xref: /qemu/hw/arm/virt.c (revision d0bf492f)
1 /*
2  * ARM mach-virt emulation
3  *
4  * Copyright (c) 2013 Linaro Limited
5  *
6  * This program is free software; you can redistribute it and/or modify it
7  * under the terms and conditions of the GNU General Public License,
8  * version 2 or later, as published by the Free Software Foundation.
9  *
10  * This program is distributed in the hope it will be useful, but WITHOUT
11  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
12  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
13  * more details.
14  *
15  * You should have received a copy of the GNU General Public License along with
16  * this program.  If not, see <http://www.gnu.org/licenses/>.
17  *
18  * Emulate a virtual board which works by passing Linux all the information
19  * it needs about what devices are present via the device tree.
20  * There are some restrictions about what we can do here:
21  *  + we can only present devices whose Linux drivers will work based
22  *    purely on the device tree with no platform data at all
23  *  + we want to present a very stripped-down minimalist platform,
24  *    both because this reduces the security attack surface from the guest
25  *    and also because it reduces our exposure to being broken when
26  *    the kernel updates its device tree bindings and requires further
27  *    information in a device binding that we aren't providing.
28  * This is essentially the same approach kvmtool uses.
29  */
30 
31 #include "qemu/osdep.h"
32 #include "qemu/datadir.h"
33 #include "qemu/units.h"
34 #include "qemu/option.h"
35 #include "monitor/qdev.h"
36 #include "hw/sysbus.h"
37 #include "hw/arm/boot.h"
38 #include "hw/arm/primecell.h"
39 #include "hw/arm/virt.h"
40 #include "hw/block/flash.h"
41 #include "hw/vfio/vfio-calxeda-xgmac.h"
42 #include "hw/vfio/vfio-amd-xgbe.h"
43 #include "hw/display/ramfb.h"
44 #include "net/net.h"
45 #include "sysemu/device_tree.h"
46 #include "sysemu/numa.h"
47 #include "sysemu/runstate.h"
48 #include "sysemu/tpm.h"
49 #include "sysemu/tcg.h"
50 #include "sysemu/kvm.h"
51 #include "sysemu/hvf.h"
52 #include "sysemu/qtest.h"
53 #include "hw/loader.h"
54 #include "qapi/error.h"
55 #include "qemu/bitops.h"
56 #include "qemu/error-report.h"
57 #include "qemu/module.h"
58 #include "hw/pci-host/gpex.h"
59 #include "hw/virtio/virtio-pci.h"
60 #include "hw/core/sysbus-fdt.h"
61 #include "hw/platform-bus.h"
62 #include "hw/qdev-properties.h"
63 #include "hw/arm/fdt.h"
64 #include "hw/intc/arm_gic.h"
65 #include "hw/intc/arm_gicv3_common.h"
66 #include "hw/intc/arm_gicv3_its_common.h"
67 #include "hw/irq.h"
68 #include "kvm_arm.h"
69 #include "hw/firmware/smbios.h"
70 #include "qapi/visitor.h"
71 #include "qapi/qapi-visit-common.h"
72 #include "standard-headers/linux/input.h"
73 #include "hw/arm/smmuv3.h"
74 #include "hw/acpi/acpi.h"
75 #include "target/arm/internals.h"
76 #include "hw/mem/memory-device.h"
77 #include "hw/mem/pc-dimm.h"
78 #include "hw/mem/nvdimm.h"
79 #include "hw/acpi/generic_event_device.h"
80 #include "hw/virtio/virtio-mem-pci.h"
81 #include "hw/virtio/virtio-iommu.h"
82 #include "hw/char/pl011.h"
83 #include "qemu/guest-random.h"
84 
85 #define DEFINE_VIRT_MACHINE_LATEST(major, minor, latest) \
86     static void virt_##major##_##minor##_class_init(ObjectClass *oc, \
87                                                     void *data) \
88     { \
89         MachineClass *mc = MACHINE_CLASS(oc); \
90         virt_machine_##major##_##minor##_options(mc); \
91         mc->desc = "QEMU " # major "." # minor " ARM Virtual Machine"; \
92         if (latest) { \
93             mc->alias = "virt"; \
94         } \
95     } \
96     static const TypeInfo machvirt_##major##_##minor##_info = { \
97         .name = MACHINE_TYPE_NAME("virt-" # major "." # minor), \
98         .parent = TYPE_VIRT_MACHINE, \
99         .class_init = virt_##major##_##minor##_class_init, \
100     }; \
101     static void machvirt_machine_##major##_##minor##_init(void) \
102     { \
103         type_register_static(&machvirt_##major##_##minor##_info); \
104     } \
105     type_init(machvirt_machine_##major##_##minor##_init);
106 
107 #define DEFINE_VIRT_MACHINE_AS_LATEST(major, minor) \
108     DEFINE_VIRT_MACHINE_LATEST(major, minor, true)
109 #define DEFINE_VIRT_MACHINE(major, minor) \
110     DEFINE_VIRT_MACHINE_LATEST(major, minor, false)
111 
112 
113 /* Number of external interrupt lines to configure the GIC with */
114 #define NUM_IRQS 256
115 
116 #define PLATFORM_BUS_NUM_IRQS 64
117 
118 /* Legacy RAM limit in GB (< version 4.0) */
119 #define LEGACY_RAMLIMIT_GB 255
120 #define LEGACY_RAMLIMIT_BYTES (LEGACY_RAMLIMIT_GB * GiB)
121 
122 /* Addresses and sizes of our components.
123  * 0..128MB is space for a flash device so we can run bootrom code such as UEFI.
124  * 128MB..256MB is used for miscellaneous device I/O.
125  * 256MB..1GB is reserved for possible future PCI support (ie where the
126  * PCI memory window will go if we add a PCI host controller).
127  * 1GB and up is RAM (which may happily spill over into the
128  * high memory region beyond 4GB).
129  * This represents a compromise between how much RAM can be given to
130  * a 32 bit VM and leaving space for expansion and in particular for PCI.
131  * Note that devices should generally be placed at multiples of 0x10000,
132  * to accommodate guests using 64K pages.
133  */
134 static const MemMapEntry base_memmap[] = {
135     /* Space up to 0x8000000 is reserved for a boot ROM */
136     [VIRT_FLASH] =              {          0, 0x08000000 },
137     [VIRT_CPUPERIPHS] =         { 0x08000000, 0x00020000 },
138     /* GIC distributor and CPU interfaces sit inside the CPU peripheral space */
139     [VIRT_GIC_DIST] =           { 0x08000000, 0x00010000 },
140     [VIRT_GIC_CPU] =            { 0x08010000, 0x00010000 },
141     [VIRT_GIC_V2M] =            { 0x08020000, 0x00001000 },
142     [VIRT_GIC_HYP] =            { 0x08030000, 0x00010000 },
143     [VIRT_GIC_VCPU] =           { 0x08040000, 0x00010000 },
144     /* The space in between here is reserved for GICv3 CPU/vCPU/HYP */
145     [VIRT_GIC_ITS] =            { 0x08080000, 0x00020000 },
146     /* This redistributor space allows up to 2*64kB*123 CPUs */
147     [VIRT_GIC_REDIST] =         { 0x080A0000, 0x00F60000 },
148     [VIRT_UART] =               { 0x09000000, 0x00001000 },
149     [VIRT_RTC] =                { 0x09010000, 0x00001000 },
150     [VIRT_FW_CFG] =             { 0x09020000, 0x00000018 },
151     [VIRT_GPIO] =               { 0x09030000, 0x00001000 },
152     [VIRT_SECURE_UART] =        { 0x09040000, 0x00001000 },
153     [VIRT_SMMU] =               { 0x09050000, 0x00020000 },
154     [VIRT_PCDIMM_ACPI] =        { 0x09070000, MEMORY_HOTPLUG_IO_LEN },
155     [VIRT_ACPI_GED] =           { 0x09080000, ACPI_GED_EVT_SEL_LEN },
156     [VIRT_NVDIMM_ACPI] =        { 0x09090000, NVDIMM_ACPI_IO_LEN},
157     [VIRT_PVTIME] =             { 0x090a0000, 0x00010000 },
158     [VIRT_SECURE_GPIO] =        { 0x090b0000, 0x00001000 },
159     [VIRT_MMIO] =               { 0x0a000000, 0x00000200 },
160     /* ...repeating for a total of NUM_VIRTIO_TRANSPORTS, each of that size */
161     [VIRT_PLATFORM_BUS] =       { 0x0c000000, 0x02000000 },
162     [VIRT_SECURE_MEM] =         { 0x0e000000, 0x01000000 },
163     [VIRT_PCIE_MMIO] =          { 0x10000000, 0x2eff0000 },
164     [VIRT_PCIE_PIO] =           { 0x3eff0000, 0x00010000 },
165     [VIRT_PCIE_ECAM] =          { 0x3f000000, 0x01000000 },
166     /* Actual RAM size depends on initial RAM and device memory settings */
167     [VIRT_MEM] =                { GiB, LEGACY_RAMLIMIT_BYTES },
168 };
169 
170 /*
171  * Highmem IO Regions: This memory map is floating, located after the RAM.
172  * Each MemMapEntry base (GPA) will be dynamically computed, depending on the
173  * top of the RAM, so that its base get the same alignment as the size,
174  * ie. a 512GiB entry will be aligned on a 512GiB boundary. If there is
175  * less than 256GiB of RAM, the floating area starts at the 256GiB mark.
176  * Note the extended_memmap is sized so that it eventually also includes the
177  * base_memmap entries (VIRT_HIGH_GIC_REDIST2 index is greater than the last
178  * index of base_memmap).
179  *
180  * The memory map for these Highmem IO Regions can be in legacy or compact
181  * layout, depending on 'compact-highmem' property. With legacy layout, the
182  * PA space for one specific region is always reserved, even if the region
183  * has been disabled or doesn't fit into the PA space. However, the PA space
184  * for the region won't be reserved in these circumstances with compact layout.
185  */
186 static MemMapEntry extended_memmap[] = {
187     /* Additional 64 MB redist region (can contain up to 512 redistributors) */
188     [VIRT_HIGH_GIC_REDIST2] =   { 0x0, 64 * MiB },
189     [VIRT_HIGH_PCIE_ECAM] =     { 0x0, 256 * MiB },
190     /* Second PCIe window */
191     [VIRT_HIGH_PCIE_MMIO] =     { 0x0, 512 * GiB },
192 };
193 
194 static const int a15irqmap[] = {
195     [VIRT_UART] = 1,
196     [VIRT_RTC] = 2,
197     [VIRT_PCIE] = 3, /* ... to 6 */
198     [VIRT_GPIO] = 7,
199     [VIRT_SECURE_UART] = 8,
200     [VIRT_ACPI_GED] = 9,
201     [VIRT_MMIO] = 16, /* ...to 16 + NUM_VIRTIO_TRANSPORTS - 1 */
202     [VIRT_GIC_V2M] = 48, /* ...to 48 + NUM_GICV2M_SPIS - 1 */
203     [VIRT_SMMU] = 74,    /* ...to 74 + NUM_SMMU_IRQS - 1 */
204     [VIRT_PLATFORM_BUS] = 112, /* ...to 112 + PLATFORM_BUS_NUM_IRQS -1 */
205 };
206 
207 static const char *valid_cpus[] = {
208 #ifdef CONFIG_TCG
209     ARM_CPU_TYPE_NAME("cortex-a7"),
210     ARM_CPU_TYPE_NAME("cortex-a15"),
211     ARM_CPU_TYPE_NAME("cortex-a35"),
212     ARM_CPU_TYPE_NAME("cortex-a55"),
213     ARM_CPU_TYPE_NAME("cortex-a72"),
214     ARM_CPU_TYPE_NAME("cortex-a76"),
215     ARM_CPU_TYPE_NAME("a64fx"),
216     ARM_CPU_TYPE_NAME("neoverse-n1"),
217     ARM_CPU_TYPE_NAME("neoverse-v1"),
218 #endif
219     ARM_CPU_TYPE_NAME("cortex-a53"),
220     ARM_CPU_TYPE_NAME("cortex-a57"),
221     ARM_CPU_TYPE_NAME("host"),
222     ARM_CPU_TYPE_NAME("max"),
223 };
224 
225 static bool cpu_type_valid(const char *cpu)
226 {
227     int i;
228 
229     for (i = 0; i < ARRAY_SIZE(valid_cpus); i++) {
230         if (strcmp(cpu, valid_cpus[i]) == 0) {
231             return true;
232         }
233     }
234     return false;
235 }
236 
237 static void create_randomness(MachineState *ms, const char *node)
238 {
239     struct {
240         uint64_t kaslr;
241         uint8_t rng[32];
242     } seed;
243 
244     if (qemu_guest_getrandom(&seed, sizeof(seed), NULL)) {
245         return;
246     }
247     qemu_fdt_setprop_u64(ms->fdt, node, "kaslr-seed", seed.kaslr);
248     qemu_fdt_setprop(ms->fdt, node, "rng-seed", seed.rng, sizeof(seed.rng));
249 }
250 
251 static void create_fdt(VirtMachineState *vms)
252 {
253     MachineState *ms = MACHINE(vms);
254     int nb_numa_nodes = ms->numa_state->num_nodes;
255     void *fdt = create_device_tree(&vms->fdt_size);
256 
257     if (!fdt) {
258         error_report("create_device_tree() failed");
259         exit(1);
260     }
261 
262     ms->fdt = fdt;
263 
264     /* Header */
265     qemu_fdt_setprop_string(fdt, "/", "compatible", "linux,dummy-virt");
266     qemu_fdt_setprop_cell(fdt, "/", "#address-cells", 0x2);
267     qemu_fdt_setprop_cell(fdt, "/", "#size-cells", 0x2);
268     qemu_fdt_setprop_string(fdt, "/", "model", "linux,dummy-virt");
269 
270     /* /chosen must exist for load_dtb to fill in necessary properties later */
271     qemu_fdt_add_subnode(fdt, "/chosen");
272     if (vms->dtb_randomness) {
273         create_randomness(ms, "/chosen");
274     }
275 
276     if (vms->secure) {
277         qemu_fdt_add_subnode(fdt, "/secure-chosen");
278         if (vms->dtb_randomness) {
279             create_randomness(ms, "/secure-chosen");
280         }
281     }
282 
283     /* Clock node, for the benefit of the UART. The kernel device tree
284      * binding documentation claims the PL011 node clock properties are
285      * optional but in practice if you omit them the kernel refuses to
286      * probe for the device.
287      */
288     vms->clock_phandle = qemu_fdt_alloc_phandle(fdt);
289     qemu_fdt_add_subnode(fdt, "/apb-pclk");
290     qemu_fdt_setprop_string(fdt, "/apb-pclk", "compatible", "fixed-clock");
291     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "#clock-cells", 0x0);
292     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "clock-frequency", 24000000);
293     qemu_fdt_setprop_string(fdt, "/apb-pclk", "clock-output-names",
294                                 "clk24mhz");
295     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "phandle", vms->clock_phandle);
296 
297     if (nb_numa_nodes > 0 && ms->numa_state->have_numa_distance) {
298         int size = nb_numa_nodes * nb_numa_nodes * 3 * sizeof(uint32_t);
299         uint32_t *matrix = g_malloc0(size);
300         int idx, i, j;
301 
302         for (i = 0; i < nb_numa_nodes; i++) {
303             for (j = 0; j < nb_numa_nodes; j++) {
304                 idx = (i * nb_numa_nodes + j) * 3;
305                 matrix[idx + 0] = cpu_to_be32(i);
306                 matrix[idx + 1] = cpu_to_be32(j);
307                 matrix[idx + 2] =
308                     cpu_to_be32(ms->numa_state->nodes[i].distance[j]);
309             }
310         }
311 
312         qemu_fdt_add_subnode(fdt, "/distance-map");
313         qemu_fdt_setprop_string(fdt, "/distance-map", "compatible",
314                                 "numa-distance-map-v1");
315         qemu_fdt_setprop(fdt, "/distance-map", "distance-matrix",
316                          matrix, size);
317         g_free(matrix);
318     }
319 }
320 
321 static void fdt_add_timer_nodes(const VirtMachineState *vms)
322 {
323     /* On real hardware these interrupts are level-triggered.
324      * On KVM they were edge-triggered before host kernel version 4.4,
325      * and level-triggered afterwards.
326      * On emulated QEMU they are level-triggered.
327      *
328      * Getting the DTB info about them wrong is awkward for some
329      * guest kernels:
330      *  pre-4.8 ignore the DT and leave the interrupt configured
331      *   with whatever the GIC reset value (or the bootloader) left it at
332      *  4.8 before rc6 honour the incorrect data by programming it back
333      *   into the GIC, causing problems
334      *  4.8rc6 and later ignore the DT and always write "level triggered"
335      *   into the GIC
336      *
337      * For backwards-compatibility, virt-2.8 and earlier will continue
338      * to say these are edge-triggered, but later machines will report
339      * the correct information.
340      */
341     ARMCPU *armcpu;
342     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
343     uint32_t irqflags = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
344     MachineState *ms = MACHINE(vms);
345 
346     if (vmc->claim_edge_triggered_timers) {
347         irqflags = GIC_FDT_IRQ_FLAGS_EDGE_LO_HI;
348     }
349 
350     if (vms->gic_version == VIRT_GIC_VERSION_2) {
351         irqflags = deposit32(irqflags, GIC_FDT_IRQ_PPI_CPU_START,
352                              GIC_FDT_IRQ_PPI_CPU_WIDTH,
353                              (1 << MACHINE(vms)->smp.cpus) - 1);
354     }
355 
356     qemu_fdt_add_subnode(ms->fdt, "/timer");
357 
358     armcpu = ARM_CPU(qemu_get_cpu(0));
359     if (arm_feature(&armcpu->env, ARM_FEATURE_V8)) {
360         const char compat[] = "arm,armv8-timer\0arm,armv7-timer";
361         qemu_fdt_setprop(ms->fdt, "/timer", "compatible",
362                          compat, sizeof(compat));
363     } else {
364         qemu_fdt_setprop_string(ms->fdt, "/timer", "compatible",
365                                 "arm,armv7-timer");
366     }
367     qemu_fdt_setprop(ms->fdt, "/timer", "always-on", NULL, 0);
368     qemu_fdt_setprop_cells(ms->fdt, "/timer", "interrupts",
369                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_S_EL1_IRQ, irqflags,
370                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_NS_EL1_IRQ, irqflags,
371                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_VIRT_IRQ, irqflags,
372                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_NS_EL2_IRQ, irqflags);
373 }
374 
375 static void fdt_add_cpu_nodes(const VirtMachineState *vms)
376 {
377     int cpu;
378     int addr_cells = 1;
379     const MachineState *ms = MACHINE(vms);
380     const VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
381     int smp_cpus = ms->smp.cpus;
382 
383     /*
384      * See Linux Documentation/devicetree/bindings/arm/cpus.yaml
385      * On ARM v8 64-bit systems value should be set to 2,
386      * that corresponds to the MPIDR_EL1 register size.
387      * If MPIDR_EL1[63:32] value is equal to 0 on all CPUs
388      * in the system, #address-cells can be set to 1, since
389      * MPIDR_EL1[63:32] bits are not used for CPUs
390      * identification.
391      *
392      * Here we actually don't know whether our system is 32- or 64-bit one.
393      * The simplest way to go is to examine affinity IDs of all our CPUs. If
394      * at least one of them has Aff3 populated, we set #address-cells to 2.
395      */
396     for (cpu = 0; cpu < smp_cpus; cpu++) {
397         ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(cpu));
398 
399         if (armcpu->mp_affinity & ARM_AFF3_MASK) {
400             addr_cells = 2;
401             break;
402         }
403     }
404 
405     qemu_fdt_add_subnode(ms->fdt, "/cpus");
406     qemu_fdt_setprop_cell(ms->fdt, "/cpus", "#address-cells", addr_cells);
407     qemu_fdt_setprop_cell(ms->fdt, "/cpus", "#size-cells", 0x0);
408 
409     for (cpu = smp_cpus - 1; cpu >= 0; cpu--) {
410         char *nodename = g_strdup_printf("/cpus/cpu@%d", cpu);
411         ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(cpu));
412         CPUState *cs = CPU(armcpu);
413 
414         qemu_fdt_add_subnode(ms->fdt, nodename);
415         qemu_fdt_setprop_string(ms->fdt, nodename, "device_type", "cpu");
416         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
417                                     armcpu->dtb_compatible);
418 
419         if (vms->psci_conduit != QEMU_PSCI_CONDUIT_DISABLED && smp_cpus > 1) {
420             qemu_fdt_setprop_string(ms->fdt, nodename,
421                                         "enable-method", "psci");
422         }
423 
424         if (addr_cells == 2) {
425             qemu_fdt_setprop_u64(ms->fdt, nodename, "reg",
426                                  armcpu->mp_affinity);
427         } else {
428             qemu_fdt_setprop_cell(ms->fdt, nodename, "reg",
429                                   armcpu->mp_affinity);
430         }
431 
432         if (ms->possible_cpus->cpus[cs->cpu_index].props.has_node_id) {
433             qemu_fdt_setprop_cell(ms->fdt, nodename, "numa-node-id",
434                 ms->possible_cpus->cpus[cs->cpu_index].props.node_id);
435         }
436 
437         if (!vmc->no_cpu_topology) {
438             qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle",
439                                   qemu_fdt_alloc_phandle(ms->fdt));
440         }
441 
442         g_free(nodename);
443     }
444 
445     if (!vmc->no_cpu_topology) {
446         /*
447          * Add vCPU topology description through fdt node cpu-map.
448          *
449          * See Linux Documentation/devicetree/bindings/cpu/cpu-topology.txt
450          * In a SMP system, the hierarchy of CPUs can be defined through
451          * four entities that are used to describe the layout of CPUs in
452          * the system: socket/cluster/core/thread.
453          *
454          * A socket node represents the boundary of system physical package
455          * and its child nodes must be one or more cluster nodes. A system
456          * can contain several layers of clustering within a single physical
457          * package and cluster nodes can be contained in parent cluster nodes.
458          *
459          * Note: currently we only support one layer of clustering within
460          * each physical package.
461          */
462         qemu_fdt_add_subnode(ms->fdt, "/cpus/cpu-map");
463 
464         for (cpu = smp_cpus - 1; cpu >= 0; cpu--) {
465             char *cpu_path = g_strdup_printf("/cpus/cpu@%d", cpu);
466             char *map_path;
467 
468             if (ms->smp.threads > 1) {
469                 map_path = g_strdup_printf(
470                     "/cpus/cpu-map/socket%d/cluster%d/core%d/thread%d",
471                     cpu / (ms->smp.clusters * ms->smp.cores * ms->smp.threads),
472                     (cpu / (ms->smp.cores * ms->smp.threads)) % ms->smp.clusters,
473                     (cpu / ms->smp.threads) % ms->smp.cores,
474                     cpu % ms->smp.threads);
475             } else {
476                 map_path = g_strdup_printf(
477                     "/cpus/cpu-map/socket%d/cluster%d/core%d",
478                     cpu / (ms->smp.clusters * ms->smp.cores),
479                     (cpu / ms->smp.cores) % ms->smp.clusters,
480                     cpu % ms->smp.cores);
481             }
482             qemu_fdt_add_path(ms->fdt, map_path);
483             qemu_fdt_setprop_phandle(ms->fdt, map_path, "cpu", cpu_path);
484 
485             g_free(map_path);
486             g_free(cpu_path);
487         }
488     }
489 }
490 
491 static void fdt_add_its_gic_node(VirtMachineState *vms)
492 {
493     char *nodename;
494     MachineState *ms = MACHINE(vms);
495 
496     vms->msi_phandle = qemu_fdt_alloc_phandle(ms->fdt);
497     nodename = g_strdup_printf("/intc/its@%" PRIx64,
498                                vms->memmap[VIRT_GIC_ITS].base);
499     qemu_fdt_add_subnode(ms->fdt, nodename);
500     qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
501                             "arm,gic-v3-its");
502     qemu_fdt_setprop(ms->fdt, nodename, "msi-controller", NULL, 0);
503     qemu_fdt_setprop_cell(ms->fdt, nodename, "#msi-cells", 1);
504     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
505                                  2, vms->memmap[VIRT_GIC_ITS].base,
506                                  2, vms->memmap[VIRT_GIC_ITS].size);
507     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", vms->msi_phandle);
508     g_free(nodename);
509 }
510 
511 static void fdt_add_v2m_gic_node(VirtMachineState *vms)
512 {
513     MachineState *ms = MACHINE(vms);
514     char *nodename;
515 
516     nodename = g_strdup_printf("/intc/v2m@%" PRIx64,
517                                vms->memmap[VIRT_GIC_V2M].base);
518     vms->msi_phandle = qemu_fdt_alloc_phandle(ms->fdt);
519     qemu_fdt_add_subnode(ms->fdt, nodename);
520     qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
521                             "arm,gic-v2m-frame");
522     qemu_fdt_setprop(ms->fdt, nodename, "msi-controller", NULL, 0);
523     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
524                                  2, vms->memmap[VIRT_GIC_V2M].base,
525                                  2, vms->memmap[VIRT_GIC_V2M].size);
526     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", vms->msi_phandle);
527     g_free(nodename);
528 }
529 
530 static void fdt_add_gic_node(VirtMachineState *vms)
531 {
532     MachineState *ms = MACHINE(vms);
533     char *nodename;
534 
535     vms->gic_phandle = qemu_fdt_alloc_phandle(ms->fdt);
536     qemu_fdt_setprop_cell(ms->fdt, "/", "interrupt-parent", vms->gic_phandle);
537 
538     nodename = g_strdup_printf("/intc@%" PRIx64,
539                                vms->memmap[VIRT_GIC_DIST].base);
540     qemu_fdt_add_subnode(ms->fdt, nodename);
541     qemu_fdt_setprop_cell(ms->fdt, nodename, "#interrupt-cells", 3);
542     qemu_fdt_setprop(ms->fdt, nodename, "interrupt-controller", NULL, 0);
543     qemu_fdt_setprop_cell(ms->fdt, nodename, "#address-cells", 0x2);
544     qemu_fdt_setprop_cell(ms->fdt, nodename, "#size-cells", 0x2);
545     qemu_fdt_setprop(ms->fdt, nodename, "ranges", NULL, 0);
546     if (vms->gic_version != VIRT_GIC_VERSION_2) {
547         int nb_redist_regions = virt_gicv3_redist_region_count(vms);
548 
549         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
550                                 "arm,gic-v3");
551 
552         qemu_fdt_setprop_cell(ms->fdt, nodename,
553                               "#redistributor-regions", nb_redist_regions);
554 
555         if (nb_redist_regions == 1) {
556             qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
557                                          2, vms->memmap[VIRT_GIC_DIST].base,
558                                          2, vms->memmap[VIRT_GIC_DIST].size,
559                                          2, vms->memmap[VIRT_GIC_REDIST].base,
560                                          2, vms->memmap[VIRT_GIC_REDIST].size);
561         } else {
562             qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
563                                  2, vms->memmap[VIRT_GIC_DIST].base,
564                                  2, vms->memmap[VIRT_GIC_DIST].size,
565                                  2, vms->memmap[VIRT_GIC_REDIST].base,
566                                  2, vms->memmap[VIRT_GIC_REDIST].size,
567                                  2, vms->memmap[VIRT_HIGH_GIC_REDIST2].base,
568                                  2, vms->memmap[VIRT_HIGH_GIC_REDIST2].size);
569         }
570 
571         if (vms->virt) {
572             qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
573                                    GIC_FDT_IRQ_TYPE_PPI, ARCH_GIC_MAINT_IRQ,
574                                    GIC_FDT_IRQ_FLAGS_LEVEL_HI);
575         }
576     } else {
577         /* 'cortex-a15-gic' means 'GIC v2' */
578         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
579                                 "arm,cortex-a15-gic");
580         if (!vms->virt) {
581             qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
582                                          2, vms->memmap[VIRT_GIC_DIST].base,
583                                          2, vms->memmap[VIRT_GIC_DIST].size,
584                                          2, vms->memmap[VIRT_GIC_CPU].base,
585                                          2, vms->memmap[VIRT_GIC_CPU].size);
586         } else {
587             qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
588                                          2, vms->memmap[VIRT_GIC_DIST].base,
589                                          2, vms->memmap[VIRT_GIC_DIST].size,
590                                          2, vms->memmap[VIRT_GIC_CPU].base,
591                                          2, vms->memmap[VIRT_GIC_CPU].size,
592                                          2, vms->memmap[VIRT_GIC_HYP].base,
593                                          2, vms->memmap[VIRT_GIC_HYP].size,
594                                          2, vms->memmap[VIRT_GIC_VCPU].base,
595                                          2, vms->memmap[VIRT_GIC_VCPU].size);
596             qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
597                                    GIC_FDT_IRQ_TYPE_PPI, ARCH_GIC_MAINT_IRQ,
598                                    GIC_FDT_IRQ_FLAGS_LEVEL_HI);
599         }
600     }
601 
602     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", vms->gic_phandle);
603     g_free(nodename);
604 }
605 
606 static void fdt_add_pmu_nodes(const VirtMachineState *vms)
607 {
608     ARMCPU *armcpu = ARM_CPU(first_cpu);
609     uint32_t irqflags = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
610     MachineState *ms = MACHINE(vms);
611 
612     if (!arm_feature(&armcpu->env, ARM_FEATURE_PMU)) {
613         assert(!object_property_get_bool(OBJECT(armcpu), "pmu", NULL));
614         return;
615     }
616 
617     if (vms->gic_version == VIRT_GIC_VERSION_2) {
618         irqflags = deposit32(irqflags, GIC_FDT_IRQ_PPI_CPU_START,
619                              GIC_FDT_IRQ_PPI_CPU_WIDTH,
620                              (1 << MACHINE(vms)->smp.cpus) - 1);
621     }
622 
623     qemu_fdt_add_subnode(ms->fdt, "/pmu");
624     if (arm_feature(&armcpu->env, ARM_FEATURE_V8)) {
625         const char compat[] = "arm,armv8-pmuv3";
626         qemu_fdt_setprop(ms->fdt, "/pmu", "compatible",
627                          compat, sizeof(compat));
628         qemu_fdt_setprop_cells(ms->fdt, "/pmu", "interrupts",
629                                GIC_FDT_IRQ_TYPE_PPI, VIRTUAL_PMU_IRQ, irqflags);
630     }
631 }
632 
633 static inline DeviceState *create_acpi_ged(VirtMachineState *vms)
634 {
635     DeviceState *dev;
636     MachineState *ms = MACHINE(vms);
637     int irq = vms->irqmap[VIRT_ACPI_GED];
638     uint32_t event = ACPI_GED_PWR_DOWN_EVT;
639 
640     if (ms->ram_slots) {
641         event |= ACPI_GED_MEM_HOTPLUG_EVT;
642     }
643 
644     if (ms->nvdimms_state->is_enabled) {
645         event |= ACPI_GED_NVDIMM_HOTPLUG_EVT;
646     }
647 
648     dev = qdev_new(TYPE_ACPI_GED);
649     qdev_prop_set_uint32(dev, "ged-event", event);
650 
651     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_ACPI_GED].base);
652     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 1, vms->memmap[VIRT_PCDIMM_ACPI].base);
653     sysbus_connect_irq(SYS_BUS_DEVICE(dev), 0, qdev_get_gpio_in(vms->gic, irq));
654 
655     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
656 
657     return dev;
658 }
659 
660 static void create_its(VirtMachineState *vms)
661 {
662     const char *itsclass = its_class_name();
663     DeviceState *dev;
664 
665     if (!strcmp(itsclass, "arm-gicv3-its")) {
666         if (!vms->tcg_its) {
667             itsclass = NULL;
668         }
669     }
670 
671     if (!itsclass) {
672         /* Do nothing if not supported */
673         return;
674     }
675 
676     dev = qdev_new(itsclass);
677 
678     object_property_set_link(OBJECT(dev), "parent-gicv3", OBJECT(vms->gic),
679                              &error_abort);
680     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
681     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_GIC_ITS].base);
682 
683     fdt_add_its_gic_node(vms);
684     vms->msi_controller = VIRT_MSI_CTRL_ITS;
685 }
686 
687 static void create_v2m(VirtMachineState *vms)
688 {
689     int i;
690     int irq = vms->irqmap[VIRT_GIC_V2M];
691     DeviceState *dev;
692 
693     dev = qdev_new("arm-gicv2m");
694     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_GIC_V2M].base);
695     qdev_prop_set_uint32(dev, "base-spi", irq);
696     qdev_prop_set_uint32(dev, "num-spi", NUM_GICV2M_SPIS);
697     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
698 
699     for (i = 0; i < NUM_GICV2M_SPIS; i++) {
700         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
701                            qdev_get_gpio_in(vms->gic, irq + i));
702     }
703 
704     fdt_add_v2m_gic_node(vms);
705     vms->msi_controller = VIRT_MSI_CTRL_GICV2M;
706 }
707 
708 static void create_gic(VirtMachineState *vms, MemoryRegion *mem)
709 {
710     MachineState *ms = MACHINE(vms);
711     /* We create a standalone GIC */
712     SysBusDevice *gicbusdev;
713     const char *gictype;
714     int i;
715     unsigned int smp_cpus = ms->smp.cpus;
716     uint32_t nb_redist_regions = 0;
717     int revision;
718 
719     if (vms->gic_version == VIRT_GIC_VERSION_2) {
720         gictype = gic_class_name();
721     } else {
722         gictype = gicv3_class_name();
723     }
724 
725     switch (vms->gic_version) {
726     case VIRT_GIC_VERSION_2:
727         revision = 2;
728         break;
729     case VIRT_GIC_VERSION_3:
730         revision = 3;
731         break;
732     case VIRT_GIC_VERSION_4:
733         revision = 4;
734         break;
735     default:
736         g_assert_not_reached();
737     }
738     vms->gic = qdev_new(gictype);
739     qdev_prop_set_uint32(vms->gic, "revision", revision);
740     qdev_prop_set_uint32(vms->gic, "num-cpu", smp_cpus);
741     /* Note that the num-irq property counts both internal and external
742      * interrupts; there are always 32 of the former (mandated by GIC spec).
743      */
744     qdev_prop_set_uint32(vms->gic, "num-irq", NUM_IRQS + 32);
745     if (!kvm_irqchip_in_kernel()) {
746         qdev_prop_set_bit(vms->gic, "has-security-extensions", vms->secure);
747     }
748 
749     if (vms->gic_version != VIRT_GIC_VERSION_2) {
750         uint32_t redist0_capacity = virt_redist_capacity(vms, VIRT_GIC_REDIST);
751         uint32_t redist0_count = MIN(smp_cpus, redist0_capacity);
752 
753         nb_redist_regions = virt_gicv3_redist_region_count(vms);
754 
755         qdev_prop_set_uint32(vms->gic, "len-redist-region-count",
756                              nb_redist_regions);
757         qdev_prop_set_uint32(vms->gic, "redist-region-count[0]", redist0_count);
758 
759         if (!kvm_irqchip_in_kernel()) {
760             if (vms->tcg_its) {
761                 object_property_set_link(OBJECT(vms->gic), "sysmem",
762                                          OBJECT(mem), &error_fatal);
763                 qdev_prop_set_bit(vms->gic, "has-lpi", true);
764             }
765         }
766 
767         if (nb_redist_regions == 2) {
768             uint32_t redist1_capacity =
769                 virt_redist_capacity(vms, VIRT_HIGH_GIC_REDIST2);
770 
771             qdev_prop_set_uint32(vms->gic, "redist-region-count[1]",
772                 MIN(smp_cpus - redist0_count, redist1_capacity));
773         }
774     } else {
775         if (!kvm_irqchip_in_kernel()) {
776             qdev_prop_set_bit(vms->gic, "has-virtualization-extensions",
777                               vms->virt);
778         }
779     }
780     gicbusdev = SYS_BUS_DEVICE(vms->gic);
781     sysbus_realize_and_unref(gicbusdev, &error_fatal);
782     sysbus_mmio_map(gicbusdev, 0, vms->memmap[VIRT_GIC_DIST].base);
783     if (vms->gic_version != VIRT_GIC_VERSION_2) {
784         sysbus_mmio_map(gicbusdev, 1, vms->memmap[VIRT_GIC_REDIST].base);
785         if (nb_redist_regions == 2) {
786             sysbus_mmio_map(gicbusdev, 2,
787                             vms->memmap[VIRT_HIGH_GIC_REDIST2].base);
788         }
789     } else {
790         sysbus_mmio_map(gicbusdev, 1, vms->memmap[VIRT_GIC_CPU].base);
791         if (vms->virt) {
792             sysbus_mmio_map(gicbusdev, 2, vms->memmap[VIRT_GIC_HYP].base);
793             sysbus_mmio_map(gicbusdev, 3, vms->memmap[VIRT_GIC_VCPU].base);
794         }
795     }
796 
797     /* Wire the outputs from each CPU's generic timer and the GICv3
798      * maintenance interrupt signal to the appropriate GIC PPI inputs,
799      * and the GIC's IRQ/FIQ/VIRQ/VFIQ interrupt outputs to the CPU's inputs.
800      */
801     for (i = 0; i < smp_cpus; i++) {
802         DeviceState *cpudev = DEVICE(qemu_get_cpu(i));
803         int ppibase = NUM_IRQS + i * GIC_INTERNAL + GIC_NR_SGIS;
804         int irq;
805         /* Mapping from the output timer irq lines from the CPU to the
806          * GIC PPI inputs we use for the virt board.
807          */
808         const int timer_irq[] = {
809             [GTIMER_PHYS] = ARCH_TIMER_NS_EL1_IRQ,
810             [GTIMER_VIRT] = ARCH_TIMER_VIRT_IRQ,
811             [GTIMER_HYP]  = ARCH_TIMER_NS_EL2_IRQ,
812             [GTIMER_SEC]  = ARCH_TIMER_S_EL1_IRQ,
813         };
814 
815         for (irq = 0; irq < ARRAY_SIZE(timer_irq); irq++) {
816             qdev_connect_gpio_out(cpudev, irq,
817                                   qdev_get_gpio_in(vms->gic,
818                                                    ppibase + timer_irq[irq]));
819         }
820 
821         if (vms->gic_version != VIRT_GIC_VERSION_2) {
822             qemu_irq irq = qdev_get_gpio_in(vms->gic,
823                                             ppibase + ARCH_GIC_MAINT_IRQ);
824             qdev_connect_gpio_out_named(cpudev, "gicv3-maintenance-interrupt",
825                                         0, irq);
826         } else if (vms->virt) {
827             qemu_irq irq = qdev_get_gpio_in(vms->gic,
828                                             ppibase + ARCH_GIC_MAINT_IRQ);
829             sysbus_connect_irq(gicbusdev, i + 4 * smp_cpus, irq);
830         }
831 
832         qdev_connect_gpio_out_named(cpudev, "pmu-interrupt", 0,
833                                     qdev_get_gpio_in(vms->gic, ppibase
834                                                      + VIRTUAL_PMU_IRQ));
835 
836         sysbus_connect_irq(gicbusdev, i, qdev_get_gpio_in(cpudev, ARM_CPU_IRQ));
837         sysbus_connect_irq(gicbusdev, i + smp_cpus,
838                            qdev_get_gpio_in(cpudev, ARM_CPU_FIQ));
839         sysbus_connect_irq(gicbusdev, i + 2 * smp_cpus,
840                            qdev_get_gpio_in(cpudev, ARM_CPU_VIRQ));
841         sysbus_connect_irq(gicbusdev, i + 3 * smp_cpus,
842                            qdev_get_gpio_in(cpudev, ARM_CPU_VFIQ));
843     }
844 
845     fdt_add_gic_node(vms);
846 
847     if (vms->gic_version != VIRT_GIC_VERSION_2 && vms->its) {
848         create_its(vms);
849     } else if (vms->gic_version == VIRT_GIC_VERSION_2) {
850         create_v2m(vms);
851     }
852 }
853 
854 static void create_uart(const VirtMachineState *vms, int uart,
855                         MemoryRegion *mem, Chardev *chr)
856 {
857     char *nodename;
858     hwaddr base = vms->memmap[uart].base;
859     hwaddr size = vms->memmap[uart].size;
860     int irq = vms->irqmap[uart];
861     const char compat[] = "arm,pl011\0arm,primecell";
862     const char clocknames[] = "uartclk\0apb_pclk";
863     DeviceState *dev = qdev_new(TYPE_PL011);
864     SysBusDevice *s = SYS_BUS_DEVICE(dev);
865     MachineState *ms = MACHINE(vms);
866 
867     qdev_prop_set_chr(dev, "chardev", chr);
868     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
869     memory_region_add_subregion(mem, base,
870                                 sysbus_mmio_get_region(s, 0));
871     sysbus_connect_irq(s, 0, qdev_get_gpio_in(vms->gic, irq));
872 
873     nodename = g_strdup_printf("/pl011@%" PRIx64, base);
874     qemu_fdt_add_subnode(ms->fdt, nodename);
875     /* Note that we can't use setprop_string because of the embedded NUL */
876     qemu_fdt_setprop(ms->fdt, nodename, "compatible",
877                          compat, sizeof(compat));
878     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
879                                      2, base, 2, size);
880     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
881                                GIC_FDT_IRQ_TYPE_SPI, irq,
882                                GIC_FDT_IRQ_FLAGS_LEVEL_HI);
883     qemu_fdt_setprop_cells(ms->fdt, nodename, "clocks",
884                                vms->clock_phandle, vms->clock_phandle);
885     qemu_fdt_setprop(ms->fdt, nodename, "clock-names",
886                          clocknames, sizeof(clocknames));
887 
888     if (uart == VIRT_UART) {
889         qemu_fdt_setprop_string(ms->fdt, "/chosen", "stdout-path", nodename);
890     } else {
891         /* Mark as not usable by the normal world */
892         qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
893         qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
894 
895         qemu_fdt_setprop_string(ms->fdt, "/secure-chosen", "stdout-path",
896                                 nodename);
897     }
898 
899     g_free(nodename);
900 }
901 
902 static void create_rtc(const VirtMachineState *vms)
903 {
904     char *nodename;
905     hwaddr base = vms->memmap[VIRT_RTC].base;
906     hwaddr size = vms->memmap[VIRT_RTC].size;
907     int irq = vms->irqmap[VIRT_RTC];
908     const char compat[] = "arm,pl031\0arm,primecell";
909     MachineState *ms = MACHINE(vms);
910 
911     sysbus_create_simple("pl031", base, qdev_get_gpio_in(vms->gic, irq));
912 
913     nodename = g_strdup_printf("/pl031@%" PRIx64, base);
914     qemu_fdt_add_subnode(ms->fdt, nodename);
915     qemu_fdt_setprop(ms->fdt, nodename, "compatible", compat, sizeof(compat));
916     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
917                                  2, base, 2, size);
918     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
919                            GIC_FDT_IRQ_TYPE_SPI, irq,
920                            GIC_FDT_IRQ_FLAGS_LEVEL_HI);
921     qemu_fdt_setprop_cell(ms->fdt, nodename, "clocks", vms->clock_phandle);
922     qemu_fdt_setprop_string(ms->fdt, nodename, "clock-names", "apb_pclk");
923     g_free(nodename);
924 }
925 
926 static DeviceState *gpio_key_dev;
927 static void virt_powerdown_req(Notifier *n, void *opaque)
928 {
929     VirtMachineState *s = container_of(n, VirtMachineState, powerdown_notifier);
930 
931     if (s->acpi_dev) {
932         acpi_send_event(s->acpi_dev, ACPI_POWER_DOWN_STATUS);
933     } else {
934         /* use gpio Pin 3 for power button event */
935         qemu_set_irq(qdev_get_gpio_in(gpio_key_dev, 0), 1);
936     }
937 }
938 
939 static void create_gpio_keys(char *fdt, DeviceState *pl061_dev,
940                              uint32_t phandle)
941 {
942     gpio_key_dev = sysbus_create_simple("gpio-key", -1,
943                                         qdev_get_gpio_in(pl061_dev, 3));
944 
945     qemu_fdt_add_subnode(fdt, "/gpio-keys");
946     qemu_fdt_setprop_string(fdt, "/gpio-keys", "compatible", "gpio-keys");
947 
948     qemu_fdt_add_subnode(fdt, "/gpio-keys/poweroff");
949     qemu_fdt_setprop_string(fdt, "/gpio-keys/poweroff",
950                             "label", "GPIO Key Poweroff");
951     qemu_fdt_setprop_cell(fdt, "/gpio-keys/poweroff", "linux,code",
952                           KEY_POWER);
953     qemu_fdt_setprop_cells(fdt, "/gpio-keys/poweroff",
954                            "gpios", phandle, 3, 0);
955 }
956 
957 #define SECURE_GPIO_POWEROFF 0
958 #define SECURE_GPIO_RESET    1
959 
960 static void create_secure_gpio_pwr(char *fdt, DeviceState *pl061_dev,
961                                    uint32_t phandle)
962 {
963     DeviceState *gpio_pwr_dev;
964 
965     /* gpio-pwr */
966     gpio_pwr_dev = sysbus_create_simple("gpio-pwr", -1, NULL);
967 
968     /* connect secure pl061 to gpio-pwr */
969     qdev_connect_gpio_out(pl061_dev, SECURE_GPIO_RESET,
970                           qdev_get_gpio_in_named(gpio_pwr_dev, "reset", 0));
971     qdev_connect_gpio_out(pl061_dev, SECURE_GPIO_POWEROFF,
972                           qdev_get_gpio_in_named(gpio_pwr_dev, "shutdown", 0));
973 
974     qemu_fdt_add_subnode(fdt, "/gpio-poweroff");
975     qemu_fdt_setprop_string(fdt, "/gpio-poweroff", "compatible",
976                             "gpio-poweroff");
977     qemu_fdt_setprop_cells(fdt, "/gpio-poweroff",
978                            "gpios", phandle, SECURE_GPIO_POWEROFF, 0);
979     qemu_fdt_setprop_string(fdt, "/gpio-poweroff", "status", "disabled");
980     qemu_fdt_setprop_string(fdt, "/gpio-poweroff", "secure-status",
981                             "okay");
982 
983     qemu_fdt_add_subnode(fdt, "/gpio-restart");
984     qemu_fdt_setprop_string(fdt, "/gpio-restart", "compatible",
985                             "gpio-restart");
986     qemu_fdt_setprop_cells(fdt, "/gpio-restart",
987                            "gpios", phandle, SECURE_GPIO_RESET, 0);
988     qemu_fdt_setprop_string(fdt, "/gpio-restart", "status", "disabled");
989     qemu_fdt_setprop_string(fdt, "/gpio-restart", "secure-status",
990                             "okay");
991 }
992 
993 static void create_gpio_devices(const VirtMachineState *vms, int gpio,
994                                 MemoryRegion *mem)
995 {
996     char *nodename;
997     DeviceState *pl061_dev;
998     hwaddr base = vms->memmap[gpio].base;
999     hwaddr size = vms->memmap[gpio].size;
1000     int irq = vms->irqmap[gpio];
1001     const char compat[] = "arm,pl061\0arm,primecell";
1002     SysBusDevice *s;
1003     MachineState *ms = MACHINE(vms);
1004 
1005     pl061_dev = qdev_new("pl061");
1006     /* Pull lines down to 0 if not driven by the PL061 */
1007     qdev_prop_set_uint32(pl061_dev, "pullups", 0);
1008     qdev_prop_set_uint32(pl061_dev, "pulldowns", 0xff);
1009     s = SYS_BUS_DEVICE(pl061_dev);
1010     sysbus_realize_and_unref(s, &error_fatal);
1011     memory_region_add_subregion(mem, base, sysbus_mmio_get_region(s, 0));
1012     sysbus_connect_irq(s, 0, qdev_get_gpio_in(vms->gic, irq));
1013 
1014     uint32_t phandle = qemu_fdt_alloc_phandle(ms->fdt);
1015     nodename = g_strdup_printf("/pl061@%" PRIx64, base);
1016     qemu_fdt_add_subnode(ms->fdt, nodename);
1017     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1018                                  2, base, 2, size);
1019     qemu_fdt_setprop(ms->fdt, nodename, "compatible", compat, sizeof(compat));
1020     qemu_fdt_setprop_cell(ms->fdt, nodename, "#gpio-cells", 2);
1021     qemu_fdt_setprop(ms->fdt, nodename, "gpio-controller", NULL, 0);
1022     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
1023                            GIC_FDT_IRQ_TYPE_SPI, irq,
1024                            GIC_FDT_IRQ_FLAGS_LEVEL_HI);
1025     qemu_fdt_setprop_cell(ms->fdt, nodename, "clocks", vms->clock_phandle);
1026     qemu_fdt_setprop_string(ms->fdt, nodename, "clock-names", "apb_pclk");
1027     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", phandle);
1028 
1029     if (gpio != VIRT_GPIO) {
1030         /* Mark as not usable by the normal world */
1031         qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
1032         qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
1033     }
1034     g_free(nodename);
1035 
1036     /* Child gpio devices */
1037     if (gpio == VIRT_GPIO) {
1038         create_gpio_keys(ms->fdt, pl061_dev, phandle);
1039     } else {
1040         create_secure_gpio_pwr(ms->fdt, pl061_dev, phandle);
1041     }
1042 }
1043 
1044 static void create_virtio_devices(const VirtMachineState *vms)
1045 {
1046     int i;
1047     hwaddr size = vms->memmap[VIRT_MMIO].size;
1048     MachineState *ms = MACHINE(vms);
1049 
1050     /* We create the transports in forwards order. Since qbus_realize()
1051      * prepends (not appends) new child buses, the incrementing loop below will
1052      * create a list of virtio-mmio buses with decreasing base addresses.
1053      *
1054      * When a -device option is processed from the command line,
1055      * qbus_find_recursive() picks the next free virtio-mmio bus in forwards
1056      * order. The upshot is that -device options in increasing command line
1057      * order are mapped to virtio-mmio buses with decreasing base addresses.
1058      *
1059      * When this code was originally written, that arrangement ensured that the
1060      * guest Linux kernel would give the lowest "name" (/dev/vda, eth0, etc) to
1061      * the first -device on the command line. (The end-to-end order is a
1062      * function of this loop, qbus_realize(), qbus_find_recursive(), and the
1063      * guest kernel's name-to-address assignment strategy.)
1064      *
1065      * Meanwhile, the kernel's traversal seems to have been reversed; see eg.
1066      * the message, if not necessarily the code, of commit 70161ff336.
1067      * Therefore the loop now establishes the inverse of the original intent.
1068      *
1069      * Unfortunately, we can't counteract the kernel change by reversing the
1070      * loop; it would break existing command lines.
1071      *
1072      * In any case, the kernel makes no guarantee about the stability of
1073      * enumeration order of virtio devices (as demonstrated by it changing
1074      * between kernel versions). For reliable and stable identification
1075      * of disks users must use UUIDs or similar mechanisms.
1076      */
1077     for (i = 0; i < NUM_VIRTIO_TRANSPORTS; i++) {
1078         int irq = vms->irqmap[VIRT_MMIO] + i;
1079         hwaddr base = vms->memmap[VIRT_MMIO].base + i * size;
1080 
1081         sysbus_create_simple("virtio-mmio", base,
1082                              qdev_get_gpio_in(vms->gic, irq));
1083     }
1084 
1085     /* We add dtb nodes in reverse order so that they appear in the finished
1086      * device tree lowest address first.
1087      *
1088      * Note that this mapping is independent of the loop above. The previous
1089      * loop influences virtio device to virtio transport assignment, whereas
1090      * this loop controls how virtio transports are laid out in the dtb.
1091      */
1092     for (i = NUM_VIRTIO_TRANSPORTS - 1; i >= 0; i--) {
1093         char *nodename;
1094         int irq = vms->irqmap[VIRT_MMIO] + i;
1095         hwaddr base = vms->memmap[VIRT_MMIO].base + i * size;
1096 
1097         nodename = g_strdup_printf("/virtio_mmio@%" PRIx64, base);
1098         qemu_fdt_add_subnode(ms->fdt, nodename);
1099         qemu_fdt_setprop_string(ms->fdt, nodename,
1100                                 "compatible", "virtio,mmio");
1101         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1102                                      2, base, 2, size);
1103         qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
1104                                GIC_FDT_IRQ_TYPE_SPI, irq,
1105                                GIC_FDT_IRQ_FLAGS_EDGE_LO_HI);
1106         qemu_fdt_setprop(ms->fdt, nodename, "dma-coherent", NULL, 0);
1107         g_free(nodename);
1108     }
1109 }
1110 
1111 #define VIRT_FLASH_SECTOR_SIZE (256 * KiB)
1112 
1113 static PFlashCFI01 *virt_flash_create1(VirtMachineState *vms,
1114                                         const char *name,
1115                                         const char *alias_prop_name)
1116 {
1117     /*
1118      * Create a single flash device.  We use the same parameters as
1119      * the flash devices on the Versatile Express board.
1120      */
1121     DeviceState *dev = qdev_new(TYPE_PFLASH_CFI01);
1122 
1123     qdev_prop_set_uint64(dev, "sector-length", VIRT_FLASH_SECTOR_SIZE);
1124     qdev_prop_set_uint8(dev, "width", 4);
1125     qdev_prop_set_uint8(dev, "device-width", 2);
1126     qdev_prop_set_bit(dev, "big-endian", false);
1127     qdev_prop_set_uint16(dev, "id0", 0x89);
1128     qdev_prop_set_uint16(dev, "id1", 0x18);
1129     qdev_prop_set_uint16(dev, "id2", 0x00);
1130     qdev_prop_set_uint16(dev, "id3", 0x00);
1131     qdev_prop_set_string(dev, "name", name);
1132     object_property_add_child(OBJECT(vms), name, OBJECT(dev));
1133     object_property_add_alias(OBJECT(vms), alias_prop_name,
1134                               OBJECT(dev), "drive");
1135     return PFLASH_CFI01(dev);
1136 }
1137 
1138 static void virt_flash_create(VirtMachineState *vms)
1139 {
1140     vms->flash[0] = virt_flash_create1(vms, "virt.flash0", "pflash0");
1141     vms->flash[1] = virt_flash_create1(vms, "virt.flash1", "pflash1");
1142 }
1143 
1144 static void virt_flash_map1(PFlashCFI01 *flash,
1145                             hwaddr base, hwaddr size,
1146                             MemoryRegion *sysmem)
1147 {
1148     DeviceState *dev = DEVICE(flash);
1149 
1150     assert(QEMU_IS_ALIGNED(size, VIRT_FLASH_SECTOR_SIZE));
1151     assert(size / VIRT_FLASH_SECTOR_SIZE <= UINT32_MAX);
1152     qdev_prop_set_uint32(dev, "num-blocks", size / VIRT_FLASH_SECTOR_SIZE);
1153     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1154 
1155     memory_region_add_subregion(sysmem, base,
1156                                 sysbus_mmio_get_region(SYS_BUS_DEVICE(dev),
1157                                                        0));
1158 }
1159 
1160 static void virt_flash_map(VirtMachineState *vms,
1161                            MemoryRegion *sysmem,
1162                            MemoryRegion *secure_sysmem)
1163 {
1164     /*
1165      * Map two flash devices to fill the VIRT_FLASH space in the memmap.
1166      * sysmem is the system memory space. secure_sysmem is the secure view
1167      * of the system, and the first flash device should be made visible only
1168      * there. The second flash device is visible to both secure and nonsecure.
1169      * If sysmem == secure_sysmem this means there is no separate Secure
1170      * address space and both flash devices are generally visible.
1171      */
1172     hwaddr flashsize = vms->memmap[VIRT_FLASH].size / 2;
1173     hwaddr flashbase = vms->memmap[VIRT_FLASH].base;
1174 
1175     virt_flash_map1(vms->flash[0], flashbase, flashsize,
1176                     secure_sysmem);
1177     virt_flash_map1(vms->flash[1], flashbase + flashsize, flashsize,
1178                     sysmem);
1179 }
1180 
1181 static void virt_flash_fdt(VirtMachineState *vms,
1182                            MemoryRegion *sysmem,
1183                            MemoryRegion *secure_sysmem)
1184 {
1185     hwaddr flashsize = vms->memmap[VIRT_FLASH].size / 2;
1186     hwaddr flashbase = vms->memmap[VIRT_FLASH].base;
1187     MachineState *ms = MACHINE(vms);
1188     char *nodename;
1189 
1190     if (sysmem == secure_sysmem) {
1191         /* Report both flash devices as a single node in the DT */
1192         nodename = g_strdup_printf("/flash@%" PRIx64, flashbase);
1193         qemu_fdt_add_subnode(ms->fdt, nodename);
1194         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible", "cfi-flash");
1195         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1196                                      2, flashbase, 2, flashsize,
1197                                      2, flashbase + flashsize, 2, flashsize);
1198         qemu_fdt_setprop_cell(ms->fdt, nodename, "bank-width", 4);
1199         g_free(nodename);
1200     } else {
1201         /*
1202          * Report the devices as separate nodes so we can mark one as
1203          * only visible to the secure world.
1204          */
1205         nodename = g_strdup_printf("/secflash@%" PRIx64, flashbase);
1206         qemu_fdt_add_subnode(ms->fdt, nodename);
1207         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible", "cfi-flash");
1208         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1209                                      2, flashbase, 2, flashsize);
1210         qemu_fdt_setprop_cell(ms->fdt, nodename, "bank-width", 4);
1211         qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
1212         qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
1213         g_free(nodename);
1214 
1215         nodename = g_strdup_printf("/flash@%" PRIx64, flashbase + flashsize);
1216         qemu_fdt_add_subnode(ms->fdt, nodename);
1217         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible", "cfi-flash");
1218         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1219                                      2, flashbase + flashsize, 2, flashsize);
1220         qemu_fdt_setprop_cell(ms->fdt, nodename, "bank-width", 4);
1221         g_free(nodename);
1222     }
1223 }
1224 
1225 static bool virt_firmware_init(VirtMachineState *vms,
1226                                MemoryRegion *sysmem,
1227                                MemoryRegion *secure_sysmem)
1228 {
1229     int i;
1230     const char *bios_name;
1231     BlockBackend *pflash_blk0;
1232 
1233     /* Map legacy -drive if=pflash to machine properties */
1234     for (i = 0; i < ARRAY_SIZE(vms->flash); i++) {
1235         pflash_cfi01_legacy_drive(vms->flash[i],
1236                                   drive_get(IF_PFLASH, 0, i));
1237     }
1238 
1239     virt_flash_map(vms, sysmem, secure_sysmem);
1240 
1241     pflash_blk0 = pflash_cfi01_get_blk(vms->flash[0]);
1242 
1243     bios_name = MACHINE(vms)->firmware;
1244     if (bios_name) {
1245         char *fname;
1246         MemoryRegion *mr;
1247         int image_size;
1248 
1249         if (pflash_blk0) {
1250             error_report("The contents of the first flash device may be "
1251                          "specified with -bios or with -drive if=pflash... "
1252                          "but you cannot use both options at once");
1253             exit(1);
1254         }
1255 
1256         /* Fall back to -bios */
1257 
1258         fname = qemu_find_file(QEMU_FILE_TYPE_BIOS, bios_name);
1259         if (!fname) {
1260             error_report("Could not find ROM image '%s'", bios_name);
1261             exit(1);
1262         }
1263         mr = sysbus_mmio_get_region(SYS_BUS_DEVICE(vms->flash[0]), 0);
1264         image_size = load_image_mr(fname, mr);
1265         g_free(fname);
1266         if (image_size < 0) {
1267             error_report("Could not load ROM image '%s'", bios_name);
1268             exit(1);
1269         }
1270     }
1271 
1272     return pflash_blk0 || bios_name;
1273 }
1274 
1275 static FWCfgState *create_fw_cfg(const VirtMachineState *vms, AddressSpace *as)
1276 {
1277     MachineState *ms = MACHINE(vms);
1278     hwaddr base = vms->memmap[VIRT_FW_CFG].base;
1279     hwaddr size = vms->memmap[VIRT_FW_CFG].size;
1280     FWCfgState *fw_cfg;
1281     char *nodename;
1282 
1283     fw_cfg = fw_cfg_init_mem_wide(base + 8, base, 8, base + 16, as);
1284     fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, (uint16_t)ms->smp.cpus);
1285 
1286     nodename = g_strdup_printf("/fw-cfg@%" PRIx64, base);
1287     qemu_fdt_add_subnode(ms->fdt, nodename);
1288     qemu_fdt_setprop_string(ms->fdt, nodename,
1289                             "compatible", "qemu,fw-cfg-mmio");
1290     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1291                                  2, base, 2, size);
1292     qemu_fdt_setprop(ms->fdt, nodename, "dma-coherent", NULL, 0);
1293     g_free(nodename);
1294     return fw_cfg;
1295 }
1296 
1297 static void create_pcie_irq_map(const MachineState *ms,
1298                                 uint32_t gic_phandle,
1299                                 int first_irq, const char *nodename)
1300 {
1301     int devfn, pin;
1302     uint32_t full_irq_map[4 * 4 * 10] = { 0 };
1303     uint32_t *irq_map = full_irq_map;
1304 
1305     for (devfn = 0; devfn <= 0x18; devfn += 0x8) {
1306         for (pin = 0; pin < 4; pin++) {
1307             int irq_type = GIC_FDT_IRQ_TYPE_SPI;
1308             int irq_nr = first_irq + ((pin + PCI_SLOT(devfn)) % PCI_NUM_PINS);
1309             int irq_level = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
1310             int i;
1311 
1312             uint32_t map[] = {
1313                 devfn << 8, 0, 0,                           /* devfn */
1314                 pin + 1,                                    /* PCI pin */
1315                 gic_phandle, 0, 0, irq_type, irq_nr, irq_level }; /* GIC irq */
1316 
1317             /* Convert map to big endian */
1318             for (i = 0; i < 10; i++) {
1319                 irq_map[i] = cpu_to_be32(map[i]);
1320             }
1321             irq_map += 10;
1322         }
1323     }
1324 
1325     qemu_fdt_setprop(ms->fdt, nodename, "interrupt-map",
1326                      full_irq_map, sizeof(full_irq_map));
1327 
1328     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupt-map-mask",
1329                            cpu_to_be16(PCI_DEVFN(3, 0)), /* Slot 3 */
1330                            0, 0,
1331                            0x7           /* PCI irq */);
1332 }
1333 
1334 static void create_smmu(const VirtMachineState *vms,
1335                         PCIBus *bus)
1336 {
1337     char *node;
1338     const char compat[] = "arm,smmu-v3";
1339     int irq =  vms->irqmap[VIRT_SMMU];
1340     int i;
1341     hwaddr base = vms->memmap[VIRT_SMMU].base;
1342     hwaddr size = vms->memmap[VIRT_SMMU].size;
1343     const char irq_names[] = "eventq\0priq\0cmdq-sync\0gerror";
1344     DeviceState *dev;
1345     MachineState *ms = MACHINE(vms);
1346 
1347     if (vms->iommu != VIRT_IOMMU_SMMUV3 || !vms->iommu_phandle) {
1348         return;
1349     }
1350 
1351     dev = qdev_new(TYPE_ARM_SMMUV3);
1352 
1353     object_property_set_link(OBJECT(dev), "primary-bus", OBJECT(bus),
1354                              &error_abort);
1355     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1356     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, base);
1357     for (i = 0; i < NUM_SMMU_IRQS; i++) {
1358         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
1359                            qdev_get_gpio_in(vms->gic, irq + i));
1360     }
1361 
1362     node = g_strdup_printf("/smmuv3@%" PRIx64, base);
1363     qemu_fdt_add_subnode(ms->fdt, node);
1364     qemu_fdt_setprop(ms->fdt, node, "compatible", compat, sizeof(compat));
1365     qemu_fdt_setprop_sized_cells(ms->fdt, node, "reg", 2, base, 2, size);
1366 
1367     qemu_fdt_setprop_cells(ms->fdt, node, "interrupts",
1368             GIC_FDT_IRQ_TYPE_SPI, irq    , GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1369             GIC_FDT_IRQ_TYPE_SPI, irq + 1, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1370             GIC_FDT_IRQ_TYPE_SPI, irq + 2, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1371             GIC_FDT_IRQ_TYPE_SPI, irq + 3, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI);
1372 
1373     qemu_fdt_setprop(ms->fdt, node, "interrupt-names", irq_names,
1374                      sizeof(irq_names));
1375 
1376     qemu_fdt_setprop(ms->fdt, node, "dma-coherent", NULL, 0);
1377 
1378     qemu_fdt_setprop_cell(ms->fdt, node, "#iommu-cells", 1);
1379 
1380     qemu_fdt_setprop_cell(ms->fdt, node, "phandle", vms->iommu_phandle);
1381     g_free(node);
1382 }
1383 
1384 static void create_virtio_iommu_dt_bindings(VirtMachineState *vms)
1385 {
1386     const char compat[] = "virtio,pci-iommu\0pci1af4,1057";
1387     uint16_t bdf = vms->virtio_iommu_bdf;
1388     MachineState *ms = MACHINE(vms);
1389     char *node;
1390 
1391     vms->iommu_phandle = qemu_fdt_alloc_phandle(ms->fdt);
1392 
1393     node = g_strdup_printf("%s/virtio_iommu@%x,%x", vms->pciehb_nodename,
1394                            PCI_SLOT(bdf), PCI_FUNC(bdf));
1395     qemu_fdt_add_subnode(ms->fdt, node);
1396     qemu_fdt_setprop(ms->fdt, node, "compatible", compat, sizeof(compat));
1397     qemu_fdt_setprop_sized_cells(ms->fdt, node, "reg",
1398                                  1, bdf << 8, 1, 0, 1, 0,
1399                                  1, 0, 1, 0);
1400 
1401     qemu_fdt_setprop_cell(ms->fdt, node, "#iommu-cells", 1);
1402     qemu_fdt_setprop_cell(ms->fdt, node, "phandle", vms->iommu_phandle);
1403     g_free(node);
1404 
1405     qemu_fdt_setprop_cells(ms->fdt, vms->pciehb_nodename, "iommu-map",
1406                            0x0, vms->iommu_phandle, 0x0, bdf,
1407                            bdf + 1, vms->iommu_phandle, bdf + 1, 0xffff - bdf);
1408 }
1409 
1410 static void create_pcie(VirtMachineState *vms)
1411 {
1412     hwaddr base_mmio = vms->memmap[VIRT_PCIE_MMIO].base;
1413     hwaddr size_mmio = vms->memmap[VIRT_PCIE_MMIO].size;
1414     hwaddr base_mmio_high = vms->memmap[VIRT_HIGH_PCIE_MMIO].base;
1415     hwaddr size_mmio_high = vms->memmap[VIRT_HIGH_PCIE_MMIO].size;
1416     hwaddr base_pio = vms->memmap[VIRT_PCIE_PIO].base;
1417     hwaddr size_pio = vms->memmap[VIRT_PCIE_PIO].size;
1418     hwaddr base_ecam, size_ecam;
1419     hwaddr base = base_mmio;
1420     int nr_pcie_buses;
1421     int irq = vms->irqmap[VIRT_PCIE];
1422     MemoryRegion *mmio_alias;
1423     MemoryRegion *mmio_reg;
1424     MemoryRegion *ecam_alias;
1425     MemoryRegion *ecam_reg;
1426     DeviceState *dev;
1427     char *nodename;
1428     int i, ecam_id;
1429     PCIHostState *pci;
1430     MachineState *ms = MACHINE(vms);
1431     MachineClass *mc = MACHINE_GET_CLASS(ms);
1432 
1433     dev = qdev_new(TYPE_GPEX_HOST);
1434     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1435 
1436     ecam_id = VIRT_ECAM_ID(vms->highmem_ecam);
1437     base_ecam = vms->memmap[ecam_id].base;
1438     size_ecam = vms->memmap[ecam_id].size;
1439     nr_pcie_buses = size_ecam / PCIE_MMCFG_SIZE_MIN;
1440     /* Map only the first size_ecam bytes of ECAM space */
1441     ecam_alias = g_new0(MemoryRegion, 1);
1442     ecam_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 0);
1443     memory_region_init_alias(ecam_alias, OBJECT(dev), "pcie-ecam",
1444                              ecam_reg, 0, size_ecam);
1445     memory_region_add_subregion(get_system_memory(), base_ecam, ecam_alias);
1446 
1447     /* Map the MMIO window into system address space so as to expose
1448      * the section of PCI MMIO space which starts at the same base address
1449      * (ie 1:1 mapping for that part of PCI MMIO space visible through
1450      * the window).
1451      */
1452     mmio_alias = g_new0(MemoryRegion, 1);
1453     mmio_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 1);
1454     memory_region_init_alias(mmio_alias, OBJECT(dev), "pcie-mmio",
1455                              mmio_reg, base_mmio, size_mmio);
1456     memory_region_add_subregion(get_system_memory(), base_mmio, mmio_alias);
1457 
1458     if (vms->highmem_mmio) {
1459         /* Map high MMIO space */
1460         MemoryRegion *high_mmio_alias = g_new0(MemoryRegion, 1);
1461 
1462         memory_region_init_alias(high_mmio_alias, OBJECT(dev), "pcie-mmio-high",
1463                                  mmio_reg, base_mmio_high, size_mmio_high);
1464         memory_region_add_subregion(get_system_memory(), base_mmio_high,
1465                                     high_mmio_alias);
1466     }
1467 
1468     /* Map IO port space */
1469     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 2, base_pio);
1470 
1471     for (i = 0; i < GPEX_NUM_IRQS; i++) {
1472         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
1473                            qdev_get_gpio_in(vms->gic, irq + i));
1474         gpex_set_irq_num(GPEX_HOST(dev), i, irq + i);
1475     }
1476 
1477     pci = PCI_HOST_BRIDGE(dev);
1478     pci->bypass_iommu = vms->default_bus_bypass_iommu;
1479     vms->bus = pci->bus;
1480     if (vms->bus) {
1481         for (i = 0; i < nb_nics; i++) {
1482             pci_nic_init_nofail(&nd_table[i], pci->bus, mc->default_nic, NULL);
1483         }
1484     }
1485 
1486     nodename = vms->pciehb_nodename = g_strdup_printf("/pcie@%" PRIx64, base);
1487     qemu_fdt_add_subnode(ms->fdt, nodename);
1488     qemu_fdt_setprop_string(ms->fdt, nodename,
1489                             "compatible", "pci-host-ecam-generic");
1490     qemu_fdt_setprop_string(ms->fdt, nodename, "device_type", "pci");
1491     qemu_fdt_setprop_cell(ms->fdt, nodename, "#address-cells", 3);
1492     qemu_fdt_setprop_cell(ms->fdt, nodename, "#size-cells", 2);
1493     qemu_fdt_setprop_cell(ms->fdt, nodename, "linux,pci-domain", 0);
1494     qemu_fdt_setprop_cells(ms->fdt, nodename, "bus-range", 0,
1495                            nr_pcie_buses - 1);
1496     qemu_fdt_setprop(ms->fdt, nodename, "dma-coherent", NULL, 0);
1497 
1498     if (vms->msi_phandle) {
1499         qemu_fdt_setprop_cells(ms->fdt, nodename, "msi-map",
1500                                0, vms->msi_phandle, 0, 0x10000);
1501     }
1502 
1503     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1504                                  2, base_ecam, 2, size_ecam);
1505 
1506     if (vms->highmem_mmio) {
1507         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "ranges",
1508                                      1, FDT_PCI_RANGE_IOPORT, 2, 0,
1509                                      2, base_pio, 2, size_pio,
1510                                      1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
1511                                      2, base_mmio, 2, size_mmio,
1512                                      1, FDT_PCI_RANGE_MMIO_64BIT,
1513                                      2, base_mmio_high,
1514                                      2, base_mmio_high, 2, size_mmio_high);
1515     } else {
1516         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "ranges",
1517                                      1, FDT_PCI_RANGE_IOPORT, 2, 0,
1518                                      2, base_pio, 2, size_pio,
1519                                      1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
1520                                      2, base_mmio, 2, size_mmio);
1521     }
1522 
1523     qemu_fdt_setprop_cell(ms->fdt, nodename, "#interrupt-cells", 1);
1524     create_pcie_irq_map(ms, vms->gic_phandle, irq, nodename);
1525 
1526     if (vms->iommu) {
1527         vms->iommu_phandle = qemu_fdt_alloc_phandle(ms->fdt);
1528 
1529         switch (vms->iommu) {
1530         case VIRT_IOMMU_SMMUV3:
1531             create_smmu(vms, vms->bus);
1532             qemu_fdt_setprop_cells(ms->fdt, nodename, "iommu-map",
1533                                    0x0, vms->iommu_phandle, 0x0, 0x10000);
1534             break;
1535         default:
1536             g_assert_not_reached();
1537         }
1538     }
1539 }
1540 
1541 static void create_platform_bus(VirtMachineState *vms)
1542 {
1543     DeviceState *dev;
1544     SysBusDevice *s;
1545     int i;
1546     MemoryRegion *sysmem = get_system_memory();
1547 
1548     dev = qdev_new(TYPE_PLATFORM_BUS_DEVICE);
1549     dev->id = g_strdup(TYPE_PLATFORM_BUS_DEVICE);
1550     qdev_prop_set_uint32(dev, "num_irqs", PLATFORM_BUS_NUM_IRQS);
1551     qdev_prop_set_uint32(dev, "mmio_size", vms->memmap[VIRT_PLATFORM_BUS].size);
1552     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1553     vms->platform_bus_dev = dev;
1554 
1555     s = SYS_BUS_DEVICE(dev);
1556     for (i = 0; i < PLATFORM_BUS_NUM_IRQS; i++) {
1557         int irq = vms->irqmap[VIRT_PLATFORM_BUS] + i;
1558         sysbus_connect_irq(s, i, qdev_get_gpio_in(vms->gic, irq));
1559     }
1560 
1561     memory_region_add_subregion(sysmem,
1562                                 vms->memmap[VIRT_PLATFORM_BUS].base,
1563                                 sysbus_mmio_get_region(s, 0));
1564 }
1565 
1566 static void create_tag_ram(MemoryRegion *tag_sysmem,
1567                            hwaddr base, hwaddr size,
1568                            const char *name)
1569 {
1570     MemoryRegion *tagram = g_new(MemoryRegion, 1);
1571 
1572     memory_region_init_ram(tagram, NULL, name, size / 32, &error_fatal);
1573     memory_region_add_subregion(tag_sysmem, base / 32, tagram);
1574 }
1575 
1576 static void create_secure_ram(VirtMachineState *vms,
1577                               MemoryRegion *secure_sysmem,
1578                               MemoryRegion *secure_tag_sysmem)
1579 {
1580     MemoryRegion *secram = g_new(MemoryRegion, 1);
1581     char *nodename;
1582     hwaddr base = vms->memmap[VIRT_SECURE_MEM].base;
1583     hwaddr size = vms->memmap[VIRT_SECURE_MEM].size;
1584     MachineState *ms = MACHINE(vms);
1585 
1586     memory_region_init_ram(secram, NULL, "virt.secure-ram", size,
1587                            &error_fatal);
1588     memory_region_add_subregion(secure_sysmem, base, secram);
1589 
1590     nodename = g_strdup_printf("/secram@%" PRIx64, base);
1591     qemu_fdt_add_subnode(ms->fdt, nodename);
1592     qemu_fdt_setprop_string(ms->fdt, nodename, "device_type", "memory");
1593     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg", 2, base, 2, size);
1594     qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
1595     qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
1596 
1597     if (secure_tag_sysmem) {
1598         create_tag_ram(secure_tag_sysmem, base, size, "mach-virt.secure-tag");
1599     }
1600 
1601     g_free(nodename);
1602 }
1603 
1604 static void *machvirt_dtb(const struct arm_boot_info *binfo, int *fdt_size)
1605 {
1606     const VirtMachineState *board = container_of(binfo, VirtMachineState,
1607                                                  bootinfo);
1608     MachineState *ms = MACHINE(board);
1609 
1610 
1611     *fdt_size = board->fdt_size;
1612     return ms->fdt;
1613 }
1614 
1615 static void virt_build_smbios(VirtMachineState *vms)
1616 {
1617     MachineClass *mc = MACHINE_GET_CLASS(vms);
1618     MachineState *ms = MACHINE(vms);
1619     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
1620     uint8_t *smbios_tables, *smbios_anchor;
1621     size_t smbios_tables_len, smbios_anchor_len;
1622     struct smbios_phys_mem_area mem_array;
1623     const char *product = "QEMU Virtual Machine";
1624 
1625     if (kvm_enabled()) {
1626         product = "KVM Virtual Machine";
1627     }
1628 
1629     smbios_set_defaults("QEMU", product,
1630                         vmc->smbios_old_sys_ver ? "1.0" : mc->name, false,
1631                         true, SMBIOS_ENTRY_POINT_TYPE_64);
1632 
1633     /* build the array of physical mem area from base_memmap */
1634     mem_array.address = vms->memmap[VIRT_MEM].base;
1635     mem_array.length = ms->ram_size;
1636 
1637     smbios_get_tables(ms, &mem_array, 1,
1638                       &smbios_tables, &smbios_tables_len,
1639                       &smbios_anchor, &smbios_anchor_len,
1640                       &error_fatal);
1641 
1642     if (smbios_anchor) {
1643         fw_cfg_add_file(vms->fw_cfg, "etc/smbios/smbios-tables",
1644                         smbios_tables, smbios_tables_len);
1645         fw_cfg_add_file(vms->fw_cfg, "etc/smbios/smbios-anchor",
1646                         smbios_anchor, smbios_anchor_len);
1647     }
1648 }
1649 
1650 static
1651 void virt_machine_done(Notifier *notifier, void *data)
1652 {
1653     VirtMachineState *vms = container_of(notifier, VirtMachineState,
1654                                          machine_done);
1655     MachineState *ms = MACHINE(vms);
1656     ARMCPU *cpu = ARM_CPU(first_cpu);
1657     struct arm_boot_info *info = &vms->bootinfo;
1658     AddressSpace *as = arm_boot_address_space(cpu, info);
1659 
1660     /*
1661      * If the user provided a dtb, we assume the dynamic sysbus nodes
1662      * already are integrated there. This corresponds to a use case where
1663      * the dynamic sysbus nodes are complex and their generation is not yet
1664      * supported. In that case the user can take charge of the guest dt
1665      * while qemu takes charge of the qom stuff.
1666      */
1667     if (info->dtb_filename == NULL) {
1668         platform_bus_add_all_fdt_nodes(ms->fdt, "/intc",
1669                                        vms->memmap[VIRT_PLATFORM_BUS].base,
1670                                        vms->memmap[VIRT_PLATFORM_BUS].size,
1671                                        vms->irqmap[VIRT_PLATFORM_BUS]);
1672     }
1673     if (arm_load_dtb(info->dtb_start, info, info->dtb_limit, as, ms) < 0) {
1674         exit(1);
1675     }
1676 
1677     fw_cfg_add_extra_pci_roots(vms->bus, vms->fw_cfg);
1678 
1679     virt_acpi_setup(vms);
1680     virt_build_smbios(vms);
1681 }
1682 
1683 static uint64_t virt_cpu_mp_affinity(VirtMachineState *vms, int idx)
1684 {
1685     uint8_t clustersz = ARM_DEFAULT_CPUS_PER_CLUSTER;
1686     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
1687 
1688     if (!vmc->disallow_affinity_adjustment) {
1689         /* Adjust MPIDR like 64-bit KVM hosts, which incorporate the
1690          * GIC's target-list limitations. 32-bit KVM hosts currently
1691          * always create clusters of 4 CPUs, but that is expected to
1692          * change when they gain support for gicv3. When KVM is enabled
1693          * it will override the changes we make here, therefore our
1694          * purposes are to make TCG consistent (with 64-bit KVM hosts)
1695          * and to improve SGI efficiency.
1696          */
1697         if (vms->gic_version == VIRT_GIC_VERSION_2) {
1698             clustersz = GIC_TARGETLIST_BITS;
1699         } else {
1700             clustersz = GICV3_TARGETLIST_BITS;
1701         }
1702     }
1703     return arm_cpu_mp_affinity(idx, clustersz);
1704 }
1705 
1706 static inline bool *virt_get_high_memmap_enabled(VirtMachineState *vms,
1707                                                  int index)
1708 {
1709     bool *enabled_array[] = {
1710         &vms->highmem_redists,
1711         &vms->highmem_ecam,
1712         &vms->highmem_mmio,
1713     };
1714 
1715     assert(ARRAY_SIZE(extended_memmap) - VIRT_LOWMEMMAP_LAST ==
1716            ARRAY_SIZE(enabled_array));
1717     assert(index - VIRT_LOWMEMMAP_LAST < ARRAY_SIZE(enabled_array));
1718 
1719     return enabled_array[index - VIRT_LOWMEMMAP_LAST];
1720 }
1721 
1722 static void virt_set_high_memmap(VirtMachineState *vms,
1723                                  hwaddr base, int pa_bits)
1724 {
1725     hwaddr region_base, region_size;
1726     bool *region_enabled, fits;
1727     int i;
1728 
1729     for (i = VIRT_LOWMEMMAP_LAST; i < ARRAY_SIZE(extended_memmap); i++) {
1730         region_enabled = virt_get_high_memmap_enabled(vms, i);
1731         region_base = ROUND_UP(base, extended_memmap[i].size);
1732         region_size = extended_memmap[i].size;
1733 
1734         vms->memmap[i].base = region_base;
1735         vms->memmap[i].size = region_size;
1736 
1737         /*
1738          * Check each device to see if it fits in the PA space,
1739          * moving highest_gpa as we go. For compatibility, move
1740          * highest_gpa for disabled fitting devices as well, if
1741          * the compact layout has been disabled.
1742          *
1743          * For each device that doesn't fit, disable it.
1744          */
1745         fits = (region_base + region_size) <= BIT_ULL(pa_bits);
1746         *region_enabled &= fits;
1747         if (vms->highmem_compact && !*region_enabled) {
1748             continue;
1749         }
1750 
1751         base = region_base + region_size;
1752         if (fits) {
1753             vms->highest_gpa = base - 1;
1754         }
1755     }
1756 }
1757 
1758 static void virt_set_memmap(VirtMachineState *vms, int pa_bits)
1759 {
1760     MachineState *ms = MACHINE(vms);
1761     hwaddr base, device_memory_base, device_memory_size, memtop;
1762     int i;
1763 
1764     vms->memmap = extended_memmap;
1765 
1766     for (i = 0; i < ARRAY_SIZE(base_memmap); i++) {
1767         vms->memmap[i] = base_memmap[i];
1768     }
1769 
1770     if (ms->ram_slots > ACPI_MAX_RAM_SLOTS) {
1771         error_report("unsupported number of memory slots: %"PRIu64,
1772                      ms->ram_slots);
1773         exit(EXIT_FAILURE);
1774     }
1775 
1776     /*
1777      * !highmem is exactly the same as limiting the PA space to 32bit,
1778      * irrespective of the underlying capabilities of the HW.
1779      */
1780     if (!vms->highmem) {
1781         pa_bits = 32;
1782     }
1783 
1784     /*
1785      * We compute the base of the high IO region depending on the
1786      * amount of initial and device memory. The device memory start/size
1787      * is aligned on 1GiB. We never put the high IO region below 256GiB
1788      * so that if maxram_size is < 255GiB we keep the legacy memory map.
1789      * The device region size assumes 1GiB page max alignment per slot.
1790      */
1791     device_memory_base =
1792         ROUND_UP(vms->memmap[VIRT_MEM].base + ms->ram_size, GiB);
1793     device_memory_size = ms->maxram_size - ms->ram_size + ms->ram_slots * GiB;
1794 
1795     /* Base address of the high IO region */
1796     memtop = base = device_memory_base + ROUND_UP(device_memory_size, GiB);
1797     if (memtop > BIT_ULL(pa_bits)) {
1798 	    error_report("Addressing limited to %d bits, but memory exceeds it by %llu bytes\n",
1799 			 pa_bits, memtop - BIT_ULL(pa_bits));
1800         exit(EXIT_FAILURE);
1801     }
1802     if (base < device_memory_base) {
1803         error_report("maxmem/slots too huge");
1804         exit(EXIT_FAILURE);
1805     }
1806     if (base < vms->memmap[VIRT_MEM].base + LEGACY_RAMLIMIT_BYTES) {
1807         base = vms->memmap[VIRT_MEM].base + LEGACY_RAMLIMIT_BYTES;
1808     }
1809 
1810     /* We know for sure that at least the memory fits in the PA space */
1811     vms->highest_gpa = memtop - 1;
1812 
1813     virt_set_high_memmap(vms, base, pa_bits);
1814 
1815     if (device_memory_size > 0) {
1816         ms->device_memory = g_malloc0(sizeof(*ms->device_memory));
1817         ms->device_memory->base = device_memory_base;
1818         memory_region_init(&ms->device_memory->mr, OBJECT(vms),
1819                            "device-memory", device_memory_size);
1820     }
1821 }
1822 
1823 static VirtGICType finalize_gic_version_do(const char *accel_name,
1824                                            VirtGICType gic_version,
1825                                            int gics_supported,
1826                                            unsigned int max_cpus)
1827 {
1828     /* Convert host/max/nosel to GIC version number */
1829     switch (gic_version) {
1830     case VIRT_GIC_VERSION_HOST:
1831         if (!kvm_enabled()) {
1832             error_report("gic-version=host requires KVM");
1833             exit(1);
1834         }
1835 
1836         /* For KVM, gic-version=host means gic-version=max */
1837         return finalize_gic_version_do(accel_name, VIRT_GIC_VERSION_MAX,
1838                                        gics_supported, max_cpus);
1839     case VIRT_GIC_VERSION_MAX:
1840         if (gics_supported & VIRT_GIC_VERSION_4_MASK) {
1841             gic_version = VIRT_GIC_VERSION_4;
1842         } else if (gics_supported & VIRT_GIC_VERSION_3_MASK) {
1843             gic_version = VIRT_GIC_VERSION_3;
1844         } else {
1845             gic_version = VIRT_GIC_VERSION_2;
1846         }
1847         break;
1848     case VIRT_GIC_VERSION_NOSEL:
1849         if ((gics_supported & VIRT_GIC_VERSION_2_MASK) &&
1850             max_cpus <= GIC_NCPU) {
1851             gic_version = VIRT_GIC_VERSION_2;
1852         } else if (gics_supported & VIRT_GIC_VERSION_3_MASK) {
1853             /*
1854              * in case the host does not support v2 emulation or
1855              * the end-user requested more than 8 VCPUs we now default
1856              * to v3. In any case defaulting to v2 would be broken.
1857              */
1858             gic_version = VIRT_GIC_VERSION_3;
1859         } else if (max_cpus > GIC_NCPU) {
1860             error_report("%s only supports GICv2 emulation but more than 8 "
1861                          "vcpus are requested", accel_name);
1862             exit(1);
1863         }
1864         break;
1865     case VIRT_GIC_VERSION_2:
1866     case VIRT_GIC_VERSION_3:
1867     case VIRT_GIC_VERSION_4:
1868         break;
1869     }
1870 
1871     /* Check chosen version is effectively supported */
1872     switch (gic_version) {
1873     case VIRT_GIC_VERSION_2:
1874         if (!(gics_supported & VIRT_GIC_VERSION_2_MASK)) {
1875             error_report("%s does not support GICv2 emulation", accel_name);
1876             exit(1);
1877         }
1878         break;
1879     case VIRT_GIC_VERSION_3:
1880         if (!(gics_supported & VIRT_GIC_VERSION_3_MASK)) {
1881             error_report("%s does not support GICv3 emulation", accel_name);
1882             exit(1);
1883         }
1884         break;
1885     case VIRT_GIC_VERSION_4:
1886         if (!(gics_supported & VIRT_GIC_VERSION_4_MASK)) {
1887             error_report("%s does not support GICv4 emulation, is virtualization=on?",
1888                          accel_name);
1889             exit(1);
1890         }
1891         break;
1892     default:
1893         error_report("logic error in finalize_gic_version");
1894         exit(1);
1895         break;
1896     }
1897 
1898     return gic_version;
1899 }
1900 
1901 /*
1902  * finalize_gic_version - Determines the final gic_version
1903  * according to the gic-version property
1904  *
1905  * Default GIC type is v2
1906  */
1907 static void finalize_gic_version(VirtMachineState *vms)
1908 {
1909     const char *accel_name = current_accel_name();
1910     unsigned int max_cpus = MACHINE(vms)->smp.max_cpus;
1911     int gics_supported = 0;
1912 
1913     /* Determine which GIC versions the current environment supports */
1914     if (kvm_enabled() && kvm_irqchip_in_kernel()) {
1915         int probe_bitmap = kvm_arm_vgic_probe();
1916 
1917         if (!probe_bitmap) {
1918             error_report("Unable to determine GIC version supported by host");
1919             exit(1);
1920         }
1921 
1922         if (probe_bitmap & KVM_ARM_VGIC_V2) {
1923             gics_supported |= VIRT_GIC_VERSION_2_MASK;
1924         }
1925         if (probe_bitmap & KVM_ARM_VGIC_V3) {
1926             gics_supported |= VIRT_GIC_VERSION_3_MASK;
1927         }
1928     } else if (kvm_enabled() && !kvm_irqchip_in_kernel()) {
1929         /* KVM w/o kernel irqchip can only deal with GICv2 */
1930         gics_supported |= VIRT_GIC_VERSION_2_MASK;
1931         accel_name = "KVM with kernel-irqchip=off";
1932     } else if (tcg_enabled() || hvf_enabled() || qtest_enabled())  {
1933         gics_supported |= VIRT_GIC_VERSION_2_MASK;
1934         if (module_object_class_by_name("arm-gicv3")) {
1935             gics_supported |= VIRT_GIC_VERSION_3_MASK;
1936             if (vms->virt) {
1937                 /* GICv4 only makes sense if CPU has EL2 */
1938                 gics_supported |= VIRT_GIC_VERSION_4_MASK;
1939             }
1940         }
1941     } else {
1942         error_report("Unsupported accelerator, can not determine GIC support");
1943         exit(1);
1944     }
1945 
1946     /*
1947      * Then convert helpers like host/max to concrete GIC versions and ensure
1948      * the desired version is supported
1949      */
1950     vms->gic_version = finalize_gic_version_do(accel_name, vms->gic_version,
1951                                                gics_supported, max_cpus);
1952 }
1953 
1954 /*
1955  * virt_cpu_post_init() must be called after the CPUs have
1956  * been realized and the GIC has been created.
1957  */
1958 static void virt_cpu_post_init(VirtMachineState *vms, MemoryRegion *sysmem)
1959 {
1960     int max_cpus = MACHINE(vms)->smp.max_cpus;
1961     bool aarch64, pmu, steal_time;
1962     CPUState *cpu;
1963 
1964     aarch64 = object_property_get_bool(OBJECT(first_cpu), "aarch64", NULL);
1965     pmu = object_property_get_bool(OBJECT(first_cpu), "pmu", NULL);
1966     steal_time = object_property_get_bool(OBJECT(first_cpu),
1967                                           "kvm-steal-time", NULL);
1968 
1969     if (kvm_enabled()) {
1970         hwaddr pvtime_reg_base = vms->memmap[VIRT_PVTIME].base;
1971         hwaddr pvtime_reg_size = vms->memmap[VIRT_PVTIME].size;
1972 
1973         if (steal_time) {
1974             MemoryRegion *pvtime = g_new(MemoryRegion, 1);
1975             hwaddr pvtime_size = max_cpus * PVTIME_SIZE_PER_CPU;
1976 
1977             /* The memory region size must be a multiple of host page size. */
1978             pvtime_size = REAL_HOST_PAGE_ALIGN(pvtime_size);
1979 
1980             if (pvtime_size > pvtime_reg_size) {
1981                 error_report("pvtime requires a %" HWADDR_PRId
1982                              " byte memory region for %d CPUs,"
1983                              " but only %" HWADDR_PRId " has been reserved",
1984                              pvtime_size, max_cpus, pvtime_reg_size);
1985                 exit(1);
1986             }
1987 
1988             memory_region_init_ram(pvtime, NULL, "pvtime", pvtime_size, NULL);
1989             memory_region_add_subregion(sysmem, pvtime_reg_base, pvtime);
1990         }
1991 
1992         CPU_FOREACH(cpu) {
1993             if (pmu) {
1994                 assert(arm_feature(&ARM_CPU(cpu)->env, ARM_FEATURE_PMU));
1995                 if (kvm_irqchip_in_kernel()) {
1996                     kvm_arm_pmu_set_irq(cpu, PPI(VIRTUAL_PMU_IRQ));
1997                 }
1998                 kvm_arm_pmu_init(cpu);
1999             }
2000             if (steal_time) {
2001                 kvm_arm_pvtime_init(cpu, pvtime_reg_base +
2002                                          cpu->cpu_index * PVTIME_SIZE_PER_CPU);
2003             }
2004         }
2005     } else {
2006         if (aarch64 && vms->highmem) {
2007             int requested_pa_size = 64 - clz64(vms->highest_gpa);
2008             int pamax = arm_pamax(ARM_CPU(first_cpu));
2009 
2010             if (pamax < requested_pa_size) {
2011                 error_report("VCPU supports less PA bits (%d) than "
2012                              "requested by the memory map (%d)",
2013                              pamax, requested_pa_size);
2014                 exit(1);
2015             }
2016         }
2017     }
2018 }
2019 
2020 static void machvirt_init(MachineState *machine)
2021 {
2022     VirtMachineState *vms = VIRT_MACHINE(machine);
2023     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(machine);
2024     MachineClass *mc = MACHINE_GET_CLASS(machine);
2025     const CPUArchIdList *possible_cpus;
2026     MemoryRegion *sysmem = get_system_memory();
2027     MemoryRegion *secure_sysmem = NULL;
2028     MemoryRegion *tag_sysmem = NULL;
2029     MemoryRegion *secure_tag_sysmem = NULL;
2030     int n, virt_max_cpus;
2031     bool firmware_loaded;
2032     bool aarch64 = true;
2033     bool has_ged = !vmc->no_ged;
2034     unsigned int smp_cpus = machine->smp.cpus;
2035     unsigned int max_cpus = machine->smp.max_cpus;
2036 
2037     if (!cpu_type_valid(machine->cpu_type)) {
2038         error_report("mach-virt: CPU type %s not supported", machine->cpu_type);
2039         exit(1);
2040     }
2041 
2042     possible_cpus = mc->possible_cpu_arch_ids(machine);
2043 
2044     /*
2045      * In accelerated mode, the memory map is computed earlier in kvm_type()
2046      * to create a VM with the right number of IPA bits.
2047      */
2048     if (!vms->memmap) {
2049         Object *cpuobj;
2050         ARMCPU *armcpu;
2051         int pa_bits;
2052 
2053         /*
2054          * Instantiate a temporary CPU object to find out about what
2055          * we are about to deal with. Once this is done, get rid of
2056          * the object.
2057          */
2058         cpuobj = object_new(possible_cpus->cpus[0].type);
2059         armcpu = ARM_CPU(cpuobj);
2060 
2061         pa_bits = arm_pamax(armcpu);
2062 
2063         object_unref(cpuobj);
2064 
2065         virt_set_memmap(vms, pa_bits);
2066     }
2067 
2068     /* We can probe only here because during property set
2069      * KVM is not available yet
2070      */
2071     finalize_gic_version(vms);
2072 
2073     if (vms->secure) {
2074         /*
2075          * The Secure view of the world is the same as the NonSecure,
2076          * but with a few extra devices. Create it as a container region
2077          * containing the system memory at low priority; any secure-only
2078          * devices go in at higher priority and take precedence.
2079          */
2080         secure_sysmem = g_new(MemoryRegion, 1);
2081         memory_region_init(secure_sysmem, OBJECT(machine), "secure-memory",
2082                            UINT64_MAX);
2083         memory_region_add_subregion_overlap(secure_sysmem, 0, sysmem, -1);
2084     }
2085 
2086     firmware_loaded = virt_firmware_init(vms, sysmem,
2087                                          secure_sysmem ?: sysmem);
2088 
2089     /* If we have an EL3 boot ROM then the assumption is that it will
2090      * implement PSCI itself, so disable QEMU's internal implementation
2091      * so it doesn't get in the way. Instead of starting secondary
2092      * CPUs in PSCI powerdown state we will start them all running and
2093      * let the boot ROM sort them out.
2094      * The usual case is that we do use QEMU's PSCI implementation;
2095      * if the guest has EL2 then we will use SMC as the conduit,
2096      * and otherwise we will use HVC (for backwards compatibility and
2097      * because if we're using KVM then we must use HVC).
2098      */
2099     if (vms->secure && firmware_loaded) {
2100         vms->psci_conduit = QEMU_PSCI_CONDUIT_DISABLED;
2101     } else if (vms->virt) {
2102         vms->psci_conduit = QEMU_PSCI_CONDUIT_SMC;
2103     } else {
2104         vms->psci_conduit = QEMU_PSCI_CONDUIT_HVC;
2105     }
2106 
2107     /*
2108      * The maximum number of CPUs depends on the GIC version, or on how
2109      * many redistributors we can fit into the memory map (which in turn
2110      * depends on whether this is a GICv3 or v4).
2111      */
2112     if (vms->gic_version == VIRT_GIC_VERSION_2) {
2113         virt_max_cpus = GIC_NCPU;
2114     } else {
2115         virt_max_cpus = virt_redist_capacity(vms, VIRT_GIC_REDIST);
2116         if (vms->highmem_redists) {
2117             virt_max_cpus += virt_redist_capacity(vms, VIRT_HIGH_GIC_REDIST2);
2118         }
2119     }
2120 
2121     if (max_cpus > virt_max_cpus) {
2122         error_report("Number of SMP CPUs requested (%d) exceeds max CPUs "
2123                      "supported by machine 'mach-virt' (%d)",
2124                      max_cpus, virt_max_cpus);
2125         if (vms->gic_version != VIRT_GIC_VERSION_2 && !vms->highmem_redists) {
2126             error_printf("Try 'highmem-redists=on' for more CPUs\n");
2127         }
2128 
2129         exit(1);
2130     }
2131 
2132     if (vms->secure && (kvm_enabled() || hvf_enabled())) {
2133         error_report("mach-virt: %s does not support providing "
2134                      "Security extensions (TrustZone) to the guest CPU",
2135                      current_accel_name());
2136         exit(1);
2137     }
2138 
2139     if (vms->virt && (kvm_enabled() || hvf_enabled())) {
2140         error_report("mach-virt: %s does not support providing "
2141                      "Virtualization extensions to the guest CPU",
2142                      current_accel_name());
2143         exit(1);
2144     }
2145 
2146     if (vms->mte && (kvm_enabled() || hvf_enabled())) {
2147         error_report("mach-virt: %s does not support providing "
2148                      "MTE to the guest CPU",
2149                      current_accel_name());
2150         exit(1);
2151     }
2152 
2153     create_fdt(vms);
2154 
2155     assert(possible_cpus->len == max_cpus);
2156     for (n = 0; n < possible_cpus->len; n++) {
2157         Object *cpuobj;
2158         CPUState *cs;
2159 
2160         if (n >= smp_cpus) {
2161             break;
2162         }
2163 
2164         cpuobj = object_new(possible_cpus->cpus[n].type);
2165         object_property_set_int(cpuobj, "mp-affinity",
2166                                 possible_cpus->cpus[n].arch_id, NULL);
2167 
2168         cs = CPU(cpuobj);
2169         cs->cpu_index = n;
2170 
2171         numa_cpu_pre_plug(&possible_cpus->cpus[cs->cpu_index], DEVICE(cpuobj),
2172                           &error_fatal);
2173 
2174         aarch64 &= object_property_get_bool(cpuobj, "aarch64", NULL);
2175 
2176         if (!vms->secure) {
2177             object_property_set_bool(cpuobj, "has_el3", false, NULL);
2178         }
2179 
2180         if (!vms->virt && object_property_find(cpuobj, "has_el2")) {
2181             object_property_set_bool(cpuobj, "has_el2", false, NULL);
2182         }
2183 
2184         if (vmc->kvm_no_adjvtime &&
2185             object_property_find(cpuobj, "kvm-no-adjvtime")) {
2186             object_property_set_bool(cpuobj, "kvm-no-adjvtime", true, NULL);
2187         }
2188 
2189         if (vmc->no_kvm_steal_time &&
2190             object_property_find(cpuobj, "kvm-steal-time")) {
2191             object_property_set_bool(cpuobj, "kvm-steal-time", false, NULL);
2192         }
2193 
2194         if (vmc->no_pmu && object_property_find(cpuobj, "pmu")) {
2195             object_property_set_bool(cpuobj, "pmu", false, NULL);
2196         }
2197 
2198         if (vmc->no_tcg_lpa2 && object_property_find(cpuobj, "lpa2")) {
2199             object_property_set_bool(cpuobj, "lpa2", false, NULL);
2200         }
2201 
2202         if (object_property_find(cpuobj, "reset-cbar")) {
2203             object_property_set_int(cpuobj, "reset-cbar",
2204                                     vms->memmap[VIRT_CPUPERIPHS].base,
2205                                     &error_abort);
2206         }
2207 
2208         object_property_set_link(cpuobj, "memory", OBJECT(sysmem),
2209                                  &error_abort);
2210         if (vms->secure) {
2211             object_property_set_link(cpuobj, "secure-memory",
2212                                      OBJECT(secure_sysmem), &error_abort);
2213         }
2214 
2215         if (vms->mte) {
2216             /* Create the memory region only once, but link to all cpus. */
2217             if (!tag_sysmem) {
2218                 /*
2219                  * The property exists only if MemTag is supported.
2220                  * If it is, we must allocate the ram to back that up.
2221                  */
2222                 if (!object_property_find(cpuobj, "tag-memory")) {
2223                     error_report("MTE requested, but not supported "
2224                                  "by the guest CPU");
2225                     exit(1);
2226                 }
2227 
2228                 tag_sysmem = g_new(MemoryRegion, 1);
2229                 memory_region_init(tag_sysmem, OBJECT(machine),
2230                                    "tag-memory", UINT64_MAX / 32);
2231 
2232                 if (vms->secure) {
2233                     secure_tag_sysmem = g_new(MemoryRegion, 1);
2234                     memory_region_init(secure_tag_sysmem, OBJECT(machine),
2235                                        "secure-tag-memory", UINT64_MAX / 32);
2236 
2237                     /* As with ram, secure-tag takes precedence over tag.  */
2238                     memory_region_add_subregion_overlap(secure_tag_sysmem, 0,
2239                                                         tag_sysmem, -1);
2240                 }
2241             }
2242 
2243             object_property_set_link(cpuobj, "tag-memory", OBJECT(tag_sysmem),
2244                                      &error_abort);
2245             if (vms->secure) {
2246                 object_property_set_link(cpuobj, "secure-tag-memory",
2247                                          OBJECT(secure_tag_sysmem),
2248                                          &error_abort);
2249             }
2250         }
2251 
2252         qdev_realize(DEVICE(cpuobj), NULL, &error_fatal);
2253         object_unref(cpuobj);
2254     }
2255     fdt_add_timer_nodes(vms);
2256     fdt_add_cpu_nodes(vms);
2257 
2258     memory_region_add_subregion(sysmem, vms->memmap[VIRT_MEM].base,
2259                                 machine->ram);
2260     if (machine->device_memory) {
2261         memory_region_add_subregion(sysmem, machine->device_memory->base,
2262                                     &machine->device_memory->mr);
2263     }
2264 
2265     virt_flash_fdt(vms, sysmem, secure_sysmem ?: sysmem);
2266 
2267     create_gic(vms, sysmem);
2268 
2269     virt_cpu_post_init(vms, sysmem);
2270 
2271     fdt_add_pmu_nodes(vms);
2272 
2273     create_uart(vms, VIRT_UART, sysmem, serial_hd(0));
2274 
2275     if (vms->secure) {
2276         create_secure_ram(vms, secure_sysmem, secure_tag_sysmem);
2277         create_uart(vms, VIRT_SECURE_UART, secure_sysmem, serial_hd(1));
2278     }
2279 
2280     if (tag_sysmem) {
2281         create_tag_ram(tag_sysmem, vms->memmap[VIRT_MEM].base,
2282                        machine->ram_size, "mach-virt.tag");
2283     }
2284 
2285     vms->highmem_ecam &= (!firmware_loaded || aarch64);
2286 
2287     create_rtc(vms);
2288 
2289     create_pcie(vms);
2290 
2291     if (has_ged && aarch64 && firmware_loaded && virt_is_acpi_enabled(vms)) {
2292         vms->acpi_dev = create_acpi_ged(vms);
2293     } else {
2294         create_gpio_devices(vms, VIRT_GPIO, sysmem);
2295     }
2296 
2297     if (vms->secure && !vmc->no_secure_gpio) {
2298         create_gpio_devices(vms, VIRT_SECURE_GPIO, secure_sysmem);
2299     }
2300 
2301      /* connect powerdown request */
2302      vms->powerdown_notifier.notify = virt_powerdown_req;
2303      qemu_register_powerdown_notifier(&vms->powerdown_notifier);
2304 
2305     /* Create mmio transports, so the user can create virtio backends
2306      * (which will be automatically plugged in to the transports). If
2307      * no backend is created the transport will just sit harmlessly idle.
2308      */
2309     create_virtio_devices(vms);
2310 
2311     vms->fw_cfg = create_fw_cfg(vms, &address_space_memory);
2312     rom_set_fw(vms->fw_cfg);
2313 
2314     create_platform_bus(vms);
2315 
2316     if (machine->nvdimms_state->is_enabled) {
2317         const struct AcpiGenericAddress arm_virt_nvdimm_acpi_dsmio = {
2318             .space_id = AML_AS_SYSTEM_MEMORY,
2319             .address = vms->memmap[VIRT_NVDIMM_ACPI].base,
2320             .bit_width = NVDIMM_ACPI_IO_LEN << 3
2321         };
2322 
2323         nvdimm_init_acpi_state(machine->nvdimms_state, sysmem,
2324                                arm_virt_nvdimm_acpi_dsmio,
2325                                vms->fw_cfg, OBJECT(vms));
2326     }
2327 
2328     vms->bootinfo.ram_size = machine->ram_size;
2329     vms->bootinfo.board_id = -1;
2330     vms->bootinfo.loader_start = vms->memmap[VIRT_MEM].base;
2331     vms->bootinfo.get_dtb = machvirt_dtb;
2332     vms->bootinfo.skip_dtb_autoload = true;
2333     vms->bootinfo.firmware_loaded = firmware_loaded;
2334     vms->bootinfo.psci_conduit = vms->psci_conduit;
2335     arm_load_kernel(ARM_CPU(first_cpu), machine, &vms->bootinfo);
2336 
2337     vms->machine_done.notify = virt_machine_done;
2338     qemu_add_machine_init_done_notifier(&vms->machine_done);
2339 }
2340 
2341 static bool virt_get_secure(Object *obj, Error **errp)
2342 {
2343     VirtMachineState *vms = VIRT_MACHINE(obj);
2344 
2345     return vms->secure;
2346 }
2347 
2348 static void virt_set_secure(Object *obj, bool value, Error **errp)
2349 {
2350     VirtMachineState *vms = VIRT_MACHINE(obj);
2351 
2352     vms->secure = value;
2353 }
2354 
2355 static bool virt_get_virt(Object *obj, Error **errp)
2356 {
2357     VirtMachineState *vms = VIRT_MACHINE(obj);
2358 
2359     return vms->virt;
2360 }
2361 
2362 static void virt_set_virt(Object *obj, bool value, Error **errp)
2363 {
2364     VirtMachineState *vms = VIRT_MACHINE(obj);
2365 
2366     vms->virt = value;
2367 }
2368 
2369 static bool virt_get_highmem(Object *obj, Error **errp)
2370 {
2371     VirtMachineState *vms = VIRT_MACHINE(obj);
2372 
2373     return vms->highmem;
2374 }
2375 
2376 static void virt_set_highmem(Object *obj, bool value, Error **errp)
2377 {
2378     VirtMachineState *vms = VIRT_MACHINE(obj);
2379 
2380     vms->highmem = value;
2381 }
2382 
2383 static bool virt_get_compact_highmem(Object *obj, Error **errp)
2384 {
2385     VirtMachineState *vms = VIRT_MACHINE(obj);
2386 
2387     return vms->highmem_compact;
2388 }
2389 
2390 static void virt_set_compact_highmem(Object *obj, bool value, Error **errp)
2391 {
2392     VirtMachineState *vms = VIRT_MACHINE(obj);
2393 
2394     vms->highmem_compact = value;
2395 }
2396 
2397 static bool virt_get_highmem_redists(Object *obj, Error **errp)
2398 {
2399     VirtMachineState *vms = VIRT_MACHINE(obj);
2400 
2401     return vms->highmem_redists;
2402 }
2403 
2404 static void virt_set_highmem_redists(Object *obj, bool value, Error **errp)
2405 {
2406     VirtMachineState *vms = VIRT_MACHINE(obj);
2407 
2408     vms->highmem_redists = value;
2409 }
2410 
2411 static bool virt_get_highmem_ecam(Object *obj, Error **errp)
2412 {
2413     VirtMachineState *vms = VIRT_MACHINE(obj);
2414 
2415     return vms->highmem_ecam;
2416 }
2417 
2418 static void virt_set_highmem_ecam(Object *obj, bool value, Error **errp)
2419 {
2420     VirtMachineState *vms = VIRT_MACHINE(obj);
2421 
2422     vms->highmem_ecam = value;
2423 }
2424 
2425 static bool virt_get_highmem_mmio(Object *obj, Error **errp)
2426 {
2427     VirtMachineState *vms = VIRT_MACHINE(obj);
2428 
2429     return vms->highmem_mmio;
2430 }
2431 
2432 static void virt_set_highmem_mmio(Object *obj, bool value, Error **errp)
2433 {
2434     VirtMachineState *vms = VIRT_MACHINE(obj);
2435 
2436     vms->highmem_mmio = value;
2437 }
2438 
2439 
2440 static bool virt_get_its(Object *obj, Error **errp)
2441 {
2442     VirtMachineState *vms = VIRT_MACHINE(obj);
2443 
2444     return vms->its;
2445 }
2446 
2447 static void virt_set_its(Object *obj, bool value, Error **errp)
2448 {
2449     VirtMachineState *vms = VIRT_MACHINE(obj);
2450 
2451     vms->its = value;
2452 }
2453 
2454 static bool virt_get_dtb_randomness(Object *obj, Error **errp)
2455 {
2456     VirtMachineState *vms = VIRT_MACHINE(obj);
2457 
2458     return vms->dtb_randomness;
2459 }
2460 
2461 static void virt_set_dtb_randomness(Object *obj, bool value, Error **errp)
2462 {
2463     VirtMachineState *vms = VIRT_MACHINE(obj);
2464 
2465     vms->dtb_randomness = value;
2466 }
2467 
2468 static char *virt_get_oem_id(Object *obj, Error **errp)
2469 {
2470     VirtMachineState *vms = VIRT_MACHINE(obj);
2471 
2472     return g_strdup(vms->oem_id);
2473 }
2474 
2475 static void virt_set_oem_id(Object *obj, const char *value, Error **errp)
2476 {
2477     VirtMachineState *vms = VIRT_MACHINE(obj);
2478     size_t len = strlen(value);
2479 
2480     if (len > 6) {
2481         error_setg(errp,
2482                    "User specified oem-id value is bigger than 6 bytes in size");
2483         return;
2484     }
2485 
2486     strncpy(vms->oem_id, value, 6);
2487 }
2488 
2489 static char *virt_get_oem_table_id(Object *obj, Error **errp)
2490 {
2491     VirtMachineState *vms = VIRT_MACHINE(obj);
2492 
2493     return g_strdup(vms->oem_table_id);
2494 }
2495 
2496 static void virt_set_oem_table_id(Object *obj, const char *value,
2497                                   Error **errp)
2498 {
2499     VirtMachineState *vms = VIRT_MACHINE(obj);
2500     size_t len = strlen(value);
2501 
2502     if (len > 8) {
2503         error_setg(errp,
2504                    "User specified oem-table-id value is bigger than 8 bytes in size");
2505         return;
2506     }
2507     strncpy(vms->oem_table_id, value, 8);
2508 }
2509 
2510 
2511 bool virt_is_acpi_enabled(VirtMachineState *vms)
2512 {
2513     if (vms->acpi == ON_OFF_AUTO_OFF) {
2514         return false;
2515     }
2516     return true;
2517 }
2518 
2519 static void virt_get_acpi(Object *obj, Visitor *v, const char *name,
2520                           void *opaque, Error **errp)
2521 {
2522     VirtMachineState *vms = VIRT_MACHINE(obj);
2523     OnOffAuto acpi = vms->acpi;
2524 
2525     visit_type_OnOffAuto(v, name, &acpi, errp);
2526 }
2527 
2528 static void virt_set_acpi(Object *obj, Visitor *v, const char *name,
2529                           void *opaque, Error **errp)
2530 {
2531     VirtMachineState *vms = VIRT_MACHINE(obj);
2532 
2533     visit_type_OnOffAuto(v, name, &vms->acpi, errp);
2534 }
2535 
2536 static bool virt_get_ras(Object *obj, Error **errp)
2537 {
2538     VirtMachineState *vms = VIRT_MACHINE(obj);
2539 
2540     return vms->ras;
2541 }
2542 
2543 static void virt_set_ras(Object *obj, bool value, Error **errp)
2544 {
2545     VirtMachineState *vms = VIRT_MACHINE(obj);
2546 
2547     vms->ras = value;
2548 }
2549 
2550 static bool virt_get_mte(Object *obj, Error **errp)
2551 {
2552     VirtMachineState *vms = VIRT_MACHINE(obj);
2553 
2554     return vms->mte;
2555 }
2556 
2557 static void virt_set_mte(Object *obj, bool value, Error **errp)
2558 {
2559     VirtMachineState *vms = VIRT_MACHINE(obj);
2560 
2561     vms->mte = value;
2562 }
2563 
2564 static char *virt_get_gic_version(Object *obj, Error **errp)
2565 {
2566     VirtMachineState *vms = VIRT_MACHINE(obj);
2567     const char *val;
2568 
2569     switch (vms->gic_version) {
2570     case VIRT_GIC_VERSION_4:
2571         val = "4";
2572         break;
2573     case VIRT_GIC_VERSION_3:
2574         val = "3";
2575         break;
2576     default:
2577         val = "2";
2578         break;
2579     }
2580     return g_strdup(val);
2581 }
2582 
2583 static void virt_set_gic_version(Object *obj, const char *value, Error **errp)
2584 {
2585     VirtMachineState *vms = VIRT_MACHINE(obj);
2586 
2587     if (!strcmp(value, "4")) {
2588         vms->gic_version = VIRT_GIC_VERSION_4;
2589     } else if (!strcmp(value, "3")) {
2590         vms->gic_version = VIRT_GIC_VERSION_3;
2591     } else if (!strcmp(value, "2")) {
2592         vms->gic_version = VIRT_GIC_VERSION_2;
2593     } else if (!strcmp(value, "host")) {
2594         vms->gic_version = VIRT_GIC_VERSION_HOST; /* Will probe later */
2595     } else if (!strcmp(value, "max")) {
2596         vms->gic_version = VIRT_GIC_VERSION_MAX; /* Will probe later */
2597     } else {
2598         error_setg(errp, "Invalid gic-version value");
2599         error_append_hint(errp, "Valid values are 3, 2, host, max.\n");
2600     }
2601 }
2602 
2603 static char *virt_get_iommu(Object *obj, Error **errp)
2604 {
2605     VirtMachineState *vms = VIRT_MACHINE(obj);
2606 
2607     switch (vms->iommu) {
2608     case VIRT_IOMMU_NONE:
2609         return g_strdup("none");
2610     case VIRT_IOMMU_SMMUV3:
2611         return g_strdup("smmuv3");
2612     default:
2613         g_assert_not_reached();
2614     }
2615 }
2616 
2617 static void virt_set_iommu(Object *obj, const char *value, Error **errp)
2618 {
2619     VirtMachineState *vms = VIRT_MACHINE(obj);
2620 
2621     if (!strcmp(value, "smmuv3")) {
2622         vms->iommu = VIRT_IOMMU_SMMUV3;
2623     } else if (!strcmp(value, "none")) {
2624         vms->iommu = VIRT_IOMMU_NONE;
2625     } else {
2626         error_setg(errp, "Invalid iommu value");
2627         error_append_hint(errp, "Valid values are none, smmuv3.\n");
2628     }
2629 }
2630 
2631 static bool virt_get_default_bus_bypass_iommu(Object *obj, Error **errp)
2632 {
2633     VirtMachineState *vms = VIRT_MACHINE(obj);
2634 
2635     return vms->default_bus_bypass_iommu;
2636 }
2637 
2638 static void virt_set_default_bus_bypass_iommu(Object *obj, bool value,
2639                                               Error **errp)
2640 {
2641     VirtMachineState *vms = VIRT_MACHINE(obj);
2642 
2643     vms->default_bus_bypass_iommu = value;
2644 }
2645 
2646 static CpuInstanceProperties
2647 virt_cpu_index_to_props(MachineState *ms, unsigned cpu_index)
2648 {
2649     MachineClass *mc = MACHINE_GET_CLASS(ms);
2650     const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(ms);
2651 
2652     assert(cpu_index < possible_cpus->len);
2653     return possible_cpus->cpus[cpu_index].props;
2654 }
2655 
2656 static int64_t virt_get_default_cpu_node_id(const MachineState *ms, int idx)
2657 {
2658     int64_t socket_id = ms->possible_cpus->cpus[idx].props.socket_id;
2659 
2660     return socket_id % ms->numa_state->num_nodes;
2661 }
2662 
2663 static const CPUArchIdList *virt_possible_cpu_arch_ids(MachineState *ms)
2664 {
2665     int n;
2666     unsigned int max_cpus = ms->smp.max_cpus;
2667     VirtMachineState *vms = VIRT_MACHINE(ms);
2668     MachineClass *mc = MACHINE_GET_CLASS(vms);
2669 
2670     if (ms->possible_cpus) {
2671         assert(ms->possible_cpus->len == max_cpus);
2672         return ms->possible_cpus;
2673     }
2674 
2675     ms->possible_cpus = g_malloc0(sizeof(CPUArchIdList) +
2676                                   sizeof(CPUArchId) * max_cpus);
2677     ms->possible_cpus->len = max_cpus;
2678     for (n = 0; n < ms->possible_cpus->len; n++) {
2679         ms->possible_cpus->cpus[n].type = ms->cpu_type;
2680         ms->possible_cpus->cpus[n].arch_id =
2681             virt_cpu_mp_affinity(vms, n);
2682 
2683         assert(!mc->smp_props.dies_supported);
2684         ms->possible_cpus->cpus[n].props.has_socket_id = true;
2685         ms->possible_cpus->cpus[n].props.socket_id =
2686             n / (ms->smp.clusters * ms->smp.cores * ms->smp.threads);
2687         ms->possible_cpus->cpus[n].props.has_cluster_id = true;
2688         ms->possible_cpus->cpus[n].props.cluster_id =
2689             (n / (ms->smp.cores * ms->smp.threads)) % ms->smp.clusters;
2690         ms->possible_cpus->cpus[n].props.has_core_id = true;
2691         ms->possible_cpus->cpus[n].props.core_id =
2692             (n / ms->smp.threads) % ms->smp.cores;
2693         ms->possible_cpus->cpus[n].props.has_thread_id = true;
2694         ms->possible_cpus->cpus[n].props.thread_id =
2695             n % ms->smp.threads;
2696     }
2697     return ms->possible_cpus;
2698 }
2699 
2700 static void virt_memory_pre_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
2701                                  Error **errp)
2702 {
2703     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2704     const MachineState *ms = MACHINE(hotplug_dev);
2705     const bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2706 
2707     if (!vms->acpi_dev) {
2708         error_setg(errp,
2709                    "memory hotplug is not enabled: missing acpi-ged device");
2710         return;
2711     }
2712 
2713     if (vms->mte) {
2714         error_setg(errp, "memory hotplug is not enabled: MTE is enabled");
2715         return;
2716     }
2717 
2718     if (is_nvdimm && !ms->nvdimms_state->is_enabled) {
2719         error_setg(errp, "nvdimm is not enabled: add 'nvdimm=on' to '-M'");
2720         return;
2721     }
2722 
2723     pc_dimm_pre_plug(PC_DIMM(dev), MACHINE(hotplug_dev), NULL, errp);
2724 }
2725 
2726 static void virt_memory_plug(HotplugHandler *hotplug_dev,
2727                              DeviceState *dev, Error **errp)
2728 {
2729     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2730     MachineState *ms = MACHINE(hotplug_dev);
2731     bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2732 
2733     pc_dimm_plug(PC_DIMM(dev), MACHINE(vms));
2734 
2735     if (is_nvdimm) {
2736         nvdimm_plug(ms->nvdimms_state);
2737     }
2738 
2739     hotplug_handler_plug(HOTPLUG_HANDLER(vms->acpi_dev),
2740                          dev, &error_abort);
2741 }
2742 
2743 static void virt_virtio_md_pci_pre_plug(HotplugHandler *hotplug_dev,
2744                                         DeviceState *dev, Error **errp)
2745 {
2746     HotplugHandler *hotplug_dev2 = qdev_get_bus_hotplug_handler(dev);
2747     Error *local_err = NULL;
2748 
2749     if (!hotplug_dev2 && dev->hotplugged) {
2750         /*
2751          * Without a bus hotplug handler, we cannot control the plug/unplug
2752          * order. We should never reach this point when hotplugging on ARM.
2753          * However, it's nice to add a safety net, similar to what we have
2754          * on x86.
2755          */
2756         error_setg(errp, "hotplug of virtio based memory devices not supported"
2757                    " on this bus.");
2758         return;
2759     }
2760     /*
2761      * First, see if we can plug this memory device at all. If that
2762      * succeeds, branch of to the actual hotplug handler.
2763      */
2764     memory_device_pre_plug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev), NULL,
2765                            &local_err);
2766     if (!local_err && hotplug_dev2) {
2767         hotplug_handler_pre_plug(hotplug_dev2, dev, &local_err);
2768     }
2769     error_propagate(errp, local_err);
2770 }
2771 
2772 static void virt_virtio_md_pci_plug(HotplugHandler *hotplug_dev,
2773                                     DeviceState *dev, Error **errp)
2774 {
2775     HotplugHandler *hotplug_dev2 = qdev_get_bus_hotplug_handler(dev);
2776     Error *local_err = NULL;
2777 
2778     /*
2779      * Plug the memory device first and then branch off to the actual
2780      * hotplug handler. If that one fails, we can easily undo the memory
2781      * device bits.
2782      */
2783     memory_device_plug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev));
2784     if (hotplug_dev2) {
2785         hotplug_handler_plug(hotplug_dev2, dev, &local_err);
2786         if (local_err) {
2787             memory_device_unplug(MEMORY_DEVICE(dev), MACHINE(hotplug_dev));
2788         }
2789     }
2790     error_propagate(errp, local_err);
2791 }
2792 
2793 static void virt_virtio_md_pci_unplug_request(HotplugHandler *hotplug_dev,
2794                                               DeviceState *dev, Error **errp)
2795 {
2796     /* We don't support hot unplug of virtio based memory devices */
2797     error_setg(errp, "virtio based memory devices cannot be unplugged.");
2798 }
2799 
2800 
2801 static void virt_machine_device_pre_plug_cb(HotplugHandler *hotplug_dev,
2802                                             DeviceState *dev, Error **errp)
2803 {
2804     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2805 
2806     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2807         virt_memory_pre_plug(hotplug_dev, dev, errp);
2808     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI)) {
2809         virt_virtio_md_pci_pre_plug(hotplug_dev, dev, errp);
2810     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2811         hwaddr db_start = 0, db_end = 0;
2812         char *resv_prop_str;
2813 
2814         if (vms->iommu != VIRT_IOMMU_NONE) {
2815             error_setg(errp, "virt machine does not support multiple IOMMUs");
2816             return;
2817         }
2818 
2819         switch (vms->msi_controller) {
2820         case VIRT_MSI_CTRL_NONE:
2821             return;
2822         case VIRT_MSI_CTRL_ITS:
2823             /* GITS_TRANSLATER page */
2824             db_start = base_memmap[VIRT_GIC_ITS].base + 0x10000;
2825             db_end = base_memmap[VIRT_GIC_ITS].base +
2826                      base_memmap[VIRT_GIC_ITS].size - 1;
2827             break;
2828         case VIRT_MSI_CTRL_GICV2M:
2829             /* MSI_SETSPI_NS page */
2830             db_start = base_memmap[VIRT_GIC_V2M].base;
2831             db_end = db_start + base_memmap[VIRT_GIC_V2M].size - 1;
2832             break;
2833         }
2834         resv_prop_str = g_strdup_printf("0x%"PRIx64":0x%"PRIx64":%u",
2835                                         db_start, db_end,
2836                                         VIRTIO_IOMMU_RESV_MEM_T_MSI);
2837 
2838         object_property_set_uint(OBJECT(dev), "len-reserved-regions", 1, errp);
2839         object_property_set_str(OBJECT(dev), "reserved-regions[0]",
2840                                 resv_prop_str, errp);
2841         g_free(resv_prop_str);
2842     }
2843 }
2844 
2845 static void virt_machine_device_plug_cb(HotplugHandler *hotplug_dev,
2846                                         DeviceState *dev, Error **errp)
2847 {
2848     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2849 
2850     if (vms->platform_bus_dev) {
2851         MachineClass *mc = MACHINE_GET_CLASS(vms);
2852 
2853         if (device_is_dynamic_sysbus(mc, dev)) {
2854             platform_bus_link_device(PLATFORM_BUS_DEVICE(vms->platform_bus_dev),
2855                                      SYS_BUS_DEVICE(dev));
2856         }
2857     }
2858     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2859         virt_memory_plug(hotplug_dev, dev, errp);
2860     }
2861 
2862     if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI)) {
2863         virt_virtio_md_pci_plug(hotplug_dev, dev, errp);
2864     }
2865 
2866     if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2867         PCIDevice *pdev = PCI_DEVICE(dev);
2868 
2869         vms->iommu = VIRT_IOMMU_VIRTIO;
2870         vms->virtio_iommu_bdf = pci_get_bdf(pdev);
2871         create_virtio_iommu_dt_bindings(vms);
2872     }
2873 }
2874 
2875 static void virt_dimm_unplug_request(HotplugHandler *hotplug_dev,
2876                                      DeviceState *dev, Error **errp)
2877 {
2878     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2879 
2880     if (!vms->acpi_dev) {
2881         error_setg(errp,
2882                    "memory hotplug is not enabled: missing acpi-ged device");
2883         return;
2884     }
2885 
2886     if (object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM)) {
2887         error_setg(errp, "nvdimm device hot unplug is not supported yet.");
2888         return;
2889     }
2890 
2891     hotplug_handler_unplug_request(HOTPLUG_HANDLER(vms->acpi_dev), dev,
2892                                    errp);
2893 }
2894 
2895 static void virt_dimm_unplug(HotplugHandler *hotplug_dev,
2896                              DeviceState *dev, Error **errp)
2897 {
2898     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2899     Error *local_err = NULL;
2900 
2901     hotplug_handler_unplug(HOTPLUG_HANDLER(vms->acpi_dev), dev, &local_err);
2902     if (local_err) {
2903         goto out;
2904     }
2905 
2906     pc_dimm_unplug(PC_DIMM(dev), MACHINE(vms));
2907     qdev_unrealize(dev);
2908 
2909 out:
2910     error_propagate(errp, local_err);
2911 }
2912 
2913 static void virt_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
2914                                           DeviceState *dev, Error **errp)
2915 {
2916     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2917         virt_dimm_unplug_request(hotplug_dev, dev, errp);
2918     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI)) {
2919         virt_virtio_md_pci_unplug_request(hotplug_dev, dev, errp);
2920     } else {
2921         error_setg(errp, "device unplug request for unsupported device"
2922                    " type: %s", object_get_typename(OBJECT(dev)));
2923     }
2924 }
2925 
2926 static void virt_machine_device_unplug_cb(HotplugHandler *hotplug_dev,
2927                                           DeviceState *dev, Error **errp)
2928 {
2929     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2930         virt_dimm_unplug(hotplug_dev, dev, errp);
2931     } else {
2932         error_setg(errp, "virt: device unplug for unsupported device"
2933                    " type: %s", object_get_typename(OBJECT(dev)));
2934     }
2935 }
2936 
2937 static HotplugHandler *virt_machine_get_hotplug_handler(MachineState *machine,
2938                                                         DeviceState *dev)
2939 {
2940     MachineClass *mc = MACHINE_GET_CLASS(machine);
2941 
2942     if (device_is_dynamic_sysbus(mc, dev) ||
2943         object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) ||
2944         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MEM_PCI) ||
2945         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2946         return HOTPLUG_HANDLER(machine);
2947     }
2948     return NULL;
2949 }
2950 
2951 /*
2952  * for arm64 kvm_type [7-0] encodes the requested number of bits
2953  * in the IPA address space
2954  */
2955 static int virt_kvm_type(MachineState *ms, const char *type_str)
2956 {
2957     VirtMachineState *vms = VIRT_MACHINE(ms);
2958     int max_vm_pa_size, requested_pa_size;
2959     bool fixed_ipa;
2960 
2961     max_vm_pa_size = kvm_arm_get_max_vm_ipa_size(ms, &fixed_ipa);
2962 
2963     /* we freeze the memory map to compute the highest gpa */
2964     virt_set_memmap(vms, max_vm_pa_size);
2965 
2966     requested_pa_size = 64 - clz64(vms->highest_gpa);
2967 
2968     /*
2969      * KVM requires the IPA size to be at least 32 bits.
2970      */
2971     if (requested_pa_size < 32) {
2972         requested_pa_size = 32;
2973     }
2974 
2975     if (requested_pa_size > max_vm_pa_size) {
2976         error_report("-m and ,maxmem option values "
2977                      "require an IPA range (%d bits) larger than "
2978                      "the one supported by the host (%d bits)",
2979                      requested_pa_size, max_vm_pa_size);
2980         exit(1);
2981     }
2982     /*
2983      * We return the requested PA log size, unless KVM only supports
2984      * the implicit legacy 40b IPA setting, in which case the kvm_type
2985      * must be 0.
2986      */
2987     return fixed_ipa ? 0 : requested_pa_size;
2988 }
2989 
2990 static void virt_machine_class_init(ObjectClass *oc, void *data)
2991 {
2992     MachineClass *mc = MACHINE_CLASS(oc);
2993     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
2994 
2995     mc->init = machvirt_init;
2996     /* Start with max_cpus set to 512, which is the maximum supported by KVM.
2997      * The value may be reduced later when we have more information about the
2998      * configuration of the particular instance.
2999      */
3000     mc->max_cpus = 512;
3001     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_CALXEDA_XGMAC);
3002     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_AMD_XGBE);
3003     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_RAMFB_DEVICE);
3004     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_PLATFORM);
3005 #ifdef CONFIG_TPM
3006     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_TPM_TIS_SYSBUS);
3007 #endif
3008     mc->block_default_type = IF_VIRTIO;
3009     mc->no_cdrom = 1;
3010     mc->pci_allow_0_address = true;
3011     /* We know we will never create a pre-ARMv7 CPU which needs 1K pages */
3012     mc->minimum_page_bits = 12;
3013     mc->possible_cpu_arch_ids = virt_possible_cpu_arch_ids;
3014     mc->cpu_index_to_instance_props = virt_cpu_index_to_props;
3015 #ifdef CONFIG_TCG
3016     mc->default_cpu_type = ARM_CPU_TYPE_NAME("cortex-a15");
3017 #else
3018     mc->default_cpu_type = ARM_CPU_TYPE_NAME("max");
3019 #endif
3020     mc->get_default_cpu_node_id = virt_get_default_cpu_node_id;
3021     mc->kvm_type = virt_kvm_type;
3022     assert(!mc->get_hotplug_handler);
3023     mc->get_hotplug_handler = virt_machine_get_hotplug_handler;
3024     hc->pre_plug = virt_machine_device_pre_plug_cb;
3025     hc->plug = virt_machine_device_plug_cb;
3026     hc->unplug_request = virt_machine_device_unplug_request_cb;
3027     hc->unplug = virt_machine_device_unplug_cb;
3028     mc->nvdimm_supported = true;
3029     mc->smp_props.clusters_supported = true;
3030     mc->auto_enable_numa_with_memhp = true;
3031     mc->auto_enable_numa_with_memdev = true;
3032     /* platform instead of architectural choice */
3033     mc->cpu_cluster_has_numa_boundary = true;
3034     mc->default_ram_id = "mach-virt.ram";
3035     mc->default_nic = "virtio-net-pci";
3036 
3037     object_class_property_add(oc, "acpi", "OnOffAuto",
3038         virt_get_acpi, virt_set_acpi,
3039         NULL, NULL);
3040     object_class_property_set_description(oc, "acpi",
3041         "Enable ACPI");
3042     object_class_property_add_bool(oc, "secure", virt_get_secure,
3043                                    virt_set_secure);
3044     object_class_property_set_description(oc, "secure",
3045                                                 "Set on/off to enable/disable the ARM "
3046                                                 "Security Extensions (TrustZone)");
3047 
3048     object_class_property_add_bool(oc, "virtualization", virt_get_virt,
3049                                    virt_set_virt);
3050     object_class_property_set_description(oc, "virtualization",
3051                                           "Set on/off to enable/disable emulating a "
3052                                           "guest CPU which implements the ARM "
3053                                           "Virtualization Extensions");
3054 
3055     object_class_property_add_bool(oc, "highmem", virt_get_highmem,
3056                                    virt_set_highmem);
3057     object_class_property_set_description(oc, "highmem",
3058                                           "Set on/off to enable/disable using "
3059                                           "physical address space above 32 bits");
3060 
3061     object_class_property_add_bool(oc, "compact-highmem",
3062                                    virt_get_compact_highmem,
3063                                    virt_set_compact_highmem);
3064     object_class_property_set_description(oc, "compact-highmem",
3065                                           "Set on/off to enable/disable compact "
3066                                           "layout for high memory regions");
3067 
3068     object_class_property_add_bool(oc, "highmem-redists",
3069                                    virt_get_highmem_redists,
3070                                    virt_set_highmem_redists);
3071     object_class_property_set_description(oc, "highmem-redists",
3072                                           "Set on/off to enable/disable high "
3073                                           "memory region for GICv3 or GICv4 "
3074                                           "redistributor");
3075 
3076     object_class_property_add_bool(oc, "highmem-ecam",
3077                                    virt_get_highmem_ecam,
3078                                    virt_set_highmem_ecam);
3079     object_class_property_set_description(oc, "highmem-ecam",
3080                                           "Set on/off to enable/disable high "
3081                                           "memory region for PCI ECAM");
3082 
3083     object_class_property_add_bool(oc, "highmem-mmio",
3084                                    virt_get_highmem_mmio,
3085                                    virt_set_highmem_mmio);
3086     object_class_property_set_description(oc, "highmem-mmio",
3087                                           "Set on/off to enable/disable high "
3088                                           "memory region for PCI MMIO");
3089 
3090     object_class_property_add_str(oc, "gic-version", virt_get_gic_version,
3091                                   virt_set_gic_version);
3092     object_class_property_set_description(oc, "gic-version",
3093                                           "Set GIC version. "
3094                                           "Valid values are 2, 3, 4, host and max");
3095 
3096     object_class_property_add_str(oc, "iommu", virt_get_iommu, virt_set_iommu);
3097     object_class_property_set_description(oc, "iommu",
3098                                           "Set the IOMMU type. "
3099                                           "Valid values are none and smmuv3");
3100 
3101     object_class_property_add_bool(oc, "default-bus-bypass-iommu",
3102                                    virt_get_default_bus_bypass_iommu,
3103                                    virt_set_default_bus_bypass_iommu);
3104     object_class_property_set_description(oc, "default-bus-bypass-iommu",
3105                                           "Set on/off to enable/disable "
3106                                           "bypass_iommu for default root bus");
3107 
3108     object_class_property_add_bool(oc, "ras", virt_get_ras,
3109                                    virt_set_ras);
3110     object_class_property_set_description(oc, "ras",
3111                                           "Set on/off to enable/disable reporting host memory errors "
3112                                           "to a KVM guest using ACPI and guest external abort exceptions");
3113 
3114     object_class_property_add_bool(oc, "mte", virt_get_mte, virt_set_mte);
3115     object_class_property_set_description(oc, "mte",
3116                                           "Set on/off to enable/disable emulating a "
3117                                           "guest CPU which implements the ARM "
3118                                           "Memory Tagging Extension");
3119 
3120     object_class_property_add_bool(oc, "its", virt_get_its,
3121                                    virt_set_its);
3122     object_class_property_set_description(oc, "its",
3123                                           "Set on/off to enable/disable "
3124                                           "ITS instantiation");
3125 
3126     object_class_property_add_bool(oc, "dtb-randomness",
3127                                    virt_get_dtb_randomness,
3128                                    virt_set_dtb_randomness);
3129     object_class_property_set_description(oc, "dtb-randomness",
3130                                           "Set off to disable passing random or "
3131                                           "non-deterministic dtb nodes to guest");
3132 
3133     object_class_property_add_bool(oc, "dtb-kaslr-seed",
3134                                    virt_get_dtb_randomness,
3135                                    virt_set_dtb_randomness);
3136     object_class_property_set_description(oc, "dtb-kaslr-seed",
3137                                           "Deprecated synonym of dtb-randomness");
3138 
3139     object_class_property_add_str(oc, "x-oem-id",
3140                                   virt_get_oem_id,
3141                                   virt_set_oem_id);
3142     object_class_property_set_description(oc, "x-oem-id",
3143                                           "Override the default value of field OEMID "
3144                                           "in ACPI table header."
3145                                           "The string may be up to 6 bytes in size");
3146 
3147 
3148     object_class_property_add_str(oc, "x-oem-table-id",
3149                                   virt_get_oem_table_id,
3150                                   virt_set_oem_table_id);
3151     object_class_property_set_description(oc, "x-oem-table-id",
3152                                           "Override the default value of field OEM Table ID "
3153                                           "in ACPI table header."
3154                                           "The string may be up to 8 bytes in size");
3155 
3156 }
3157 
3158 static void virt_instance_init(Object *obj)
3159 {
3160     VirtMachineState *vms = VIRT_MACHINE(obj);
3161     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
3162 
3163     /* EL3 is disabled by default on virt: this makes us consistent
3164      * between KVM and TCG for this board, and it also allows us to
3165      * boot UEFI blobs which assume no TrustZone support.
3166      */
3167     vms->secure = false;
3168 
3169     /* EL2 is also disabled by default, for similar reasons */
3170     vms->virt = false;
3171 
3172     /* High memory is enabled by default */
3173     vms->highmem = true;
3174     vms->highmem_compact = !vmc->no_highmem_compact;
3175     vms->gic_version = VIRT_GIC_VERSION_NOSEL;
3176 
3177     vms->highmem_ecam = !vmc->no_highmem_ecam;
3178     vms->highmem_mmio = true;
3179     vms->highmem_redists = true;
3180 
3181     if (vmc->no_its) {
3182         vms->its = false;
3183     } else {
3184         /* Default allows ITS instantiation */
3185         vms->its = true;
3186 
3187         if (vmc->no_tcg_its) {
3188             vms->tcg_its = false;
3189         } else {
3190             vms->tcg_its = true;
3191         }
3192     }
3193 
3194     /* Default disallows iommu instantiation */
3195     vms->iommu = VIRT_IOMMU_NONE;
3196 
3197     /* The default root bus is attached to iommu by default */
3198     vms->default_bus_bypass_iommu = false;
3199 
3200     /* Default disallows RAS instantiation */
3201     vms->ras = false;
3202 
3203     /* MTE is disabled by default.  */
3204     vms->mte = false;
3205 
3206     /* Supply kaslr-seed and rng-seed by default */
3207     vms->dtb_randomness = true;
3208 
3209     vms->irqmap = a15irqmap;
3210 
3211     virt_flash_create(vms);
3212 
3213     vms->oem_id = g_strndup(ACPI_BUILD_APPNAME6, 6);
3214     vms->oem_table_id = g_strndup(ACPI_BUILD_APPNAME8, 8);
3215 }
3216 
3217 static const TypeInfo virt_machine_info = {
3218     .name          = TYPE_VIRT_MACHINE,
3219     .parent        = TYPE_MACHINE,
3220     .abstract      = true,
3221     .instance_size = sizeof(VirtMachineState),
3222     .class_size    = sizeof(VirtMachineClass),
3223     .class_init    = virt_machine_class_init,
3224     .instance_init = virt_instance_init,
3225     .interfaces = (InterfaceInfo[]) {
3226          { TYPE_HOTPLUG_HANDLER },
3227          { }
3228     },
3229 };
3230 
3231 static void machvirt_machine_init(void)
3232 {
3233     type_register_static(&virt_machine_info);
3234 }
3235 type_init(machvirt_machine_init);
3236 
3237 static void virt_machine_8_1_options(MachineClass *mc)
3238 {
3239 }
3240 DEFINE_VIRT_MACHINE_AS_LATEST(8, 1)
3241 
3242 static void virt_machine_8_0_options(MachineClass *mc)
3243 {
3244     virt_machine_8_1_options(mc);
3245     compat_props_add(mc->compat_props, hw_compat_8_0, hw_compat_8_0_len);
3246 }
3247 DEFINE_VIRT_MACHINE(8, 0)
3248 
3249 static void virt_machine_7_2_options(MachineClass *mc)
3250 {
3251     virt_machine_8_0_options(mc);
3252     compat_props_add(mc->compat_props, hw_compat_7_2, hw_compat_7_2_len);
3253 }
3254 DEFINE_VIRT_MACHINE(7, 2)
3255 
3256 static void virt_machine_7_1_options(MachineClass *mc)
3257 {
3258     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3259 
3260     virt_machine_7_2_options(mc);
3261     compat_props_add(mc->compat_props, hw_compat_7_1, hw_compat_7_1_len);
3262     /* Compact layout for high memory regions was introduced with 7.2 */
3263     vmc->no_highmem_compact = true;
3264 }
3265 DEFINE_VIRT_MACHINE(7, 1)
3266 
3267 static void virt_machine_7_0_options(MachineClass *mc)
3268 {
3269     virt_machine_7_1_options(mc);
3270     compat_props_add(mc->compat_props, hw_compat_7_0, hw_compat_7_0_len);
3271 }
3272 DEFINE_VIRT_MACHINE(7, 0)
3273 
3274 static void virt_machine_6_2_options(MachineClass *mc)
3275 {
3276     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3277 
3278     virt_machine_7_0_options(mc);
3279     compat_props_add(mc->compat_props, hw_compat_6_2, hw_compat_6_2_len);
3280     vmc->no_tcg_lpa2 = true;
3281 }
3282 DEFINE_VIRT_MACHINE(6, 2)
3283 
3284 static void virt_machine_6_1_options(MachineClass *mc)
3285 {
3286     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3287 
3288     virt_machine_6_2_options(mc);
3289     compat_props_add(mc->compat_props, hw_compat_6_1, hw_compat_6_1_len);
3290     mc->smp_props.prefer_sockets = true;
3291     vmc->no_cpu_topology = true;
3292 
3293     /* qemu ITS was introduced with 6.2 */
3294     vmc->no_tcg_its = true;
3295 }
3296 DEFINE_VIRT_MACHINE(6, 1)
3297 
3298 static void virt_machine_6_0_options(MachineClass *mc)
3299 {
3300     virt_machine_6_1_options(mc);
3301     compat_props_add(mc->compat_props, hw_compat_6_0, hw_compat_6_0_len);
3302 }
3303 DEFINE_VIRT_MACHINE(6, 0)
3304 
3305 static void virt_machine_5_2_options(MachineClass *mc)
3306 {
3307     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3308 
3309     virt_machine_6_0_options(mc);
3310     compat_props_add(mc->compat_props, hw_compat_5_2, hw_compat_5_2_len);
3311     vmc->no_secure_gpio = true;
3312 }
3313 DEFINE_VIRT_MACHINE(5, 2)
3314 
3315 static void virt_machine_5_1_options(MachineClass *mc)
3316 {
3317     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3318 
3319     virt_machine_5_2_options(mc);
3320     compat_props_add(mc->compat_props, hw_compat_5_1, hw_compat_5_1_len);
3321     vmc->no_kvm_steal_time = true;
3322 }
3323 DEFINE_VIRT_MACHINE(5, 1)
3324 
3325 static void virt_machine_5_0_options(MachineClass *mc)
3326 {
3327     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3328 
3329     virt_machine_5_1_options(mc);
3330     compat_props_add(mc->compat_props, hw_compat_5_0, hw_compat_5_0_len);
3331     mc->numa_mem_supported = true;
3332     vmc->acpi_expose_flash = true;
3333     mc->auto_enable_numa_with_memdev = false;
3334 }
3335 DEFINE_VIRT_MACHINE(5, 0)
3336 
3337 static void virt_machine_4_2_options(MachineClass *mc)
3338 {
3339     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3340 
3341     virt_machine_5_0_options(mc);
3342     compat_props_add(mc->compat_props, hw_compat_4_2, hw_compat_4_2_len);
3343     vmc->kvm_no_adjvtime = true;
3344 }
3345 DEFINE_VIRT_MACHINE(4, 2)
3346 
3347 static void virt_machine_4_1_options(MachineClass *mc)
3348 {
3349     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3350 
3351     virt_machine_4_2_options(mc);
3352     compat_props_add(mc->compat_props, hw_compat_4_1, hw_compat_4_1_len);
3353     vmc->no_ged = true;
3354     mc->auto_enable_numa_with_memhp = false;
3355 }
3356 DEFINE_VIRT_MACHINE(4, 1)
3357 
3358 static void virt_machine_4_0_options(MachineClass *mc)
3359 {
3360     virt_machine_4_1_options(mc);
3361     compat_props_add(mc->compat_props, hw_compat_4_0, hw_compat_4_0_len);
3362 }
3363 DEFINE_VIRT_MACHINE(4, 0)
3364 
3365 static void virt_machine_3_1_options(MachineClass *mc)
3366 {
3367     virt_machine_4_0_options(mc);
3368     compat_props_add(mc->compat_props, hw_compat_3_1, hw_compat_3_1_len);
3369 }
3370 DEFINE_VIRT_MACHINE(3, 1)
3371 
3372 static void virt_machine_3_0_options(MachineClass *mc)
3373 {
3374     virt_machine_3_1_options(mc);
3375     compat_props_add(mc->compat_props, hw_compat_3_0, hw_compat_3_0_len);
3376 }
3377 DEFINE_VIRT_MACHINE(3, 0)
3378 
3379 static void virt_machine_2_12_options(MachineClass *mc)
3380 {
3381     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3382 
3383     virt_machine_3_0_options(mc);
3384     compat_props_add(mc->compat_props, hw_compat_2_12, hw_compat_2_12_len);
3385     vmc->no_highmem_ecam = true;
3386     mc->max_cpus = 255;
3387 }
3388 DEFINE_VIRT_MACHINE(2, 12)
3389 
3390 static void virt_machine_2_11_options(MachineClass *mc)
3391 {
3392     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3393 
3394     virt_machine_2_12_options(mc);
3395     compat_props_add(mc->compat_props, hw_compat_2_11, hw_compat_2_11_len);
3396     vmc->smbios_old_sys_ver = true;
3397 }
3398 DEFINE_VIRT_MACHINE(2, 11)
3399 
3400 static void virt_machine_2_10_options(MachineClass *mc)
3401 {
3402     virt_machine_2_11_options(mc);
3403     compat_props_add(mc->compat_props, hw_compat_2_10, hw_compat_2_10_len);
3404     /* before 2.11 we never faulted accesses to bad addresses */
3405     mc->ignore_memory_transaction_failures = true;
3406 }
3407 DEFINE_VIRT_MACHINE(2, 10)
3408 
3409 static void virt_machine_2_9_options(MachineClass *mc)
3410 {
3411     virt_machine_2_10_options(mc);
3412     compat_props_add(mc->compat_props, hw_compat_2_9, hw_compat_2_9_len);
3413 }
3414 DEFINE_VIRT_MACHINE(2, 9)
3415 
3416 static void virt_machine_2_8_options(MachineClass *mc)
3417 {
3418     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3419 
3420     virt_machine_2_9_options(mc);
3421     compat_props_add(mc->compat_props, hw_compat_2_8, hw_compat_2_8_len);
3422     /* For 2.8 and earlier we falsely claimed in the DT that
3423      * our timers were edge-triggered, not level-triggered.
3424      */
3425     vmc->claim_edge_triggered_timers = true;
3426 }
3427 DEFINE_VIRT_MACHINE(2, 8)
3428 
3429 static void virt_machine_2_7_options(MachineClass *mc)
3430 {
3431     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3432 
3433     virt_machine_2_8_options(mc);
3434     compat_props_add(mc->compat_props, hw_compat_2_7, hw_compat_2_7_len);
3435     /* ITS was introduced with 2.8 */
3436     vmc->no_its = true;
3437     /* Stick with 1K pages for migration compatibility */
3438     mc->minimum_page_bits = 0;
3439 }
3440 DEFINE_VIRT_MACHINE(2, 7)
3441 
3442 static void virt_machine_2_6_options(MachineClass *mc)
3443 {
3444     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3445 
3446     virt_machine_2_7_options(mc);
3447     compat_props_add(mc->compat_props, hw_compat_2_6, hw_compat_2_6_len);
3448     vmc->disallow_affinity_adjustment = true;
3449     /* Disable PMU for 2.6 as PMU support was first introduced in 2.7 */
3450     vmc->no_pmu = true;
3451 }
3452 DEFINE_VIRT_MACHINE(2, 6)
3453