodename, "dma-coherent", NULL, 0);
1250     g_free(nodename);
1251     return fw_cfg;
1252 }
1253 
1254 static void create_pcie_irq_map(const MachineState *ms,
1255                                 uint32_t gic_phandle,
1256                                 int first_irq, const char *nodename)
1257 {
1258     int devfn, pin;
1259     uint32_t full_irq_map[4 * 4 * 10] = { 0 };
1260     uint32_t *irq_map = full_irq_map;
1261 
1262     for (devfn = 0; devfn <= 0x18; devfn += 0x8) {
1263         for (pin = 0; pin < 4; pin++) {
1264             int irq_type = GIC_FDT_IRQ_TYPE_SPI;
1265             int irq_nr = first_irq + ((pin + PCI_SLOT(devfn)) % PCI_NUM_PINS);
1266             int irq_level = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
1267             int i;
1268 
1269             uint32_t map[] = {
1270                 devfn << 8, 0, 0,                           /* devfn */
1271                 pin + 1,                                    /* PCI pin */
1272                 gic_phandle, 0, 0, irq_type, irq_nr, irq_level }; /* GIC irq */
1273 
1274             /* Convert map to big endian */
1275             for (i = 0; i < 10; i++) {
1276                 irq_map[i] = cpu_to_be32(map[i]);
1277             }
1278             irq_map += 10;
1279         }
1280     }
1281 
1282     qemu_fdt_setprop(ms->fdt, nodename, "interrupt-map",
1283                      full_irq_map, sizeof(full_irq_map));
1284 
1285     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupt-map-mask",
1286                            cpu_to_be16(PCI_DEVFN(3, 0)), /* Slot 3 */
1287                            0, 0,
1288                            0x7           /* PCI irq */);
1289 }
1290 
1291 static void create_smmu(const VirtMachineState *vms,
1292                         PCIBus *bus)
1293 {
1294     char *node;
1295     const char compat[] = "arm,smmu-v3";
1296     int irq =  vms->irqmap[VIRT_SMMU];
1297     int i;
1298     hwaddr base = vms->memmap[VIRT_SMMU].base;
1299     hwaddr size = vms->memmap[VIRT_SMMU].size;
1300     const char irq_names[] = "eventq\0priq\0cmdq-sync\0gerror";
1301     DeviceState *dev;
1302     MachineState *ms = MACHINE(vms);
1303 
1304     if (vms->iommu != VIRT_IOMMU_SMMUV3 || !vms->iommu_phandle) {
1305         return;
1306     }
1307 
1308     dev = qdev_new("arm-smmuv3");
1309 
1310     object_property_set_link(OBJECT(dev), "primary-bus", OBJECT(bus),
1311                              &error_abort);
1312     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1313     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, base);
1314     for (i = 0; i < NUM_SMMU_IRQS; i++) {
1315         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
1316                            qdev_get_gpio_in(vms->gic, irq + i));
1317     }
1318 
1319     node = g_strdup_printf("/smmuv3@%" PRIx64, base);
1320     qemu_fdt_add_subnode(ms->fdt, node);
1321     qemu_fdt_setprop(ms->fdt, node, "compatible", compat, sizeof(compat));
1322     qemu_fdt_setprop_sized_cells(ms->fdt, node, "reg", 2, base, 2, size);
1323 
1324     qemu_fdt_setprop_cells(ms->fdt, node, "interrupts",
1325             GIC_FDT_IRQ_TYPE_SPI, irq    , GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1326             GIC_FDT_IRQ_TYPE_SPI, irq + 1, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1327             GIC_FDT_IRQ_TYPE_SPI, irq + 2, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1328             GIC_FDT_IRQ_TYPE_SPI, irq + 3, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI);
1329 
1330     qemu_fdt_setprop(ms->fdt, node, "interrupt-names", irq_names,
1331                      sizeof(irq_names));
1332 
1333     qemu_fdt_setprop_cell(ms->fdt, node, "clocks", vms->clock_phandle);
1334     qemu_fdt_setprop_string(ms->fdt, node, "clock-names", "apb_pclk");
1335     qemu_fdt_setprop(ms->fdt, node, "dma-coherent", NULL, 0);
1336 
1337     qemu_fdt_setprop_cell(ms->fdt, node, "#iommu-cells", 1);
1338 
1339     qemu_fdt_setprop_cell(ms->fdt, node, "phandle", vms->iommu_phandle);
1340     g_free(node);
1341 }
1342 
1343 static void create_virtio_iommu_dt_bindings(VirtMachineState *vms)
1344 {
1345     const char compat[] = "virtio,pci-iommu";
1346     uint16_t bdf = vms->virtio_iommu_bdf;
1347     MachineState *ms = MACHINE(vms);
1348     char *node;
1349 
1350     vms->iommu_phandle = qemu_fdt_alloc_phandle(ms->fdt);
1351 
1352     node = g_strdup_printf("%s/virtio_iommu@%d", vms->pciehb_nodename, bdf);
1353     qemu_fdt_add_subnode(ms->fdt, node);
1354     qemu_fdt_setprop(ms->fdt, node, "compatible", compat, sizeof(compat));
1355     qemu_fdt_setprop_sized_cells(ms->fdt, node, "reg",
1356                                  1, bdf << 8, 1, 0, 1, 0,
1357                                  1, 0, 1, 0);
1358 
1359     qemu_fdt_setprop_cell(ms->fdt, node, "#iommu-cells", 1);
1360     qemu_fdt_setprop_cell(ms->fdt, node, "phandle", vms->iommu_phandle);
1361     g_free(node);
1362 
1363     qemu_fdt_setprop_cells(ms->fdt, vms->pciehb_nodename, "iommu-map",
1364                            0x0, vms->iommu_phandle, 0x0, bdf,
1365                            bdf + 1, vms->iommu_phandle, bdf + 1, 0xffff - bdf);
1366 }
1367 
1368 static void create_pcie(VirtMachineState *vms)
1369 {
1370     hwaddr base_mmio = vms->memmap[VIRT_PCIE_MMIO].base;
1371     hwaddr size_mmio = vms->memmap[VIRT_PCIE_MMIO].size;
1372     hwaddr base_mmio_high = vms->memmap[VIRT_HIGH_PCIE_MMIO].base;
1373     hwaddr size_mmio_high = vms->memmap[VIRT_HIGH_PCIE_MMIO].size;
1374     hwaddr base_pio = vms->memmap[VIRT_PCIE_PIO].base;
1375     hwaddr size_pio = vms->memmap[VIRT_PCIE_PIO].size;
1376     hwaddr base_ecam, size_ecam;
1377     hwaddr base = base_mmio;
1378     int nr_pcie_buses;
1379     int irq = vms->irqmap[VIRT_PCIE];
1380     MemoryRegion *mmio_alias;
1381     MemoryRegion *mmio_reg;
1382     MemoryRegion *ecam_alias;
1383     MemoryRegion *ecam_reg;
1384     DeviceState *dev;
1385     char *nodename;
1386     int i, ecam_id;
1387     PCIHostState *pci;
1388     MachineState *ms = MACHINE(vms);
1389 
1390     dev = qdev_new(TYPE_GPEX_HOST);
1391     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1392 
1393     ecam_id = VIRT_ECAM_ID(vms->highmem_ecam);
1394     base_ecam = vms->memmap[ecam_id].base;
1395     size_ecam = vms->memmap[ecam_id].size;
1396     nr_pcie_buses = size_ecam / PCIE_MMCFG_SIZE_MIN;
1397     /* Map only the first size_ecam bytes of ECAM space */
1398     ecam_alias = g_new0(MemoryRegion, 1);
1399     ecam_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 0);
1400     memory_region_init_alias(ecam_alias, OBJECT(dev), "pcie-ecam",
1401                              ecam_reg, 0, size_ecam);
1402     memory_region_add_subregion(get_system_memory(), base_ecam, ecam_alias);
1403 
1404     /* Map the MMIO window into system address space so as to expose
1405      * the section of PCI MMIO space which starts at the same base address
1406      * (ie 1:1 mapping for that part of PCI MMIO space visible through
1407      * the window).
1408      */
1409     mmio_alias = g_new0(MemoryRegion, 1);
1410     mmio_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 1);
1411     memory_region_init_alias(mmio_alias, OBJECT(dev), "pcie-mmio",
1412                              mmio_reg, base_mmio, size_mmio);
1413     memory_region_add_subregion(get_system_memory(), base_mmio, mmio_alias);
1414 
1415     if (vms->highmem) {
1416         /* Map high MMIO space */
1417         MemoryRegion *high_mmio_alias = g_new0(MemoryRegion, 1);
1418 
1419         memory_region_init_alias(high_mmio_alias, OBJECT(dev), "pcie-mmio-high",
1420                                  mmio_reg, base_mmio_high, size_mmio_high);
1421         memory_region_add_subregion(get_system_memory(), base_mmio_high,
1422                                     high_mmio_alias);
1423     }
1424 
1425     /* Map IO port space */
1426     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 2, base_pio);
1427 
1428     for (i = 0; i < GPEX_NUM_IRQS; i++) {
1429         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
1430                            qdev_get_gpio_in(vms->gic, irq + i));
1431         gpex_set_irq_num(GPEX_HOST(dev), i, irq + i);
1432     }
1433 
1434     pci = PCI_HOST_BRIDGE(dev);
1435     pci->bypass_iommu = vms->default_bus_bypass_iommu;
1436     vms->bus = pci->bus;
1437     if (vms->bus) {
1438         for (i = 0; i < nb_nics; i++) {
1439             NICInfo *nd = &nd_table[i];
1440 
1441             if (!nd->model) {
1442                 nd->model = g_strdup("virtio");
1443             }
1444 
1445             pci_nic_init_nofail(nd, pci->bus, nd->model, NULL);
1446         }
1447     }
1448 
1449     nodename = vms->pciehb_nodename = g_strdup_printf("/pcie@%" PRIx64, base);
1450     qemu_fdt_add_subnode(ms->fdt, nodename);
1451     qemu_fdt_setprop_string(ms->fdt, nodename,
1452                             "compatible", "pci-host-ecam-generic");
1453     qemu_fdt_setprop_string(ms->fdt, nodename, "device_type", "pci");
1454     qemu_fdt_setprop_cell(ms->fdt, nodename, "#address-cells", 3);
1455     qemu_fdt_setprop_cell(ms->fdt, nodename, "#size-cells", 2);
1456     qemu_fdt_setprop_cell(ms->fdt, nodename, "linux,pci-domain", 0);
1457     qemu_fdt_setprop_cells(ms->fdt, nodename, "bus-range", 0,
1458                            nr_pcie_buses - 1);
1459     qemu_fdt_setprop(ms->fdt, nodename, "dma-coherent", NULL, 0);
1460 
1461     if (vms->msi_phandle) {
1462         qemu_fdt_setprop_cells(ms->fdt, nodename, "msi-parent",
1463                                vms->msi_phandle);
1464     }
1465 
1466     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1467                                  2, base_ecam, 2, size_ecam);
1468 
1469     if (vms->highmem) {
1470         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "ranges",
1471                                      1, FDT_PCI_RANGE_IOPORT, 2, 0,
1472                                      2, base_pio, 2, size_pio,
1473                                      1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
1474                                      2, base_mmio, 2, size_mmio,
1475                                      1, FDT_PCI_RANGE_MMIO_64BIT,
1476                                      2, base_mmio_high,
1477                                      2, base_mmio_high, 2, size_mmio_high);
1478     } else {
1479         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "ranges",
1480                                      1, FDT_PCI_RANGE_IOPORT, 2, 0,
1481                                      2, base_pio, 2, size_pio,
1482                                      1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
1483                                      2, base_mmio, 2, size_mmio);
1484     }
1485 
1486     qemu_fdt_setprop_cell(ms->fdt, nodename, "#interrupt-cells", 1);
1487     create_pcie_irq_map(ms, vms->gic_phandle, irq, nodename);
1488 
1489     if (vms->iommu) {
1490         vms->iommu_phandle = qemu_fdt_alloc_phandle(ms->fdt);
1491 
1492         switch (vms->iommu) {
1493         case VIRT_IOMMU_SMMUV3:
1494             create_smmu(vms, vms->bus);
1495             qemu_fdt_setprop_cells(ms->fdt, nodename, "iommu-map",
1496                                    0x0, vms->iommu_phandle, 0x0, 0x10000);
1497             break;
1498         default:
1499             g_assert_not_reached();
1500         }
1501     }
1502 }
1503 
1504 static void create_platform_bus(VirtMachineState *vms)
1505 {
1506     DeviceState *dev;
1507     SysBusDevice *s;
1508     int i;
1509     MemoryRegion *sysmem = get_system_memory();
1510 
1511     dev = qdev_new(TYPE_PLATFORM_BUS_DEVICE);
1512     dev->id = g_strdup(TYPE_PLATFORM_BUS_DEVICE);
1513     qdev_prop_set_uint32(dev, "num_irqs", PLATFORM_BUS_NUM_IRQS);
1514     qdev_prop_set_uint32(dev, "mmio_size", vms->memmap[VIRT_PLATFORM_BUS].size);
1515     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1516     vms->platform_bus_dev = dev;
1517 
1518     s = SYS_BUS_DEVICE(dev);
1519     for (i = 0; i < PLATFORM_BUS_NUM_IRQS; i++) {
1520         int irq = vms->irqmap[VIRT_PLATFORM_BUS] + i;
1521         sysbus_connect_irq(s, i, qdev_get_gpio_in(vms->gic, irq));
1522     }
1523 
1524     memory_region_add_subregion(sysmem,
1525                                 vms->memmap[VIRT_PLATFORM_BUS].base,
1526                                 sysbus_mmio_get_region(s, 0));
1527 }
1528 
1529 static void create_tag_ram(MemoryRegion *tag_sysmem,
1530                            hwaddr base, hwaddr size,
1531                            const char *name)
1532 {
1533     MemoryRegion *tagram = g_new(MemoryRegion, 1);
1534 
1535     memory_region_init_ram(tagram, NULL, name, size / 32, &error_fatal);
1536     memory_region_add_subregion(tag_sysmem, base / 32, tagram);
1537 }
1538 
1539 static void create_secure_ram(VirtMachineState *vms,
1540                               MemoryRegion *secure_sysmem,
1541                               MemoryRegion *secure_tag_sysmem)
1542 {
1543     MemoryRegion *secram = g_new(MemoryRegion, 1);
1544     char *nodename;
1545     hwaddr base = vms->memmap[VIRT_SECURE_MEM].base;
1546     hwaddr size = vms->memmap[VIRT_SECURE_MEM].size;
1547     MachineState *ms = MACHINE(vms);
1548 
1549     memory_region_init_ram(secram, NULL, "virt.secure-ram", size,
1550                            &error_fatal);
1551     memory_region_add_subregion(secure_sysmem, base, secram);
1552 
1553     nodename = g_strdup_printf("/secram@%" PRIx64, base);
1554     qemu_fdt_add_subnode(ms->fdt, nodename);
1555     qemu_fdt_setprop_string(ms->fdt, nodename, "device_type", "memory");
1556     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg", 2, base, 2, size);
1557     qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
1558     qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
1559 
1560     if (secure_tag_sysmem) {
1561         create_tag_ram(secure_tag_sysmem, base, size, "mach-virt.secure-tag");
1562     }
1563 
1564     g_free(nodename);
1565 }
1566 
1567 static void *machvirt_dtb(const struct arm_boot_info *binfo, int *fdt_size)
1568 {
1569     const VirtMachineState *board = container_of(binfo, VirtMachineState,
1570                                                  bootinfo);
1571     MachineState *ms = MACHINE(board);
1572 
1573 
1574     *fdt_size = board->fdt_size;
1575     return ms->fdt;
1576 }
1577 
1578 static void virt_build_smbios(VirtMachineState *vms)
1579 {
1580     MachineClass *mc = MACHINE_GET_CLASS(vms);
1581     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
1582     uint8_t *smbios_tables, *smbios_anchor;
1583     size_t smbios_tables_len, smbios_anchor_len;
1584     const char *product = "QEMU Virtual Machine";
1585 
1586     if (kvm_enabled()) {
1587         product = "KVM Virtual Machine";
1588     }
1589 
1590     smbios_set_defaults("QEMU", product,
1591                         vmc->smbios_old_sys_ver ? "1.0" : mc->name, false,
1592                         true, SMBIOS_ENTRY_POINT_30);
1593 
1594     smbios_get_tables(MACHINE(vms), NULL, 0,
1595                       &smbios_tables, &smbios_tables_len,
1596                       &smbios_anchor, &smbios_anchor_len,
1597                       &error_fatal);
1598 
1599     if (smbios_anchor) {
1600         fw_cfg_add_file(vms->fw_cfg, "etc/smbios/smbios-tables",
1601                         smbios_tables, smbios_tables_len);
1602         fw_cfg_add_file(vms->fw_cfg, "etc/smbios/smbios-anchor",
1603                         smbios_anchor, smbios_anchor_len);
1604     }
1605 }
1606 
1607 static
1608 void virt_machine_done(Notifier *notifier, void *data)
1609 {
1610     VirtMachineState *vms = container_of(notifier, VirtMachineState,
1611                                          machine_done);
1612     MachineState *ms = MACHINE(vms);
1613     ARMCPU *cpu = ARM_CPU(first_cpu);
1614     struct arm_boot_info *info = &vms->bootinfo;
1615     AddressSpace *as = arm_boot_address_space(cpu, info);
1616 
1617     /*
1618      * If the user provided a dtb, we assume the dynamic sysbus nodes
1619      * already are integrated there. This corresponds to a use case where
1620      * the dynamic sysbus nodes are complex and their generation is not yet
1621      * supported. In that case the user can take charge of the guest dt
1622      * while qemu takes charge of the qom stuff.
1623      */
1624     if (info->dtb_filename == NULL) {
1625         platform_bus_add_all_fdt_nodes(ms->fdt, "/intc",
1626                                        vms->memmap[VIRT_PLATFORM_BUS].base,
1627                                        vms->memmap[VIRT_PLATFORM_BUS].size,
1628                                        vms->irqmap[VIRT_PLATFORM_BUS]);
1629     }
1630     if (arm_load_dtb(info->dtb_start, info, info->dtb_limit, as, ms) < 0) {
1631         exit(1);
1632     }
1633 
1634     fw_cfg_add_extra_pci_roots(vms->bus, vms->fw_cfg);
1635 
1636     virt_acpi_setup(vms);
1637     virt_build_smbios(vms);
1638 }
1639 
1640 static uint64_t virt_cpu_mp_affinity(VirtMachineState *vms, int idx)
1641 {
1642     uint8_t clustersz = ARM_DEFAULT_CPUS_PER_CLUSTER;
1643     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
1644 
1645     if (!vmc->disallow_affinity_adjustment) {
1646         /* Adjust MPIDR like 64-bit KVM hosts, which incorporate the
1647          * GIC's target-list limitations. 32-bit KVM hosts currently
1648          * always create clusters of 4 CPUs, but that is expected to
1649          * change when they gain support for gicv3. When KVM is enabled
1650          * it will override the changes we make here, therefore our
1651          * purposes are to make TCG consistent (with 64-bit KVM hosts)
1652          * and to improve SGI efficiency.
1653          */
1654         if (vms->gic_version == VIRT_GIC_VERSION_3) {
1655             clustersz = GICV3_TARGETLIST_BITS;
1656         } else {
1657             clustersz = GIC_TARGETLIST_BITS;
1658         }
1659     }
1660     return arm_cpu_mp_affinity(idx, clustersz);
1661 }
1662 
1663 static void virt_set_memmap(VirtMachineState *vms)
1664 {
1665     MachineState *ms = MACHINE(vms);
1666     hwaddr base, device_memory_base, device_memory_size;
1667     int i;
1668 
1669     vms->memmap = extended_memmap;
1670 
1671     for (i = 0; i < ARRAY_SIZE(base_memmap); i++) {
1672         vms->memmap[i] = base_memmap[i];
1673     }
1674 
1675     if (ms->ram_slots > ACPI_MAX_RAM_SLOTS) {
1676         error_report("unsupported number of memory slots: %"PRIu64,
1677                      ms->ram_slots);
1678         exit(EXIT_FAILURE);
1679     }
1680 
1681     /*
1682      * We compute the base of the high IO region depending on the
1683      * amount of initial and device memory. The device memory start/size
1684      * is aligned on 1GiB. We never put the high IO region below 256GiB
1685      * so that if maxram_size is < 255GiB we keep the legacy memory map.
1686      * The device region size assumes 1GiB page max alignment per slot.
1687      */
1688     device_memory_base =
1689         ROUND_UP(vms->memmap[VIRT_MEM].base + ms->ram_size, GiB);
1690     device_memory_size = ms->maxram_size - ms->ram_size + ms->ram_slots * GiB;
1691 
1692     /* Base address of the high IO region */
1693     base = device_memory_base + ROUND_UP(device_memory_size, GiB);
1694     if (base < device_memory_base) {
1695         error_report("maxmem/slots too huge");
1696         exit(EXIT_FAILURE);
1697     }
1698     if (base < vms->memmap[VIRT_MEM].base + LEGACY_RAMLIMIT_BYTES) {
1699         base = vms->memmap[VIRT_MEM].base + LEGACY_RAMLIMIT_BYTES;
1700     }
1701 
1702     for (i = VIRT_LOWMEMMAP_LAST; i < ARRAY_SIZE(extended_memmap); i++) {
1703         hwaddr size = extended_memmap[i].size;
1704 
1705         base = ROUND_UP(base, size);
1706         vms->memmap[i].base = base;
1707         vms->memmap[i].size = size;
1708         base += size;
1709     }
1710     vms->highest_gpa = base - 1;
1711     if (device_memory_size > 0) {
1712         ms->device_memory = g_malloc0(sizeof(*ms->device_memory));
1713         ms->device_memory->base = device_memory_base;
1714         memory_region_init(&ms->device_memory->mr, OBJECT(vms),
1715                            "device-memory", device_memory_size);
1716     }
1717 }
1718 
1719 /*
1720  * finalize_gic_version - Determines the final gic_version
1721  * according to the gic-version property
1722  *
1723  * Default GIC type is v2
1724  */
1725 static void finalize_gic_version(VirtMachineState *vms)
1726 {
1727     unsigned int max_cpus = MACHINE(vms)->smp.max_cpus;
1728 
1729     if (kvm_enabled()) {
1730         int probe_bitmap;
1731 
1732         if (!kvm_irqchip_in_kernel()) {
1733             switch (vms->gic_version) {
1734             case VIRT_GIC_VERSION_HOST:
1735                 warn_report(
1736                     "gic-version=host not relevant with kernel-irqchip=off "
1737                      "as only userspace GICv2 is supported. Using v2 ...");
1738                 return;
1739             case VIRT_GIC_VERSION_MAX:
1740             case VIRT_GIC_VERSION_NOSEL:
1741                 vms->gic_version = VIRT_GIC_VERSION_2;
1742                 return;
1743             case VIRT_GIC_VERSION_2:
1744                 return;
1745             case VIRT_GIC_VERSION_3:
1746                 error_report(
1747                     "gic-version=3 is not supported with kernel-irqchip=off");
1748                 exit(1);
1749             }
1750         }
1751 
1752         probe_bitmap = kvm_arm_vgic_probe();
1753         if (!probe_bitmap) {
1754             error_report("Unable to determine GIC version supported by host");
1755             exit(1);
1756         }
1757 
1758         switch (vms->gic_version) {
1759         case VIRT_GIC_VERSION_HOST:
1760         case VIRT_GIC_VERSION_MAX:
1761             if (probe_bitmap & KVM_ARM_VGIC_V3) {
1762                 vms->gic_version = VIRT_GIC_VERSION_3;
1763             } else {
1764                 vms->gic_version = VIRT_GIC_VERSION_2;
1765             }
1766             return;
1767         case VIRT_GIC_VERSION_NOSEL:
1768             if ((probe_bitmap & KVM_ARM_VGIC_V2) && max_cpus <= GIC_NCPU) {
1769                 vms->gic_version = VIRT_GIC_VERSION_2;
1770             } else if (probe_bitmap & KVM_ARM_VGIC_V3) {
1771                 /*
1772                  * in case the host does not support v2 in-kernel emulation or
1773                  * the end-user requested more than 8 VCPUs we now default
1774                  * to v3. In any case defaulting to v2 would be broken.
1775                  */
1776                 vms->gic_version = VIRT_GIC_VERSION_3;
1777             } else if (max_cpus > GIC_NCPU) {
1778                 error_report("host only supports in-kernel GICv2 emulation "
1779                              "but more than 8 vcpus are requested");
1780                 exit(1);
1781             }
1782             break;
1783         case VIRT_GIC_VERSION_2:
1784         case VIRT_GIC_VERSION_3:
1785             break;
1786         }
1787 
1788         /* Check chosen version is effectively supported by the host */
1789         if (vms->gic_version == VIRT_GIC_VERSION_2 &&
1790             !(probe_bitmap & KVM_ARM_VGIC_V2)) {
1791             error_report("host does not support in-kernel GICv2 emulation");
1792             exit(1);
1793         } else if (vms->gic_version == VIRT_GIC_VERSION_3 &&
1794                    !(probe_bitmap & KVM_ARM_VGIC_V3)) {
1795             error_report("host does not support in-kernel GICv3 emulation");
1796             exit(1);
1797         }
1798         return;
1799     }
1800 
1801     /* TCG mode */
1802     switch (vms->gic_version) {
1803     case VIRT_GIC_VERSION_NOSEL:
1804         vms->gic_version = VIRT_GIC_VERSION_2;
1805         break;
1806     case VIRT_GIC_VERSION_MAX:
1807         vms->gic_version = VIRT_GIC_VERSION_3;
1808         break;
1809     case VIRT_GIC_VERSION_HOST:
1810         error_report("gic-version=host requires KVM");
1811         exit(1);
1812     case VIRT_GIC_VERSION_2:
1813     case VIRT_GIC_VERSION_3:
1814         break;
1815     }
1816 }
1817 
1818 /*
1819  * virt_cpu_post_init() must be called after the CPUs have
1820  * been realized and the GIC has been created.
1821  */
1822 static void virt_cpu_post_init(VirtMachineState *vms, MemoryRegion *sysmem)
1823 {
1824     int max_cpus = MACHINE(vms)->smp.max_cpus;
1825     bool aarch64, pmu, steal_time;
1826     CPUState *cpu;
1827 
1828     aarch64 = object_property_get_bool(OBJECT(first_cpu), "aarch64", NULL);
1829     pmu = object_property_get_bool(OBJECT(first_cpu), "pmu", NULL);
1830     steal_time = object_property_get_bool(OBJECT(first_cpu),
1831                                           "kvm-steal-time", NULL);
1832 
1833     if (kvm_enabled()) {
1834         hwaddr pvtime_reg_base = vms->memmap[VIRT_PVTIME].base;
1835         hwaddr pvtime_reg_size = vms->memmap[VIRT_PVTIME].size;
1836 
1837         if (steal_time) {
1838             MemoryRegion *pvtime = g_new(MemoryRegion, 1);
1839             hwaddr pvtime_size = max_cpus * PVTIME_SIZE_PER_CPU;
1840 
1841             /* The memory region size must be a multiple of host page size. */
1842             pvtime_size = REAL_HOST_PAGE_ALIGN(pvtime_size);
1843 
1844             if (pvtime_size > pvtime_reg_size) {
1845                 error_report("pvtime requires a %" HWADDR_PRId
1846                              " byte memory region for %d CPUs,"
1847                              " but only %" HWADDR_PRId " has been reserved",
1848                              pvtime_size, max_cpus, pvtime_reg_size);
1849                 exit(1);
1850             }
1851 
1852             memory_region_init_ram(pvtime, NULL, "pvtime", pvtime_size, NULL);
1853             memory_region_add_subregion(sysmem, pvtime_reg_base, pvtime);
1854         }
1855 
1856         CPU_FOREACH(cpu) {
1857             if (pmu) {
1858                 assert(arm_feature(&ARM_CPU(cpu)->env, ARM_FEATURE_PMU));
1859                 if (kvm_irqchip_in_kernel()) {
1860                     kvm_arm_pmu_set_irq(cpu, PPI(VIRTUAL_PMU_IRQ));
1861                 }
1862                 kvm_arm_pmu_init(cpu);
1863             }
1864             if (steal_time) {
1865                 kvm_arm_pvtime_init(cpu, pvtime_reg_base +
1866                                          cpu->cpu_index * PVTIME_SIZE_PER_CPU);
1867             }
1868         }
1869     } else {
1870         if (aarch64 && vms->highmem) {
1871             int requested_pa_size = 64 - clz64(vms->highest_gpa);
1872             int pamax = arm_pamax(ARM_CPU(first_cpu));
1873 
1874             if (pamax < requested_pa_size) {
1875                 error_report("VCPU supports less PA bits (%d) than "
1876                              "requested by the memory map (%d)",
1877                              pamax, requested_pa_size);
1878                 exit(1);
1879             }
1880         }
1881     }
1882 }
1883 
1884 static void machvirt_init(MachineState *machine)
1885 {
1886     VirtMachineState *vms = VIRT_MACHINE(machine);
1887     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(machine);
1888     MachineClass *mc = MACHINE_GET_CLASS(machine);
1889     const CPUArchIdList *possible_cpus;
1890     MemoryRegion *sysmem = get_system_memory();
1891     MemoryRegion *secure_sysmem = NULL;
1892     MemoryRegion *tag_sysmem = NULL;
1893     MemoryRegion *secure_tag_sysmem = NULL;
1894     int n, virt_max_cpus;
1895     bool firmware_loaded;
1896     bool aarch64 = true;
1897     bool has_ged = !vmc->no_ged;
1898     unsigned int smp_cpus = machine->smp.cpus;
1899     unsigned int max_cpus = machine->smp.max_cpus;
1900 
1901     /*
1902      * In accelerated mode, the memory map is computed earlier in kvm_type()
1903      * to create a VM with the right number of IPA bits.
1904      */
1905     if (!vms->memmap) {
1906         virt_set_memmap(vms);
1907     }
1908 
1909     /* We can probe only here because during property set
1910      * KVM is not available yet
1911      */
1912     finalize_gic_version(vms);
1913 
1914     if (!cpu_type_valid(machine->cpu_type)) {
1915         error_report("mach-virt: CPU type %s not supported", machine->cpu_type);
1916         exit(1);
1917     }
1918 
1919     if (vms->secure) {
1920         /*
1921          * The Secure view of the world is the same as the NonSecure,
1922          * but with a few extra devices. Create it as a container region
1923          * containing the system memory at low priority; any secure-only
1924          * devices go in at higher priority and take precedence.
1925          */
1926         secure_sysmem = g_new(MemoryRegion, 1);
1927         memory_region_init(secure_sysmem, OBJECT(machine), "secure-memory",
1928                            UINT64_MAX);
1929         memory_region_add_subregion_overlap(secure_sysmem, 0, sysmem, -1);
1930     }
1931 
1932     firmware_loaded = virt_firmware_init(vms, sysmem,
1933                                          secure_sysmem ?: sysmem);
1934 
1935     /* If we have an EL3 boot ROM then the assumption is that it will
1936      * implement PSCI itself, so disable QEMU's internal implementation
1937      * so it doesn't get in the way. Instead of starting secondary
1938      * CPUs in PSCI powerdown state we will start them all running and
1939      * let the boot ROM sort them out.
1940      * The usual case is that we do use QEMU's PSCI implementation;
1941      * if the guest has EL2 then we will use SMC as the conduit,
1942      * and otherwise we will use HVC (for backwards compatibility and
1943      * because if we're using KVM then we must use HVC).
1944      */
1945     if (vms->secure && firmware_loaded) {
1946         vms->psci_conduit = QEMU_PSCI_CONDUIT_DISABLED;
1947     } else if (vms->virt) {
1948         vms->psci_conduit = QEMU_PSCI_CONDUIT_SMC;
1949     } else {
1950         vms->psci_conduit = QEMU_PSCI_CONDUIT_HVC;
1951     }
1952 
1953     /* The maximum number of CPUs depends on the GIC version, or on how
1954      * many redistributors we can fit into the memory map.
1955      */
1956     if (vms->gic_version == VIRT_GIC_VERSION_3) {
1957         virt_max_cpus =
1958             vms->memmap[VIRT_GIC_REDIST].size / GICV3_REDIST_SIZE;
1959         virt_max_cpus +=
1960             vms->memmap[VIRT_HIGH_GIC_REDIST2].size / GICV3_REDIST_SIZE;
1961     } else {
1962         virt_max_cpus = GIC_NCPU;
1963     }
1964 
1965     if (max_cpus > virt_max_cpus) {
1966         error_report("Number of SMP CPUs requested (%d) exceeds max CPUs "
1967                      "supported by machine 'mach-virt' (%d)",
1968                      max_cpus, virt_max_cpus);
1969         exit(1);
1970     }
1971 
1972     if (vms->virt && (kvm_enabled() || hvf_enabled())) {
1973         error_report("mach-virt: %s does not support providing "
1974                      "Virtualization extensions to the guest CPU",
1975                      kvm_enabled() ? "KVM" : "HVF");
1976         exit(1);
1977     }
1978 
1979     if (vms->mte && (kvm_enabled() || hvf_enabled())) {
1980         error_report("mach-virt: %s does not support providing "
1981                      "MTE to the guest CPU",
1982                      kvm_enabled() ? "KVM" : "HVF");
1983         exit(1);
1984     }
1985 
1986     create_fdt(vms);
1987 
1988     possible_cpus = mc->possible_cpu_arch_ids(machine);
1989     assert(possible_cpus->len == max_cpus);
1990     for (n = 0; n < possible_cpus->len; n++) {
1991         Object *cpuobj;
1992         CPUState *cs;
1993 
1994         if (n >= smp_cpus) {
1995             break;
1996         }
1997 
1998         cpuobj = object_new(possible_cpus->cpus[n].type);
1999         object_property_set_int(cpuobj, "mp-affinity",
2000                                 possible_cpus->cpus[n].arch_id, NULL);
2001 
2002         cs = CPU(cpuobj);
2003         cs->cpu_index = n;
2004 
2005         numa_cpu_pre_plug(&possible_cpus->cpus[cs->cpu_index], DEVICE(cpuobj),
2006                           &error_fatal);
2007 
2008         aarch64 &= object_property_get_bool(cpuobj, "aarch64", NULL);
2009 
2010         if (!vms->secure) {
2011             object_property_set_bool(cpuobj, "has_el3", false, NULL);
2012         }
2013 
2014         if (!vms->virt && object_property_find(cpuobj, "has_el2")) {
2015             object_property_set_bool(cpuobj, "has_el2", false, NULL);
2016         }
2017 
2018         if (vms->psci_conduit != QEMU_PSCI_CONDUIT_DISABLED) {
2019             object_property_set_int(cpuobj, "psci-conduit", vms->psci_conduit,
2020                                     NULL);
2021 
2022             /* Secondary CPUs start in PSCI powered-down state */
2023             if (n > 0) {
2024                 object_property_set_bool(cpuobj, "start-powered-off", true,
2025                                          NULL);
2026             }
2027         }
2028 
2029         if (vmc->kvm_no_adjvtime &&
2030             object_property_find(cpuobj, "kvm-no-adjvtime")) {
2031             object_property_set_bool(cpuobj, "kvm-no-adjvtime", true, NULL);
2032         }
2033 
2034         if (vmc->no_kvm_steal_time &&
2035             object_property_find(cpuobj, "kvm-steal-time")) {
2036             object_property_set_bool(cpuobj, "kvm-steal-time", false, NULL);
2037         }
2038 
2039         if (vmc->no_pmu && object_property_find(cpuobj, "pmu")) {
2040             object_property_set_bool(cpuobj, "pmu", false, NULL);
2041         }
2042 
2043         if (object_property_find(cpuobj, "reset-cbar")) {
2044             object_property_set_int(cpuobj, "reset-cbar",
2045                                     vms->memmap[VIRT_CPUPERIPHS].base,
2046                                     &error_abort);
2047         }
2048 
2049         object_property_set_link(cpuobj, "memory", OBJECT(sysmem),
2050                                  &error_abort);
2051         if (vms->secure) {
2052             object_property_set_link(cpuobj, "secure-memory",
2053                                      OBJECT(secure_sysmem), &error_abort);
2054         }
2055 
2056         if (vms->mte) {
2057             /* Create the memory region only once, but link to all cpus. */
2058             if (!tag_sysmem) {
2059                 /*
2060                  * The property exists only if MemTag is supported.
2061                  * If it is, we must allocate the ram to back that up.
2062                  */
2063                 if (!object_property_find(cpuobj, "tag-memory")) {
2064                     error_report("MTE requested, but not supported "
2065                                  "by the guest CPU");
2066                     exit(1);
2067                 }
2068 
2069                 tag_sysmem = g_new(MemoryRegion, 1);
2070                 memory_region_init(tag_sysmem, OBJECT(machine),
2071                                    "tag-memory", UINT64_MAX / 32);
2072 
2073                 if (vms->secure) {
2074                     secure_tag_sysmem = g_new(MemoryRegion, 1);
2075                     memory_region_init(secure_tag_sysmem, OBJECT(machine),
2076                                        "secure-tag-memory", UINT64_MAX / 32);
2077 
2078                     /* As with ram, secure-tag takes precedence over tag.  */
2079                     memory_region_add_subregion_overlap(secure_tag_sysmem, 0,
2080                                                         tag_sysmem, -1);
2081                 }
2082             }
2083 
2084             object_property_set_link(cpuobj, "tag-memory", OBJECT(tag_sysmem),
2085                                      &error_abort);
2086             if (vms->secure) {
2087                 object_property_set_link(cpuobj, "secure-tag-memory",
2088                                          OBJECT(secure_tag_sysmem),
2089                                          &error_abort);
2090             }
2091         }
2092 
2093         qdev_realize(DEVICE(cpuobj), NULL, &error_fatal);
2094         object_unref(cpuobj);
2095     }
2096     fdt_add_timer_nodes(vms);
2097     fdt_add_cpu_nodes(vms);
2098 
2099     memory_region_add_subregion(sysmem, vms->memmap[VIRT_MEM].base,
2100                                 machine->ram);
2101     if (machine->device_memory) {
2102         memory_region_add_subregion(sysmem, machine->device_memory->base,
2103                                     &machine->device_memory->mr);
2104     }
2105 
2106     virt_flash_fdt(vms, sysmem, secure_sysmem ?: sysmem);
2107 
2108     create_gic(vms, sysmem);
2109 
2110     virt_cpu_post_init(vms, sysmem);
2111 
2112     fdt_add_pmu_nodes(vms);
2113 
2114     create_uart(vms, VIRT_UART, sysmem, serial_hd(0));
2115 
2116     if (vms->secure) {
2117         create_secure_ram(vms, secure_sysmem, secure_tag_sysmem);
2118         create_uart(vms, VIRT_SECURE_UART, secure_sysmem, serial_hd(1));
2119     }
2120 
2121     if (tag_sysmem) {
2122         create_tag_ram(tag_sysmem, vms->memmap[VIRT_MEM].base,
2123                        machine->ram_size, "mach-virt.tag");
2124     }
2125 
2126     vms->highmem_ecam &= vms->highmem && (!firmware_loaded || aarch64);
2127 
2128     create_rtc(vms);
2129 
2130     create_pcie(vms);
2131 
2132     if (has_ged && aarch64 && firmware_loaded && virt_is_acpi_enabled(vms)) {
2133         vms->acpi_dev = create_acpi_ged(vms);
2134     } else {
2135         create_gpio_devices(vms, VIRT_GPIO, sysmem);
2136     }
2137 
2138     if (vms->secure && !vmc->no_secure_gpio) {
2139         create_gpio_devices(vms, VIRT_SECURE_GPIO, secure_sysmem);
2140     }
2141 
2142      /* connect powerdown request */
2143      vms->powerdown_notifier.notify = virt_powerdown_req;
2144      qemu_register_powerdown_notifier(&vms->powerdown_notifier);
2145 
2146     /* Create mmio transports, so the user can create virtio backends
2147      * (which will be automatically plugged in to the transports). If
2148      * no backend is created the transport will just sit harmlessly idle.
2149      */
2150     create_virtio_devices(vms);
2151 
2152     vms->fw_cfg = create_fw_cfg(vms, &address_space_memory);
2153     rom_set_fw(vms->fw_cfg);
2154 
2155     create_platform_bus(vms);
2156 
2157     if (machine->nvdimms_state->is_enabled) {
2158         const struct AcpiGenericAddress arm_virt_nvdimm_acpi_dsmio = {
2159             .space_id = AML_AS_SYSTEM_MEMORY,
2160             .address = vms->memmap[VIRT_NVDIMM_ACPI].base,
2161             .bit_width = NVDIMM_ACPI_IO_LEN << 3
2162         };
2163 
2164         nvdimm_init_acpi_state(machine->nvdimms_state, sysmem,
2165                                arm_virt_nvdimm_acpi_dsmio,
2166                                vms->fw_cfg, OBJECT(vms));
2167     }
2168 
2169     vms->bootinfo.ram_size = machine->ram_size;
2170     vms->bootinfo.nb_cpus = smp_cpus;
2171     vms->bootinfo.board_id = -1;
2172     vms->bootinfo.loader_start = vms->memmap[VIRT_MEM].base;
2173     vms->bootinfo.get_dtb = machvirt_dtb;
2174     vms->bootinfo.skip_dtb_autoload = true;
2175     vms->bootinfo.firmware_loaded = firmware_loaded;
2176     arm_load_kernel(ARM_CPU(first_cpu), machine, &vms->bootinfo);
2177 
2178     vms->machine_done.notify = virt_machine_done;
2179     qemu_add_machine_init_done_notifier(&vms->machine_done);
2180 }
2181 
2182 static bool virt_get_secure(Object *obj, Error **errp)
2183 {
2184     VirtMachineState *vms = VIRT_MACHINE(obj);
2185 
2186     return vms->secure;
2187 }
2188 
2189 static void virt_set_secure(Object *obj, bool value, Error **errp)
2190 {
2191     VirtMachineState *vms = VIRT_MACHINE(obj);
2192 
2193     vms->secure = value;
2194 }
2195 
2196 static bool virt_get_virt(Object *obj, Error **errp)
2197 {
2198     VirtMachineState *vms = VIRT_MACHINE(obj);
2199 
2200     return vms->virt;
2201 }
2202 
2203 static void virt_set_virt(Object *obj, bool value, Error **errp)
2204 {
2205     VirtMachineState *vms = VIRT_MACHINE(obj);
2206 
2207     vms->virt = value;
2208 }
2209 
2210 static bool virt_get_highmem(Object *obj, Error **errp)
2211 {
2212     VirtMachineState *vms = VIRT_MACHINE(obj);
2213 
2214     return vms->highmem;
2215 }
2216 
2217 static void virt_set_highmem(Object *obj, bool value, Error **errp)
2218 {
2219     VirtMachineState *vms = VIRT_MACHINE(obj);
2220 
2221     vms->highmem = value;
2222 }
2223 
2224 static bool virt_get_its(Object *obj, Error **errp)
2225 {
2226     VirtMachineState *vms = VIRT_MACHINE(obj);
2227 
2228     return vms->its;
2229 }
2230 
2231 static void virt_set_its(Object *obj, bool value, Error **errp)
2232 {
2233     VirtMachineState *vms = VIRT_MACHINE(obj);
2234 
2235     vms->its = value;
2236 }
2237 
2238 static char *virt_get_oem_id(Object *obj, Error **errp)
2239 {
2240     VirtMachineState *vms = VIRT_MACHINE(obj);
2241 
2242     return g_strdup(vms->oem_id);
2243 }
2244 
2245 static void virt_set_oem_id(Object *obj, const char *value, Error **errp)
2246 {
2247     VirtMachineState *vms = VIRT_MACHINE(obj);
2248     size_t len = strlen(value);
2249 
2250     if (len > 6) {
2251         error_setg(errp,
2252                    "User specified oem-id value is bigger than 6 bytes in size");
2253         return;
2254     }
2255 
2256     strncpy(vms->oem_id, value, 6);
2257 }
2258 
2259 static char *virt_get_oem_table_id(Object *obj, Error **errp)
2260 {
2261     VirtMachineState *vms = VIRT_MACHINE(obj);
2262 
2263     return g_strdup(vms->oem_table_id);
2264 }
2265 
2266 static void virt_set_oem_table_id(Object *obj, const char *value,
2267                                   Error **errp)
2268 {
2269     VirtMachineState *vms = VIRT_MACHINE(obj);
2270     size_t len = strlen(value);
2271 
2272     if (len > 8) {
2273         error_setg(errp,
2274                    "User specified oem-table-id value is bigger than 8 bytes in size");
2275         return;
2276     }
2277     strncpy(vms->oem_table_id, value, 8);
2278 }
2279 
2280 
2281 bool virt_is_acpi_enabled(VirtMachineState *vms)
2282 {
2283     if (vms->acpi == ON_OFF_AUTO_OFF) {
2284         return false;
2285     }
2286     return true;
2287 }
2288 
2289 static void virt_get_acpi(Object *obj, Visitor *v, const char *name,
2290                           void *opaque, Error **errp)
2291 {
2292     VirtMachineState *vms = VIRT_MACHINE(obj);
2293     OnOffAuto acpi = vms->acpi;
2294 
2295     visit_type_OnOffAuto(v, name, &acpi, errp);
2296 }
2297 
2298 static void virt_set_acpi(Object *obj, Visitor *v, const char *name,
2299                           void *opaque, Error **errp)
2300 {
2301     VirtMachineState *vms = VIRT_MACHINE(obj);
2302 
2303     visit_type_OnOffAuto(v, name, &vms->acpi, errp);
2304 }
2305 
2306 static bool virt_get_ras(Object *obj, Error **errp)
2307 {
2308     VirtMachineState *vms = VIRT_MACHINE(obj);
2309 
2310     return vms->ras;
2311 }
2312 
2313 static void virt_set_ras(Object *obj, bool value, Error **errp)
2314 {
2315     VirtMachineState *vms = VIRT_MACHINE(obj);
2316 
2317     vms->ras = value;
2318 }
2319 
2320 static bool virt_get_mte(Object *obj, Error **errp)
2321 {
2322     VirtMachineState *vms = VIRT_MACHINE(obj);
2323 
2324     return vms->mte;
2325 }
2326 
2327 static void virt_set_mte(Object *obj, bool value, Error **errp)
2328 {
2329     VirtMachineState *vms = VIRT_MACHINE(obj);
2330 
2331     vms->mte = value;
2332 }
2333 
2334 static char *virt_get_gic_version(Object *obj, Error **errp)
2335 {
2336     VirtMachineState *vms = VIRT_MACHINE(obj);
2337     const char *val = vms->gic_version == VIRT_GIC_VERSION_3 ? "3" : "2";
2338 
2339     return g_strdup(val);
2340 }
2341 
2342 static void virt_set_gic_version(Object *obj, const char *value, Error **errp)
2343 {
2344     VirtMachineState *vms = VIRT_MACHINE(obj);
2345 
2346     if (!strcmp(value, "3")) {
2347         vms->gic_version = VIRT_GIC_VERSION_3;
2348     } else if (!strcmp(value, "2")) {
2349         vms->gic_version = VIRT_GIC_VERSION_2;
2350     } else if (!strcmp(value, "host")) {
2351         vms->gic_version = VIRT_GIC_VERSION_HOST; /* Will probe later */
2352     } else if (!strcmp(value, "max")) {
2353         vms->gic_version = VIRT_GIC_VERSION_MAX; /* Will probe later */
2354     } else {
2355         error_setg(errp, "Invalid gic-version value");
2356         error_append_hint(errp, "Valid values are 3, 2, host, max.\n");
2357     }
2358 }
2359 
2360 static char *virt_get_iommu(Object *obj, Error **errp)
2361 {
2362     VirtMachineState *vms = VIRT_MACHINE(obj);
2363 
2364     switch (vms->iommu) {
2365     case VIRT_IOMMU_NONE:
2366         return g_strdup("none");
2367     case VIRT_IOMMU_SMMUV3:
2368         return g_strdup("smmuv3");
2369     default:
2370         g_assert_not_reached();
2371     }
2372 }
2373 
2374 static void virt_set_iommu(Object *obj, const char *value, Error **errp)
2375 {
2376     VirtMachineState *vms = VIRT_MACHINE(obj);
2377 
2378     if (!strcmp(value, "smmuv3")) {
2379         vms->iommu = VIRT_IOMMU_SMMUV3;
2380     } else if (!strcmp(value, "none")) {
2381         vms->iommu = VIRT_IOMMU_NONE;
2382     } else {
2383         error_setg(errp, "Invalid iommu value");
2384         error_append_hint(errp, "Valid values are none, smmuv3.\n");
2385     }
2386 }
2387 
2388 static bool virt_get_default_bus_bypass_iommu(Object *obj, Error **errp)
2389 {
2390     VirtMachineState *vms = VIRT_MACHINE(obj);
2391 
2392     return vms->default_bus_bypass_iommu;
2393 }
2394 
2395 static void virt_set_default_bus_bypass_iommu(Object *obj, bool value,
2396                                               Error **errp)
2397 {
2398     VirtMachineState *vms = VIRT_MACHINE(obj);
2399 
2400     vms->default_bus_bypass_iommu = value;
2401 }
2402 
2403 static CpuInstanceProperties
2404 virt_cpu_index_to_props(MachineState *ms, unsigned cpu_index)
2405 {
2406     MachineClass *mc = MACHINE_GET_CLASS(ms);
2407     const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(ms);
2408 
2409     assert(cpu_index < possible_cpus->len);
2410     return possible_cpus->cpus[cpu_index].props;
2411 }
2412 
2413 static int64_t virt_get_default_cpu_node_id(const MachineState *ms, int idx)
2414 {
2415     return idx % ms->numa_state->num_nodes;
2416 }
2417 
2418 static const CPUArchIdList *virt_possible_cpu_arch_ids(MachineState *ms)
2419 {
2420     int n;
2421     unsigned int max_cpus = ms->smp.max_cpus;
2422     VirtMachineState *vms = VIRT_MACHINE(ms);
2423 
2424     if (ms->possible_cpus) {
2425         assert(ms->possible_cpus->len == max_cpus);
2426         return ms->possible_cpus;
2427     }
2428 
2429     ms->possible_cpus = g_malloc0(sizeof(CPUArchIdList) +
2430                                   sizeof(CPUArchId) * max_cpus);
2431     ms->possible_cpus->len = max_cpus;
2432     for (n = 0; n < ms->possible_cpus->len; n++) {
2433         ms->possible_cpus->cpus[n].type = ms->cpu_type;
2434         ms->possible_cpus->cpus[n].arch_id =
2435             virt_cpu_mp_affinity(vms, n);
2436         ms->possible_cpus->cpus[n].props.has_thread_id = true;
2437         ms->possible_cpus->cpus[n].props.thread_id = n;
2438     }
2439     return ms->possible_cpus;
2440 }
2441 
2442 static void virt_memory_pre_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
2443                                  Error **errp)
2444 {
2445     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2446     const MachineState *ms = MACHINE(hotplug_dev);
2447     const bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2448 
2449     if (!vms->acpi_dev) {
2450         error_setg(errp,
2451                    "memory hotplug is not enabled: missing acpi-ged device");
2452         return;
2453     }
2454 
2455     if (vms->mte) {
2456         error_setg(errp, "memory hotplug is not enabled: MTE is enabled");
2457         return;
2458     }
2459 
2460     if (is_nvdimm && !ms->nvdimms_state->is_enabled) {
2461         error_setg(errp, "nvdimm is not enabled: add 'nvdimm=on' to '-M'");
2462         return;
2463     }
2464 
2465     pc_dimm_pre_plug(PC_DIMM(dev), MACHINE(hotplug_dev), NULL, errp);
2466 }
2467 
2468 static void virt_memory_plug(HotplugHandler *hotplug_dev,
2469                              DeviceState *dev, Error **errp)
2470 {
2471     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2472     MachineState *ms = MACHINE(hotplug_dev);
2473     bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2474 
2475     pc_dimm_plug(PC_DIMM(dev), MACHINE(vms));
2476 
2477     if (is_nvdimm) {
2478         nvdimm_plug(ms->nvdimms_state);
2479     }
2480 
2481     hotplug_handler_plug(HOTPLUG_HANDLER(vms->acpi_dev),
2482                          dev, &error_abort);
2483 }
2484 
2485 static void virt_machine_device_pre_plug_cb(HotplugHandler *hotplug_dev,
2486                                             DeviceState *dev, Error **errp)
2487 {
2488     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2489 
2490     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2491         virt_memory_pre_plug(hotplug_dev, dev, errp);
2492     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2493         hwaddr db_start = 0, db_end = 0;
2494         char *resv_prop_str;
2495 
2496         if (vms->iommu != VIRT_IOMMU_NONE) {
2497             error_setg(errp, "virt machine does not support multiple IOMMUs");
2498             return;
2499         }
2500 
2501         switch (vms->msi_controller) {
2502         case VIRT_MSI_CTRL_NONE:
2503             return;
2504         case VIRT_MSI_CTRL_ITS:
2505             /* GITS_TRANSLATER page */
2506             db_start = base_memmap[VIRT_GIC_ITS].base + 0x10000;
2507             db_end = base_memmap[VIRT_GIC_ITS].base +
2508                      base_memmap[VIRT_GIC_ITS].size - 1;
2509             break;
2510         case VIRT_MSI_CTRL_GICV2M:
2511             /* MSI_SETSPI_NS page */
2512             db_start = base_memmap[VIRT_GIC_V2M].base;
2513             db_end = db_start + base_memmap[VIRT_GIC_V2M].size - 1;
2514             break;
2515         }
2516         resv_prop_str = g_strdup_printf("0x%"PRIx64":0x%"PRIx64":%u",
2517                                         db_start, db_end,
2518                                         VIRTIO_IOMMU_RESV_MEM_T_MSI);
2519 
2520         object_property_set_uint(OBJECT(dev), "len-reserved-regions", 1, errp);
2521         object_property_set_str(OBJECT(dev), "reserved-regions[0]",
2522                                 resv_prop_str, errp);
2523         g_free(resv_prop_str);
2524     }
2525 }
2526 
2527 static void virt_machine_device_plug_cb(HotplugHandler *hotplug_dev,
2528                                         DeviceState *dev, Error **errp)
2529 {
2530     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2531 
2532     if (vms->platform_bus_dev) {
2533         MachineClass *mc = MACHINE_GET_CLASS(vms);
2534 
2535         if (device_is_dynamic_sysbus(mc, dev)) {
2536             platform_bus_link_device(PLATFORM_BUS_DEVICE(vms->platform_bus_dev),
2537                                      SYS_BUS_DEVICE(dev));
2538         }
2539     }
2540     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2541         virt_memory_plug(hotplug_dev, dev, errp);
2542     }
2543     if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2544         PCIDevice *pdev = PCI_DEVICE(dev);
2545 
2546         vms->iommu = VIRT_IOMMU_VIRTIO;
2547         vms->virtio_iommu_bdf = pci_get_bdf(pdev);
2548         create_virtio_iommu_dt_bindings(vms);
2549     }
2550 }
2551 
2552 static void virt_dimm_unplug_request(HotplugHandler *hotplug_dev,
2553                                      DeviceState *dev, Error **errp)
2554 {
2555     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2556     Error *local_err = NULL;
2557 
2558     if (!vms->acpi_dev) {
2559         error_setg(&local_err,
2560                    "memory hotplug is not enabled: missing acpi-ged device");
2561         goto out;
2562     }
2563 
2564     if (object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM)) {
2565         error_setg(&local_err,
2566                    "nvdimm device hot unplug is not supported yet.");
2567         goto out;
2568     }
2569 
2570     hotplug_handler_unplug_request(HOTPLUG_HANDLER(vms->acpi_dev), dev,
2571                                    &local_err);
2572 out:
2573     error_propagate(errp, local_err);
2574 }
2575 
2576 static void virt_dimm_unplug(HotplugHandler *hotplug_dev,
2577                              DeviceState *dev, Error **errp)
2578 {
2579     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2580     Error *local_err = NULL;
2581 
2582     hotplug_handler_unplug(HOTPLUG_HANDLER(vms->acpi_dev), dev, &local_err);
2583     if (local_err) {
2584         goto out;
2585     }
2586 
2587     pc_dimm_unplug(PC_DIMM(dev), MACHINE(vms));
2588     qdev_unrealize(dev);
2589 
2590 out:
2591     error_propagate(errp, local_err);
2592 }
2593 
2594 static void virt_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
2595                                           DeviceState *dev, Error **errp)
2596 {
2597     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2598         virt_dimm_unplug_request(hotplug_dev, dev, errp);
2599     } else {
2600         error_setg(errp, "device unplug request for unsupported device"
2601                    " type: %s", object_get_typename(OBJECT(dev)));
2602     }
2603 }
2604 
2605 static void virt_machine_device_unplug_cb(HotplugHandler *hotplug_dev,
2606                                           DeviceState *dev, Error **errp)
2607 {
2608     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2609         virt_dimm_unplug(hotplug_dev, dev, errp);
2610     } else {
2611         error_setg(errp, "virt: device unplug for unsupported device"
2612                    " type: %s", object_get_typename(OBJECT(dev)));
2613     }
2614 }
2615 
2616 static HotplugHandler *virt_machine_get_hotplug_handler(MachineState *machine,
2617                                                         DeviceState *dev)
2618 {
2619     MachineClass *mc = MACHINE_GET_CLASS(machine);
2620 
2621     if (device_is_dynamic_sysbus(mc, dev) ||
2622         object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) ||
2623         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2624         return HOTPLUG_HANDLER(machine);
2625     }
2626     return NULL;
2627 }
2628 
2629 /*
2630  * for arm64 kvm_type [7-0] encodes the requested number of bits
2631  * in the IPA address space
2632  */
2633 static int virt_kvm_type(MachineState *ms, const char *type_str)
2634 {
2635     VirtMachineState *vms = VIRT_MACHINE(ms);
2636     int max_vm_pa_size, requested_pa_size;
2637     bool fixed_ipa;
2638 
2639     max_vm_pa_size = kvm_arm_get_max_vm_ipa_size(ms, &fixed_ipa);
2640 
2641     /* we freeze the memory map to compute the highest gpa */
2642     virt_set_memmap(vms);
2643 
2644     requested_pa_size = 64 - clz64(vms->highest_gpa);
2645 
2646     /*
2647      * KVM requires the IPA size to be at least 32 bits.
2648      */
2649     if (requested_pa_size < 32) {
2650         requested_pa_size = 32;
2651     }
2652 
2653     if (requested_pa_size > max_vm_pa_size) {
2654         error_report("-m and ,maxmem option values "
2655                      "require an IPA range (%d bits) larger than "
2656                      "the one supported by the host (%d bits)",
2657                      requested_pa_size, max_vm_pa_size);
2658         exit(1);
2659     }
2660     /*
2661      * We return the requested PA log size, unless KVM only supports
2662      * the implicit legacy 40b IPA setting, in which case the kvm_type
2663      * must be 0.
2664      */
2665     return fixed_ipa ? 0 : requested_pa_size;
2666 }
2667 
2668 static void virt_machine_class_init(ObjectClass *oc, void *data)
2669 {
2670     MachineClass *mc = MACHINE_CLASS(oc);
2671     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
2672 
2673     mc->init = machvirt_init;
2674     /* Start with max_cpus set to 512, which is the maximum supported by KVM.
2675      * The value may be reduced later when we have more information about the
2676      * configuration of the particular instance.
2677      */
2678     mc->max_cpus = 512;
2679     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_CALXEDA_XGMAC);
2680     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_AMD_XGBE);
2681     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_RAMFB_DEVICE);
2682     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_PLATFORM);
2683 #ifdef CONFIG_TPM
2684     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_TPM_TIS_SYSBUS);
2685 #endif
2686     mc->block_default_type = IF_VIRTIO;
2687     mc->no_cdrom = 1;
2688     mc->pci_allow_0_address = true;
2689     /* We know we will never create a pre-ARMv7 CPU which needs 1K pages */
2690     mc->minimum_page_bits = 12;
2691     mc->possible_cpu_arch_ids = virt_possible_cpu_arch_ids;
2692     mc->cpu_index_to_instance_props = virt_cpu_index_to_props;
2693     mc->default_cpu_type = ARM_CPU_TYPE_NAME("cortex-a15");
2694     mc->get_default_cpu_node_id = virt_get_default_cpu_node_id;
2695     mc->kvm_type = virt_kvm_type;
2696     assert(!mc->get_hotplug_handler);
2697     mc->get_hotplug_handler = virt_machine_get_hotplug_handler;
2698     hc->pre_plug = virt_machine_device_pre_plug_cb;
2699     hc->plug = virt_machine_device_plug_cb;
2700     hc->unplug_request = virt_machine_device_unplug_request_cb;
2701     hc->unplug = virt_machine_device_unplug_cb;
2702     mc->nvdimm_supported = true;
2703     mc->auto_enable_numa_with_memhp = true;
2704     mc->auto_enable_numa_with_memdev = true;
2705     mc->default_ram_id = "mach-virt.ram";
2706 
2707     object_class_property_add(oc, "acpi", "OnOffAuto",
2708         virt_get_acpi, virt_set_acpi,
2709         NULL, NULL);
2710     object_class_property_set_description(oc, "acpi",
2711         "Enable ACPI");
2712     object_class_property_add_bool(oc, "secure", virt_get_secure,
2713                                    virt_set_secure);
2714     object_class_property_set_description(oc, "secure",
2715                                                 "Set on/off to enable/disable the ARM "
2716                                                 "Security Extensions (TrustZone)");
2717 
2718     object_class_property_add_bool(oc, "virtualization", virt_get_virt,
2719                                    virt_set_virt);
2720     object_class_property_set_description(oc, "virtualization",
2721                                           "Set on/off to enable/disable emulating a "
2722                                           "guest CPU which implements the ARM "
2723                                           "Virtualization Extensions");
2724 
2725     object_class_property_add_bool(oc, "highmem", virt_get_highmem,
2726                                    virt_set_highmem);
2727     object_class_property_set_description(oc, "highmem",
2728                                           "Set on/off to enable/disable using "
2729                                           "physical address space above 32 bits");
2730 
2731     object_class_property_add_str(oc, "gic-version", virt_get_gic_version,
2732                                   virt_set_gic_version);
2733     object_class_property_set_description(oc, "gic-version",
2734                                           "Set GIC version. "
2735                                           "Valid values are 2, 3, host and max");
2736 
2737     object_class_property_add_str(oc, "iommu", virt_get_iommu, virt_set_iommu);
2738     object_class_property_set_description(oc, "iommu",
2739                                           "Set the IOMMU type. "
2740                                           "Valid values are none and smmuv3");
2741 
2742     object_class_property_add_bool(oc, "default-bus-bypass-iommu",
2743                                    virt_get_default_bus_bypass_iommu,
2744                                    virt_set_default_bus_bypass_iommu);
2745     object_class_property_set_description(oc, "default-bus-bypass-iommu",
2746                                           "Set on/off to enable/disable "
2747                                           "bypass_iommu for default root bus");
2748 
2749     object_class_property_add_bool(oc, "ras", virt_get_ras,
2750                                    virt_set_ras);
2751     object_class_property_set_description(oc, "ras",
2752                                           "Set on/off to enable/disable reporting host memory errors "
2753                                           "to a KVM guest using ACPI and guest external abort exceptions");
2754 
2755     object_class_property_add_bool(oc, "mte", virt_get_mte, virt_set_mte);
2756     object_class_property_set_description(oc, "mte",
2757                                           "Set on/off to enable/disable emulating a "
2758                                           "guest CPU which implements the ARM "
2759                                           "Memory Tagging Extension");
2760 
2761     object_class_property_add_bool(oc, "its", virt_get_its,
2762                                    virt_set_its);
2763     object_class_property_set_description(oc, "its",
2764                                           "Set on/off to enable/disable "
2765                                           "ITS instantiation");
2766 
2767     object_class_property_add_str(oc, "x-oem-id",
2768                                   virt_get_oem_id,
2769                                   virt_set_oem_id);
2770     object_class_property_set_description(oc, "x-oem-id",
2771                                           "Override the default value of field OEMID "
2772                                           "in ACPI table header."
2773                                           "The string may be up to 6 bytes in size");
2774 
2775 
2776     object_class_property_add_str(oc, "x-oem-table-id",
2777                                   virt_get_oem_table_id,
2778                                   virt_set_oem_table_id);
2779     object_class_property_set_description(oc, "x-oem-table-id",
2780                                           "Override the default value of field OEM Table ID "
2781                                           "in ACPI table header."
2782                                           "The string may be up to 8 bytes in size");
2783 
2784 }
2785 
2786 static void virt_instance_init(Object *obj)
2787 {
2788     VirtMachineState *vms = VIRT_MACHINE(obj);
2789     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
2790 
2791     /* EL3 is disabled by default on virt: this makes us consistent
2792      * between KVM and TCG for this board, and it also allows us to
2793      * boot UEFI blobs which assume no TrustZone support.
2794      */
2795     vms->secure = false;
2796 
2797     /* EL2 is also disabled by default, for similar reasons */
2798     vms->virt = false;
2799 
2800     /* High memory is enabled by default */
2801     vms->highmem = true;
2802     vms->gic_version = VIRT_GIC_VERSION_NOSEL;
2803 
2804     vms->highmem_ecam = !vmc->no_highmem_ecam;
2805 
2806     if (vmc->no_its) {
2807         vms->its = false;
2808     } else {
2809         /* Default allows ITS instantiation */
2810         vms->its = true;
2811 
2812         if (vmc->no_tcg_its) {
2813             vms->tcg_its = false;
2814         } else {
2815             vms->tcg_its = true;
2816         }
2817     }
2818 
2819     /* Default disallows iommu instantiation */
2820     vms->iommu = VIRT_IOMMU_NONE;
2821 
2822     /* The default root bus is attached to iommu by default */
2823     vms->default_bus_bypass_iommu = false;
2824 
2825     /* Default disallows RAS instantiation */
2826     vms->ras = false;
2827 
2828     /* MTE is disabled by default.  */
2829     vms->mte = false;
2830 
2831     vms->irqmap = a15irqmap;
2832 
2833     virt_flash_create(vms);
2834 
2835     vms->oem_id = g_strndup(ACPI_BUILD_APPNAME6, 6);
2836     vms->oem_table_id = g_strndup(ACPI_BUILD_APPNAME8, 8);
2837 }
2838 
2839 static const TypeInfo virt_machine_info = {
2840     .name          = TYPE_VIRT_MACHINE,
2841     .parent        = TYPE_MACHINE,
2842     .abstract      = true,
2843     .instance_size = sizeof(VirtMachineState),
2844     .class_size    = sizeof(VirtMachineClass),
2845     .class_init    = virt_machine_class_init,
2846     .instance_init = virt_instance_init,
2847     .interfaces = (InterfaceInfo[]) {
2848          { TYPE_HOTPLUG_HANDLER },
2849          { }
2850     },
2851 };
2852 
2853 static void machvirt_machine_init(void)
2854 {
2855     type_register_static(&virt_machine_info);
2856 }
2857 type_init(machvirt_machine_init);
2858 
2859 static void virt_machine_6_2_options(MachineClass *mc)
2860 {
2861 }
2862 DEFINE_VIRT_MACHINE_AS_LATEST(6, 2)
2863 
2864 static void virt_machine_6_1_options(MachineClass *mc)
2865 {
2866     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2867 
2868     virt_machine_6_2_options(mc);
2869     compat_props_add(mc->compat_props, hw_compat_6_1, hw_compat_6_1_len);
2870     mc->smp_props.prefer_sockets = true;
2871     vmc->no_cpu_topology = true;
2872 
2873     /* qemu ITS was introduced with 6.2 */
2874     vmc->no_tcg_its = true;
2875 }
2876 DEFINE_VIRT_MACHINE(6, 1)
2877 
2878 static void virt_machine_6_0_options(MachineClass *mc)
2879 {
2880     virt_machine_6_1_options(mc);
2881     compat_props_add(mc->compat_props, hw_compat_6_0, hw_compat_6_0_len);
2882 }
2883 DEFINE_VIRT_MACHINE(6, 0)
2884 
2885 static void virt_machine_5_2_options(MachineClass *mc)
2886 {
2887     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2888 
2889     virt_machine_6_0_options(mc);
2890     compat_props_add(mc->compat_props, hw_compat_5_2, hw_compat_5_2_len);
2891     vmc->no_secure_gpio = true;
2892 }
2893 DEFINE_VIRT_MACHINE(5, 2)
2894 
2895 static void virt_machine_5_1_options(MachineClass *mc)
2896 {
2897     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2898 
2899     virt_machine_5_2_options(mc);
2900     compat_props_add(mc->compat_props, hw_compat_5_1, hw_compat_5_1_len);
2901     vmc->no_kvm_steal_time = true;
2902 }
2903 DEFINE_VIRT_MACHINE(5, 1)
2904 
2905 static void virt_machine_5_0_options(MachineClass *mc)
2906 {
2907     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2908 
2909     virt_machine_5_1_options(mc);
2910     compat_props_add(mc->compat_props, hw_compat_5_0, hw_compat_5_0_len);
2911     mc->numa_mem_supported = true;
2912     vmc->acpi_expose_flash = true;
2913     mc->auto_enable_numa_with_memdev = false;
2914 }
2915 DEFINE_VIRT_MACHINE(5, 0)
2916 
2917 static void virt_machine_4_2_options(MachineClass *mc)
2918 {
2919     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2920 
2921     virt_machine_5_0_options(mc);
2922     compat_props_add(mc->compat_props, hw_compat_4_2, hw_compat_4_2_len);
2923     vmc->kvm_no_adjvtime = true;
2924 }
2925 DEFINE_VIRT_MACHINE(4, 2)
2926 
2927 static void virt_machine_4_1_options(MachineClass *mc)
2928 {
2929     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2930 
2931     virt_machine_4_2_options(mc);
2932     compat_props_add(mc->compat_props, hw_compat_4_1, hw_compat_4_1_len);
2933     vmc->no_ged = true;
2934     mc->auto_enable_numa_with_memhp = false;
2935 }
2936 DEFINE_VIRT_MACHINE(4, 1)
2937 
2938 static void virt_machine_4_0_options(MachineClass *mc)
2939 {
2940     virt_machine_4_1_options(mc);
2941     compat_props_add(mc->compat_props, hw_compat_4_0, hw_compat_4_0_len);
2942 }
2943 DEFINE_VIRT_MACHINE(4, 0)
2944 
2945 static void virt_machine_3_1_options(MachineClass *mc)
2946 {
2947     virt_machine_4_0_options(mc);
2948     compat_props_add(mc->compat_props, hw_compat_3_1, hw_compat_3_1_len);
2949 }
2950 DEFINE_VIRT_MACHINE(3, 1)
2951 
2952 static void virt_machine_3_0_options(MachineClass *mc)
2953 {
2954     virt_machine_3_1_options(mc);
2955     compat_props_add(mc->compat_props, hw_compat_3_0, hw_compat_3_0_len);
2956 }
2957 DEFINE_VIRT_MACHINE(3, 0)
2958 
2959 static void virt_machine_2_12_options(MachineClass *mc)
2960 {
2961     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2962 
2963     virt_machine_3_0_options(mc);
2964     compat_props_add(mc->compat_props, hw_compat_2_12, hw_compat_2_12_len);
2965     vmc->no_highmem_ecam = true;
2966     mc->max_cpus = 255;
2967 }
2968 DEFINE_VIRT_MACHINE(2, 12)
2969 
2970 static void virt_machine_2_11_options(MachineClass *mc)
2971 {
2972     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2973 
2974     virt_machine_2_12_options(mc);
2975     compat_props_add(mc->compat_props, hw_compat_2_11, hw_compat_2_11_len);
2976     vmc->smbios_old_sys_ver = true;
2977 }
2978 DEFINE_VIRT_MACHINE(2, 11)
2979 
2980 static void virt_machine_2_10_options(MachineClass *mc)
2981 {
2982     virt_machine_2_11_options(mc);
2983     compat_props_add(mc->compat_props, hw_compat_2_10, hw_compat_2_10_len);
2984     /* before 2.11 we never faulted accesses to bad addresses */
2985     mc->ignore_memory_transaction_failures = true;
2986 }
2987 DEFINE_VIRT_MACHINE(2, 10)
2988 
2989 static void virt_machine_2_9_options(MachineClass *mc)
2990 {
2991     virt_machine_2_10_options(mc);
2992     compat_props_add(mc->compat_props, hw_compat_2_9, hw_compat_2_9_len);
2993 }
2994 DEFINE_VIRT_MACHINE(2, 9)
2995 
2996 static void virt_machine_2_8_options(MachineClass *mc)
2997 {
2998     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2999 
3000     virt_machine_2_9_options(mc);
3001     compat_props_add(mc->compat_props, hw_compat_2_8, hw_compat_2_8_len);
3002     /* For 2.8 and earlier we falsely claimed in the DT that
3003      * our timers were edge-triggered, not level-triggered.
3004      */
3005     vmc->claim_edge_triggered_timers = true;
3006 }
3007 DEFINE_VIRT_MACHINE(2, 8)
3008 
3009 static void virt_machine_2_7_options(MachineClass *mc)
3010 {
3011     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3012 
3013     virt_machine_2_8_options(mc);
3014     compat_props_add(mc->compat_props, hw_compat_2_7, hw_compat_2_7_len);
3015     /* ITS was introduced with 2.8 */
3016     vmc->no_its = true;
3017     /* Stick with 1K pages for migration compatibility */
3018     mc->minimum_page_bits = 0;
3019 }
3020 DEFINE_VIRT_MACHINE(2, 7)
3021 
3022 static void virt_machine_2_6_options(MachineClass *mc)
3023 {
3024     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3025 
3026     virt_machine_2_7_options(mc);
3027     compat_props_add(mc->compat_props, hw_compat_2_6, hw_compat_2_6_len);
3028     vmc->disallow_affinity_adjustment = true;
3029     /* Disable PMU for 2.6 as PMU support was first introduced in 2.7 */
3030     vmc->no_pmu = true;
3031 }
3032 DEFINE_VIRT_MACHINE(2, 6)
3033