xref: /openbmc/qemu/hw/arm/virt.c (revision 6c1ebe75)
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-common.h"
33 #include "qemu/datadir.h"
34 #include "qemu/units.h"
35 #include "qemu/option.h"
36 #include "monitor/qdev.h"
37 #include "qapi/error.h"
38 #include "hw/sysbus.h"
39 #include "hw/arm/boot.h"
40 #include "hw/arm/primecell.h"
41 #include "hw/arm/virt.h"
42 #include "hw/block/flash.h"
43 #include "hw/vfio/vfio-calxeda-xgmac.h"
44 #include "hw/vfio/vfio-amd-xgbe.h"
45 #include "hw/display/ramfb.h"
46 #include "net/net.h"
47 #include "sysemu/device_tree.h"
48 #include "sysemu/numa.h"
49 #include "sysemu/runstate.h"
50 #include "sysemu/tpm.h"
51 #include "sysemu/kvm.h"
52 #include "hw/loader.h"
53 #include "qapi/error.h"
54 #include "qemu/bitops.h"
55 #include "qemu/error-report.h"
56 #include "qemu/module.h"
57 #include "hw/pci-host/gpex.h"
58 #include "hw/virtio/virtio-pci.h"
59 #include "hw/arm/sysbus-fdt.h"
60 #include "hw/platform-bus.h"
61 #include "hw/qdev-properties.h"
62 #include "hw/arm/fdt.h"
63 #include "hw/intc/arm_gic.h"
64 #include "hw/intc/arm_gicv3_common.h"
65 #include "hw/irq.h"
66 #include "kvm_arm.h"
67 #include "hw/firmware/smbios.h"
68 #include "qapi/visitor.h"
69 #include "qapi/qapi-visit-common.h"
70 #include "standard-headers/linux/input.h"
71 #include "hw/arm/smmuv3.h"
72 #include "hw/acpi/acpi.h"
73 #include "target/arm/internals.h"
74 #include "hw/mem/pc-dimm.h"
75 #include "hw/mem/nvdimm.h"
76 #include "hw/acpi/generic_event_device.h"
77 #include "hw/virtio/virtio-iommu.h"
78 #include "hw/char/pl011.h"
79 #include "qemu/guest-random.h"
80 
81 #define DEFINE_VIRT_MACHINE_LATEST(major, minor, latest) \
82     static void virt_##major##_##minor##_class_init(ObjectClass *oc, \
83                                                     void *data) \
84     { \
85         MachineClass *mc = MACHINE_CLASS(oc); \
86         virt_machine_##major##_##minor##_options(mc); \
87         mc->desc = "QEMU " # major "." # minor " ARM Virtual Machine"; \
88         if (latest) { \
89             mc->alias = "virt"; \
90         } \
91     } \
92     static const TypeInfo machvirt_##major##_##minor##_info = { \
93         .name = MACHINE_TYPE_NAME("virt-" # major "." # minor), \
94         .parent = TYPE_VIRT_MACHINE, \
95         .class_init = virt_##major##_##minor##_class_init, \
96     }; \
97     static void machvirt_machine_##major##_##minor##_init(void) \
98     { \
99         type_register_static(&machvirt_##major##_##minor##_info); \
100     } \
101     type_init(machvirt_machine_##major##_##minor##_init);
102 
103 #define DEFINE_VIRT_MACHINE_AS_LATEST(major, minor) \
104     DEFINE_VIRT_MACHINE_LATEST(major, minor, true)
105 #define DEFINE_VIRT_MACHINE(major, minor) \
106     DEFINE_VIRT_MACHINE_LATEST(major, minor, false)
107 
108 
109 /* Number of external interrupt lines to configure the GIC with */
110 #define NUM_IRQS 256
111 
112 #define PLATFORM_BUS_NUM_IRQS 64
113 
114 /* Legacy RAM limit in GB (< version 4.0) */
115 #define LEGACY_RAMLIMIT_GB 255
116 #define LEGACY_RAMLIMIT_BYTES (LEGACY_RAMLIMIT_GB * GiB)
117 
118 /* Addresses and sizes of our components.
119  * 0..128MB is space for a flash device so we can run bootrom code such as UEFI.
120  * 128MB..256MB is used for miscellaneous device I/O.
121  * 256MB..1GB is reserved for possible future PCI support (ie where the
122  * PCI memory window will go if we add a PCI host controller).
123  * 1GB and up is RAM (which may happily spill over into the
124  * high memory region beyond 4GB).
125  * This represents a compromise between how much RAM can be given to
126  * a 32 bit VM and leaving space for expansion and in particular for PCI.
127  * Note that devices should generally be placed at multiples of 0x10000,
128  * to accommodate guests using 64K pages.
129  */
130 static const MemMapEntry base_memmap[] = {
131     /* Space up to 0x8000000 is reserved for a boot ROM */
132     [VIRT_FLASH] =              {          0, 0x08000000 },
133     [VIRT_CPUPERIPHS] =         { 0x08000000, 0x00020000 },
134     /* GIC distributor and CPU interfaces sit inside the CPU peripheral space */
135     [VIRT_GIC_DIST] =           { 0x08000000, 0x00010000 },
136     [VIRT_GIC_CPU] =            { 0x08010000, 0x00010000 },
137     [VIRT_GIC_V2M] =            { 0x08020000, 0x00001000 },
138     [VIRT_GIC_HYP] =            { 0x08030000, 0x00010000 },
139     [VIRT_GIC_VCPU] =           { 0x08040000, 0x00010000 },
140     /* The space in between here is reserved for GICv3 CPU/vCPU/HYP */
141     [VIRT_GIC_ITS] =            { 0x08080000, 0x00020000 },
142     /* This redistributor space allows up to 2*64kB*123 CPUs */
143     [VIRT_GIC_REDIST] =         { 0x080A0000, 0x00F60000 },
144     [VIRT_UART] =               { 0x09000000, 0x00001000 },
145     [VIRT_RTC] =                { 0x09010000, 0x00001000 },
146     [VIRT_FW_CFG] =             { 0x09020000, 0x00000018 },
147     [VIRT_GPIO] =               { 0x09030000, 0x00001000 },
148     [VIRT_SECURE_UART] =        { 0x09040000, 0x00001000 },
149     [VIRT_SMMU] =               { 0x09050000, 0x00020000 },
150     [VIRT_PCDIMM_ACPI] =        { 0x09070000, MEMORY_HOTPLUG_IO_LEN },
151     [VIRT_ACPI_GED] =           { 0x09080000, ACPI_GED_EVT_SEL_LEN },
152     [VIRT_NVDIMM_ACPI] =        { 0x09090000, NVDIMM_ACPI_IO_LEN},
153     [VIRT_PVTIME] =             { 0x090a0000, 0x00010000 },
154     [VIRT_SECURE_GPIO] =        { 0x090b0000, 0x00001000 },
155     [VIRT_MMIO] =               { 0x0a000000, 0x00000200 },
156     /* ...repeating for a total of NUM_VIRTIO_TRANSPORTS, each of that size */
157     [VIRT_PLATFORM_BUS] =       { 0x0c000000, 0x02000000 },
158     [VIRT_SECURE_MEM] =         { 0x0e000000, 0x01000000 },
159     [VIRT_PCIE_MMIO] =          { 0x10000000, 0x2eff0000 },
160     [VIRT_PCIE_PIO] =           { 0x3eff0000, 0x00010000 },
161     [VIRT_PCIE_ECAM] =          { 0x3f000000, 0x01000000 },
162     /* Actual RAM size depends on initial RAM and device memory settings */
163     [VIRT_MEM] =                { GiB, LEGACY_RAMLIMIT_BYTES },
164 };
165 
166 /*
167  * Highmem IO Regions: This memory map is floating, located after the RAM.
168  * Each MemMapEntry base (GPA) will be dynamically computed, depending on the
169  * top of the RAM, so that its base get the same alignment as the size,
170  * ie. a 512GiB entry will be aligned on a 512GiB boundary. If there is
171  * less than 256GiB of RAM, the floating area starts at the 256GiB mark.
172  * Note the extended_memmap is sized so that it eventually also includes the
173  * base_memmap entries (VIRT_HIGH_GIC_REDIST2 index is greater than the last
174  * index of base_memmap).
175  */
176 static MemMapEntry extended_memmap[] = {
177     /* Additional 64 MB redist region (can contain up to 512 redistributors) */
178     [VIRT_HIGH_GIC_REDIST2] =   { 0x0, 64 * MiB },
179     [VIRT_HIGH_PCIE_ECAM] =     { 0x0, 256 * MiB },
180     /* Second PCIe window */
181     [VIRT_HIGH_PCIE_MMIO] =     { 0x0, 512 * GiB },
182 };
183 
184 static const int a15irqmap[] = {
185     [VIRT_UART] = 1,
186     [VIRT_RTC] = 2,
187     [VIRT_PCIE] = 3, /* ... to 6 */
188     [VIRT_GPIO] = 7,
189     [VIRT_SECURE_UART] = 8,
190     [VIRT_ACPI_GED] = 9,
191     [VIRT_MMIO] = 16, /* ...to 16 + NUM_VIRTIO_TRANSPORTS - 1 */
192     [VIRT_GIC_V2M] = 48, /* ...to 48 + NUM_GICV2M_SPIS - 1 */
193     [VIRT_SMMU] = 74,    /* ...to 74 + NUM_SMMU_IRQS - 1 */
194     [VIRT_PLATFORM_BUS] = 112, /* ...to 112 + PLATFORM_BUS_NUM_IRQS -1 */
195 };
196 
197 static const char *valid_cpus[] = {
198     ARM_CPU_TYPE_NAME("cortex-a7"),
199     ARM_CPU_TYPE_NAME("cortex-a15"),
200     ARM_CPU_TYPE_NAME("cortex-a53"),
201     ARM_CPU_TYPE_NAME("cortex-a57"),
202     ARM_CPU_TYPE_NAME("cortex-a72"),
203     ARM_CPU_TYPE_NAME("a64fx"),
204     ARM_CPU_TYPE_NAME("host"),
205     ARM_CPU_TYPE_NAME("max"),
206 };
207 
208 static bool cpu_type_valid(const char *cpu)
209 {
210     int i;
211 
212     for (i = 0; i < ARRAY_SIZE(valid_cpus); i++) {
213         if (strcmp(cpu, valid_cpus[i]) == 0) {
214             return true;
215         }
216     }
217     return false;
218 }
219 
220 static void create_kaslr_seed(MachineState *ms, const char *node)
221 {
222     uint64_t seed;
223 
224     if (qemu_guest_getrandom(&seed, sizeof(seed), NULL)) {
225         return;
226     }
227     qemu_fdt_setprop_u64(ms->fdt, node, "kaslr-seed", seed);
228 }
229 
230 static void create_fdt(VirtMachineState *vms)
231 {
232     MachineState *ms = MACHINE(vms);
233     int nb_numa_nodes = ms->numa_state->num_nodes;
234     void *fdt = create_device_tree(&vms->fdt_size);
235 
236     if (!fdt) {
237         error_report("create_device_tree() failed");
238         exit(1);
239     }
240 
241     ms->fdt = fdt;
242 
243     /* Header */
244     qemu_fdt_setprop_string(fdt, "/", "compatible", "linux,dummy-virt");
245     qemu_fdt_setprop_cell(fdt, "/", "#address-cells", 0x2);
246     qemu_fdt_setprop_cell(fdt, "/", "#size-cells", 0x2);
247 
248     /* /chosen must exist for load_dtb to fill in necessary properties later */
249     qemu_fdt_add_subnode(fdt, "/chosen");
250     create_kaslr_seed(ms, "/chosen");
251 
252     if (vms->secure) {
253         qemu_fdt_add_subnode(fdt, "/secure-chosen");
254         create_kaslr_seed(ms, "/secure-chosen");
255     }
256 
257     /* Clock node, for the benefit of the UART. The kernel device tree
258      * binding documentation claims the PL011 node clock properties are
259      * optional but in practice if you omit them the kernel refuses to
260      * probe for the device.
261      */
262     vms->clock_phandle = qemu_fdt_alloc_phandle(fdt);
263     qemu_fdt_add_subnode(fdt, "/apb-pclk");
264     qemu_fdt_setprop_string(fdt, "/apb-pclk", "compatible", "fixed-clock");
265     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "#clock-cells", 0x0);
266     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "clock-frequency", 24000000);
267     qemu_fdt_setprop_string(fdt, "/apb-pclk", "clock-output-names",
268                                 "clk24mhz");
269     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "phandle", vms->clock_phandle);
270 
271     if (nb_numa_nodes > 0 && ms->numa_state->have_numa_distance) {
272         int size = nb_numa_nodes * nb_numa_nodes * 3 * sizeof(uint32_t);
273         uint32_t *matrix = g_malloc0(size);
274         int idx, i, j;
275 
276         for (i = 0; i < nb_numa_nodes; i++) {
277             for (j = 0; j < nb_numa_nodes; j++) {
278                 idx = (i * nb_numa_nodes + j) * 3;
279                 matrix[idx + 0] = cpu_to_be32(i);
280                 matrix[idx + 1] = cpu_to_be32(j);
281                 matrix[idx + 2] =
282                     cpu_to_be32(ms->numa_state->nodes[i].distance[j]);
283             }
284         }
285 
286         qemu_fdt_add_subnode(fdt, "/distance-map");
287         qemu_fdt_setprop_string(fdt, "/distance-map", "compatible",
288                                 "numa-distance-map-v1");
289         qemu_fdt_setprop(fdt, "/distance-map", "distance-matrix",
290                          matrix, size);
291         g_free(matrix);
292     }
293 }
294 
295 static void fdt_add_timer_nodes(const VirtMachineState *vms)
296 {
297     /* On real hardware these interrupts are level-triggered.
298      * On KVM they were edge-triggered before host kernel version 4.4,
299      * and level-triggered afterwards.
300      * On emulated QEMU they are level-triggered.
301      *
302      * Getting the DTB info about them wrong is awkward for some
303      * guest kernels:
304      *  pre-4.8 ignore the DT and leave the interrupt configured
305      *   with whatever the GIC reset value (or the bootloader) left it at
306      *  4.8 before rc6 honour the incorrect data by programming it back
307      *   into the GIC, causing problems
308      *  4.8rc6 and later ignore the DT and always write "level triggered"
309      *   into the GIC
310      *
311      * For backwards-compatibility, virt-2.8 and earlier will continue
312      * to say these are edge-triggered, but later machines will report
313      * the correct information.
314      */
315     ARMCPU *armcpu;
316     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
317     uint32_t irqflags = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
318     MachineState *ms = MACHINE(vms);
319 
320     if (vmc->claim_edge_triggered_timers) {
321         irqflags = GIC_FDT_IRQ_FLAGS_EDGE_LO_HI;
322     }
323 
324     if (vms->gic_version == VIRT_GIC_VERSION_2) {
325         irqflags = deposit32(irqflags, GIC_FDT_IRQ_PPI_CPU_START,
326                              GIC_FDT_IRQ_PPI_CPU_WIDTH,
327                              (1 << MACHINE(vms)->smp.cpus) - 1);
328     }
329 
330     qemu_fdt_add_subnode(ms->fdt, "/timer");
331 
332     armcpu = ARM_CPU(qemu_get_cpu(0));
333     if (arm_feature(&armcpu->env, ARM_FEATURE_V8)) {
334         const char compat[] = "arm,armv8-timer\0arm,armv7-timer";
335         qemu_fdt_setprop(ms->fdt, "/timer", "compatible",
336                          compat, sizeof(compat));
337     } else {
338         qemu_fdt_setprop_string(ms->fdt, "/timer", "compatible",
339                                 "arm,armv7-timer");
340     }
341     qemu_fdt_setprop(ms->fdt, "/timer", "always-on", NULL, 0);
342     qemu_fdt_setprop_cells(ms->fdt, "/timer", "interrupts",
343                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_S_EL1_IRQ, irqflags,
344                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_NS_EL1_IRQ, irqflags,
345                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_VIRT_IRQ, irqflags,
346                        GIC_FDT_IRQ_TYPE_PPI, ARCH_TIMER_NS_EL2_IRQ, irqflags);
347 }
348 
349 static void fdt_add_cpu_nodes(const VirtMachineState *vms)
350 {
351     int cpu;
352     int addr_cells = 1;
353     const MachineState *ms = MACHINE(vms);
354     int smp_cpus = ms->smp.cpus;
355 
356     /*
357      * From Documentation/devicetree/bindings/arm/cpus.txt
358      *  On ARM v8 64-bit systems value should be set to 2,
359      *  that corresponds to the MPIDR_EL1 register size.
360      *  If MPIDR_EL1[63:32] value is equal to 0 on all CPUs
361      *  in the system, #address-cells can be set to 1, since
362      *  MPIDR_EL1[63:32] bits are not used for CPUs
363      *  identification.
364      *
365      *  Here we actually don't know whether our system is 32- or 64-bit one.
366      *  The simplest way to go is to examine affinity IDs of all our CPUs. If
367      *  at least one of them has Aff3 populated, we set #address-cells to 2.
368      */
369     for (cpu = 0; cpu < smp_cpus; cpu++) {
370         ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(cpu));
371 
372         if (armcpu->mp_affinity & ARM_AFF3_MASK) {
373             addr_cells = 2;
374             break;
375         }
376     }
377 
378     qemu_fdt_add_subnode(ms->fdt, "/cpus");
379     qemu_fdt_setprop_cell(ms->fdt, "/cpus", "#address-cells", addr_cells);
380     qemu_fdt_setprop_cell(ms->fdt, "/cpus", "#size-cells", 0x0);
381 
382     for (cpu = smp_cpus - 1; cpu >= 0; cpu--) {
383         char *nodename = g_strdup_printf("/cpus/cpu@%d", cpu);
384         ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(cpu));
385         CPUState *cs = CPU(armcpu);
386 
387         qemu_fdt_add_subnode(ms->fdt, nodename);
388         qemu_fdt_setprop_string(ms->fdt, nodename, "device_type", "cpu");
389         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
390                                     armcpu->dtb_compatible);
391 
392         if (vms->psci_conduit != QEMU_PSCI_CONDUIT_DISABLED && smp_cpus > 1) {
393             qemu_fdt_setprop_string(ms->fdt, nodename,
394                                         "enable-method", "psci");
395         }
396 
397         if (addr_cells == 2) {
398             qemu_fdt_setprop_u64(ms->fdt, nodename, "reg",
399                                  armcpu->mp_affinity);
400         } else {
401             qemu_fdt_setprop_cell(ms->fdt, nodename, "reg",
402                                   armcpu->mp_affinity);
403         }
404 
405         if (ms->possible_cpus->cpus[cs->cpu_index].props.has_node_id) {
406             qemu_fdt_setprop_cell(ms->fdt, nodename, "numa-node-id",
407                 ms->possible_cpus->cpus[cs->cpu_index].props.node_id);
408         }
409 
410         g_free(nodename);
411     }
412 }
413 
414 static void fdt_add_its_gic_node(VirtMachineState *vms)
415 {
416     char *nodename;
417     MachineState *ms = MACHINE(vms);
418 
419     vms->msi_phandle = qemu_fdt_alloc_phandle(ms->fdt);
420     nodename = g_strdup_printf("/intc/its@%" PRIx64,
421                                vms->memmap[VIRT_GIC_ITS].base);
422     qemu_fdt_add_subnode(ms->fdt, nodename);
423     qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
424                             "arm,gic-v3-its");
425     qemu_fdt_setprop(ms->fdt, nodename, "msi-controller", NULL, 0);
426     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
427                                  2, vms->memmap[VIRT_GIC_ITS].base,
428                                  2, vms->memmap[VIRT_GIC_ITS].size);
429     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", vms->msi_phandle);
430     g_free(nodename);
431 }
432 
433 static void fdt_add_v2m_gic_node(VirtMachineState *vms)
434 {
435     MachineState *ms = MACHINE(vms);
436     char *nodename;
437 
438     nodename = g_strdup_printf("/intc/v2m@%" PRIx64,
439                                vms->memmap[VIRT_GIC_V2M].base);
440     vms->msi_phandle = qemu_fdt_alloc_phandle(ms->fdt);
441     qemu_fdt_add_subnode(ms->fdt, nodename);
442     qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
443                             "arm,gic-v2m-frame");
444     qemu_fdt_setprop(ms->fdt, nodename, "msi-controller", NULL, 0);
445     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
446                                  2, vms->memmap[VIRT_GIC_V2M].base,
447                                  2, vms->memmap[VIRT_GIC_V2M].size);
448     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", vms->msi_phandle);
449     g_free(nodename);
450 }
451 
452 static void fdt_add_gic_node(VirtMachineState *vms)
453 {
454     MachineState *ms = MACHINE(vms);
455     char *nodename;
456 
457     vms->gic_phandle = qemu_fdt_alloc_phandle(ms->fdt);
458     qemu_fdt_setprop_cell(ms->fdt, "/", "interrupt-parent", vms->gic_phandle);
459 
460     nodename = g_strdup_printf("/intc@%" PRIx64,
461                                vms->memmap[VIRT_GIC_DIST].base);
462     qemu_fdt_add_subnode(ms->fdt, nodename);
463     qemu_fdt_setprop_cell(ms->fdt, nodename, "#interrupt-cells", 3);
464     qemu_fdt_setprop(ms->fdt, nodename, "interrupt-controller", NULL, 0);
465     qemu_fdt_setprop_cell(ms->fdt, nodename, "#address-cells", 0x2);
466     qemu_fdt_setprop_cell(ms->fdt, nodename, "#size-cells", 0x2);
467     qemu_fdt_setprop(ms->fdt, nodename, "ranges", NULL, 0);
468     if (vms->gic_version == VIRT_GIC_VERSION_3) {
469         int nb_redist_regions = virt_gicv3_redist_region_count(vms);
470 
471         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
472                                 "arm,gic-v3");
473 
474         qemu_fdt_setprop_cell(ms->fdt, nodename,
475                               "#redistributor-regions", nb_redist_regions);
476 
477         if (nb_redist_regions == 1) {
478             qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
479                                          2, vms->memmap[VIRT_GIC_DIST].base,
480                                          2, vms->memmap[VIRT_GIC_DIST].size,
481                                          2, vms->memmap[VIRT_GIC_REDIST].base,
482                                          2, vms->memmap[VIRT_GIC_REDIST].size);
483         } else {
484             qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
485                                  2, vms->memmap[VIRT_GIC_DIST].base,
486                                  2, vms->memmap[VIRT_GIC_DIST].size,
487                                  2, vms->memmap[VIRT_GIC_REDIST].base,
488                                  2, vms->memmap[VIRT_GIC_REDIST].size,
489                                  2, vms->memmap[VIRT_HIGH_GIC_REDIST2].base,
490                                  2, vms->memmap[VIRT_HIGH_GIC_REDIST2].size);
491         }
492 
493         if (vms->virt) {
494             qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
495                                    GIC_FDT_IRQ_TYPE_PPI, ARCH_GIC_MAINT_IRQ,
496                                    GIC_FDT_IRQ_FLAGS_LEVEL_HI);
497         }
498     } else {
499         /* 'cortex-a15-gic' means 'GIC v2' */
500         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
501                                 "arm,cortex-a15-gic");
502         if (!vms->virt) {
503             qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
504                                          2, vms->memmap[VIRT_GIC_DIST].base,
505                                          2, vms->memmap[VIRT_GIC_DIST].size,
506                                          2, vms->memmap[VIRT_GIC_CPU].base,
507                                          2, vms->memmap[VIRT_GIC_CPU].size);
508         } else {
509             qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
510                                          2, vms->memmap[VIRT_GIC_DIST].base,
511                                          2, vms->memmap[VIRT_GIC_DIST].size,
512                                          2, vms->memmap[VIRT_GIC_CPU].base,
513                                          2, vms->memmap[VIRT_GIC_CPU].size,
514                                          2, vms->memmap[VIRT_GIC_HYP].base,
515                                          2, vms->memmap[VIRT_GIC_HYP].size,
516                                          2, vms->memmap[VIRT_GIC_VCPU].base,
517                                          2, vms->memmap[VIRT_GIC_VCPU].size);
518             qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
519                                    GIC_FDT_IRQ_TYPE_PPI, ARCH_GIC_MAINT_IRQ,
520                                    GIC_FDT_IRQ_FLAGS_LEVEL_HI);
521         }
522     }
523 
524     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", vms->gic_phandle);
525     g_free(nodename);
526 }
527 
528 static void fdt_add_pmu_nodes(const VirtMachineState *vms)
529 {
530     ARMCPU *armcpu = ARM_CPU(first_cpu);
531     uint32_t irqflags = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
532     MachineState *ms = MACHINE(vms);
533 
534     if (!arm_feature(&armcpu->env, ARM_FEATURE_PMU)) {
535         assert(!object_property_get_bool(OBJECT(armcpu), "pmu", NULL));
536         return;
537     }
538 
539     if (vms->gic_version == VIRT_GIC_VERSION_2) {
540         irqflags = deposit32(irqflags, GIC_FDT_IRQ_PPI_CPU_START,
541                              GIC_FDT_IRQ_PPI_CPU_WIDTH,
542                              (1 << MACHINE(vms)->smp.cpus) - 1);
543     }
544 
545     qemu_fdt_add_subnode(ms->fdt, "/pmu");
546     if (arm_feature(&armcpu->env, ARM_FEATURE_V8)) {
547         const char compat[] = "arm,armv8-pmuv3";
548         qemu_fdt_setprop(ms->fdt, "/pmu", "compatible",
549                          compat, sizeof(compat));
550         qemu_fdt_setprop_cells(ms->fdt, "/pmu", "interrupts",
551                                GIC_FDT_IRQ_TYPE_PPI, VIRTUAL_PMU_IRQ, irqflags);
552     }
553 }
554 
555 static inline DeviceState *create_acpi_ged(VirtMachineState *vms)
556 {
557     DeviceState *dev;
558     MachineState *ms = MACHINE(vms);
559     int irq = vms->irqmap[VIRT_ACPI_GED];
560     uint32_t event = ACPI_GED_PWR_DOWN_EVT;
561 
562     if (ms->ram_slots) {
563         event |= ACPI_GED_MEM_HOTPLUG_EVT;
564     }
565 
566     if (ms->nvdimms_state->is_enabled) {
567         event |= ACPI_GED_NVDIMM_HOTPLUG_EVT;
568     }
569 
570     dev = qdev_new(TYPE_ACPI_GED);
571     qdev_prop_set_uint32(dev, "ged-event", event);
572 
573     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_ACPI_GED].base);
574     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 1, vms->memmap[VIRT_PCDIMM_ACPI].base);
575     sysbus_connect_irq(SYS_BUS_DEVICE(dev), 0, qdev_get_gpio_in(vms->gic, irq));
576 
577     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
578 
579     return dev;
580 }
581 
582 static void create_its(VirtMachineState *vms)
583 {
584     const char *itsclass = its_class_name();
585     DeviceState *dev;
586 
587     if (!itsclass) {
588         /* Do nothing if not supported */
589         return;
590     }
591 
592     dev = qdev_new(itsclass);
593 
594     object_property_set_link(OBJECT(dev), "parent-gicv3", OBJECT(vms->gic),
595                              &error_abort);
596     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
597     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_GIC_ITS].base);
598 
599     fdt_add_its_gic_node(vms);
600     vms->msi_controller = VIRT_MSI_CTRL_ITS;
601 }
602 
603 static void create_v2m(VirtMachineState *vms)
604 {
605     int i;
606     int irq = vms->irqmap[VIRT_GIC_V2M];
607     DeviceState *dev;
608 
609     dev = qdev_new("arm-gicv2m");
610     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_GIC_V2M].base);
611     qdev_prop_set_uint32(dev, "base-spi", irq);
612     qdev_prop_set_uint32(dev, "num-spi", NUM_GICV2M_SPIS);
613     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
614 
615     for (i = 0; i < NUM_GICV2M_SPIS; i++) {
616         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
617                            qdev_get_gpio_in(vms->gic, irq + i));
618     }
619 
620     fdt_add_v2m_gic_node(vms);
621     vms->msi_controller = VIRT_MSI_CTRL_GICV2M;
622 }
623 
624 static void create_gic(VirtMachineState *vms)
625 {
626     MachineState *ms = MACHINE(vms);
627     /* We create a standalone GIC */
628     SysBusDevice *gicbusdev;
629     const char *gictype;
630     int type = vms->gic_version, i;
631     unsigned int smp_cpus = ms->smp.cpus;
632     uint32_t nb_redist_regions = 0;
633 
634     gictype = (type == 3) ? gicv3_class_name() : gic_class_name();
635 
636     vms->gic = qdev_new(gictype);
637     qdev_prop_set_uint32(vms->gic, "revision", type);
638     qdev_prop_set_uint32(vms->gic, "num-cpu", smp_cpus);
639     /* Note that the num-irq property counts both internal and external
640      * interrupts; there are always 32 of the former (mandated by GIC spec).
641      */
642     qdev_prop_set_uint32(vms->gic, "num-irq", NUM_IRQS + 32);
643     if (!kvm_irqchip_in_kernel()) {
644         qdev_prop_set_bit(vms->gic, "has-security-extensions", vms->secure);
645     }
646 
647     if (type == 3) {
648         uint32_t redist0_capacity =
649                     vms->memmap[VIRT_GIC_REDIST].size / GICV3_REDIST_SIZE;
650         uint32_t redist0_count = MIN(smp_cpus, redist0_capacity);
651 
652         nb_redist_regions = virt_gicv3_redist_region_count(vms);
653 
654         qdev_prop_set_uint32(vms->gic, "len-redist-region-count",
655                              nb_redist_regions);
656         qdev_prop_set_uint32(vms->gic, "redist-region-count[0]", redist0_count);
657 
658         if (nb_redist_regions == 2) {
659             uint32_t redist1_capacity =
660                     vms->memmap[VIRT_HIGH_GIC_REDIST2].size / GICV3_REDIST_SIZE;
661 
662             qdev_prop_set_uint32(vms->gic, "redist-region-count[1]",
663                 MIN(smp_cpus - redist0_count, redist1_capacity));
664         }
665     } else {
666         if (!kvm_irqchip_in_kernel()) {
667             qdev_prop_set_bit(vms->gic, "has-virtualization-extensions",
668                               vms->virt);
669         }
670     }
671     gicbusdev = SYS_BUS_DEVICE(vms->gic);
672     sysbus_realize_and_unref(gicbusdev, &error_fatal);
673     sysbus_mmio_map(gicbusdev, 0, vms->memmap[VIRT_GIC_DIST].base);
674     if (type == 3) {
675         sysbus_mmio_map(gicbusdev, 1, vms->memmap[VIRT_GIC_REDIST].base);
676         if (nb_redist_regions == 2) {
677             sysbus_mmio_map(gicbusdev, 2,
678                             vms->memmap[VIRT_HIGH_GIC_REDIST2].base);
679         }
680     } else {
681         sysbus_mmio_map(gicbusdev, 1, vms->memmap[VIRT_GIC_CPU].base);
682         if (vms->virt) {
683             sysbus_mmio_map(gicbusdev, 2, vms->memmap[VIRT_GIC_HYP].base);
684             sysbus_mmio_map(gicbusdev, 3, vms->memmap[VIRT_GIC_VCPU].base);
685         }
686     }
687 
688     /* Wire the outputs from each CPU's generic timer and the GICv3
689      * maintenance interrupt signal to the appropriate GIC PPI inputs,
690      * and the GIC's IRQ/FIQ/VIRQ/VFIQ interrupt outputs to the CPU's inputs.
691      */
692     for (i = 0; i < smp_cpus; i++) {
693         DeviceState *cpudev = DEVICE(qemu_get_cpu(i));
694         int ppibase = NUM_IRQS + i * GIC_INTERNAL + GIC_NR_SGIS;
695         int irq;
696         /* Mapping from the output timer irq lines from the CPU to the
697          * GIC PPI inputs we use for the virt board.
698          */
699         const int timer_irq[] = {
700             [GTIMER_PHYS] = ARCH_TIMER_NS_EL1_IRQ,
701             [GTIMER_VIRT] = ARCH_TIMER_VIRT_IRQ,
702             [GTIMER_HYP]  = ARCH_TIMER_NS_EL2_IRQ,
703             [GTIMER_SEC]  = ARCH_TIMER_S_EL1_IRQ,
704         };
705 
706         for (irq = 0; irq < ARRAY_SIZE(timer_irq); irq++) {
707             qdev_connect_gpio_out(cpudev, irq,
708                                   qdev_get_gpio_in(vms->gic,
709                                                    ppibase + timer_irq[irq]));
710         }
711 
712         if (type == 3) {
713             qemu_irq irq = qdev_get_gpio_in(vms->gic,
714                                             ppibase + ARCH_GIC_MAINT_IRQ);
715             qdev_connect_gpio_out_named(cpudev, "gicv3-maintenance-interrupt",
716                                         0, irq);
717         } else if (vms->virt) {
718             qemu_irq irq = qdev_get_gpio_in(vms->gic,
719                                             ppibase + ARCH_GIC_MAINT_IRQ);
720             sysbus_connect_irq(gicbusdev, i + 4 * smp_cpus, irq);
721         }
722 
723         qdev_connect_gpio_out_named(cpudev, "pmu-interrupt", 0,
724                                     qdev_get_gpio_in(vms->gic, ppibase
725                                                      + VIRTUAL_PMU_IRQ));
726 
727         sysbus_connect_irq(gicbusdev, i, qdev_get_gpio_in(cpudev, ARM_CPU_IRQ));
728         sysbus_connect_irq(gicbusdev, i + smp_cpus,
729                            qdev_get_gpio_in(cpudev, ARM_CPU_FIQ));
730         sysbus_connect_irq(gicbusdev, i + 2 * smp_cpus,
731                            qdev_get_gpio_in(cpudev, ARM_CPU_VIRQ));
732         sysbus_connect_irq(gicbusdev, i + 3 * smp_cpus,
733                            qdev_get_gpio_in(cpudev, ARM_CPU_VFIQ));
734     }
735 
736     fdt_add_gic_node(vms);
737 
738     if (type == 3 && vms->its) {
739         create_its(vms);
740     } else if (type == 2) {
741         create_v2m(vms);
742     }
743 }
744 
745 static void create_uart(const VirtMachineState *vms, int uart,
746                         MemoryRegion *mem, Chardev *chr)
747 {
748     char *nodename;
749     hwaddr base = vms->memmap[uart].base;
750     hwaddr size = vms->memmap[uart].size;
751     int irq = vms->irqmap[uart];
752     const char compat[] = "arm,pl011\0arm,primecell";
753     const char clocknames[] = "uartclk\0apb_pclk";
754     DeviceState *dev = qdev_new(TYPE_PL011);
755     SysBusDevice *s = SYS_BUS_DEVICE(dev);
756     MachineState *ms = MACHINE(vms);
757 
758     qdev_prop_set_chr(dev, "chardev", chr);
759     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
760     memory_region_add_subregion(mem, base,
761                                 sysbus_mmio_get_region(s, 0));
762     sysbus_connect_irq(s, 0, qdev_get_gpio_in(vms->gic, irq));
763 
764     nodename = g_strdup_printf("/pl011@%" PRIx64, base);
765     qemu_fdt_add_subnode(ms->fdt, nodename);
766     /* Note that we can't use setprop_string because of the embedded NUL */
767     qemu_fdt_setprop(ms->fdt, nodename, "compatible",
768                          compat, sizeof(compat));
769     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
770                                      2, base, 2, size);
771     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
772                                GIC_FDT_IRQ_TYPE_SPI, irq,
773                                GIC_FDT_IRQ_FLAGS_LEVEL_HI);
774     qemu_fdt_setprop_cells(ms->fdt, nodename, "clocks",
775                                vms->clock_phandle, vms->clock_phandle);
776     qemu_fdt_setprop(ms->fdt, nodename, "clock-names",
777                          clocknames, sizeof(clocknames));
778 
779     if (uart == VIRT_UART) {
780         qemu_fdt_setprop_string(ms->fdt, "/chosen", "stdout-path", nodename);
781     } else {
782         /* Mark as not usable by the normal world */
783         qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
784         qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
785 
786         qemu_fdt_setprop_string(ms->fdt, "/secure-chosen", "stdout-path",
787                                 nodename);
788     }
789 
790     g_free(nodename);
791 }
792 
793 static void create_rtc(const VirtMachineState *vms)
794 {
795     char *nodename;
796     hwaddr base = vms->memmap[VIRT_RTC].base;
797     hwaddr size = vms->memmap[VIRT_RTC].size;
798     int irq = vms->irqmap[VIRT_RTC];
799     const char compat[] = "arm,pl031\0arm,primecell";
800     MachineState *ms = MACHINE(vms);
801 
802     sysbus_create_simple("pl031", base, qdev_get_gpio_in(vms->gic, irq));
803 
804     nodename = g_strdup_printf("/pl031@%" PRIx64, base);
805     qemu_fdt_add_subnode(ms->fdt, nodename);
806     qemu_fdt_setprop(ms->fdt, nodename, "compatible", compat, sizeof(compat));
807     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
808                                  2, base, 2, size);
809     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
810                            GIC_FDT_IRQ_TYPE_SPI, irq,
811                            GIC_FDT_IRQ_FLAGS_LEVEL_HI);
812     qemu_fdt_setprop_cell(ms->fdt, nodename, "clocks", vms->clock_phandle);
813     qemu_fdt_setprop_string(ms->fdt, nodename, "clock-names", "apb_pclk");
814     g_free(nodename);
815 }
816 
817 static DeviceState *gpio_key_dev;
818 static void virt_powerdown_req(Notifier *n, void *opaque)
819 {
820     VirtMachineState *s = container_of(n, VirtMachineState, powerdown_notifier);
821 
822     if (s->acpi_dev) {
823         acpi_send_event(s->acpi_dev, ACPI_POWER_DOWN_STATUS);
824     } else {
825         /* use gpio Pin 3 for power button event */
826         qemu_set_irq(qdev_get_gpio_in(gpio_key_dev, 0), 1);
827     }
828 }
829 
830 static void create_gpio_keys(char *fdt, DeviceState *pl061_dev,
831                              uint32_t phandle)
832 {
833     gpio_key_dev = sysbus_create_simple("gpio-key", -1,
834                                         qdev_get_gpio_in(pl061_dev, 3));
835 
836     qemu_fdt_add_subnode(fdt, "/gpio-keys");
837     qemu_fdt_setprop_string(fdt, "/gpio-keys", "compatible", "gpio-keys");
838     qemu_fdt_setprop_cell(fdt, "/gpio-keys", "#size-cells", 0);
839     qemu_fdt_setprop_cell(fdt, "/gpio-keys", "#address-cells", 1);
840 
841     qemu_fdt_add_subnode(fdt, "/gpio-keys/poweroff");
842     qemu_fdt_setprop_string(fdt, "/gpio-keys/poweroff",
843                             "label", "GPIO Key Poweroff");
844     qemu_fdt_setprop_cell(fdt, "/gpio-keys/poweroff", "linux,code",
845                           KEY_POWER);
846     qemu_fdt_setprop_cells(fdt, "/gpio-keys/poweroff",
847                            "gpios", phandle, 3, 0);
848 }
849 
850 #define SECURE_GPIO_POWEROFF 0
851 #define SECURE_GPIO_RESET    1
852 
853 static void create_secure_gpio_pwr(char *fdt, DeviceState *pl061_dev,
854                                    uint32_t phandle)
855 {
856     DeviceState *gpio_pwr_dev;
857 
858     /* gpio-pwr */
859     gpio_pwr_dev = sysbus_create_simple("gpio-pwr", -1, NULL);
860 
861     /* connect secure pl061 to gpio-pwr */
862     qdev_connect_gpio_out(pl061_dev, SECURE_GPIO_RESET,
863                           qdev_get_gpio_in_named(gpio_pwr_dev, "reset", 0));
864     qdev_connect_gpio_out(pl061_dev, SECURE_GPIO_POWEROFF,
865                           qdev_get_gpio_in_named(gpio_pwr_dev, "shutdown", 0));
866 
867     qemu_fdt_add_subnode(fdt, "/gpio-poweroff");
868     qemu_fdt_setprop_string(fdt, "/gpio-poweroff", "compatible",
869                             "gpio-poweroff");
870     qemu_fdt_setprop_cells(fdt, "/gpio-poweroff",
871                            "gpios", phandle, SECURE_GPIO_POWEROFF, 0);
872     qemu_fdt_setprop_string(fdt, "/gpio-poweroff", "status", "disabled");
873     qemu_fdt_setprop_string(fdt, "/gpio-poweroff", "secure-status",
874                             "okay");
875 
876     qemu_fdt_add_subnode(fdt, "/gpio-restart");
877     qemu_fdt_setprop_string(fdt, "/gpio-restart", "compatible",
878                             "gpio-restart");
879     qemu_fdt_setprop_cells(fdt, "/gpio-restart",
880                            "gpios", phandle, SECURE_GPIO_RESET, 0);
881     qemu_fdt_setprop_string(fdt, "/gpio-restart", "status", "disabled");
882     qemu_fdt_setprop_string(fdt, "/gpio-restart", "secure-status",
883                             "okay");
884 }
885 
886 static void create_gpio_devices(const VirtMachineState *vms, int gpio,
887                                 MemoryRegion *mem)
888 {
889     char *nodename;
890     DeviceState *pl061_dev;
891     hwaddr base = vms->memmap[gpio].base;
892     hwaddr size = vms->memmap[gpio].size;
893     int irq = vms->irqmap[gpio];
894     const char compat[] = "arm,pl061\0arm,primecell";
895     SysBusDevice *s;
896     MachineState *ms = MACHINE(vms);
897 
898     pl061_dev = qdev_new("pl061");
899     /* Pull lines down to 0 if not driven by the PL061 */
900     qdev_prop_set_uint32(pl061_dev, "pullups", 0);
901     qdev_prop_set_uint32(pl061_dev, "pulldowns", 0xff);
902     s = SYS_BUS_DEVICE(pl061_dev);
903     sysbus_realize_and_unref(s, &error_fatal);
904     memory_region_add_subregion(mem, base, sysbus_mmio_get_region(s, 0));
905     sysbus_connect_irq(s, 0, qdev_get_gpio_in(vms->gic, irq));
906 
907     uint32_t phandle = qemu_fdt_alloc_phandle(ms->fdt);
908     nodename = g_strdup_printf("/pl061@%" PRIx64, base);
909     qemu_fdt_add_subnode(ms->fdt, nodename);
910     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
911                                  2, base, 2, size);
912     qemu_fdt_setprop(ms->fdt, nodename, "compatible", compat, sizeof(compat));
913     qemu_fdt_setprop_cell(ms->fdt, nodename, "#gpio-cells", 2);
914     qemu_fdt_setprop(ms->fdt, nodename, "gpio-controller", NULL, 0);
915     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
916                            GIC_FDT_IRQ_TYPE_SPI, irq,
917                            GIC_FDT_IRQ_FLAGS_LEVEL_HI);
918     qemu_fdt_setprop_cell(ms->fdt, nodename, "clocks", vms->clock_phandle);
919     qemu_fdt_setprop_string(ms->fdt, nodename, "clock-names", "apb_pclk");
920     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", phandle);
921 
922     if (gpio != VIRT_GPIO) {
923         /* Mark as not usable by the normal world */
924         qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
925         qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
926     }
927     g_free(nodename);
928 
929     /* Child gpio devices */
930     if (gpio == VIRT_GPIO) {
931         create_gpio_keys(ms->fdt, pl061_dev, phandle);
932     } else {
933         create_secure_gpio_pwr(ms->fdt, pl061_dev, phandle);
934     }
935 }
936 
937 static void create_virtio_devices(const VirtMachineState *vms)
938 {
939     int i;
940     hwaddr size = vms->memmap[VIRT_MMIO].size;
941     MachineState *ms = MACHINE(vms);
942 
943     /* We create the transports in forwards order. Since qbus_realize()
944      * prepends (not appends) new child buses, the incrementing loop below will
945      * create a list of virtio-mmio buses with decreasing base addresses.
946      *
947      * When a -device option is processed from the command line,
948      * qbus_find_recursive() picks the next free virtio-mmio bus in forwards
949      * order. The upshot is that -device options in increasing command line
950      * order are mapped to virtio-mmio buses with decreasing base addresses.
951      *
952      * When this code was originally written, that arrangement ensured that the
953      * guest Linux kernel would give the lowest "name" (/dev/vda, eth0, etc) to
954      * the first -device on the command line. (The end-to-end order is a
955      * function of this loop, qbus_realize(), qbus_find_recursive(), and the
956      * guest kernel's name-to-address assignment strategy.)
957      *
958      * Meanwhile, the kernel's traversal seems to have been reversed; see eg.
959      * the message, if not necessarily the code, of commit 70161ff336.
960      * Therefore the loop now establishes the inverse of the original intent.
961      *
962      * Unfortunately, we can't counteract the kernel change by reversing the
963      * loop; it would break existing command lines.
964      *
965      * In any case, the kernel makes no guarantee about the stability of
966      * enumeration order of virtio devices (as demonstrated by it changing
967      * between kernel versions). For reliable and stable identification
968      * of disks users must use UUIDs or similar mechanisms.
969      */
970     for (i = 0; i < NUM_VIRTIO_TRANSPORTS; i++) {
971         int irq = vms->irqmap[VIRT_MMIO] + i;
972         hwaddr base = vms->memmap[VIRT_MMIO].base + i * size;
973 
974         sysbus_create_simple("virtio-mmio", base,
975                              qdev_get_gpio_in(vms->gic, irq));
976     }
977 
978     /* We add dtb nodes in reverse order so that they appear in the finished
979      * device tree lowest address first.
980      *
981      * Note that this mapping is independent of the loop above. The previous
982      * loop influences virtio device to virtio transport assignment, whereas
983      * this loop controls how virtio transports are laid out in the dtb.
984      */
985     for (i = NUM_VIRTIO_TRANSPORTS - 1; i >= 0; i--) {
986         char *nodename;
987         int irq = vms->irqmap[VIRT_MMIO] + i;
988         hwaddr base = vms->memmap[VIRT_MMIO].base + i * size;
989 
990         nodename = g_strdup_printf("/virtio_mmio@%" PRIx64, base);
991         qemu_fdt_add_subnode(ms->fdt, nodename);
992         qemu_fdt_setprop_string(ms->fdt, nodename,
993                                 "compatible", "virtio,mmio");
994         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
995                                      2, base, 2, size);
996         qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
997                                GIC_FDT_IRQ_TYPE_SPI, irq,
998                                GIC_FDT_IRQ_FLAGS_EDGE_LO_HI);
999         qemu_fdt_setprop(ms->fdt, nodename, "dma-coherent", NULL, 0);
1000         g_free(nodename);
1001     }
1002 }
1003 
1004 #define VIRT_FLASH_SECTOR_SIZE (256 * KiB)
1005 
1006 static PFlashCFI01 *virt_flash_create1(VirtMachineState *vms,
1007                                         const char *name,
1008                                         const char *alias_prop_name)
1009 {
1010     /*
1011      * Create a single flash device.  We use the same parameters as
1012      * the flash devices on the Versatile Express board.
1013      */
1014     DeviceState *dev = qdev_new(TYPE_PFLASH_CFI01);
1015 
1016     qdev_prop_set_uint64(dev, "sector-length", VIRT_FLASH_SECTOR_SIZE);
1017     qdev_prop_set_uint8(dev, "width", 4);
1018     qdev_prop_set_uint8(dev, "device-width", 2);
1019     qdev_prop_set_bit(dev, "big-endian", false);
1020     qdev_prop_set_uint16(dev, "id0", 0x89);
1021     qdev_prop_set_uint16(dev, "id1", 0x18);
1022     qdev_prop_set_uint16(dev, "id2", 0x00);
1023     qdev_prop_set_uint16(dev, "id3", 0x00);
1024     qdev_prop_set_string(dev, "name", name);
1025     object_property_add_child(OBJECT(vms), name, OBJECT(dev));
1026     object_property_add_alias(OBJECT(vms), alias_prop_name,
1027                               OBJECT(dev), "drive");
1028     return PFLASH_CFI01(dev);
1029 }
1030 
1031 static void virt_flash_create(VirtMachineState *vms)
1032 {
1033     vms->flash[0] = virt_flash_create1(vms, "virt.flash0", "pflash0");
1034     vms->flash[1] = virt_flash_create1(vms, "virt.flash1", "pflash1");
1035 }
1036 
1037 static void virt_flash_map1(PFlashCFI01 *flash,
1038                             hwaddr base, hwaddr size,
1039                             MemoryRegion *sysmem)
1040 {
1041     DeviceState *dev = DEVICE(flash);
1042 
1043     assert(QEMU_IS_ALIGNED(size, VIRT_FLASH_SECTOR_SIZE));
1044     assert(size / VIRT_FLASH_SECTOR_SIZE <= UINT32_MAX);
1045     qdev_prop_set_uint32(dev, "num-blocks", size / VIRT_FLASH_SECTOR_SIZE);
1046     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1047 
1048     memory_region_add_subregion(sysmem, base,
1049                                 sysbus_mmio_get_region(SYS_BUS_DEVICE(dev),
1050                                                        0));
1051 }
1052 
1053 static void virt_flash_map(VirtMachineState *vms,
1054                            MemoryRegion *sysmem,
1055                            MemoryRegion *secure_sysmem)
1056 {
1057     /*
1058      * Map two flash devices to fill the VIRT_FLASH space in the memmap.
1059      * sysmem is the system memory space. secure_sysmem is the secure view
1060      * of the system, and the first flash device should be made visible only
1061      * there. The second flash device is visible to both secure and nonsecure.
1062      * If sysmem == secure_sysmem this means there is no separate Secure
1063      * address space and both flash devices are generally visible.
1064      */
1065     hwaddr flashsize = vms->memmap[VIRT_FLASH].size / 2;
1066     hwaddr flashbase = vms->memmap[VIRT_FLASH].base;
1067 
1068     virt_flash_map1(vms->flash[0], flashbase, flashsize,
1069                     secure_sysmem);
1070     virt_flash_map1(vms->flash[1], flashbase + flashsize, flashsize,
1071                     sysmem);
1072 }
1073 
1074 static void virt_flash_fdt(VirtMachineState *vms,
1075                            MemoryRegion *sysmem,
1076                            MemoryRegion *secure_sysmem)
1077 {
1078     hwaddr flashsize = vms->memmap[VIRT_FLASH].size / 2;
1079     hwaddr flashbase = vms->memmap[VIRT_FLASH].base;
1080     MachineState *ms = MACHINE(vms);
1081     char *nodename;
1082 
1083     if (sysmem == secure_sysmem) {
1084         /* Report both flash devices as a single node in the DT */
1085         nodename = g_strdup_printf("/flash@%" PRIx64, flashbase);
1086         qemu_fdt_add_subnode(ms->fdt, nodename);
1087         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible", "cfi-flash");
1088         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1089                                      2, flashbase, 2, flashsize,
1090                                      2, flashbase + flashsize, 2, flashsize);
1091         qemu_fdt_setprop_cell(ms->fdt, nodename, "bank-width", 4);
1092         g_free(nodename);
1093     } else {
1094         /*
1095          * Report the devices as separate nodes so we can mark one as
1096          * only visible to the secure world.
1097          */
1098         nodename = g_strdup_printf("/secflash@%" PRIx64, flashbase);
1099         qemu_fdt_add_subnode(ms->fdt, nodename);
1100         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible", "cfi-flash");
1101         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1102                                      2, flashbase, 2, flashsize);
1103         qemu_fdt_setprop_cell(ms->fdt, nodename, "bank-width", 4);
1104         qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
1105         qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
1106         g_free(nodename);
1107 
1108         nodename = g_strdup_printf("/flash@%" PRIx64, flashbase);
1109         qemu_fdt_add_subnode(ms->fdt, nodename);
1110         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible", "cfi-flash");
1111         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1112                                      2, flashbase + flashsize, 2, flashsize);
1113         qemu_fdt_setprop_cell(ms->fdt, nodename, "bank-width", 4);
1114         g_free(nodename);
1115     }
1116 }
1117 
1118 static bool virt_firmware_init(VirtMachineState *vms,
1119                                MemoryRegion *sysmem,
1120                                MemoryRegion *secure_sysmem)
1121 {
1122     int i;
1123     const char *bios_name;
1124     BlockBackend *pflash_blk0;
1125 
1126     /* Map legacy -drive if=pflash to machine properties */
1127     for (i = 0; i < ARRAY_SIZE(vms->flash); i++) {
1128         pflash_cfi01_legacy_drive(vms->flash[i],
1129                                   drive_get(IF_PFLASH, 0, i));
1130     }
1131 
1132     virt_flash_map(vms, sysmem, secure_sysmem);
1133 
1134     pflash_blk0 = pflash_cfi01_get_blk(vms->flash[0]);
1135 
1136     bios_name = MACHINE(vms)->firmware;
1137     if (bios_name) {
1138         char *fname;
1139         MemoryRegion *mr;
1140         int image_size;
1141 
1142         if (pflash_blk0) {
1143             error_report("The contents of the first flash device may be "
1144                          "specified with -bios or with -drive if=pflash... "
1145                          "but you cannot use both options at once");
1146             exit(1);
1147         }
1148 
1149         /* Fall back to -bios */
1150 
1151         fname = qemu_find_file(QEMU_FILE_TYPE_BIOS, bios_name);
1152         if (!fname) {
1153             error_report("Could not find ROM image '%s'", bios_name);
1154             exit(1);
1155         }
1156         mr = sysbus_mmio_get_region(SYS_BUS_DEVICE(vms->flash[0]), 0);
1157         image_size = load_image_mr(fname, mr);
1158         g_free(fname);
1159         if (image_size < 0) {
1160             error_report("Could not load ROM image '%s'", bios_name);
1161             exit(1);
1162         }
1163     }
1164 
1165     return pflash_blk0 || bios_name;
1166 }
1167 
1168 static FWCfgState *create_fw_cfg(const VirtMachineState *vms, AddressSpace *as)
1169 {
1170     MachineState *ms = MACHINE(vms);
1171     hwaddr base = vms->memmap[VIRT_FW_CFG].base;
1172     hwaddr size = vms->memmap[VIRT_FW_CFG].size;
1173     FWCfgState *fw_cfg;
1174     char *nodename;
1175 
1176     fw_cfg = fw_cfg_init_mem_wide(base + 8, base, 8, base + 16, as);
1177     fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, (uint16_t)ms->smp.cpus);
1178 
1179     nodename = g_strdup_printf("/fw-cfg@%" PRIx64, base);
1180     qemu_fdt_add_subnode(ms->fdt, nodename);
1181     qemu_fdt_setprop_string(ms->fdt, nodename,
1182                             "compatible", "qemu,fw-cfg-mmio");
1183     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1184                                  2, base, 2, size);
1185     qemu_fdt_setprop(ms->fdt, nodename, "dma-coherent", NULL, 0);
1186     g_free(nodename);
1187     return fw_cfg;
1188 }
1189 
1190 static void create_pcie_irq_map(const MachineState *ms,
1191                                 uint32_t gic_phandle,
1192                                 int first_irq, const char *nodename)
1193 {
1194     int devfn, pin;
1195     uint32_t full_irq_map[4 * 4 * 10] = { 0 };
1196     uint32_t *irq_map = full_irq_map;
1197 
1198     for (devfn = 0; devfn <= 0x18; devfn += 0x8) {
1199         for (pin = 0; pin < 4; pin++) {
1200             int irq_type = GIC_FDT_IRQ_TYPE_SPI;
1201             int irq_nr = first_irq + ((pin + PCI_SLOT(devfn)) % PCI_NUM_PINS);
1202             int irq_level = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
1203             int i;
1204 
1205             uint32_t map[] = {
1206                 devfn << 8, 0, 0,                           /* devfn */
1207                 pin + 1,                                    /* PCI pin */
1208                 gic_phandle, 0, 0, irq_type, irq_nr, irq_level }; /* GIC irq */
1209 
1210             /* Convert map to big endian */
1211             for (i = 0; i < 10; i++) {
1212                 irq_map[i] = cpu_to_be32(map[i]);
1213             }
1214             irq_map += 10;
1215         }
1216     }
1217 
1218     qemu_fdt_setprop(ms->fdt, nodename, "interrupt-map",
1219                      full_irq_map, sizeof(full_irq_map));
1220 
1221     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupt-map-mask",
1222                            cpu_to_be16(PCI_DEVFN(3, 0)), /* Slot 3 */
1223                            0, 0,
1224                            0x7           /* PCI irq */);
1225 }
1226 
1227 static void create_smmu(const VirtMachineState *vms,
1228                         PCIBus *bus)
1229 {
1230     char *node;
1231     const char compat[] = "arm,smmu-v3";
1232     int irq =  vms->irqmap[VIRT_SMMU];
1233     int i;
1234     hwaddr base = vms->memmap[VIRT_SMMU].base;
1235     hwaddr size = vms->memmap[VIRT_SMMU].size;
1236     const char irq_names[] = "eventq\0priq\0cmdq-sync\0gerror";
1237     DeviceState *dev;
1238     MachineState *ms = MACHINE(vms);
1239 
1240     if (vms->iommu != VIRT_IOMMU_SMMUV3 || !vms->iommu_phandle) {
1241         return;
1242     }
1243 
1244     dev = qdev_new("arm-smmuv3");
1245 
1246     object_property_set_link(OBJECT(dev), "primary-bus", OBJECT(bus),
1247                              &error_abort);
1248     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1249     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, base);
1250     for (i = 0; i < NUM_SMMU_IRQS; i++) {
1251         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
1252                            qdev_get_gpio_in(vms->gic, irq + i));
1253     }
1254 
1255     node = g_strdup_printf("/smmuv3@%" PRIx64, base);
1256     qemu_fdt_add_subnode(ms->fdt, node);
1257     qemu_fdt_setprop(ms->fdt, node, "compatible", compat, sizeof(compat));
1258     qemu_fdt_setprop_sized_cells(ms->fdt, node, "reg", 2, base, 2, size);
1259 
1260     qemu_fdt_setprop_cells(ms->fdt, node, "interrupts",
1261             GIC_FDT_IRQ_TYPE_SPI, irq    , GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1262             GIC_FDT_IRQ_TYPE_SPI, irq + 1, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1263             GIC_FDT_IRQ_TYPE_SPI, irq + 2, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1264             GIC_FDT_IRQ_TYPE_SPI, irq + 3, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI);
1265 
1266     qemu_fdt_setprop(ms->fdt, node, "interrupt-names", irq_names,
1267                      sizeof(irq_names));
1268 
1269     qemu_fdt_setprop_cell(ms->fdt, node, "clocks", vms->clock_phandle);
1270     qemu_fdt_setprop_string(ms->fdt, node, "clock-names", "apb_pclk");
1271     qemu_fdt_setprop(ms->fdt, node, "dma-coherent", NULL, 0);
1272 
1273     qemu_fdt_setprop_cell(ms->fdt, node, "#iommu-cells", 1);
1274 
1275     qemu_fdt_setprop_cell(ms->fdt, node, "phandle", vms->iommu_phandle);
1276     g_free(node);
1277 }
1278 
1279 static void create_virtio_iommu_dt_bindings(VirtMachineState *vms)
1280 {
1281     const char compat[] = "virtio,pci-iommu";
1282     uint16_t bdf = vms->virtio_iommu_bdf;
1283     MachineState *ms = MACHINE(vms);
1284     char *node;
1285 
1286     vms->iommu_phandle = qemu_fdt_alloc_phandle(ms->fdt);
1287 
1288     node = g_strdup_printf("%s/virtio_iommu@%d", vms->pciehb_nodename, bdf);
1289     qemu_fdt_add_subnode(ms->fdt, node);
1290     qemu_fdt_setprop(ms->fdt, node, "compatible", compat, sizeof(compat));
1291     qemu_fdt_setprop_sized_cells(ms->fdt, node, "reg",
1292                                  1, bdf << 8, 1, 0, 1, 0,
1293                                  1, 0, 1, 0);
1294 
1295     qemu_fdt_setprop_cell(ms->fdt, node, "#iommu-cells", 1);
1296     qemu_fdt_setprop_cell(ms->fdt, node, "phandle", vms->iommu_phandle);
1297     g_free(node);
1298 
1299     qemu_fdt_setprop_cells(ms->fdt, vms->pciehb_nodename, "iommu-map",
1300                            0x0, vms->iommu_phandle, 0x0, bdf,
1301                            bdf + 1, vms->iommu_phandle, bdf + 1, 0xffff - bdf);
1302 }
1303 
1304 static void create_pcie(VirtMachineState *vms)
1305 {
1306     hwaddr base_mmio = vms->memmap[VIRT_PCIE_MMIO].base;
1307     hwaddr size_mmio = vms->memmap[VIRT_PCIE_MMIO].size;
1308     hwaddr base_mmio_high = vms->memmap[VIRT_HIGH_PCIE_MMIO].base;
1309     hwaddr size_mmio_high = vms->memmap[VIRT_HIGH_PCIE_MMIO].size;
1310     hwaddr base_pio = vms->memmap[VIRT_PCIE_PIO].base;
1311     hwaddr size_pio = vms->memmap[VIRT_PCIE_PIO].size;
1312     hwaddr base_ecam, size_ecam;
1313     hwaddr base = base_mmio;
1314     int nr_pcie_buses;
1315     int irq = vms->irqmap[VIRT_PCIE];
1316     MemoryRegion *mmio_alias;
1317     MemoryRegion *mmio_reg;
1318     MemoryRegion *ecam_alias;
1319     MemoryRegion *ecam_reg;
1320     DeviceState *dev;
1321     char *nodename;
1322     int i, ecam_id;
1323     PCIHostState *pci;
1324     MachineState *ms = MACHINE(vms);
1325 
1326     dev = qdev_new(TYPE_GPEX_HOST);
1327     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1328 
1329     ecam_id = VIRT_ECAM_ID(vms->highmem_ecam);
1330     base_ecam = vms->memmap[ecam_id].base;
1331     size_ecam = vms->memmap[ecam_id].size;
1332     nr_pcie_buses = size_ecam / PCIE_MMCFG_SIZE_MIN;
1333     /* Map only the first size_ecam bytes of ECAM space */
1334     ecam_alias = g_new0(MemoryRegion, 1);
1335     ecam_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 0);
1336     memory_region_init_alias(ecam_alias, OBJECT(dev), "pcie-ecam",
1337                              ecam_reg, 0, size_ecam);
1338     memory_region_add_subregion(get_system_memory(), base_ecam, ecam_alias);
1339 
1340     /* Map the MMIO window into system address space so as to expose
1341      * the section of PCI MMIO space which starts at the same base address
1342      * (ie 1:1 mapping for that part of PCI MMIO space visible through
1343      * the window).
1344      */
1345     mmio_alias = g_new0(MemoryRegion, 1);
1346     mmio_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 1);
1347     memory_region_init_alias(mmio_alias, OBJECT(dev), "pcie-mmio",
1348                              mmio_reg, base_mmio, size_mmio);
1349     memory_region_add_subregion(get_system_memory(), base_mmio, mmio_alias);
1350 
1351     if (vms->highmem) {
1352         /* Map high MMIO space */
1353         MemoryRegion *high_mmio_alias = g_new0(MemoryRegion, 1);
1354 
1355         memory_region_init_alias(high_mmio_alias, OBJECT(dev), "pcie-mmio-high",
1356                                  mmio_reg, base_mmio_high, size_mmio_high);
1357         memory_region_add_subregion(get_system_memory(), base_mmio_high,
1358                                     high_mmio_alias);
1359     }
1360 
1361     /* Map IO port space */
1362     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 2, base_pio);
1363 
1364     for (i = 0; i < GPEX_NUM_IRQS; i++) {
1365         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
1366                            qdev_get_gpio_in(vms->gic, irq + i));
1367         gpex_set_irq_num(GPEX_HOST(dev), i, irq + i);
1368     }
1369 
1370     pci = PCI_HOST_BRIDGE(dev);
1371     pci->bypass_iommu = vms->default_bus_bypass_iommu;
1372     vms->bus = pci->bus;
1373     if (vms->bus) {
1374         for (i = 0; i < nb_nics; i++) {
1375             NICInfo *nd = &nd_table[i];
1376 
1377             if (!nd->model) {
1378                 nd->model = g_strdup("virtio");
1379             }
1380 
1381             pci_nic_init_nofail(nd, pci->bus, nd->model, NULL);
1382         }
1383     }
1384 
1385     nodename = vms->pciehb_nodename = g_strdup_printf("/pcie@%" PRIx64, base);
1386     qemu_fdt_add_subnode(ms->fdt, nodename);
1387     qemu_fdt_setprop_string(ms->fdt, nodename,
1388                             "compatible", "pci-host-ecam-generic");
1389     qemu_fdt_setprop_string(ms->fdt, nodename, "device_type", "pci");
1390     qemu_fdt_setprop_cell(ms->fdt, nodename, "#address-cells", 3);
1391     qemu_fdt_setprop_cell(ms->fdt, nodename, "#size-cells", 2);
1392     qemu_fdt_setprop_cell(ms->fdt, nodename, "linux,pci-domain", 0);
1393     qemu_fdt_setprop_cells(ms->fdt, nodename, "bus-range", 0,
1394                            nr_pcie_buses - 1);
1395     qemu_fdt_setprop(ms->fdt, nodename, "dma-coherent", NULL, 0);
1396 
1397     if (vms->msi_phandle) {
1398         qemu_fdt_setprop_cells(ms->fdt, nodename, "msi-parent",
1399                                vms->msi_phandle);
1400     }
1401 
1402     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1403                                  2, base_ecam, 2, size_ecam);
1404 
1405     if (vms->highmem) {
1406         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "ranges",
1407                                      1, FDT_PCI_RANGE_IOPORT, 2, 0,
1408                                      2, base_pio, 2, size_pio,
1409                                      1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
1410                                      2, base_mmio, 2, size_mmio,
1411                                      1, FDT_PCI_RANGE_MMIO_64BIT,
1412                                      2, base_mmio_high,
1413                                      2, base_mmio_high, 2, size_mmio_high);
1414     } else {
1415         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "ranges",
1416                                      1, FDT_PCI_RANGE_IOPORT, 2, 0,
1417                                      2, base_pio, 2, size_pio,
1418                                      1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
1419                                      2, base_mmio, 2, size_mmio);
1420     }
1421 
1422     qemu_fdt_setprop_cell(ms->fdt, nodename, "#interrupt-cells", 1);
1423     create_pcie_irq_map(ms, vms->gic_phandle, irq, nodename);
1424 
1425     if (vms->iommu) {
1426         vms->iommu_phandle = qemu_fdt_alloc_phandle(ms->fdt);
1427 
1428         switch (vms->iommu) {
1429         case VIRT_IOMMU_SMMUV3:
1430             create_smmu(vms, vms->bus);
1431             qemu_fdt_setprop_cells(ms->fdt, nodename, "iommu-map",
1432                                    0x0, vms->iommu_phandle, 0x0, 0x10000);
1433             break;
1434         default:
1435             g_assert_not_reached();
1436         }
1437     }
1438 }
1439 
1440 static void create_platform_bus(VirtMachineState *vms)
1441 {
1442     DeviceState *dev;
1443     SysBusDevice *s;
1444     int i;
1445     MemoryRegion *sysmem = get_system_memory();
1446 
1447     dev = qdev_new(TYPE_PLATFORM_BUS_DEVICE);
1448     dev->id = TYPE_PLATFORM_BUS_DEVICE;
1449     qdev_prop_set_uint32(dev, "num_irqs", PLATFORM_BUS_NUM_IRQS);
1450     qdev_prop_set_uint32(dev, "mmio_size", vms->memmap[VIRT_PLATFORM_BUS].size);
1451     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1452     vms->platform_bus_dev = dev;
1453 
1454     s = SYS_BUS_DEVICE(dev);
1455     for (i = 0; i < PLATFORM_BUS_NUM_IRQS; i++) {
1456         int irq = vms->irqmap[VIRT_PLATFORM_BUS] + i;
1457         sysbus_connect_irq(s, i, qdev_get_gpio_in(vms->gic, irq));
1458     }
1459 
1460     memory_region_add_subregion(sysmem,
1461                                 vms->memmap[VIRT_PLATFORM_BUS].base,
1462                                 sysbus_mmio_get_region(s, 0));
1463 }
1464 
1465 static void create_tag_ram(MemoryRegion *tag_sysmem,
1466                            hwaddr base, hwaddr size,
1467                            const char *name)
1468 {
1469     MemoryRegion *tagram = g_new(MemoryRegion, 1);
1470 
1471     memory_region_init_ram(tagram, NULL, name, size / 32, &error_fatal);
1472     memory_region_add_subregion(tag_sysmem, base / 32, tagram);
1473 }
1474 
1475 static void create_secure_ram(VirtMachineState *vms,
1476                               MemoryRegion *secure_sysmem,
1477                               MemoryRegion *secure_tag_sysmem)
1478 {
1479     MemoryRegion *secram = g_new(MemoryRegion, 1);
1480     char *nodename;
1481     hwaddr base = vms->memmap[VIRT_SECURE_MEM].base;
1482     hwaddr size = vms->memmap[VIRT_SECURE_MEM].size;
1483     MachineState *ms = MACHINE(vms);
1484 
1485     memory_region_init_ram(secram, NULL, "virt.secure-ram", size,
1486                            &error_fatal);
1487     memory_region_add_subregion(secure_sysmem, base, secram);
1488 
1489     nodename = g_strdup_printf("/secram@%" PRIx64, base);
1490     qemu_fdt_add_subnode(ms->fdt, nodename);
1491     qemu_fdt_setprop_string(ms->fdt, nodename, "device_type", "memory");
1492     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg", 2, base, 2, size);
1493     qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
1494     qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
1495 
1496     if (secure_tag_sysmem) {
1497         create_tag_ram(secure_tag_sysmem, base, size, "mach-virt.secure-tag");
1498     }
1499 
1500     g_free(nodename);
1501 }
1502 
1503 static void *machvirt_dtb(const struct arm_boot_info *binfo, int *fdt_size)
1504 {
1505     const VirtMachineState *board = container_of(binfo, VirtMachineState,
1506                                                  bootinfo);
1507     MachineState *ms = MACHINE(board);
1508 
1509 
1510     *fdt_size = board->fdt_size;
1511     return ms->fdt;
1512 }
1513 
1514 static void virt_build_smbios(VirtMachineState *vms)
1515 {
1516     MachineClass *mc = MACHINE_GET_CLASS(vms);
1517     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
1518     uint8_t *smbios_tables, *smbios_anchor;
1519     size_t smbios_tables_len, smbios_anchor_len;
1520     const char *product = "QEMU Virtual Machine";
1521 
1522     if (kvm_enabled()) {
1523         product = "KVM Virtual Machine";
1524     }
1525 
1526     smbios_set_defaults("QEMU", product,
1527                         vmc->smbios_old_sys_ver ? "1.0" : mc->name, false,
1528                         true, SMBIOS_ENTRY_POINT_30);
1529 
1530     smbios_get_tables(MACHINE(vms), NULL, 0,
1531                       &smbios_tables, &smbios_tables_len,
1532                       &smbios_anchor, &smbios_anchor_len,
1533                       &error_fatal);
1534 
1535     if (smbios_anchor) {
1536         fw_cfg_add_file(vms->fw_cfg, "etc/smbios/smbios-tables",
1537                         smbios_tables, smbios_tables_len);
1538         fw_cfg_add_file(vms->fw_cfg, "etc/smbios/smbios-anchor",
1539                         smbios_anchor, smbios_anchor_len);
1540     }
1541 }
1542 
1543 static
1544 void virt_machine_done(Notifier *notifier, void *data)
1545 {
1546     VirtMachineState *vms = container_of(notifier, VirtMachineState,
1547                                          machine_done);
1548     MachineState *ms = MACHINE(vms);
1549     ARMCPU *cpu = ARM_CPU(first_cpu);
1550     struct arm_boot_info *info = &vms->bootinfo;
1551     AddressSpace *as = arm_boot_address_space(cpu, info);
1552 
1553     /*
1554      * If the user provided a dtb, we assume the dynamic sysbus nodes
1555      * already are integrated there. This corresponds to a use case where
1556      * the dynamic sysbus nodes are complex and their generation is not yet
1557      * supported. In that case the user can take charge of the guest dt
1558      * while qemu takes charge of the qom stuff.
1559      */
1560     if (info->dtb_filename == NULL) {
1561         platform_bus_add_all_fdt_nodes(ms->fdt, "/intc",
1562                                        vms->memmap[VIRT_PLATFORM_BUS].base,
1563                                        vms->memmap[VIRT_PLATFORM_BUS].size,
1564                                        vms->irqmap[VIRT_PLATFORM_BUS]);
1565     }
1566     if (arm_load_dtb(info->dtb_start, info, info->dtb_limit, as, ms) < 0) {
1567         exit(1);
1568     }
1569 
1570     fw_cfg_add_extra_pci_roots(vms->bus, vms->fw_cfg);
1571 
1572     virt_acpi_setup(vms);
1573     virt_build_smbios(vms);
1574 }
1575 
1576 static uint64_t virt_cpu_mp_affinity(VirtMachineState *vms, int idx)
1577 {
1578     uint8_t clustersz = ARM_DEFAULT_CPUS_PER_CLUSTER;
1579     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
1580 
1581     if (!vmc->disallow_affinity_adjustment) {
1582         /* Adjust MPIDR like 64-bit KVM hosts, which incorporate the
1583          * GIC's target-list limitations. 32-bit KVM hosts currently
1584          * always create clusters of 4 CPUs, but that is expected to
1585          * change when they gain support for gicv3. When KVM is enabled
1586          * it will override the changes we make here, therefore our
1587          * purposes are to make TCG consistent (with 64-bit KVM hosts)
1588          * and to improve SGI efficiency.
1589          */
1590         if (vms->gic_version == VIRT_GIC_VERSION_3) {
1591             clustersz = GICV3_TARGETLIST_BITS;
1592         } else {
1593             clustersz = GIC_TARGETLIST_BITS;
1594         }
1595     }
1596     return arm_cpu_mp_affinity(idx, clustersz);
1597 }
1598 
1599 static void virt_set_memmap(VirtMachineState *vms)
1600 {
1601     MachineState *ms = MACHINE(vms);
1602     hwaddr base, device_memory_base, device_memory_size;
1603     int i;
1604 
1605     vms->memmap = extended_memmap;
1606 
1607     for (i = 0; i < ARRAY_SIZE(base_memmap); i++) {
1608         vms->memmap[i] = base_memmap[i];
1609     }
1610 
1611     if (ms->ram_slots > ACPI_MAX_RAM_SLOTS) {
1612         error_report("unsupported number of memory slots: %"PRIu64,
1613                      ms->ram_slots);
1614         exit(EXIT_FAILURE);
1615     }
1616 
1617     /*
1618      * We compute the base of the high IO region depending on the
1619      * amount of initial and device memory. The device memory start/size
1620      * is aligned on 1GiB. We never put the high IO region below 256GiB
1621      * so that if maxram_size is < 255GiB we keep the legacy memory map.
1622      * The device region size assumes 1GiB page max alignment per slot.
1623      */
1624     device_memory_base =
1625         ROUND_UP(vms->memmap[VIRT_MEM].base + ms->ram_size, GiB);
1626     device_memory_size = ms->maxram_size - ms->ram_size + ms->ram_slots * GiB;
1627 
1628     /* Base address of the high IO region */
1629     base = device_memory_base + ROUND_UP(device_memory_size, GiB);
1630     if (base < device_memory_base) {
1631         error_report("maxmem/slots too huge");
1632         exit(EXIT_FAILURE);
1633     }
1634     if (base < vms->memmap[VIRT_MEM].base + LEGACY_RAMLIMIT_BYTES) {
1635         base = vms->memmap[VIRT_MEM].base + LEGACY_RAMLIMIT_BYTES;
1636     }
1637 
1638     for (i = VIRT_LOWMEMMAP_LAST; i < ARRAY_SIZE(extended_memmap); i++) {
1639         hwaddr size = extended_memmap[i].size;
1640 
1641         base = ROUND_UP(base, size);
1642         vms->memmap[i].base = base;
1643         vms->memmap[i].size = size;
1644         base += size;
1645     }
1646     vms->highest_gpa = base - 1;
1647     if (device_memory_size > 0) {
1648         ms->device_memory = g_malloc0(sizeof(*ms->device_memory));
1649         ms->device_memory->base = device_memory_base;
1650         memory_region_init(&ms->device_memory->mr, OBJECT(vms),
1651                            "device-memory", device_memory_size);
1652     }
1653 }
1654 
1655 /*
1656  * finalize_gic_version - Determines the final gic_version
1657  * according to the gic-version property
1658  *
1659  * Default GIC type is v2
1660  */
1661 static void finalize_gic_version(VirtMachineState *vms)
1662 {
1663     unsigned int max_cpus = MACHINE(vms)->smp.max_cpus;
1664 
1665     if (kvm_enabled()) {
1666         int probe_bitmap;
1667 
1668         if (!kvm_irqchip_in_kernel()) {
1669             switch (vms->gic_version) {
1670             case VIRT_GIC_VERSION_HOST:
1671                 warn_report(
1672                     "gic-version=host not relevant with kernel-irqchip=off "
1673                      "as only userspace GICv2 is supported. Using v2 ...");
1674                 return;
1675             case VIRT_GIC_VERSION_MAX:
1676             case VIRT_GIC_VERSION_NOSEL:
1677                 vms->gic_version = VIRT_GIC_VERSION_2;
1678                 return;
1679             case VIRT_GIC_VERSION_2:
1680                 return;
1681             case VIRT_GIC_VERSION_3:
1682                 error_report(
1683                     "gic-version=3 is not supported with kernel-irqchip=off");
1684                 exit(1);
1685             }
1686         }
1687 
1688         probe_bitmap = kvm_arm_vgic_probe();
1689         if (!probe_bitmap) {
1690             error_report("Unable to determine GIC version supported by host");
1691             exit(1);
1692         }
1693 
1694         switch (vms->gic_version) {
1695         case VIRT_GIC_VERSION_HOST:
1696         case VIRT_GIC_VERSION_MAX:
1697             if (probe_bitmap & KVM_ARM_VGIC_V3) {
1698                 vms->gic_version = VIRT_GIC_VERSION_3;
1699             } else {
1700                 vms->gic_version = VIRT_GIC_VERSION_2;
1701             }
1702             return;
1703         case VIRT_GIC_VERSION_NOSEL:
1704             if ((probe_bitmap & KVM_ARM_VGIC_V2) && max_cpus <= GIC_NCPU) {
1705                 vms->gic_version = VIRT_GIC_VERSION_2;
1706             } else if (probe_bitmap & KVM_ARM_VGIC_V3) {
1707                 /*
1708                  * in case the host does not support v2 in-kernel emulation or
1709                  * the end-user requested more than 8 VCPUs we now default
1710                  * to v3. In any case defaulting to v2 would be broken.
1711                  */
1712                 vms->gic_version = VIRT_GIC_VERSION_3;
1713             } else if (max_cpus > GIC_NCPU) {
1714                 error_report("host only supports in-kernel GICv2 emulation "
1715                              "but more than 8 vcpus are requested");
1716                 exit(1);
1717             }
1718             break;
1719         case VIRT_GIC_VERSION_2:
1720         case VIRT_GIC_VERSION_3:
1721             break;
1722         }
1723 
1724         /* Check chosen version is effectively supported by the host */
1725         if (vms->gic_version == VIRT_GIC_VERSION_2 &&
1726             !(probe_bitmap & KVM_ARM_VGIC_V2)) {
1727             error_report("host does not support in-kernel GICv2 emulation");
1728             exit(1);
1729         } else if (vms->gic_version == VIRT_GIC_VERSION_3 &&
1730                    !(probe_bitmap & KVM_ARM_VGIC_V3)) {
1731             error_report("host does not support in-kernel GICv3 emulation");
1732             exit(1);
1733         }
1734         return;
1735     }
1736 
1737     /* TCG mode */
1738     switch (vms->gic_version) {
1739     case VIRT_GIC_VERSION_NOSEL:
1740         vms->gic_version = VIRT_GIC_VERSION_2;
1741         break;
1742     case VIRT_GIC_VERSION_MAX:
1743         vms->gic_version = VIRT_GIC_VERSION_3;
1744         break;
1745     case VIRT_GIC_VERSION_HOST:
1746         error_report("gic-version=host requires KVM");
1747         exit(1);
1748     case VIRT_GIC_VERSION_2:
1749     case VIRT_GIC_VERSION_3:
1750         break;
1751     }
1752 }
1753 
1754 /*
1755  * virt_cpu_post_init() must be called after the CPUs have
1756  * been realized and the GIC has been created.
1757  */
1758 static void virt_cpu_post_init(VirtMachineState *vms, MemoryRegion *sysmem)
1759 {
1760     int max_cpus = MACHINE(vms)->smp.max_cpus;
1761     bool aarch64, pmu, steal_time;
1762     CPUState *cpu;
1763 
1764     aarch64 = object_property_get_bool(OBJECT(first_cpu), "aarch64", NULL);
1765     pmu = object_property_get_bool(OBJECT(first_cpu), "pmu", NULL);
1766     steal_time = object_property_get_bool(OBJECT(first_cpu),
1767                                           "kvm-steal-time", NULL);
1768 
1769     if (kvm_enabled()) {
1770         hwaddr pvtime_reg_base = vms->memmap[VIRT_PVTIME].base;
1771         hwaddr pvtime_reg_size = vms->memmap[VIRT_PVTIME].size;
1772 
1773         if (steal_time) {
1774             MemoryRegion *pvtime = g_new(MemoryRegion, 1);
1775             hwaddr pvtime_size = max_cpus * PVTIME_SIZE_PER_CPU;
1776 
1777             /* The memory region size must be a multiple of host page size. */
1778             pvtime_size = REAL_HOST_PAGE_ALIGN(pvtime_size);
1779 
1780             if (pvtime_size > pvtime_reg_size) {
1781                 error_report("pvtime requires a %" HWADDR_PRId
1782                              " byte memory region for %d CPUs,"
1783                              " but only %" HWADDR_PRId " has been reserved",
1784                              pvtime_size, max_cpus, pvtime_reg_size);
1785                 exit(1);
1786             }
1787 
1788             memory_region_init_ram(pvtime, NULL, "pvtime", pvtime_size, NULL);
1789             memory_region_add_subregion(sysmem, pvtime_reg_base, pvtime);
1790         }
1791 
1792         CPU_FOREACH(cpu) {
1793             if (pmu) {
1794                 assert(arm_feature(&ARM_CPU(cpu)->env, ARM_FEATURE_PMU));
1795                 if (kvm_irqchip_in_kernel()) {
1796                     kvm_arm_pmu_set_irq(cpu, PPI(VIRTUAL_PMU_IRQ));
1797                 }
1798                 kvm_arm_pmu_init(cpu);
1799             }
1800             if (steal_time) {
1801                 kvm_arm_pvtime_init(cpu, pvtime_reg_base +
1802                                          cpu->cpu_index * PVTIME_SIZE_PER_CPU);
1803             }
1804         }
1805     } else {
1806         if (aarch64 && vms->highmem) {
1807             int requested_pa_size = 64 - clz64(vms->highest_gpa);
1808             int pamax = arm_pamax(ARM_CPU(first_cpu));
1809 
1810             if (pamax < requested_pa_size) {
1811                 error_report("VCPU supports less PA bits (%d) than "
1812                              "requested by the memory map (%d)",
1813                              pamax, requested_pa_size);
1814                 exit(1);
1815             }
1816         }
1817     }
1818 }
1819 
1820 static void machvirt_init(MachineState *machine)
1821 {
1822     VirtMachineState *vms = VIRT_MACHINE(machine);
1823     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(machine);
1824     MachineClass *mc = MACHINE_GET_CLASS(machine);
1825     const CPUArchIdList *possible_cpus;
1826     MemoryRegion *sysmem = get_system_memory();
1827     MemoryRegion *secure_sysmem = NULL;
1828     MemoryRegion *tag_sysmem = NULL;
1829     MemoryRegion *secure_tag_sysmem = NULL;
1830     int n, virt_max_cpus;
1831     bool firmware_loaded;
1832     bool aarch64 = true;
1833     bool has_ged = !vmc->no_ged;
1834     unsigned int smp_cpus = machine->smp.cpus;
1835     unsigned int max_cpus = machine->smp.max_cpus;
1836 
1837     /*
1838      * In accelerated mode, the memory map is computed earlier in kvm_type()
1839      * to create a VM with the right number of IPA bits.
1840      */
1841     if (!vms->memmap) {
1842         virt_set_memmap(vms);
1843     }
1844 
1845     /* We can probe only here because during property set
1846      * KVM is not available yet
1847      */
1848     finalize_gic_version(vms);
1849 
1850     if (!cpu_type_valid(machine->cpu_type)) {
1851         error_report("mach-virt: CPU type %s not supported", machine->cpu_type);
1852         exit(1);
1853     }
1854 
1855     if (vms->secure) {
1856         /*
1857          * The Secure view of the world is the same as the NonSecure,
1858          * but with a few extra devices. Create it as a container region
1859          * containing the system memory at low priority; any secure-only
1860          * devices go in at higher priority and take precedence.
1861          */
1862         secure_sysmem = g_new(MemoryRegion, 1);
1863         memory_region_init(secure_sysmem, OBJECT(machine), "secure-memory",
1864                            UINT64_MAX);
1865         memory_region_add_subregion_overlap(secure_sysmem, 0, sysmem, -1);
1866     }
1867 
1868     firmware_loaded = virt_firmware_init(vms, sysmem,
1869                                          secure_sysmem ?: sysmem);
1870 
1871     /* If we have an EL3 boot ROM then the assumption is that it will
1872      * implement PSCI itself, so disable QEMU's internal implementation
1873      * so it doesn't get in the way. Instead of starting secondary
1874      * CPUs in PSCI powerdown state we will start them all running and
1875      * let the boot ROM sort them out.
1876      * The usual case is that we do use QEMU's PSCI implementation;
1877      * if the guest has EL2 then we will use SMC as the conduit,
1878      * and otherwise we will use HVC (for backwards compatibility and
1879      * because if we're using KVM then we must use HVC).
1880      */
1881     if (vms->secure && firmware_loaded) {
1882         vms->psci_conduit = QEMU_PSCI_CONDUIT_DISABLED;
1883     } else if (vms->virt) {
1884         vms->psci_conduit = QEMU_PSCI_CONDUIT_SMC;
1885     } else {
1886         vms->psci_conduit = QEMU_PSCI_CONDUIT_HVC;
1887     }
1888 
1889     /* The maximum number of CPUs depends on the GIC version, or on how
1890      * many redistributors we can fit into the memory map.
1891      */
1892     if (vms->gic_version == VIRT_GIC_VERSION_3) {
1893         virt_max_cpus =
1894             vms->memmap[VIRT_GIC_REDIST].size / GICV3_REDIST_SIZE;
1895         virt_max_cpus +=
1896             vms->memmap[VIRT_HIGH_GIC_REDIST2].size / GICV3_REDIST_SIZE;
1897     } else {
1898         virt_max_cpus = GIC_NCPU;
1899     }
1900 
1901     if (max_cpus > virt_max_cpus) {
1902         error_report("Number of SMP CPUs requested (%d) exceeds max CPUs "
1903                      "supported by machine 'mach-virt' (%d)",
1904                      max_cpus, virt_max_cpus);
1905         exit(1);
1906     }
1907 
1908     if (vms->virt && kvm_enabled()) {
1909         error_report("mach-virt: KVM does not support providing "
1910                      "Virtualization extensions to the guest CPU");
1911         exit(1);
1912     }
1913 
1914     if (vms->mte && kvm_enabled()) {
1915         error_report("mach-virt: KVM does not support providing "
1916                      "MTE to the guest CPU");
1917         exit(1);
1918     }
1919 
1920     create_fdt(vms);
1921 
1922     possible_cpus = mc->possible_cpu_arch_ids(machine);
1923     assert(possible_cpus->len == max_cpus);
1924     for (n = 0; n < possible_cpus->len; n++) {
1925         Object *cpuobj;
1926         CPUState *cs;
1927 
1928         if (n >= smp_cpus) {
1929             break;
1930         }
1931 
1932         cpuobj = object_new(possible_cpus->cpus[n].type);
1933         object_property_set_int(cpuobj, "mp-affinity",
1934                                 possible_cpus->cpus[n].arch_id, NULL);
1935 
1936         cs = CPU(cpuobj);
1937         cs->cpu_index = n;
1938 
1939         numa_cpu_pre_plug(&possible_cpus->cpus[cs->cpu_index], DEVICE(cpuobj),
1940                           &error_fatal);
1941 
1942         aarch64 &= object_property_get_bool(cpuobj, "aarch64", NULL);
1943 
1944         if (!vms->secure) {
1945             object_property_set_bool(cpuobj, "has_el3", false, NULL);
1946         }
1947 
1948         if (!vms->virt && object_property_find(cpuobj, "has_el2")) {
1949             object_property_set_bool(cpuobj, "has_el2", false, NULL);
1950         }
1951 
1952         if (vms->psci_conduit != QEMU_PSCI_CONDUIT_DISABLED) {
1953             object_property_set_int(cpuobj, "psci-conduit", vms->psci_conduit,
1954                                     NULL);
1955 
1956             /* Secondary CPUs start in PSCI powered-down state */
1957             if (n > 0) {
1958                 object_property_set_bool(cpuobj, "start-powered-off", true,
1959                                          NULL);
1960             }
1961         }
1962 
1963         if (vmc->kvm_no_adjvtime &&
1964             object_property_find(cpuobj, "kvm-no-adjvtime")) {
1965             object_property_set_bool(cpuobj, "kvm-no-adjvtime", true, NULL);
1966         }
1967 
1968         if (vmc->no_kvm_steal_time &&
1969             object_property_find(cpuobj, "kvm-steal-time")) {
1970             object_property_set_bool(cpuobj, "kvm-steal-time", false, NULL);
1971         }
1972 
1973         if (vmc->no_pmu && object_property_find(cpuobj, "pmu")) {
1974             object_property_set_bool(cpuobj, "pmu", false, NULL);
1975         }
1976 
1977         if (object_property_find(cpuobj, "reset-cbar")) {
1978             object_property_set_int(cpuobj, "reset-cbar",
1979                                     vms->memmap[VIRT_CPUPERIPHS].base,
1980                                     &error_abort);
1981         }
1982 
1983         object_property_set_link(cpuobj, "memory", OBJECT(sysmem),
1984                                  &error_abort);
1985         if (vms->secure) {
1986             object_property_set_link(cpuobj, "secure-memory",
1987                                      OBJECT(secure_sysmem), &error_abort);
1988         }
1989 
1990         if (vms->mte) {
1991             /* Create the memory region only once, but link to all cpus. */
1992             if (!tag_sysmem) {
1993                 /*
1994                  * The property exists only if MemTag is supported.
1995                  * If it is, we must allocate the ram to back that up.
1996                  */
1997                 if (!object_property_find(cpuobj, "tag-memory")) {
1998                     error_report("MTE requested, but not supported "
1999                                  "by the guest CPU");
2000                     exit(1);
2001                 }
2002 
2003                 tag_sysmem = g_new(MemoryRegion, 1);
2004                 memory_region_init(tag_sysmem, OBJECT(machine),
2005                                    "tag-memory", UINT64_MAX / 32);
2006 
2007                 if (vms->secure) {
2008                     secure_tag_sysmem = g_new(MemoryRegion, 1);
2009                     memory_region_init(secure_tag_sysmem, OBJECT(machine),
2010                                        "secure-tag-memory", UINT64_MAX / 32);
2011 
2012                     /* As with ram, secure-tag takes precedence over tag.  */
2013                     memory_region_add_subregion_overlap(secure_tag_sysmem, 0,
2014                                                         tag_sysmem, -1);
2015                 }
2016             }
2017 
2018             object_property_set_link(cpuobj, "tag-memory", OBJECT(tag_sysmem),
2019                                      &error_abort);
2020             if (vms->secure) {
2021                 object_property_set_link(cpuobj, "secure-tag-memory",
2022                                          OBJECT(secure_tag_sysmem),
2023                                          &error_abort);
2024             }
2025         }
2026 
2027         qdev_realize(DEVICE(cpuobj), NULL, &error_fatal);
2028         object_unref(cpuobj);
2029     }
2030     fdt_add_timer_nodes(vms);
2031     fdt_add_cpu_nodes(vms);
2032 
2033     memory_region_add_subregion(sysmem, vms->memmap[VIRT_MEM].base,
2034                                 machine->ram);
2035     if (machine->device_memory) {
2036         memory_region_add_subregion(sysmem, machine->device_memory->base,
2037                                     &machine->device_memory->mr);
2038     }
2039 
2040     virt_flash_fdt(vms, sysmem, secure_sysmem ?: sysmem);
2041 
2042     create_gic(vms);
2043 
2044     virt_cpu_post_init(vms, sysmem);
2045 
2046     fdt_add_pmu_nodes(vms);
2047 
2048     create_uart(vms, VIRT_UART, sysmem, serial_hd(0));
2049 
2050     if (vms->secure) {
2051         create_secure_ram(vms, secure_sysmem, secure_tag_sysmem);
2052         create_uart(vms, VIRT_SECURE_UART, secure_sysmem, serial_hd(1));
2053     }
2054 
2055     if (tag_sysmem) {
2056         create_tag_ram(tag_sysmem, vms->memmap[VIRT_MEM].base,
2057                        machine->ram_size, "mach-virt.tag");
2058     }
2059 
2060     vms->highmem_ecam &= vms->highmem && (!firmware_loaded || aarch64);
2061 
2062     create_rtc(vms);
2063 
2064     create_pcie(vms);
2065 
2066     if (has_ged && aarch64 && firmware_loaded && virt_is_acpi_enabled(vms)) {
2067         vms->acpi_dev = create_acpi_ged(vms);
2068     } else {
2069         create_gpio_devices(vms, VIRT_GPIO, sysmem);
2070     }
2071 
2072     if (vms->secure && !vmc->no_secure_gpio) {
2073         create_gpio_devices(vms, VIRT_SECURE_GPIO, secure_sysmem);
2074     }
2075 
2076      /* connect powerdown request */
2077      vms->powerdown_notifier.notify = virt_powerdown_req;
2078      qemu_register_powerdown_notifier(&vms->powerdown_notifier);
2079 
2080     /* Create mmio transports, so the user can create virtio backends
2081      * (which will be automatically plugged in to the transports). If
2082      * no backend is created the transport will just sit harmlessly idle.
2083      */
2084     create_virtio_devices(vms);
2085 
2086     vms->fw_cfg = create_fw_cfg(vms, &address_space_memory);
2087     rom_set_fw(vms->fw_cfg);
2088 
2089     create_platform_bus(vms);
2090 
2091     if (machine->nvdimms_state->is_enabled) {
2092         const struct AcpiGenericAddress arm_virt_nvdimm_acpi_dsmio = {
2093             .space_id = AML_AS_SYSTEM_MEMORY,
2094             .address = vms->memmap[VIRT_NVDIMM_ACPI].base,
2095             .bit_width = NVDIMM_ACPI_IO_LEN << 3
2096         };
2097 
2098         nvdimm_init_acpi_state(machine->nvdimms_state, sysmem,
2099                                arm_virt_nvdimm_acpi_dsmio,
2100                                vms->fw_cfg, OBJECT(vms));
2101     }
2102 
2103     vms->bootinfo.ram_size = machine->ram_size;
2104     vms->bootinfo.nb_cpus = smp_cpus;
2105     vms->bootinfo.board_id = -1;
2106     vms->bootinfo.loader_start = vms->memmap[VIRT_MEM].base;
2107     vms->bootinfo.get_dtb = machvirt_dtb;
2108     vms->bootinfo.skip_dtb_autoload = true;
2109     vms->bootinfo.firmware_loaded = firmware_loaded;
2110     arm_load_kernel(ARM_CPU(first_cpu), machine, &vms->bootinfo);
2111 
2112     vms->machine_done.notify = virt_machine_done;
2113     qemu_add_machine_init_done_notifier(&vms->machine_done);
2114 }
2115 
2116 static bool virt_get_secure(Object *obj, Error **errp)
2117 {
2118     VirtMachineState *vms = VIRT_MACHINE(obj);
2119 
2120     return vms->secure;
2121 }
2122 
2123 static void virt_set_secure(Object *obj, bool value, Error **errp)
2124 {
2125     VirtMachineState *vms = VIRT_MACHINE(obj);
2126 
2127     vms->secure = value;
2128 }
2129 
2130 static bool virt_get_virt(Object *obj, Error **errp)
2131 {
2132     VirtMachineState *vms = VIRT_MACHINE(obj);
2133 
2134     return vms->virt;
2135 }
2136 
2137 static void virt_set_virt(Object *obj, bool value, Error **errp)
2138 {
2139     VirtMachineState *vms = VIRT_MACHINE(obj);
2140 
2141     vms->virt = value;
2142 }
2143 
2144 static bool virt_get_highmem(Object *obj, Error **errp)
2145 {
2146     VirtMachineState *vms = VIRT_MACHINE(obj);
2147 
2148     return vms->highmem;
2149 }
2150 
2151 static void virt_set_highmem(Object *obj, bool value, Error **errp)
2152 {
2153     VirtMachineState *vms = VIRT_MACHINE(obj);
2154 
2155     vms->highmem = value;
2156 }
2157 
2158 static bool virt_get_its(Object *obj, Error **errp)
2159 {
2160     VirtMachineState *vms = VIRT_MACHINE(obj);
2161 
2162     return vms->its;
2163 }
2164 
2165 static void virt_set_its(Object *obj, bool value, Error **errp)
2166 {
2167     VirtMachineState *vms = VIRT_MACHINE(obj);
2168 
2169     vms->its = value;
2170 }
2171 
2172 static char *virt_get_oem_id(Object *obj, Error **errp)
2173 {
2174     VirtMachineState *vms = VIRT_MACHINE(obj);
2175 
2176     return g_strdup(vms->oem_id);
2177 }
2178 
2179 static void virt_set_oem_id(Object *obj, const char *value, Error **errp)
2180 {
2181     VirtMachineState *vms = VIRT_MACHINE(obj);
2182     size_t len = strlen(value);
2183 
2184     if (len > 6) {
2185         error_setg(errp,
2186                    "User specified oem-id value is bigger than 6 bytes in size");
2187         return;
2188     }
2189 
2190     strncpy(vms->oem_id, value, 6);
2191 }
2192 
2193 static char *virt_get_oem_table_id(Object *obj, Error **errp)
2194 {
2195     VirtMachineState *vms = VIRT_MACHINE(obj);
2196 
2197     return g_strdup(vms->oem_table_id);
2198 }
2199 
2200 static void virt_set_oem_table_id(Object *obj, const char *value,
2201                                   Error **errp)
2202 {
2203     VirtMachineState *vms = VIRT_MACHINE(obj);
2204     size_t len = strlen(value);
2205 
2206     if (len > 8) {
2207         error_setg(errp,
2208                    "User specified oem-table-id value is bigger than 8 bytes in size");
2209         return;
2210     }
2211     strncpy(vms->oem_table_id, value, 8);
2212 }
2213 
2214 
2215 bool virt_is_acpi_enabled(VirtMachineState *vms)
2216 {
2217     if (vms->acpi == ON_OFF_AUTO_OFF) {
2218         return false;
2219     }
2220     return true;
2221 }
2222 
2223 static void virt_get_acpi(Object *obj, Visitor *v, const char *name,
2224                           void *opaque, Error **errp)
2225 {
2226     VirtMachineState *vms = VIRT_MACHINE(obj);
2227     OnOffAuto acpi = vms->acpi;
2228 
2229     visit_type_OnOffAuto(v, name, &acpi, errp);
2230 }
2231 
2232 static void virt_set_acpi(Object *obj, Visitor *v, const char *name,
2233                           void *opaque, Error **errp)
2234 {
2235     VirtMachineState *vms = VIRT_MACHINE(obj);
2236 
2237     visit_type_OnOffAuto(v, name, &vms->acpi, errp);
2238 }
2239 
2240 static bool virt_get_ras(Object *obj, Error **errp)
2241 {
2242     VirtMachineState *vms = VIRT_MACHINE(obj);
2243 
2244     return vms->ras;
2245 }
2246 
2247 static void virt_set_ras(Object *obj, bool value, Error **errp)
2248 {
2249     VirtMachineState *vms = VIRT_MACHINE(obj);
2250 
2251     vms->ras = value;
2252 }
2253 
2254 static bool virt_get_mte(Object *obj, Error **errp)
2255 {
2256     VirtMachineState *vms = VIRT_MACHINE(obj);
2257 
2258     return vms->mte;
2259 }
2260 
2261 static void virt_set_mte(Object *obj, bool value, Error **errp)
2262 {
2263     VirtMachineState *vms = VIRT_MACHINE(obj);
2264 
2265     vms->mte = value;
2266 }
2267 
2268 static char *virt_get_gic_version(Object *obj, Error **errp)
2269 {
2270     VirtMachineState *vms = VIRT_MACHINE(obj);
2271     const char *val = vms->gic_version == VIRT_GIC_VERSION_3 ? "3" : "2";
2272 
2273     return g_strdup(val);
2274 }
2275 
2276 static void virt_set_gic_version(Object *obj, const char *value, Error **errp)
2277 {
2278     VirtMachineState *vms = VIRT_MACHINE(obj);
2279 
2280     if (!strcmp(value, "3")) {
2281         vms->gic_version = VIRT_GIC_VERSION_3;
2282     } else if (!strcmp(value, "2")) {
2283         vms->gic_version = VIRT_GIC_VERSION_2;
2284     } else if (!strcmp(value, "host")) {
2285         vms->gic_version = VIRT_GIC_VERSION_HOST; /* Will probe later */
2286     } else if (!strcmp(value, "max")) {
2287         vms->gic_version = VIRT_GIC_VERSION_MAX; /* Will probe later */
2288     } else {
2289         error_setg(errp, "Invalid gic-version value");
2290         error_append_hint(errp, "Valid values are 3, 2, host, max.\n");
2291     }
2292 }
2293 
2294 static char *virt_get_iommu(Object *obj, Error **errp)
2295 {
2296     VirtMachineState *vms = VIRT_MACHINE(obj);
2297 
2298     switch (vms->iommu) {
2299     case VIRT_IOMMU_NONE:
2300         return g_strdup("none");
2301     case VIRT_IOMMU_SMMUV3:
2302         return g_strdup("smmuv3");
2303     default:
2304         g_assert_not_reached();
2305     }
2306 }
2307 
2308 static void virt_set_iommu(Object *obj, const char *value, Error **errp)
2309 {
2310     VirtMachineState *vms = VIRT_MACHINE(obj);
2311 
2312     if (!strcmp(value, "smmuv3")) {
2313         vms->iommu = VIRT_IOMMU_SMMUV3;
2314     } else if (!strcmp(value, "none")) {
2315         vms->iommu = VIRT_IOMMU_NONE;
2316     } else {
2317         error_setg(errp, "Invalid iommu value");
2318         error_append_hint(errp, "Valid values are none, smmuv3.\n");
2319     }
2320 }
2321 
2322 static bool virt_get_default_bus_bypass_iommu(Object *obj, Error **errp)
2323 {
2324     VirtMachineState *vms = VIRT_MACHINE(obj);
2325 
2326     return vms->default_bus_bypass_iommu;
2327 }
2328 
2329 static void virt_set_default_bus_bypass_iommu(Object *obj, bool value,
2330                                               Error **errp)
2331 {
2332     VirtMachineState *vms = VIRT_MACHINE(obj);
2333 
2334     vms->default_bus_bypass_iommu = value;
2335 }
2336 
2337 static CpuInstanceProperties
2338 virt_cpu_index_to_props(MachineState *ms, unsigned cpu_index)
2339 {
2340     MachineClass *mc = MACHINE_GET_CLASS(ms);
2341     const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(ms);
2342 
2343     assert(cpu_index < possible_cpus->len);
2344     return possible_cpus->cpus[cpu_index].props;
2345 }
2346 
2347 static int64_t virt_get_default_cpu_node_id(const MachineState *ms, int idx)
2348 {
2349     return idx % ms->numa_state->num_nodes;
2350 }
2351 
2352 static const CPUArchIdList *virt_possible_cpu_arch_ids(MachineState *ms)
2353 {
2354     int n;
2355     unsigned int max_cpus = ms->smp.max_cpus;
2356     VirtMachineState *vms = VIRT_MACHINE(ms);
2357 
2358     if (ms->possible_cpus) {
2359         assert(ms->possible_cpus->len == max_cpus);
2360         return ms->possible_cpus;
2361     }
2362 
2363     ms->possible_cpus = g_malloc0(sizeof(CPUArchIdList) +
2364                                   sizeof(CPUArchId) * max_cpus);
2365     ms->possible_cpus->len = max_cpus;
2366     for (n = 0; n < ms->possible_cpus->len; n++) {
2367         ms->possible_cpus->cpus[n].type = ms->cpu_type;
2368         ms->possible_cpus->cpus[n].arch_id =
2369             virt_cpu_mp_affinity(vms, n);
2370         ms->possible_cpus->cpus[n].props.has_thread_id = true;
2371         ms->possible_cpus->cpus[n].props.thread_id = n;
2372     }
2373     return ms->possible_cpus;
2374 }
2375 
2376 static void virt_memory_pre_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
2377                                  Error **errp)
2378 {
2379     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2380     const MachineState *ms = MACHINE(hotplug_dev);
2381     const bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2382 
2383     if (!vms->acpi_dev) {
2384         error_setg(errp,
2385                    "memory hotplug is not enabled: missing acpi-ged device");
2386         return;
2387     }
2388 
2389     if (vms->mte) {
2390         error_setg(errp, "memory hotplug is not enabled: MTE is enabled");
2391         return;
2392     }
2393 
2394     if (is_nvdimm && !ms->nvdimms_state->is_enabled) {
2395         error_setg(errp, "nvdimm is not enabled: add 'nvdimm=on' to '-M'");
2396         return;
2397     }
2398 
2399     pc_dimm_pre_plug(PC_DIMM(dev), MACHINE(hotplug_dev), NULL, errp);
2400 }
2401 
2402 static void virt_memory_plug(HotplugHandler *hotplug_dev,
2403                              DeviceState *dev, Error **errp)
2404 {
2405     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2406     MachineState *ms = MACHINE(hotplug_dev);
2407     bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2408 
2409     pc_dimm_plug(PC_DIMM(dev), MACHINE(vms));
2410 
2411     if (is_nvdimm) {
2412         nvdimm_plug(ms->nvdimms_state);
2413     }
2414 
2415     hotplug_handler_plug(HOTPLUG_HANDLER(vms->acpi_dev),
2416                          dev, &error_abort);
2417 }
2418 
2419 static void virt_machine_device_pre_plug_cb(HotplugHandler *hotplug_dev,
2420                                             DeviceState *dev, Error **errp)
2421 {
2422     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2423 
2424     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2425         virt_memory_pre_plug(hotplug_dev, dev, errp);
2426     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2427         hwaddr db_start = 0, db_end = 0;
2428         char *resv_prop_str;
2429 
2430         switch (vms->msi_controller) {
2431         case VIRT_MSI_CTRL_NONE:
2432             return;
2433         case VIRT_MSI_CTRL_ITS:
2434             /* GITS_TRANSLATER page */
2435             db_start = base_memmap[VIRT_GIC_ITS].base + 0x10000;
2436             db_end = base_memmap[VIRT_GIC_ITS].base +
2437                      base_memmap[VIRT_GIC_ITS].size - 1;
2438             break;
2439         case VIRT_MSI_CTRL_GICV2M:
2440             /* MSI_SETSPI_NS page */
2441             db_start = base_memmap[VIRT_GIC_V2M].base;
2442             db_end = db_start + base_memmap[VIRT_GIC_V2M].size - 1;
2443             break;
2444         }
2445         resv_prop_str = g_strdup_printf("0x%"PRIx64":0x%"PRIx64":%u",
2446                                         db_start, db_end,
2447                                         VIRTIO_IOMMU_RESV_MEM_T_MSI);
2448 
2449         qdev_prop_set_uint32(dev, "len-reserved-regions", 1);
2450         qdev_prop_set_string(dev, "reserved-regions[0]", resv_prop_str);
2451         g_free(resv_prop_str);
2452     }
2453 }
2454 
2455 static void virt_machine_device_plug_cb(HotplugHandler *hotplug_dev,
2456                                         DeviceState *dev, Error **errp)
2457 {
2458     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2459 
2460     if (vms->platform_bus_dev) {
2461         MachineClass *mc = MACHINE_GET_CLASS(vms);
2462 
2463         if (device_is_dynamic_sysbus(mc, dev)) {
2464             platform_bus_link_device(PLATFORM_BUS_DEVICE(vms->platform_bus_dev),
2465                                      SYS_BUS_DEVICE(dev));
2466         }
2467     }
2468     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2469         virt_memory_plug(hotplug_dev, dev, errp);
2470     }
2471     if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2472         PCIDevice *pdev = PCI_DEVICE(dev);
2473 
2474         vms->iommu = VIRT_IOMMU_VIRTIO;
2475         vms->virtio_iommu_bdf = pci_get_bdf(pdev);
2476         create_virtio_iommu_dt_bindings(vms);
2477     }
2478 }
2479 
2480 static void virt_dimm_unplug_request(HotplugHandler *hotplug_dev,
2481                                      DeviceState *dev, Error **errp)
2482 {
2483     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2484     Error *local_err = NULL;
2485 
2486     if (!vms->acpi_dev) {
2487         error_setg(&local_err,
2488                    "memory hotplug is not enabled: missing acpi-ged device");
2489         goto out;
2490     }
2491 
2492     if (object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM)) {
2493         error_setg(&local_err,
2494                    "nvdimm device hot unplug is not supported yet.");
2495         goto out;
2496     }
2497 
2498     hotplug_handler_unplug_request(HOTPLUG_HANDLER(vms->acpi_dev), dev,
2499                                    &local_err);
2500 out:
2501     error_propagate(errp, local_err);
2502 }
2503 
2504 static void virt_dimm_unplug(HotplugHandler *hotplug_dev,
2505                              DeviceState *dev, Error **errp)
2506 {
2507     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2508     Error *local_err = NULL;
2509 
2510     hotplug_handler_unplug(HOTPLUG_HANDLER(vms->acpi_dev), dev, &local_err);
2511     if (local_err) {
2512         goto out;
2513     }
2514 
2515     pc_dimm_unplug(PC_DIMM(dev), MACHINE(vms));
2516     qdev_unrealize(dev);
2517 
2518 out:
2519     error_propagate(errp, local_err);
2520 }
2521 
2522 static void virt_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
2523                                           DeviceState *dev, Error **errp)
2524 {
2525     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2526         virt_dimm_unplug_request(hotplug_dev, dev, errp);
2527     } else {
2528         error_setg(errp, "device unplug request for unsupported device"
2529                    " type: %s", object_get_typename(OBJECT(dev)));
2530     }
2531 }
2532 
2533 static void virt_machine_device_unplug_cb(HotplugHandler *hotplug_dev,
2534                                           DeviceState *dev, Error **errp)
2535 {
2536     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2537         virt_dimm_unplug(hotplug_dev, dev, errp);
2538     } else {
2539         error_setg(errp, "virt: device unplug for unsupported device"
2540                    " type: %s", object_get_typename(OBJECT(dev)));
2541     }
2542 }
2543 
2544 static HotplugHandler *virt_machine_get_hotplug_handler(MachineState *machine,
2545                                                         DeviceState *dev)
2546 {
2547     MachineClass *mc = MACHINE_GET_CLASS(machine);
2548 
2549     if (device_is_dynamic_sysbus(mc, dev) ||
2550        (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM))) {
2551         return HOTPLUG_HANDLER(machine);
2552     }
2553     if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2554         VirtMachineState *vms = VIRT_MACHINE(machine);
2555 
2556         if (!vms->bootinfo.firmware_loaded || !virt_is_acpi_enabled(vms)) {
2557             return HOTPLUG_HANDLER(machine);
2558         }
2559     }
2560     return NULL;
2561 }
2562 
2563 /*
2564  * for arm64 kvm_type [7-0] encodes the requested number of bits
2565  * in the IPA address space
2566  */
2567 static int virt_kvm_type(MachineState *ms, const char *type_str)
2568 {
2569     VirtMachineState *vms = VIRT_MACHINE(ms);
2570     int max_vm_pa_size, requested_pa_size;
2571     bool fixed_ipa;
2572 
2573     max_vm_pa_size = kvm_arm_get_max_vm_ipa_size(ms, &fixed_ipa);
2574 
2575     /* we freeze the memory map to compute the highest gpa */
2576     virt_set_memmap(vms);
2577 
2578     requested_pa_size = 64 - clz64(vms->highest_gpa);
2579 
2580     /*
2581      * KVM requires the IPA size to be at least 32 bits.
2582      */
2583     if (requested_pa_size < 32) {
2584         requested_pa_size = 32;
2585     }
2586 
2587     if (requested_pa_size > max_vm_pa_size) {
2588         error_report("-m and ,maxmem option values "
2589                      "require an IPA range (%d bits) larger than "
2590                      "the one supported by the host (%d bits)",
2591                      requested_pa_size, max_vm_pa_size);
2592         exit(1);
2593     }
2594     /*
2595      * We return the requested PA log size, unless KVM only supports
2596      * the implicit legacy 40b IPA setting, in which case the kvm_type
2597      * must be 0.
2598      */
2599     return fixed_ipa ? 0 : requested_pa_size;
2600 }
2601 
2602 static void virt_machine_class_init(ObjectClass *oc, void *data)
2603 {
2604     MachineClass *mc = MACHINE_CLASS(oc);
2605     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
2606 
2607     mc->init = machvirt_init;
2608     /* Start with max_cpus set to 512, which is the maximum supported by KVM.
2609      * The value may be reduced later when we have more information about the
2610      * configuration of the particular instance.
2611      */
2612     mc->max_cpus = 512;
2613     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_CALXEDA_XGMAC);
2614     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_AMD_XGBE);
2615     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_RAMFB_DEVICE);
2616     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_PLATFORM);
2617 #ifdef CONFIG_TPM
2618     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_TPM_TIS_SYSBUS);
2619 #endif
2620     mc->block_default_type = IF_VIRTIO;
2621     mc->no_cdrom = 1;
2622     mc->pci_allow_0_address = true;
2623     /* We know we will never create a pre-ARMv7 CPU which needs 1K pages */
2624     mc->minimum_page_bits = 12;
2625     mc->possible_cpu_arch_ids = virt_possible_cpu_arch_ids;
2626     mc->cpu_index_to_instance_props = virt_cpu_index_to_props;
2627     mc->default_cpu_type = ARM_CPU_TYPE_NAME("cortex-a15");
2628     mc->get_default_cpu_node_id = virt_get_default_cpu_node_id;
2629     mc->kvm_type = virt_kvm_type;
2630     assert(!mc->get_hotplug_handler);
2631     mc->get_hotplug_handler = virt_machine_get_hotplug_handler;
2632     hc->pre_plug = virt_machine_device_pre_plug_cb;
2633     hc->plug = virt_machine_device_plug_cb;
2634     hc->unplug_request = virt_machine_device_unplug_request_cb;
2635     hc->unplug = virt_machine_device_unplug_cb;
2636     mc->nvdimm_supported = true;
2637     mc->auto_enable_numa_with_memhp = true;
2638     mc->auto_enable_numa_with_memdev = true;
2639     mc->default_ram_id = "mach-virt.ram";
2640 
2641     object_class_property_add(oc, "acpi", "OnOffAuto",
2642         virt_get_acpi, virt_set_acpi,
2643         NULL, NULL);
2644     object_class_property_set_description(oc, "acpi",
2645         "Enable ACPI");
2646     object_class_property_add_bool(oc, "secure", virt_get_secure,
2647                                    virt_set_secure);
2648     object_class_property_set_description(oc, "secure",
2649                                                 "Set on/off to enable/disable the ARM "
2650                                                 "Security Extensions (TrustZone)");
2651 
2652     object_class_property_add_bool(oc, "virtualization", virt_get_virt,
2653                                    virt_set_virt);
2654     object_class_property_set_description(oc, "virtualization",
2655                                           "Set on/off to enable/disable emulating a "
2656                                           "guest CPU which implements the ARM "
2657                                           "Virtualization Extensions");
2658 
2659     object_class_property_add_bool(oc, "highmem", virt_get_highmem,
2660                                    virt_set_highmem);
2661     object_class_property_set_description(oc, "highmem",
2662                                           "Set on/off to enable/disable using "
2663                                           "physical address space above 32 bits");
2664 
2665     object_class_property_add_str(oc, "gic-version", virt_get_gic_version,
2666                                   virt_set_gic_version);
2667     object_class_property_set_description(oc, "gic-version",
2668                                           "Set GIC version. "
2669                                           "Valid values are 2, 3, host and max");
2670 
2671     object_class_property_add_str(oc, "iommu", virt_get_iommu, virt_set_iommu);
2672     object_class_property_set_description(oc, "iommu",
2673                                           "Set the IOMMU type. "
2674                                           "Valid values are none and smmuv3");
2675 
2676     object_class_property_add_bool(oc, "default_bus_bypass_iommu",
2677                                    virt_get_default_bus_bypass_iommu,
2678                                    virt_set_default_bus_bypass_iommu);
2679     object_class_property_set_description(oc, "default_bus_bypass_iommu",
2680                                           "Set on/off to enable/disable "
2681                                           "bypass_iommu for default root bus");
2682 
2683     object_class_property_add_bool(oc, "ras", virt_get_ras,
2684                                    virt_set_ras);
2685     object_class_property_set_description(oc, "ras",
2686                                           "Set on/off to enable/disable reporting host memory errors "
2687                                           "to a KVM guest using ACPI and guest external abort exceptions");
2688 
2689     object_class_property_add_bool(oc, "mte", virt_get_mte, virt_set_mte);
2690     object_class_property_set_description(oc, "mte",
2691                                           "Set on/off to enable/disable emulating a "
2692                                           "guest CPU which implements the ARM "
2693                                           "Memory Tagging Extension");
2694 
2695     object_class_property_add_bool(oc, "its", virt_get_its,
2696                                    virt_set_its);
2697     object_class_property_set_description(oc, "its",
2698                                           "Set on/off to enable/disable "
2699                                           "ITS instantiation");
2700 
2701     object_class_property_add_str(oc, "x-oem-id",
2702                                   virt_get_oem_id,
2703                                   virt_set_oem_id);
2704     object_class_property_set_description(oc, "x-oem-id",
2705                                           "Override the default value of field OEMID "
2706                                           "in ACPI table header."
2707                                           "The string may be up to 6 bytes in size");
2708 
2709 
2710     object_class_property_add_str(oc, "x-oem-table-id",
2711                                   virt_get_oem_table_id,
2712                                   virt_set_oem_table_id);
2713     object_class_property_set_description(oc, "x-oem-table-id",
2714                                           "Override the default value of field OEM Table ID "
2715                                           "in ACPI table header."
2716                                           "The string may be up to 8 bytes in size");
2717 
2718 }
2719 
2720 static void virt_instance_init(Object *obj)
2721 {
2722     VirtMachineState *vms = VIRT_MACHINE(obj);
2723     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
2724 
2725     /* EL3 is disabled by default on virt: this makes us consistent
2726      * between KVM and TCG for this board, and it also allows us to
2727      * boot UEFI blobs which assume no TrustZone support.
2728      */
2729     vms->secure = false;
2730 
2731     /* EL2 is also disabled by default, for similar reasons */
2732     vms->virt = false;
2733 
2734     /* High memory is enabled by default */
2735     vms->highmem = true;
2736     vms->gic_version = VIRT_GIC_VERSION_NOSEL;
2737 
2738     vms->highmem_ecam = !vmc->no_highmem_ecam;
2739 
2740     if (vmc->no_its) {
2741         vms->its = false;
2742     } else {
2743         /* Default allows ITS instantiation */
2744         vms->its = true;
2745     }
2746 
2747     /* Default disallows iommu instantiation */
2748     vms->iommu = VIRT_IOMMU_NONE;
2749 
2750     /* The default root bus is attached to iommu by default */
2751     vms->default_bus_bypass_iommu = false;
2752 
2753     /* Default disallows RAS instantiation */
2754     vms->ras = false;
2755 
2756     /* MTE is disabled by default.  */
2757     vms->mte = false;
2758 
2759     vms->irqmap = a15irqmap;
2760 
2761     virt_flash_create(vms);
2762 
2763     vms->oem_id = g_strndup(ACPI_BUILD_APPNAME6, 6);
2764     vms->oem_table_id = g_strndup(ACPI_BUILD_APPNAME8, 8);
2765 }
2766 
2767 static const TypeInfo virt_machine_info = {
2768     .name          = TYPE_VIRT_MACHINE,
2769     .parent        = TYPE_MACHINE,
2770     .abstract      = true,
2771     .instance_size = sizeof(VirtMachineState),
2772     .class_size    = sizeof(VirtMachineClass),
2773     .class_init    = virt_machine_class_init,
2774     .instance_init = virt_instance_init,
2775     .interfaces = (InterfaceInfo[]) {
2776          { TYPE_HOTPLUG_HANDLER },
2777          { }
2778     },
2779 };
2780 
2781 static void machvirt_machine_init(void)
2782 {
2783     type_register_static(&virt_machine_info);
2784 }
2785 type_init(machvirt_machine_init);
2786 
2787 static void virt_machine_6_2_options(MachineClass *mc)
2788 {
2789 }
2790 DEFINE_VIRT_MACHINE_AS_LATEST(6, 2)
2791 
2792 static void virt_machine_6_1_options(MachineClass *mc)
2793 {
2794     virt_machine_6_2_options(mc);
2795     compat_props_add(mc->compat_props, hw_compat_6_1, hw_compat_6_1_len);
2796 }
2797 DEFINE_VIRT_MACHINE(6, 1)
2798 
2799 static void virt_machine_6_0_options(MachineClass *mc)
2800 {
2801     virt_machine_6_1_options(mc);
2802     compat_props_add(mc->compat_props, hw_compat_6_0, hw_compat_6_0_len);
2803 }
2804 DEFINE_VIRT_MACHINE(6, 0)
2805 
2806 static void virt_machine_5_2_options(MachineClass *mc)
2807 {
2808     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2809 
2810     virt_machine_6_0_options(mc);
2811     compat_props_add(mc->compat_props, hw_compat_5_2, hw_compat_5_2_len);
2812     vmc->no_secure_gpio = true;
2813 }
2814 DEFINE_VIRT_MACHINE(5, 2)
2815 
2816 static void virt_machine_5_1_options(MachineClass *mc)
2817 {
2818     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2819 
2820     virt_machine_5_2_options(mc);
2821     compat_props_add(mc->compat_props, hw_compat_5_1, hw_compat_5_1_len);
2822     vmc->no_kvm_steal_time = true;
2823 }
2824 DEFINE_VIRT_MACHINE(5, 1)
2825 
2826 static void virt_machine_5_0_options(MachineClass *mc)
2827 {
2828     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2829 
2830     virt_machine_5_1_options(mc);
2831     compat_props_add(mc->compat_props, hw_compat_5_0, hw_compat_5_0_len);
2832     mc->numa_mem_supported = true;
2833     vmc->acpi_expose_flash = true;
2834     mc->auto_enable_numa_with_memdev = false;
2835 }
2836 DEFINE_VIRT_MACHINE(5, 0)
2837 
2838 static void virt_machine_4_2_options(MachineClass *mc)
2839 {
2840     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2841 
2842     virt_machine_5_0_options(mc);
2843     compat_props_add(mc->compat_props, hw_compat_4_2, hw_compat_4_2_len);
2844     vmc->kvm_no_adjvtime = true;
2845 }
2846 DEFINE_VIRT_MACHINE(4, 2)
2847 
2848 static void virt_machine_4_1_options(MachineClass *mc)
2849 {
2850     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2851 
2852     virt_machine_4_2_options(mc);
2853     compat_props_add(mc->compat_props, hw_compat_4_1, hw_compat_4_1_len);
2854     vmc->no_ged = true;
2855     mc->auto_enable_numa_with_memhp = false;
2856 }
2857 DEFINE_VIRT_MACHINE(4, 1)
2858 
2859 static void virt_machine_4_0_options(MachineClass *mc)
2860 {
2861     virt_machine_4_1_options(mc);
2862     compat_props_add(mc->compat_props, hw_compat_4_0, hw_compat_4_0_len);
2863 }
2864 DEFINE_VIRT_MACHINE(4, 0)
2865 
2866 static void virt_machine_3_1_options(MachineClass *mc)
2867 {
2868     virt_machine_4_0_options(mc);
2869     compat_props_add(mc->compat_props, hw_compat_3_1, hw_compat_3_1_len);
2870 }
2871 DEFINE_VIRT_MACHINE(3, 1)
2872 
2873 static void virt_machine_3_0_options(MachineClass *mc)
2874 {
2875     virt_machine_3_1_options(mc);
2876     compat_props_add(mc->compat_props, hw_compat_3_0, hw_compat_3_0_len);
2877 }
2878 DEFINE_VIRT_MACHINE(3, 0)
2879 
2880 static void virt_machine_2_12_options(MachineClass *mc)
2881 {
2882     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2883 
2884     virt_machine_3_0_options(mc);
2885     compat_props_add(mc->compat_props, hw_compat_2_12, hw_compat_2_12_len);
2886     vmc->no_highmem_ecam = true;
2887     mc->max_cpus = 255;
2888 }
2889 DEFINE_VIRT_MACHINE(2, 12)
2890 
2891 static void virt_machine_2_11_options(MachineClass *mc)
2892 {
2893     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2894 
2895     virt_machine_2_12_options(mc);
2896     compat_props_add(mc->compat_props, hw_compat_2_11, hw_compat_2_11_len);
2897     vmc->smbios_old_sys_ver = true;
2898 }
2899 DEFINE_VIRT_MACHINE(2, 11)
2900 
2901 static void virt_machine_2_10_options(MachineClass *mc)
2902 {
2903     virt_machine_2_11_options(mc);
2904     compat_props_add(mc->compat_props, hw_compat_2_10, hw_compat_2_10_len);
2905     /* before 2.11 we never faulted accesses to bad addresses */
2906     mc->ignore_memory_transaction_failures = true;
2907 }
2908 DEFINE_VIRT_MACHINE(2, 10)
2909 
2910 static void virt_machine_2_9_options(MachineClass *mc)
2911 {
2912     virt_machine_2_10_options(mc);
2913     compat_props_add(mc->compat_props, hw_compat_2_9, hw_compat_2_9_len);
2914 }
2915 DEFINE_VIRT_MACHINE(2, 9)
2916 
2917 static void virt_machine_2_8_options(MachineClass *mc)
2918 {
2919     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2920 
2921     virt_machine_2_9_options(mc);
2922     compat_props_add(mc->compat_props, hw_compat_2_8, hw_compat_2_8_len);
2923     /* For 2.8 and earlier we falsely claimed in the DT that
2924      * our timers were edge-triggered, not level-triggered.
2925      */
2926     vmc->claim_edge_triggered_timers = true;
2927 }
2928 DEFINE_VIRT_MACHINE(2, 8)
2929 
2930 static void virt_machine_2_7_options(MachineClass *mc)
2931 {
2932     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2933 
2934     virt_machine_2_8_options(mc);
2935     compat_props_add(mc->compat_props, hw_compat_2_7, hw_compat_2_7_len);
2936     /* ITS was introduced with 2.8 */
2937     vmc->no_its = true;
2938     /* Stick with 1K pages for migration compatibility */
2939     mc->minimum_page_bits = 0;
2940 }
2941 DEFINE_VIRT_MACHINE(2, 7)
2942 
2943 static void virt_machine_2_6_options(MachineClass *mc)
2944 {
2945     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2946 
2947     virt_machine_2_7_options(mc);
2948     compat_props_add(mc->compat_props, hw_compat_2_6, hw_compat_2_6_len);
2949     vmc->disallow_affinity_adjustment = true;
2950     /* Disable PMU for 2.6 as PMU support was first introduced in 2.7 */
2951     vmc->no_pmu = true;
2952 }
2953 DEFINE_VIRT_MACHINE(2, 6)
2954