xref: /qemu/hw/arm/virt.c (revision 64e5758b)
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 (!strcmp(itsclass, "arm-gicv3-its")) {
588         if (!vms->tcg_its) {
589             itsclass = NULL;
590         }
591     }
592 
593     if (!itsclass) {
594         /* Do nothing if not supported */
595         return;
596     }
597 
598     dev = qdev_new(itsclass);
599 
600     object_property_set_link(OBJECT(dev), "parent-gicv3", OBJECT(vms->gic),
601                              &error_abort);
602     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
603     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_GIC_ITS].base);
604 
605     fdt_add_its_gic_node(vms);
606     vms->msi_controller = VIRT_MSI_CTRL_ITS;
607 }
608 
609 static void create_v2m(VirtMachineState *vms)
610 {
611     int i;
612     int irq = vms->irqmap[VIRT_GIC_V2M];
613     DeviceState *dev;
614 
615     dev = qdev_new("arm-gicv2m");
616     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_GIC_V2M].base);
617     qdev_prop_set_uint32(dev, "base-spi", irq);
618     qdev_prop_set_uint32(dev, "num-spi", NUM_GICV2M_SPIS);
619     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
620 
621     for (i = 0; i < NUM_GICV2M_SPIS; i++) {
622         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
623                            qdev_get_gpio_in(vms->gic, irq + i));
624     }
625 
626     fdt_add_v2m_gic_node(vms);
627     vms->msi_controller = VIRT_MSI_CTRL_GICV2M;
628 }
629 
630 static void create_gic(VirtMachineState *vms, MemoryRegion *mem)
631 {
632     MachineState *ms = MACHINE(vms);
633     /* We create a standalone GIC */
634     SysBusDevice *gicbusdev;
635     const char *gictype;
636     int type = vms->gic_version, i;
637     unsigned int smp_cpus = ms->smp.cpus;
638     uint32_t nb_redist_regions = 0;
639 
640     gictype = (type == 3) ? gicv3_class_name() : gic_class_name();
641 
642     vms->gic = qdev_new(gictype);
643     qdev_prop_set_uint32(vms->gic, "revision", type);
644     qdev_prop_set_uint32(vms->gic, "num-cpu", smp_cpus);
645     /* Note that the num-irq property counts both internal and external
646      * interrupts; there are always 32 of the former (mandated by GIC spec).
647      */
648     qdev_prop_set_uint32(vms->gic, "num-irq", NUM_IRQS + 32);
649     if (!kvm_irqchip_in_kernel()) {
650         qdev_prop_set_bit(vms->gic, "has-security-extensions", vms->secure);
651     }
652 
653     if (type == 3) {
654         uint32_t redist0_capacity =
655                     vms->memmap[VIRT_GIC_REDIST].size / GICV3_REDIST_SIZE;
656         uint32_t redist0_count = MIN(smp_cpus, redist0_capacity);
657 
658         nb_redist_regions = virt_gicv3_redist_region_count(vms);
659 
660         qdev_prop_set_uint32(vms->gic, "len-redist-region-count",
661                              nb_redist_regions);
662         qdev_prop_set_uint32(vms->gic, "redist-region-count[0]", redist0_count);
663 
664         if (!kvm_irqchip_in_kernel()) {
665             if (vms->tcg_its) {
666                 object_property_set_link(OBJECT(vms->gic), "sysmem",
667                                          OBJECT(mem), &error_fatal);
668                 qdev_prop_set_bit(vms->gic, "has-lpi", true);
669             }
670         }
671 
672         if (nb_redist_regions == 2) {
673             uint32_t redist1_capacity =
674                     vms->memmap[VIRT_HIGH_GIC_REDIST2].size / GICV3_REDIST_SIZE;
675 
676             qdev_prop_set_uint32(vms->gic, "redist-region-count[1]",
677                 MIN(smp_cpus - redist0_count, redist1_capacity));
678         }
679     } else {
680         if (!kvm_irqchip_in_kernel()) {
681             qdev_prop_set_bit(vms->gic, "has-virtualization-extensions",
682                               vms->virt);
683         }
684     }
685     gicbusdev = SYS_BUS_DEVICE(vms->gic);
686     sysbus_realize_and_unref(gicbusdev, &error_fatal);
687     sysbus_mmio_map(gicbusdev, 0, vms->memmap[VIRT_GIC_DIST].base);
688     if (type == 3) {
689         sysbus_mmio_map(gicbusdev, 1, vms->memmap[VIRT_GIC_REDIST].base);
690         if (nb_redist_regions == 2) {
691             sysbus_mmio_map(gicbusdev, 2,
692                             vms->memmap[VIRT_HIGH_GIC_REDIST2].base);
693         }
694     } else {
695         sysbus_mmio_map(gicbusdev, 1, vms->memmap[VIRT_GIC_CPU].base);
696         if (vms->virt) {
697             sysbus_mmio_map(gicbusdev, 2, vms->memmap[VIRT_GIC_HYP].base);
698             sysbus_mmio_map(gicbusdev, 3, vms->memmap[VIRT_GIC_VCPU].base);
699         }
700     }
701 
702     /* Wire the outputs from each CPU's generic timer and the GICv3
703      * maintenance interrupt signal to the appropriate GIC PPI inputs,
704      * and the GIC's IRQ/FIQ/VIRQ/VFIQ interrupt outputs to the CPU's inputs.
705      */
706     for (i = 0; i < smp_cpus; i++) {
707         DeviceState *cpudev = DEVICE(qemu_get_cpu(i));
708         int ppibase = NUM_IRQS + i * GIC_INTERNAL + GIC_NR_SGIS;
709         int irq;
710         /* Mapping from the output timer irq lines from the CPU to the
711          * GIC PPI inputs we use for the virt board.
712          */
713         const int timer_irq[] = {
714             [GTIMER_PHYS] = ARCH_TIMER_NS_EL1_IRQ,
715             [GTIMER_VIRT] = ARCH_TIMER_VIRT_IRQ,
716             [GTIMER_HYP]  = ARCH_TIMER_NS_EL2_IRQ,
717             [GTIMER_SEC]  = ARCH_TIMER_S_EL1_IRQ,
718         };
719 
720         for (irq = 0; irq < ARRAY_SIZE(timer_irq); irq++) {
721             qdev_connect_gpio_out(cpudev, irq,
722                                   qdev_get_gpio_in(vms->gic,
723                                                    ppibase + timer_irq[irq]));
724         }
725 
726         if (type == 3) {
727             qemu_irq irq = qdev_get_gpio_in(vms->gic,
728                                             ppibase + ARCH_GIC_MAINT_IRQ);
729             qdev_connect_gpio_out_named(cpudev, "gicv3-maintenance-interrupt",
730                                         0, irq);
731         } else if (vms->virt) {
732             qemu_irq irq = qdev_get_gpio_in(vms->gic,
733                                             ppibase + ARCH_GIC_MAINT_IRQ);
734             sysbus_connect_irq(gicbusdev, i + 4 * smp_cpus, irq);
735         }
736 
737         qdev_connect_gpio_out_named(cpudev, "pmu-interrupt", 0,
738                                     qdev_get_gpio_in(vms->gic, ppibase
739                                                      + VIRTUAL_PMU_IRQ));
740 
741         sysbus_connect_irq(gicbusdev, i, qdev_get_gpio_in(cpudev, ARM_CPU_IRQ));
742         sysbus_connect_irq(gicbusdev, i + smp_cpus,
743                            qdev_get_gpio_in(cpudev, ARM_CPU_FIQ));
744         sysbus_connect_irq(gicbusdev, i + 2 * smp_cpus,
745                            qdev_get_gpio_in(cpudev, ARM_CPU_VIRQ));
746         sysbus_connect_irq(gicbusdev, i + 3 * smp_cpus,
747                            qdev_get_gpio_in(cpudev, ARM_CPU_VFIQ));
748     }
749 
750     fdt_add_gic_node(vms);
751 
752     if (type == 3 && vms->its) {
753         create_its(vms);
754     } else if (type == 2) {
755         create_v2m(vms);
756     }
757 }
758 
759 static void create_uart(const VirtMachineState *vms, int uart,
760                         MemoryRegion *mem, Chardev *chr)
761 {
762     char *nodename;
763     hwaddr base = vms->memmap[uart].base;
764     hwaddr size = vms->memmap[uart].size;
765     int irq = vms->irqmap[uart];
766     const char compat[] = "arm,pl011\0arm,primecell";
767     const char clocknames[] = "uartclk\0apb_pclk";
768     DeviceState *dev = qdev_new(TYPE_PL011);
769     SysBusDevice *s = SYS_BUS_DEVICE(dev);
770     MachineState *ms = MACHINE(vms);
771 
772     qdev_prop_set_chr(dev, "chardev", chr);
773     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
774     memory_region_add_subregion(mem, base,
775                                 sysbus_mmio_get_region(s, 0));
776     sysbus_connect_irq(s, 0, qdev_get_gpio_in(vms->gic, irq));
777 
778     nodename = g_strdup_printf("/pl011@%" PRIx64, base);
779     qemu_fdt_add_subnode(ms->fdt, nodename);
780     /* Note that we can't use setprop_string because of the embedded NUL */
781     qemu_fdt_setprop(ms->fdt, nodename, "compatible",
782                          compat, sizeof(compat));
783     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
784                                      2, base, 2, size);
785     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
786                                GIC_FDT_IRQ_TYPE_SPI, irq,
787                                GIC_FDT_IRQ_FLAGS_LEVEL_HI);
788     qemu_fdt_setprop_cells(ms->fdt, nodename, "clocks",
789                                vms->clock_phandle, vms->clock_phandle);
790     qemu_fdt_setprop(ms->fdt, nodename, "clock-names",
791                          clocknames, sizeof(clocknames));
792 
793     if (uart == VIRT_UART) {
794         qemu_fdt_setprop_string(ms->fdt, "/chosen", "stdout-path", nodename);
795     } else {
796         /* Mark as not usable by the normal world */
797         qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
798         qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
799 
800         qemu_fdt_setprop_string(ms->fdt, "/secure-chosen", "stdout-path",
801                                 nodename);
802     }
803 
804     g_free(nodename);
805 }
806 
807 static void create_rtc(const VirtMachineState *vms)
808 {
809     char *nodename;
810     hwaddr base = vms->memmap[VIRT_RTC].base;
811     hwaddr size = vms->memmap[VIRT_RTC].size;
812     int irq = vms->irqmap[VIRT_RTC];
813     const char compat[] = "arm,pl031\0arm,primecell";
814     MachineState *ms = MACHINE(vms);
815 
816     sysbus_create_simple("pl031", base, qdev_get_gpio_in(vms->gic, irq));
817 
818     nodename = g_strdup_printf("/pl031@%" PRIx64, base);
819     qemu_fdt_add_subnode(ms->fdt, nodename);
820     qemu_fdt_setprop(ms->fdt, nodename, "compatible", compat, sizeof(compat));
821     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
822                                  2, base, 2, size);
823     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
824                            GIC_FDT_IRQ_TYPE_SPI, irq,
825                            GIC_FDT_IRQ_FLAGS_LEVEL_HI);
826     qemu_fdt_setprop_cell(ms->fdt, nodename, "clocks", vms->clock_phandle);
827     qemu_fdt_setprop_string(ms->fdt, nodename, "clock-names", "apb_pclk");
828     g_free(nodename);
829 }
830 
831 static DeviceState *gpio_key_dev;
832 static void virt_powerdown_req(Notifier *n, void *opaque)
833 {
834     VirtMachineState *s = container_of(n, VirtMachineState, powerdown_notifier);
835 
836     if (s->acpi_dev) {
837         acpi_send_event(s->acpi_dev, ACPI_POWER_DOWN_STATUS);
838     } else {
839         /* use gpio Pin 3 for power button event */
840         qemu_set_irq(qdev_get_gpio_in(gpio_key_dev, 0), 1);
841     }
842 }
843 
844 static void create_gpio_keys(char *fdt, DeviceState *pl061_dev,
845                              uint32_t phandle)
846 {
847     gpio_key_dev = sysbus_create_simple("gpio-key", -1,
848                                         qdev_get_gpio_in(pl061_dev, 3));
849 
850     qemu_fdt_add_subnode(fdt, "/gpio-keys");
851     qemu_fdt_setprop_string(fdt, "/gpio-keys", "compatible", "gpio-keys");
852     qemu_fdt_setprop_cell(fdt, "/gpio-keys", "#size-cells", 0);
853     qemu_fdt_setprop_cell(fdt, "/gpio-keys", "#address-cells", 1);
854 
855     qemu_fdt_add_subnode(fdt, "/gpio-keys/poweroff");
856     qemu_fdt_setprop_string(fdt, "/gpio-keys/poweroff",
857                             "label", "GPIO Key Poweroff");
858     qemu_fdt_setprop_cell(fdt, "/gpio-keys/poweroff", "linux,code",
859                           KEY_POWER);
860     qemu_fdt_setprop_cells(fdt, "/gpio-keys/poweroff",
861                            "gpios", phandle, 3, 0);
862 }
863 
864 #define SECURE_GPIO_POWEROFF 0
865 #define SECURE_GPIO_RESET    1
866 
867 static void create_secure_gpio_pwr(char *fdt, DeviceState *pl061_dev,
868                                    uint32_t phandle)
869 {
870     DeviceState *gpio_pwr_dev;
871 
872     /* gpio-pwr */
873     gpio_pwr_dev = sysbus_create_simple("gpio-pwr", -1, NULL);
874 
875     /* connect secure pl061 to gpio-pwr */
876     qdev_connect_gpio_out(pl061_dev, SECURE_GPIO_RESET,
877                           qdev_get_gpio_in_named(gpio_pwr_dev, "reset", 0));
878     qdev_connect_gpio_out(pl061_dev, SECURE_GPIO_POWEROFF,
879                           qdev_get_gpio_in_named(gpio_pwr_dev, "shutdown", 0));
880 
881     qemu_fdt_add_subnode(fdt, "/gpio-poweroff");
882     qemu_fdt_setprop_string(fdt, "/gpio-poweroff", "compatible",
883                             "gpio-poweroff");
884     qemu_fdt_setprop_cells(fdt, "/gpio-poweroff",
885                            "gpios", phandle, SECURE_GPIO_POWEROFF, 0);
886     qemu_fdt_setprop_string(fdt, "/gpio-poweroff", "status", "disabled");
887     qemu_fdt_setprop_string(fdt, "/gpio-poweroff", "secure-status",
888                             "okay");
889 
890     qemu_fdt_add_subnode(fdt, "/gpio-restart");
891     qemu_fdt_setprop_string(fdt, "/gpio-restart", "compatible",
892                             "gpio-restart");
893     qemu_fdt_setprop_cells(fdt, "/gpio-restart",
894                            "gpios", phandle, SECURE_GPIO_RESET, 0);
895     qemu_fdt_setprop_string(fdt, "/gpio-restart", "status", "disabled");
896     qemu_fdt_setprop_string(fdt, "/gpio-restart", "secure-status",
897                             "okay");
898 }
899 
900 static void create_gpio_devices(const VirtMachineState *vms, int gpio,
901                                 MemoryRegion *mem)
902 {
903     char *nodename;
904     DeviceState *pl061_dev;
905     hwaddr base = vms->memmap[gpio].base;
906     hwaddr size = vms->memmap[gpio].size;
907     int irq = vms->irqmap[gpio];
908     const char compat[] = "arm,pl061\0arm,primecell";
909     SysBusDevice *s;
910     MachineState *ms = MACHINE(vms);
911 
912     pl061_dev = qdev_new("pl061");
913     /* Pull lines down to 0 if not driven by the PL061 */
914     qdev_prop_set_uint32(pl061_dev, "pullups", 0);
915     qdev_prop_set_uint32(pl061_dev, "pulldowns", 0xff);
916     s = SYS_BUS_DEVICE(pl061_dev);
917     sysbus_realize_and_unref(s, &error_fatal);
918     memory_region_add_subregion(mem, base, sysbus_mmio_get_region(s, 0));
919     sysbus_connect_irq(s, 0, qdev_get_gpio_in(vms->gic, irq));
920 
921     uint32_t phandle = qemu_fdt_alloc_phandle(ms->fdt);
922     nodename = g_strdup_printf("/pl061@%" PRIx64, base);
923     qemu_fdt_add_subnode(ms->fdt, nodename);
924     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
925                                  2, base, 2, size);
926     qemu_fdt_setprop(ms->fdt, nodename, "compatible", compat, sizeof(compat));
927     qemu_fdt_setprop_cell(ms->fdt, nodename, "#gpio-cells", 2);
928     qemu_fdt_setprop(ms->fdt, nodename, "gpio-controller", NULL, 0);
929     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
930                            GIC_FDT_IRQ_TYPE_SPI, irq,
931                            GIC_FDT_IRQ_FLAGS_LEVEL_HI);
932     qemu_fdt_setprop_cell(ms->fdt, nodename, "clocks", vms->clock_phandle);
933     qemu_fdt_setprop_string(ms->fdt, nodename, "clock-names", "apb_pclk");
934     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", phandle);
935 
936     if (gpio != VIRT_GPIO) {
937         /* Mark as not usable by the normal world */
938         qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
939         qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
940     }
941     g_free(nodename);
942 
943     /* Child gpio devices */
944     if (gpio == VIRT_GPIO) {
945         create_gpio_keys(ms->fdt, pl061_dev, phandle);
946     } else {
947         create_secure_gpio_pwr(ms->fdt, pl061_dev, phandle);
948     }
949 }
950 
951 static void create_virtio_devices(const VirtMachineState *vms)
952 {
953     int i;
954     hwaddr size = vms->memmap[VIRT_MMIO].size;
955     MachineState *ms = MACHINE(vms);
956 
957     /* We create the transports in forwards order. Since qbus_realize()
958      * prepends (not appends) new child buses, the incrementing loop below will
959      * create a list of virtio-mmio buses with decreasing base addresses.
960      *
961      * When a -device option is processed from the command line,
962      * qbus_find_recursive() picks the next free virtio-mmio bus in forwards
963      * order. The upshot is that -device options in increasing command line
964      * order are mapped to virtio-mmio buses with decreasing base addresses.
965      *
966      * When this code was originally written, that arrangement ensured that the
967      * guest Linux kernel would give the lowest "name" (/dev/vda, eth0, etc) to
968      * the first -device on the command line. (The end-to-end order is a
969      * function of this loop, qbus_realize(), qbus_find_recursive(), and the
970      * guest kernel's name-to-address assignment strategy.)
971      *
972      * Meanwhile, the kernel's traversal seems to have been reversed; see eg.
973      * the message, if not necessarily the code, of commit 70161ff336.
974      * Therefore the loop now establishes the inverse of the original intent.
975      *
976      * Unfortunately, we can't counteract the kernel change by reversing the
977      * loop; it would break existing command lines.
978      *
979      * In any case, the kernel makes no guarantee about the stability of
980      * enumeration order of virtio devices (as demonstrated by it changing
981      * between kernel versions). For reliable and stable identification
982      * of disks users must use UUIDs or similar mechanisms.
983      */
984     for (i = 0; i < NUM_VIRTIO_TRANSPORTS; i++) {
985         int irq = vms->irqmap[VIRT_MMIO] + i;
986         hwaddr base = vms->memmap[VIRT_MMIO].base + i * size;
987 
988         sysbus_create_simple("virtio-mmio", base,
989                              qdev_get_gpio_in(vms->gic, irq));
990     }
991 
992     /* We add dtb nodes in reverse order so that they appear in the finished
993      * device tree lowest address first.
994      *
995      * Note that this mapping is independent of the loop above. The previous
996      * loop influences virtio device to virtio transport assignment, whereas
997      * this loop controls how virtio transports are laid out in the dtb.
998      */
999     for (i = NUM_VIRTIO_TRANSPORTS - 1; i >= 0; i--) {
1000         char *nodename;
1001         int irq = vms->irqmap[VIRT_MMIO] + i;
1002         hwaddr base = vms->memmap[VIRT_MMIO].base + i * size;
1003 
1004         nodename = g_strdup_printf("/virtio_mmio@%" PRIx64, base);
1005         qemu_fdt_add_subnode(ms->fdt, nodename);
1006         qemu_fdt_setprop_string(ms->fdt, nodename,
1007                                 "compatible", "virtio,mmio");
1008         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1009                                      2, base, 2, size);
1010         qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
1011                                GIC_FDT_IRQ_TYPE_SPI, irq,
1012                                GIC_FDT_IRQ_FLAGS_EDGE_LO_HI);
1013         qemu_fdt_setprop(ms->fdt, nodename, "dma-coherent", NULL, 0);
1014         g_free(nodename);
1015     }
1016 }
1017 
1018 #define VIRT_FLASH_SECTOR_SIZE (256 * KiB)
1019 
1020 static PFlashCFI01 *virt_flash_create1(VirtMachineState *vms,
1021                                         const char *name,
1022                                         const char *alias_prop_name)
1023 {
1024     /*
1025      * Create a single flash device.  We use the same parameters as
1026      * the flash devices on the Versatile Express board.
1027      */
1028     DeviceState *dev = qdev_new(TYPE_PFLASH_CFI01);
1029 
1030     qdev_prop_set_uint64(dev, "sector-length", VIRT_FLASH_SECTOR_SIZE);
1031     qdev_prop_set_uint8(dev, "width", 4);
1032     qdev_prop_set_uint8(dev, "device-width", 2);
1033     qdev_prop_set_bit(dev, "big-endian", false);
1034     qdev_prop_set_uint16(dev, "id0", 0x89);
1035     qdev_prop_set_uint16(dev, "id1", 0x18);
1036     qdev_prop_set_uint16(dev, "id2", 0x00);
1037     qdev_prop_set_uint16(dev, "id3", 0x00);
1038     qdev_prop_set_string(dev, "name", name);
1039     object_property_add_child(OBJECT(vms), name, OBJECT(dev));
1040     object_property_add_alias(OBJECT(vms), alias_prop_name,
1041                               OBJECT(dev), "drive");
1042     return PFLASH_CFI01(dev);
1043 }
1044 
1045 static void virt_flash_create(VirtMachineState *vms)
1046 {
1047     vms->flash[0] = virt_flash_create1(vms, "virt.flash0", "pflash0");
1048     vms->flash[1] = virt_flash_create1(vms, "virt.flash1", "pflash1");
1049 }
1050 
1051 static void virt_flash_map1(PFlashCFI01 *flash,
1052                             hwaddr base, hwaddr size,
1053                             MemoryRegion *sysmem)
1054 {
1055     DeviceState *dev = DEVICE(flash);
1056 
1057     assert(QEMU_IS_ALIGNED(size, VIRT_FLASH_SECTOR_SIZE));
1058     assert(size / VIRT_FLASH_SECTOR_SIZE <= UINT32_MAX);
1059     qdev_prop_set_uint32(dev, "num-blocks", size / VIRT_FLASH_SECTOR_SIZE);
1060     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1061 
1062     memory_region_add_subregion(sysmem, base,
1063                                 sysbus_mmio_get_region(SYS_BUS_DEVICE(dev),
1064                                                        0));
1065 }
1066 
1067 static void virt_flash_map(VirtMachineState *vms,
1068                            MemoryRegion *sysmem,
1069                            MemoryRegion *secure_sysmem)
1070 {
1071     /*
1072      * Map two flash devices to fill the VIRT_FLASH space in the memmap.
1073      * sysmem is the system memory space. secure_sysmem is the secure view
1074      * of the system, and the first flash device should be made visible only
1075      * there. The second flash device is visible to both secure and nonsecure.
1076      * If sysmem == secure_sysmem this means there is no separate Secure
1077      * address space and both flash devices are generally visible.
1078      */
1079     hwaddr flashsize = vms->memmap[VIRT_FLASH].size / 2;
1080     hwaddr flashbase = vms->memmap[VIRT_FLASH].base;
1081 
1082     virt_flash_map1(vms->flash[0], flashbase, flashsize,
1083                     secure_sysmem);
1084     virt_flash_map1(vms->flash[1], flashbase + flashsize, flashsize,
1085                     sysmem);
1086 }
1087 
1088 static void virt_flash_fdt(VirtMachineState *vms,
1089                            MemoryRegion *sysmem,
1090                            MemoryRegion *secure_sysmem)
1091 {
1092     hwaddr flashsize = vms->memmap[VIRT_FLASH].size / 2;
1093     hwaddr flashbase = vms->memmap[VIRT_FLASH].base;
1094     MachineState *ms = MACHINE(vms);
1095     char *nodename;
1096 
1097     if (sysmem == secure_sysmem) {
1098         /* Report both flash devices as a single node in the DT */
1099         nodename = g_strdup_printf("/flash@%" PRIx64, flashbase);
1100         qemu_fdt_add_subnode(ms->fdt, nodename);
1101         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible", "cfi-flash");
1102         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1103                                      2, flashbase, 2, flashsize,
1104                                      2, flashbase + flashsize, 2, flashsize);
1105         qemu_fdt_setprop_cell(ms->fdt, nodename, "bank-width", 4);
1106         g_free(nodename);
1107     } else {
1108         /*
1109          * Report the devices as separate nodes so we can mark one as
1110          * only visible to the secure world.
1111          */
1112         nodename = g_strdup_printf("/secflash@%" PRIx64, flashbase);
1113         qemu_fdt_add_subnode(ms->fdt, nodename);
1114         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible", "cfi-flash");
1115         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1116                                      2, flashbase, 2, flashsize);
1117         qemu_fdt_setprop_cell(ms->fdt, nodename, "bank-width", 4);
1118         qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
1119         qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
1120         g_free(nodename);
1121 
1122         nodename = g_strdup_printf("/flash@%" PRIx64, flashbase);
1123         qemu_fdt_add_subnode(ms->fdt, nodename);
1124         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible", "cfi-flash");
1125         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1126                                      2, flashbase + flashsize, 2, flashsize);
1127         qemu_fdt_setprop_cell(ms->fdt, nodename, "bank-width", 4);
1128         g_free(nodename);
1129     }
1130 }
1131 
1132 static bool virt_firmware_init(VirtMachineState *vms,
1133                                MemoryRegion *sysmem,
1134                                MemoryRegion *secure_sysmem)
1135 {
1136     int i;
1137     const char *bios_name;
1138     BlockBackend *pflash_blk0;
1139 
1140     /* Map legacy -drive if=pflash to machine properties */
1141     for (i = 0; i < ARRAY_SIZE(vms->flash); i++) {
1142         pflash_cfi01_legacy_drive(vms->flash[i],
1143                                   drive_get(IF_PFLASH, 0, i));
1144     }
1145 
1146     virt_flash_map(vms, sysmem, secure_sysmem);
1147 
1148     pflash_blk0 = pflash_cfi01_get_blk(vms->flash[0]);
1149 
1150     bios_name = MACHINE(vms)->firmware;
1151     if (bios_name) {
1152         char *fname;
1153         MemoryRegion *mr;
1154         int image_size;
1155 
1156         if (pflash_blk0) {
1157             error_report("The contents of the first flash device may be "
1158                          "specified with -bios or with -drive if=pflash... "
1159                          "but you cannot use both options at once");
1160             exit(1);
1161         }
1162 
1163         /* Fall back to -bios */
1164 
1165         fname = qemu_find_file(QEMU_FILE_TYPE_BIOS, bios_name);
1166         if (!fname) {
1167             error_report("Could not find ROM image '%s'", bios_name);
1168             exit(1);
1169         }
1170         mr = sysbus_mmio_get_region(SYS_BUS_DEVICE(vms->flash[0]), 0);
1171         image_size = load_image_mr(fname, mr);
1172         g_free(fname);
1173         if (image_size < 0) {
1174             error_report("Could not load ROM image '%s'", bios_name);
1175             exit(1);
1176         }
1177     }
1178 
1179     return pflash_blk0 || bios_name;
1180 }
1181 
1182 static FWCfgState *create_fw_cfg(const VirtMachineState *vms, AddressSpace *as)
1183 {
1184     MachineState *ms = MACHINE(vms);
1185     hwaddr base = vms->memmap[VIRT_FW_CFG].base;
1186     hwaddr size = vms->memmap[VIRT_FW_CFG].size;
1187     FWCfgState *fw_cfg;
1188     char *nodename;
1189 
1190     fw_cfg = fw_cfg_init_mem_wide(base + 8, base, 8, base + 16, as);
1191     fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, (uint16_t)ms->smp.cpus);
1192 
1193     nodename = g_strdup_printf("/fw-cfg@%" PRIx64, base);
1194     qemu_fdt_add_subnode(ms->fdt, nodename);
1195     qemu_fdt_setprop_string(ms->fdt, nodename,
1196                             "compatible", "qemu,fw-cfg-mmio");
1197     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1198                                  2, base, 2, size);
1199     qemu_fdt_setprop(ms->fdt, nodename, "dma-coherent", NULL, 0);
1200     g_free(nodename);
1201     return fw_cfg;
1202 }
1203 
1204 static void create_pcie_irq_map(const MachineState *ms,
1205                                 uint32_t gic_phandle,
1206                                 int first_irq, const char *nodename)
1207 {
1208     int devfn, pin;
1209     uint32_t full_irq_map[4 * 4 * 10] = { 0 };
1210     uint32_t *irq_map = full_irq_map;
1211 
1212     for (devfn = 0; devfn <= 0x18; devfn += 0x8) {
1213         for (pin = 0; pin < 4; pin++) {
1214             int irq_type = GIC_FDT_IRQ_TYPE_SPI;
1215             int irq_nr = first_irq + ((pin + PCI_SLOT(devfn)) % PCI_NUM_PINS);
1216             int irq_level = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
1217             int i;
1218 
1219             uint32_t map[] = {
1220                 devfn << 8, 0, 0,                           /* devfn */
1221                 pin + 1,                                    /* PCI pin */
1222                 gic_phandle, 0, 0, irq_type, irq_nr, irq_level }; /* GIC irq */
1223 
1224             /* Convert map to big endian */
1225             for (i = 0; i < 10; i++) {
1226                 irq_map[i] = cpu_to_be32(map[i]);
1227             }
1228             irq_map += 10;
1229         }
1230     }
1231 
1232     qemu_fdt_setprop(ms->fdt, nodename, "interrupt-map",
1233                      full_irq_map, sizeof(full_irq_map));
1234 
1235     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupt-map-mask",
1236                            cpu_to_be16(PCI_DEVFN(3, 0)), /* Slot 3 */
1237                            0, 0,
1238                            0x7           /* PCI irq */);
1239 }
1240 
1241 static void create_smmu(const VirtMachineState *vms,
1242                         PCIBus *bus)
1243 {
1244     char *node;
1245     const char compat[] = "arm,smmu-v3";
1246     int irq =  vms->irqmap[VIRT_SMMU];
1247     int i;
1248     hwaddr base = vms->memmap[VIRT_SMMU].base;
1249     hwaddr size = vms->memmap[VIRT_SMMU].size;
1250     const char irq_names[] = "eventq\0priq\0cmdq-sync\0gerror";
1251     DeviceState *dev;
1252     MachineState *ms = MACHINE(vms);
1253 
1254     if (vms->iommu != VIRT_IOMMU_SMMUV3 || !vms->iommu_phandle) {
1255         return;
1256     }
1257 
1258     dev = qdev_new("arm-smmuv3");
1259 
1260     object_property_set_link(OBJECT(dev), "primary-bus", OBJECT(bus),
1261                              &error_abort);
1262     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1263     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, base);
1264     for (i = 0; i < NUM_SMMU_IRQS; i++) {
1265         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
1266                            qdev_get_gpio_in(vms->gic, irq + i));
1267     }
1268 
1269     node = g_strdup_printf("/smmuv3@%" PRIx64, base);
1270     qemu_fdt_add_subnode(ms->fdt, node);
1271     qemu_fdt_setprop(ms->fdt, node, "compatible", compat, sizeof(compat));
1272     qemu_fdt_setprop_sized_cells(ms->fdt, node, "reg", 2, base, 2, size);
1273 
1274     qemu_fdt_setprop_cells(ms->fdt, node, "interrupts",
1275             GIC_FDT_IRQ_TYPE_SPI, irq    , GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1276             GIC_FDT_IRQ_TYPE_SPI, irq + 1, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1277             GIC_FDT_IRQ_TYPE_SPI, irq + 2, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1278             GIC_FDT_IRQ_TYPE_SPI, irq + 3, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI);
1279 
1280     qemu_fdt_setprop(ms->fdt, node, "interrupt-names", irq_names,
1281                      sizeof(irq_names));
1282 
1283     qemu_fdt_setprop_cell(ms->fdt, node, "clocks", vms->clock_phandle);
1284     qemu_fdt_setprop_string(ms->fdt, node, "clock-names", "apb_pclk");
1285     qemu_fdt_setprop(ms->fdt, node, "dma-coherent", NULL, 0);
1286 
1287     qemu_fdt_setprop_cell(ms->fdt, node, "#iommu-cells", 1);
1288 
1289     qemu_fdt_setprop_cell(ms->fdt, node, "phandle", vms->iommu_phandle);
1290     g_free(node);
1291 }
1292 
1293 static void create_virtio_iommu_dt_bindings(VirtMachineState *vms)
1294 {
1295     const char compat[] = "virtio,pci-iommu";
1296     uint16_t bdf = vms->virtio_iommu_bdf;
1297     MachineState *ms = MACHINE(vms);
1298     char *node;
1299 
1300     vms->iommu_phandle = qemu_fdt_alloc_phandle(ms->fdt);
1301 
1302     node = g_strdup_printf("%s/virtio_iommu@%d", vms->pciehb_nodename, bdf);
1303     qemu_fdt_add_subnode(ms->fdt, node);
1304     qemu_fdt_setprop(ms->fdt, node, "compatible", compat, sizeof(compat));
1305     qemu_fdt_setprop_sized_cells(ms->fdt, node, "reg",
1306                                  1, bdf << 8, 1, 0, 1, 0,
1307                                  1, 0, 1, 0);
1308 
1309     qemu_fdt_setprop_cell(ms->fdt, node, "#iommu-cells", 1);
1310     qemu_fdt_setprop_cell(ms->fdt, node, "phandle", vms->iommu_phandle);
1311     g_free(node);
1312 
1313     qemu_fdt_setprop_cells(ms->fdt, vms->pciehb_nodename, "iommu-map",
1314                            0x0, vms->iommu_phandle, 0x0, bdf,
1315                            bdf + 1, vms->iommu_phandle, bdf + 1, 0xffff - bdf);
1316 }
1317 
1318 static void create_pcie(VirtMachineState *vms)
1319 {
1320     hwaddr base_mmio = vms->memmap[VIRT_PCIE_MMIO].base;
1321     hwaddr size_mmio = vms->memmap[VIRT_PCIE_MMIO].size;
1322     hwaddr base_mmio_high = vms->memmap[VIRT_HIGH_PCIE_MMIO].base;
1323     hwaddr size_mmio_high = vms->memmap[VIRT_HIGH_PCIE_MMIO].size;
1324     hwaddr base_pio = vms->memmap[VIRT_PCIE_PIO].base;
1325     hwaddr size_pio = vms->memmap[VIRT_PCIE_PIO].size;
1326     hwaddr base_ecam, size_ecam;
1327     hwaddr base = base_mmio;
1328     int nr_pcie_buses;
1329     int irq = vms->irqmap[VIRT_PCIE];
1330     MemoryRegion *mmio_alias;
1331     MemoryRegion *mmio_reg;
1332     MemoryRegion *ecam_alias;
1333     MemoryRegion *ecam_reg;
1334     DeviceState *dev;
1335     char *nodename;
1336     int i, ecam_id;
1337     PCIHostState *pci;
1338     MachineState *ms = MACHINE(vms);
1339 
1340     dev = qdev_new(TYPE_GPEX_HOST);
1341     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1342 
1343     ecam_id = VIRT_ECAM_ID(vms->highmem_ecam);
1344     base_ecam = vms->memmap[ecam_id].base;
1345     size_ecam = vms->memmap[ecam_id].size;
1346     nr_pcie_buses = size_ecam / PCIE_MMCFG_SIZE_MIN;
1347     /* Map only the first size_ecam bytes of ECAM space */
1348     ecam_alias = g_new0(MemoryRegion, 1);
1349     ecam_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 0);
1350     memory_region_init_alias(ecam_alias, OBJECT(dev), "pcie-ecam",
1351                              ecam_reg, 0, size_ecam);
1352     memory_region_add_subregion(get_system_memory(), base_ecam, ecam_alias);
1353 
1354     /* Map the MMIO window into system address space so as to expose
1355      * the section of PCI MMIO space which starts at the same base address
1356      * (ie 1:1 mapping for that part of PCI MMIO space visible through
1357      * the window).
1358      */
1359     mmio_alias = g_new0(MemoryRegion, 1);
1360     mmio_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 1);
1361     memory_region_init_alias(mmio_alias, OBJECT(dev), "pcie-mmio",
1362                              mmio_reg, base_mmio, size_mmio);
1363     memory_region_add_subregion(get_system_memory(), base_mmio, mmio_alias);
1364 
1365     if (vms->highmem) {
1366         /* Map high MMIO space */
1367         MemoryRegion *high_mmio_alias = g_new0(MemoryRegion, 1);
1368 
1369         memory_region_init_alias(high_mmio_alias, OBJECT(dev), "pcie-mmio-high",
1370                                  mmio_reg, base_mmio_high, size_mmio_high);
1371         memory_region_add_subregion(get_system_memory(), base_mmio_high,
1372                                     high_mmio_alias);
1373     }
1374 
1375     /* Map IO port space */
1376     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 2, base_pio);
1377 
1378     for (i = 0; i < GPEX_NUM_IRQS; i++) {
1379         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
1380                            qdev_get_gpio_in(vms->gic, irq + i));
1381         gpex_set_irq_num(GPEX_HOST(dev), i, irq + i);
1382     }
1383 
1384     pci = PCI_HOST_BRIDGE(dev);
1385     pci->bypass_iommu = vms->default_bus_bypass_iommu;
1386     vms->bus = pci->bus;
1387     if (vms->bus) {
1388         for (i = 0; i < nb_nics; i++) {
1389             NICInfo *nd = &nd_table[i];
1390 
1391             if (!nd->model) {
1392                 nd->model = g_strdup("virtio");
1393             }
1394 
1395             pci_nic_init_nofail(nd, pci->bus, nd->model, NULL);
1396         }
1397     }
1398 
1399     nodename = vms->pciehb_nodename = g_strdup_printf("/pcie@%" PRIx64, base);
1400     qemu_fdt_add_subnode(ms->fdt, nodename);
1401     qemu_fdt_setprop_string(ms->fdt, nodename,
1402                             "compatible", "pci-host-ecam-generic");
1403     qemu_fdt_setprop_string(ms->fdt, nodename, "device_type", "pci");
1404     qemu_fdt_setprop_cell(ms->fdt, nodename, "#address-cells", 3);
1405     qemu_fdt_setprop_cell(ms->fdt, nodename, "#size-cells", 2);
1406     qemu_fdt_setprop_cell(ms->fdt, nodename, "linux,pci-domain", 0);
1407     qemu_fdt_setprop_cells(ms->fdt, nodename, "bus-range", 0,
1408                            nr_pcie_buses - 1);
1409     qemu_fdt_setprop(ms->fdt, nodename, "dma-coherent", NULL, 0);
1410 
1411     if (vms->msi_phandle) {
1412         qemu_fdt_setprop_cells(ms->fdt, nodename, "msi-parent",
1413                                vms->msi_phandle);
1414     }
1415 
1416     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1417                                  2, base_ecam, 2, size_ecam);
1418 
1419     if (vms->highmem) {
1420         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "ranges",
1421                                      1, FDT_PCI_RANGE_IOPORT, 2, 0,
1422                                      2, base_pio, 2, size_pio,
1423                                      1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
1424                                      2, base_mmio, 2, size_mmio,
1425                                      1, FDT_PCI_RANGE_MMIO_64BIT,
1426                                      2, base_mmio_high,
1427                                      2, base_mmio_high, 2, size_mmio_high);
1428     } else {
1429         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "ranges",
1430                                      1, FDT_PCI_RANGE_IOPORT, 2, 0,
1431                                      2, base_pio, 2, size_pio,
1432                                      1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
1433                                      2, base_mmio, 2, size_mmio);
1434     }
1435 
1436     qemu_fdt_setprop_cell(ms->fdt, nodename, "#interrupt-cells", 1);
1437     create_pcie_irq_map(ms, vms->gic_phandle, irq, nodename);
1438 
1439     if (vms->iommu) {
1440         vms->iommu_phandle = qemu_fdt_alloc_phandle(ms->fdt);
1441 
1442         switch (vms->iommu) {
1443         case VIRT_IOMMU_SMMUV3:
1444             create_smmu(vms, vms->bus);
1445             qemu_fdt_setprop_cells(ms->fdt, nodename, "iommu-map",
1446                                    0x0, vms->iommu_phandle, 0x0, 0x10000);
1447             break;
1448         default:
1449             g_assert_not_reached();
1450         }
1451     }
1452 }
1453 
1454 static void create_platform_bus(VirtMachineState *vms)
1455 {
1456     DeviceState *dev;
1457     SysBusDevice *s;
1458     int i;
1459     MemoryRegion *sysmem = get_system_memory();
1460 
1461     dev = qdev_new(TYPE_PLATFORM_BUS_DEVICE);
1462     dev->id = TYPE_PLATFORM_BUS_DEVICE;
1463     qdev_prop_set_uint32(dev, "num_irqs", PLATFORM_BUS_NUM_IRQS);
1464     qdev_prop_set_uint32(dev, "mmio_size", vms->memmap[VIRT_PLATFORM_BUS].size);
1465     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1466     vms->platform_bus_dev = dev;
1467 
1468     s = SYS_BUS_DEVICE(dev);
1469     for (i = 0; i < PLATFORM_BUS_NUM_IRQS; i++) {
1470         int irq = vms->irqmap[VIRT_PLATFORM_BUS] + i;
1471         sysbus_connect_irq(s, i, qdev_get_gpio_in(vms->gic, irq));
1472     }
1473 
1474     memory_region_add_subregion(sysmem,
1475                                 vms->memmap[VIRT_PLATFORM_BUS].base,
1476                                 sysbus_mmio_get_region(s, 0));
1477 }
1478 
1479 static void create_tag_ram(MemoryRegion *tag_sysmem,
1480                            hwaddr base, hwaddr size,
1481                            const char *name)
1482 {
1483     MemoryRegion *tagram = g_new(MemoryRegion, 1);
1484 
1485     memory_region_init_ram(tagram, NULL, name, size / 32, &error_fatal);
1486     memory_region_add_subregion(tag_sysmem, base / 32, tagram);
1487 }
1488 
1489 static void create_secure_ram(VirtMachineState *vms,
1490                               MemoryRegion *secure_sysmem,
1491                               MemoryRegion *secure_tag_sysmem)
1492 {
1493     MemoryRegion *secram = g_new(MemoryRegion, 1);
1494     char *nodename;
1495     hwaddr base = vms->memmap[VIRT_SECURE_MEM].base;
1496     hwaddr size = vms->memmap[VIRT_SECURE_MEM].size;
1497     MachineState *ms = MACHINE(vms);
1498 
1499     memory_region_init_ram(secram, NULL, "virt.secure-ram", size,
1500                            &error_fatal);
1501     memory_region_add_subregion(secure_sysmem, base, secram);
1502 
1503     nodename = g_strdup_printf("/secram@%" PRIx64, base);
1504     qemu_fdt_add_subnode(ms->fdt, nodename);
1505     qemu_fdt_setprop_string(ms->fdt, nodename, "device_type", "memory");
1506     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg", 2, base, 2, size);
1507     qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
1508     qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
1509 
1510     if (secure_tag_sysmem) {
1511         create_tag_ram(secure_tag_sysmem, base, size, "mach-virt.secure-tag");
1512     }
1513 
1514     g_free(nodename);
1515 }
1516 
1517 static void *machvirt_dtb(const struct arm_boot_info *binfo, int *fdt_size)
1518 {
1519     const VirtMachineState *board = container_of(binfo, VirtMachineState,
1520                                                  bootinfo);
1521     MachineState *ms = MACHINE(board);
1522 
1523 
1524     *fdt_size = board->fdt_size;
1525     return ms->fdt;
1526 }
1527 
1528 static void virt_build_smbios(VirtMachineState *vms)
1529 {
1530     MachineClass *mc = MACHINE_GET_CLASS(vms);
1531     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
1532     uint8_t *smbios_tables, *smbios_anchor;
1533     size_t smbios_tables_len, smbios_anchor_len;
1534     const char *product = "QEMU Virtual Machine";
1535 
1536     if (kvm_enabled()) {
1537         product = "KVM Virtual Machine";
1538     }
1539 
1540     smbios_set_defaults("QEMU", product,
1541                         vmc->smbios_old_sys_ver ? "1.0" : mc->name, false,
1542                         true, SMBIOS_ENTRY_POINT_30);
1543 
1544     smbios_get_tables(MACHINE(vms), NULL, 0,
1545                       &smbios_tables, &smbios_tables_len,
1546                       &smbios_anchor, &smbios_anchor_len,
1547                       &error_fatal);
1548 
1549     if (smbios_anchor) {
1550         fw_cfg_add_file(vms->fw_cfg, "etc/smbios/smbios-tables",
1551                         smbios_tables, smbios_tables_len);
1552         fw_cfg_add_file(vms->fw_cfg, "etc/smbios/smbios-anchor",
1553                         smbios_anchor, smbios_anchor_len);
1554     }
1555 }
1556 
1557 static
1558 void virt_machine_done(Notifier *notifier, void *data)
1559 {
1560     VirtMachineState *vms = container_of(notifier, VirtMachineState,
1561                                          machine_done);
1562     MachineState *ms = MACHINE(vms);
1563     ARMCPU *cpu = ARM_CPU(first_cpu);
1564     struct arm_boot_info *info = &vms->bootinfo;
1565     AddressSpace *as = arm_boot_address_space(cpu, info);
1566 
1567     /*
1568      * If the user provided a dtb, we assume the dynamic sysbus nodes
1569      * already are integrated there. This corresponds to a use case where
1570      * the dynamic sysbus nodes are complex and their generation is not yet
1571      * supported. In that case the user can take charge of the guest dt
1572      * while qemu takes charge of the qom stuff.
1573      */
1574     if (info->dtb_filename == NULL) {
1575         platform_bus_add_all_fdt_nodes(ms->fdt, "/intc",
1576                                        vms->memmap[VIRT_PLATFORM_BUS].base,
1577                                        vms->memmap[VIRT_PLATFORM_BUS].size,
1578                                        vms->irqmap[VIRT_PLATFORM_BUS]);
1579     }
1580     if (arm_load_dtb(info->dtb_start, info, info->dtb_limit, as, ms) < 0) {
1581         exit(1);
1582     }
1583 
1584     fw_cfg_add_extra_pci_roots(vms->bus, vms->fw_cfg);
1585 
1586     virt_acpi_setup(vms);
1587     virt_build_smbios(vms);
1588 }
1589 
1590 static uint64_t virt_cpu_mp_affinity(VirtMachineState *vms, int idx)
1591 {
1592     uint8_t clustersz = ARM_DEFAULT_CPUS_PER_CLUSTER;
1593     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
1594 
1595     if (!vmc->disallow_affinity_adjustment) {
1596         /* Adjust MPIDR like 64-bit KVM hosts, which incorporate the
1597          * GIC's target-list limitations. 32-bit KVM hosts currently
1598          * always create clusters of 4 CPUs, but that is expected to
1599          * change when they gain support for gicv3. When KVM is enabled
1600          * it will override the changes we make here, therefore our
1601          * purposes are to make TCG consistent (with 64-bit KVM hosts)
1602          * and to improve SGI efficiency.
1603          */
1604         if (vms->gic_version == VIRT_GIC_VERSION_3) {
1605             clustersz = GICV3_TARGETLIST_BITS;
1606         } else {
1607             clustersz = GIC_TARGETLIST_BITS;
1608         }
1609     }
1610     return arm_cpu_mp_affinity(idx, clustersz);
1611 }
1612 
1613 static void virt_set_memmap(VirtMachineState *vms)
1614 {
1615     MachineState *ms = MACHINE(vms);
1616     hwaddr base, device_memory_base, device_memory_size;
1617     int i;
1618 
1619     vms->memmap = extended_memmap;
1620 
1621     for (i = 0; i < ARRAY_SIZE(base_memmap); i++) {
1622         vms->memmap[i] = base_memmap[i];
1623     }
1624 
1625     if (ms->ram_slots > ACPI_MAX_RAM_SLOTS) {
1626         error_report("unsupported number of memory slots: %"PRIu64,
1627                      ms->ram_slots);
1628         exit(EXIT_FAILURE);
1629     }
1630 
1631     /*
1632      * We compute the base of the high IO region depending on the
1633      * amount of initial and device memory. The device memory start/size
1634      * is aligned on 1GiB. We never put the high IO region below 256GiB
1635      * so that if maxram_size is < 255GiB we keep the legacy memory map.
1636      * The device region size assumes 1GiB page max alignment per slot.
1637      */
1638     device_memory_base =
1639         ROUND_UP(vms->memmap[VIRT_MEM].base + ms->ram_size, GiB);
1640     device_memory_size = ms->maxram_size - ms->ram_size + ms->ram_slots * GiB;
1641 
1642     /* Base address of the high IO region */
1643     base = device_memory_base + ROUND_UP(device_memory_size, GiB);
1644     if (base < device_memory_base) {
1645         error_report("maxmem/slots too huge");
1646         exit(EXIT_FAILURE);
1647     }
1648     if (base < vms->memmap[VIRT_MEM].base + LEGACY_RAMLIMIT_BYTES) {
1649         base = vms->memmap[VIRT_MEM].base + LEGACY_RAMLIMIT_BYTES;
1650     }
1651 
1652     for (i = VIRT_LOWMEMMAP_LAST; i < ARRAY_SIZE(extended_memmap); i++) {
1653         hwaddr size = extended_memmap[i].size;
1654 
1655         base = ROUND_UP(base, size);
1656         vms->memmap[i].base = base;
1657         vms->memmap[i].size = size;
1658         base += size;
1659     }
1660     vms->highest_gpa = base - 1;
1661     if (device_memory_size > 0) {
1662         ms->device_memory = g_malloc0(sizeof(*ms->device_memory));
1663         ms->device_memory->base = device_memory_base;
1664         memory_region_init(&ms->device_memory->mr, OBJECT(vms),
1665                            "device-memory", device_memory_size);
1666     }
1667 }
1668 
1669 /*
1670  * finalize_gic_version - Determines the final gic_version
1671  * according to the gic-version property
1672  *
1673  * Default GIC type is v2
1674  */
1675 static void finalize_gic_version(VirtMachineState *vms)
1676 {
1677     unsigned int max_cpus = MACHINE(vms)->smp.max_cpus;
1678 
1679     if (kvm_enabled()) {
1680         int probe_bitmap;
1681 
1682         if (!kvm_irqchip_in_kernel()) {
1683             switch (vms->gic_version) {
1684             case VIRT_GIC_VERSION_HOST:
1685                 warn_report(
1686                     "gic-version=host not relevant with kernel-irqchip=off "
1687                      "as only userspace GICv2 is supported. Using v2 ...");
1688                 return;
1689             case VIRT_GIC_VERSION_MAX:
1690             case VIRT_GIC_VERSION_NOSEL:
1691                 vms->gic_version = VIRT_GIC_VERSION_2;
1692                 return;
1693             case VIRT_GIC_VERSION_2:
1694                 return;
1695             case VIRT_GIC_VERSION_3:
1696                 error_report(
1697                     "gic-version=3 is not supported with kernel-irqchip=off");
1698                 exit(1);
1699             }
1700         }
1701 
1702         probe_bitmap = kvm_arm_vgic_probe();
1703         if (!probe_bitmap) {
1704             error_report("Unable to determine GIC version supported by host");
1705             exit(1);
1706         }
1707 
1708         switch (vms->gic_version) {
1709         case VIRT_GIC_VERSION_HOST:
1710         case VIRT_GIC_VERSION_MAX:
1711             if (probe_bitmap & KVM_ARM_VGIC_V3) {
1712                 vms->gic_version = VIRT_GIC_VERSION_3;
1713             } else {
1714                 vms->gic_version = VIRT_GIC_VERSION_2;
1715             }
1716             return;
1717         case VIRT_GIC_VERSION_NOSEL:
1718             if ((probe_bitmap & KVM_ARM_VGIC_V2) && max_cpus <= GIC_NCPU) {
1719                 vms->gic_version = VIRT_GIC_VERSION_2;
1720             } else if (probe_bitmap & KVM_ARM_VGIC_V3) {
1721                 /*
1722                  * in case the host does not support v2 in-kernel emulation or
1723                  * the end-user requested more than 8 VCPUs we now default
1724                  * to v3. In any case defaulting to v2 would be broken.
1725                  */
1726                 vms->gic_version = VIRT_GIC_VERSION_3;
1727             } else if (max_cpus > GIC_NCPU) {
1728                 error_report("host only supports in-kernel GICv2 emulation "
1729                              "but more than 8 vcpus are requested");
1730                 exit(1);
1731             }
1732             break;
1733         case VIRT_GIC_VERSION_2:
1734         case VIRT_GIC_VERSION_3:
1735             break;
1736         }
1737 
1738         /* Check chosen version is effectively supported by the host */
1739         if (vms->gic_version == VIRT_GIC_VERSION_2 &&
1740             !(probe_bitmap & KVM_ARM_VGIC_V2)) {
1741             error_report("host does not support in-kernel GICv2 emulation");
1742             exit(1);
1743         } else if (vms->gic_version == VIRT_GIC_VERSION_3 &&
1744                    !(probe_bitmap & KVM_ARM_VGIC_V3)) {
1745             error_report("host does not support in-kernel GICv3 emulation");
1746             exit(1);
1747         }
1748         return;
1749     }
1750 
1751     /* TCG mode */
1752     switch (vms->gic_version) {
1753     case VIRT_GIC_VERSION_NOSEL:
1754         vms->gic_version = VIRT_GIC_VERSION_2;
1755         break;
1756     case VIRT_GIC_VERSION_MAX:
1757         vms->gic_version = VIRT_GIC_VERSION_3;
1758         break;
1759     case VIRT_GIC_VERSION_HOST:
1760         error_report("gic-version=host requires KVM");
1761         exit(1);
1762     case VIRT_GIC_VERSION_2:
1763     case VIRT_GIC_VERSION_3:
1764         break;
1765     }
1766 }
1767 
1768 /*
1769  * virt_cpu_post_init() must be called after the CPUs have
1770  * been realized and the GIC has been created.
1771  */
1772 static void virt_cpu_post_init(VirtMachineState *vms, MemoryRegion *sysmem)
1773 {
1774     int max_cpus = MACHINE(vms)->smp.max_cpus;
1775     bool aarch64, pmu, steal_time;
1776     CPUState *cpu;
1777 
1778     aarch64 = object_property_get_bool(OBJECT(first_cpu), "aarch64", NULL);
1779     pmu = object_property_get_bool(OBJECT(first_cpu), "pmu", NULL);
1780     steal_time = object_property_get_bool(OBJECT(first_cpu),
1781                                           "kvm-steal-time", NULL);
1782 
1783     if (kvm_enabled()) {
1784         hwaddr pvtime_reg_base = vms->memmap[VIRT_PVTIME].base;
1785         hwaddr pvtime_reg_size = vms->memmap[VIRT_PVTIME].size;
1786 
1787         if (steal_time) {
1788             MemoryRegion *pvtime = g_new(MemoryRegion, 1);
1789             hwaddr pvtime_size = max_cpus * PVTIME_SIZE_PER_CPU;
1790 
1791             /* The memory region size must be a multiple of host page size. */
1792             pvtime_size = REAL_HOST_PAGE_ALIGN(pvtime_size);
1793 
1794             if (pvtime_size > pvtime_reg_size) {
1795                 error_report("pvtime requires a %" HWADDR_PRId
1796                              " byte memory region for %d CPUs,"
1797                              " but only %" HWADDR_PRId " has been reserved",
1798                              pvtime_size, max_cpus, pvtime_reg_size);
1799                 exit(1);
1800             }
1801 
1802             memory_region_init_ram(pvtime, NULL, "pvtime", pvtime_size, NULL);
1803             memory_region_add_subregion(sysmem, pvtime_reg_base, pvtime);
1804         }
1805 
1806         CPU_FOREACH(cpu) {
1807             if (pmu) {
1808                 assert(arm_feature(&ARM_CPU(cpu)->env, ARM_FEATURE_PMU));
1809                 if (kvm_irqchip_in_kernel()) {
1810                     kvm_arm_pmu_set_irq(cpu, PPI(VIRTUAL_PMU_IRQ));
1811                 }
1812                 kvm_arm_pmu_init(cpu);
1813             }
1814             if (steal_time) {
1815                 kvm_arm_pvtime_init(cpu, pvtime_reg_base +
1816                                          cpu->cpu_index * PVTIME_SIZE_PER_CPU);
1817             }
1818         }
1819     } else {
1820         if (aarch64 && vms->highmem) {
1821             int requested_pa_size = 64 - clz64(vms->highest_gpa);
1822             int pamax = arm_pamax(ARM_CPU(first_cpu));
1823 
1824             if (pamax < requested_pa_size) {
1825                 error_report("VCPU supports less PA bits (%d) than "
1826                              "requested by the memory map (%d)",
1827                              pamax, requested_pa_size);
1828                 exit(1);
1829             }
1830         }
1831     }
1832 }
1833 
1834 static void machvirt_init(MachineState *machine)
1835 {
1836     VirtMachineState *vms = VIRT_MACHINE(machine);
1837     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(machine);
1838     MachineClass *mc = MACHINE_GET_CLASS(machine);
1839     const CPUArchIdList *possible_cpus;
1840     MemoryRegion *sysmem = get_system_memory();
1841     MemoryRegion *secure_sysmem = NULL;
1842     MemoryRegion *tag_sysmem = NULL;
1843     MemoryRegion *secure_tag_sysmem = NULL;
1844     int n, virt_max_cpus;
1845     bool firmware_loaded;
1846     bool aarch64 = true;
1847     bool has_ged = !vmc->no_ged;
1848     unsigned int smp_cpus = machine->smp.cpus;
1849     unsigned int max_cpus = machine->smp.max_cpus;
1850 
1851     /*
1852      * In accelerated mode, the memory map is computed earlier in kvm_type()
1853      * to create a VM with the right number of IPA bits.
1854      */
1855     if (!vms->memmap) {
1856         virt_set_memmap(vms);
1857     }
1858 
1859     /* We can probe only here because during property set
1860      * KVM is not available yet
1861      */
1862     finalize_gic_version(vms);
1863 
1864     if (!cpu_type_valid(machine->cpu_type)) {
1865         error_report("mach-virt: CPU type %s not supported", machine->cpu_type);
1866         exit(1);
1867     }
1868 
1869     if (vms->secure) {
1870         /*
1871          * The Secure view of the world is the same as the NonSecure,
1872          * but with a few extra devices. Create it as a container region
1873          * containing the system memory at low priority; any secure-only
1874          * devices go in at higher priority and take precedence.
1875          */
1876         secure_sysmem = g_new(MemoryRegion, 1);
1877         memory_region_init(secure_sysmem, OBJECT(machine), "secure-memory",
1878                            UINT64_MAX);
1879         memory_region_add_subregion_overlap(secure_sysmem, 0, sysmem, -1);
1880     }
1881 
1882     firmware_loaded = virt_firmware_init(vms, sysmem,
1883                                          secure_sysmem ?: sysmem);
1884 
1885     /* If we have an EL3 boot ROM then the assumption is that it will
1886      * implement PSCI itself, so disable QEMU's internal implementation
1887      * so it doesn't get in the way. Instead of starting secondary
1888      * CPUs in PSCI powerdown state we will start them all running and
1889      * let the boot ROM sort them out.
1890      * The usual case is that we do use QEMU's PSCI implementation;
1891      * if the guest has EL2 then we will use SMC as the conduit,
1892      * and otherwise we will use HVC (for backwards compatibility and
1893      * because if we're using KVM then we must use HVC).
1894      */
1895     if (vms->secure && firmware_loaded) {
1896         vms->psci_conduit = QEMU_PSCI_CONDUIT_DISABLED;
1897     } else if (vms->virt) {
1898         vms->psci_conduit = QEMU_PSCI_CONDUIT_SMC;
1899     } else {
1900         vms->psci_conduit = QEMU_PSCI_CONDUIT_HVC;
1901     }
1902 
1903     /* The maximum number of CPUs depends on the GIC version, or on how
1904      * many redistributors we can fit into the memory map.
1905      */
1906     if (vms->gic_version == VIRT_GIC_VERSION_3) {
1907         virt_max_cpus =
1908             vms->memmap[VIRT_GIC_REDIST].size / GICV3_REDIST_SIZE;
1909         virt_max_cpus +=
1910             vms->memmap[VIRT_HIGH_GIC_REDIST2].size / GICV3_REDIST_SIZE;
1911     } else {
1912         virt_max_cpus = GIC_NCPU;
1913     }
1914 
1915     if (max_cpus > virt_max_cpus) {
1916         error_report("Number of SMP CPUs requested (%d) exceeds max CPUs "
1917                      "supported by machine 'mach-virt' (%d)",
1918                      max_cpus, virt_max_cpus);
1919         exit(1);
1920     }
1921 
1922     if (vms->virt && kvm_enabled()) {
1923         error_report("mach-virt: KVM does not support providing "
1924                      "Virtualization extensions to the guest CPU");
1925         exit(1);
1926     }
1927 
1928     if (vms->mte && kvm_enabled()) {
1929         error_report("mach-virt: KVM does not support providing "
1930                      "MTE to the guest CPU");
1931         exit(1);
1932     }
1933 
1934     create_fdt(vms);
1935 
1936     possible_cpus = mc->possible_cpu_arch_ids(machine);
1937     assert(possible_cpus->len == max_cpus);
1938     for (n = 0; n < possible_cpus->len; n++) {
1939         Object *cpuobj;
1940         CPUState *cs;
1941 
1942         if (n >= smp_cpus) {
1943             break;
1944         }
1945 
1946         cpuobj = object_new(possible_cpus->cpus[n].type);
1947         object_property_set_int(cpuobj, "mp-affinity",
1948                                 possible_cpus->cpus[n].arch_id, NULL);
1949 
1950         cs = CPU(cpuobj);
1951         cs->cpu_index = n;
1952 
1953         numa_cpu_pre_plug(&possible_cpus->cpus[cs->cpu_index], DEVICE(cpuobj),
1954                           &error_fatal);
1955 
1956         aarch64 &= object_property_get_bool(cpuobj, "aarch64", NULL);
1957 
1958         if (!vms->secure) {
1959             object_property_set_bool(cpuobj, "has_el3", false, NULL);
1960         }
1961 
1962         if (!vms->virt && object_property_find(cpuobj, "has_el2")) {
1963             object_property_set_bool(cpuobj, "has_el2", false, NULL);
1964         }
1965 
1966         if (vms->psci_conduit != QEMU_PSCI_CONDUIT_DISABLED) {
1967             object_property_set_int(cpuobj, "psci-conduit", vms->psci_conduit,
1968                                     NULL);
1969 
1970             /* Secondary CPUs start in PSCI powered-down state */
1971             if (n > 0) {
1972                 object_property_set_bool(cpuobj, "start-powered-off", true,
1973                                          NULL);
1974             }
1975         }
1976 
1977         if (vmc->kvm_no_adjvtime &&
1978             object_property_find(cpuobj, "kvm-no-adjvtime")) {
1979             object_property_set_bool(cpuobj, "kvm-no-adjvtime", true, NULL);
1980         }
1981 
1982         if (vmc->no_kvm_steal_time &&
1983             object_property_find(cpuobj, "kvm-steal-time")) {
1984             object_property_set_bool(cpuobj, "kvm-steal-time", false, NULL);
1985         }
1986 
1987         if (vmc->no_pmu && object_property_find(cpuobj, "pmu")) {
1988             object_property_set_bool(cpuobj, "pmu", false, NULL);
1989         }
1990 
1991         if (object_property_find(cpuobj, "reset-cbar")) {
1992             object_property_set_int(cpuobj, "reset-cbar",
1993                                     vms->memmap[VIRT_CPUPERIPHS].base,
1994                                     &error_abort);
1995         }
1996 
1997         object_property_set_link(cpuobj, "memory", OBJECT(sysmem),
1998                                  &error_abort);
1999         if (vms->secure) {
2000             object_property_set_link(cpuobj, "secure-memory",
2001                                      OBJECT(secure_sysmem), &error_abort);
2002         }
2003 
2004         if (vms->mte) {
2005             /* Create the memory region only once, but link to all cpus. */
2006             if (!tag_sysmem) {
2007                 /*
2008                  * The property exists only if MemTag is supported.
2009                  * If it is, we must allocate the ram to back that up.
2010                  */
2011                 if (!object_property_find(cpuobj, "tag-memory")) {
2012                     error_report("MTE requested, but not supported "
2013                                  "by the guest CPU");
2014                     exit(1);
2015                 }
2016 
2017                 tag_sysmem = g_new(MemoryRegion, 1);
2018                 memory_region_init(tag_sysmem, OBJECT(machine),
2019                                    "tag-memory", UINT64_MAX / 32);
2020 
2021                 if (vms->secure) {
2022                     secure_tag_sysmem = g_new(MemoryRegion, 1);
2023                     memory_region_init(secure_tag_sysmem, OBJECT(machine),
2024                                        "secure-tag-memory", UINT64_MAX / 32);
2025 
2026                     /* As with ram, secure-tag takes precedence over tag.  */
2027                     memory_region_add_subregion_overlap(secure_tag_sysmem, 0,
2028                                                         tag_sysmem, -1);
2029                 }
2030             }
2031 
2032             object_property_set_link(cpuobj, "tag-memory", OBJECT(tag_sysmem),
2033                                      &error_abort);
2034             if (vms->secure) {
2035                 object_property_set_link(cpuobj, "secure-tag-memory",
2036                                          OBJECT(secure_tag_sysmem),
2037                                          &error_abort);
2038             }
2039         }
2040 
2041         qdev_realize(DEVICE(cpuobj), NULL, &error_fatal);
2042         object_unref(cpuobj);
2043     }
2044     fdt_add_timer_nodes(vms);
2045     fdt_add_cpu_nodes(vms);
2046 
2047     memory_region_add_subregion(sysmem, vms->memmap[VIRT_MEM].base,
2048                                 machine->ram);
2049     if (machine->device_memory) {
2050         memory_region_add_subregion(sysmem, machine->device_memory->base,
2051                                     &machine->device_memory->mr);
2052     }
2053 
2054     virt_flash_fdt(vms, sysmem, secure_sysmem ?: sysmem);
2055 
2056     create_gic(vms, sysmem);
2057 
2058     virt_cpu_post_init(vms, sysmem);
2059 
2060     fdt_add_pmu_nodes(vms);
2061 
2062     create_uart(vms, VIRT_UART, sysmem, serial_hd(0));
2063 
2064     if (vms->secure) {
2065         create_secure_ram(vms, secure_sysmem, secure_tag_sysmem);
2066         create_uart(vms, VIRT_SECURE_UART, secure_sysmem, serial_hd(1));
2067     }
2068 
2069     if (tag_sysmem) {
2070         create_tag_ram(tag_sysmem, vms->memmap[VIRT_MEM].base,
2071                        machine->ram_size, "mach-virt.tag");
2072     }
2073 
2074     vms->highmem_ecam &= vms->highmem && (!firmware_loaded || aarch64);
2075 
2076     create_rtc(vms);
2077 
2078     create_pcie(vms);
2079 
2080     if (has_ged && aarch64 && firmware_loaded && virt_is_acpi_enabled(vms)) {
2081         vms->acpi_dev = create_acpi_ged(vms);
2082     } else {
2083         create_gpio_devices(vms, VIRT_GPIO, sysmem);
2084     }
2085 
2086     if (vms->secure && !vmc->no_secure_gpio) {
2087         create_gpio_devices(vms, VIRT_SECURE_GPIO, secure_sysmem);
2088     }
2089 
2090      /* connect powerdown request */
2091      vms->powerdown_notifier.notify = virt_powerdown_req;
2092      qemu_register_powerdown_notifier(&vms->powerdown_notifier);
2093 
2094     /* Create mmio transports, so the user can create virtio backends
2095      * (which will be automatically plugged in to the transports). If
2096      * no backend is created the transport will just sit harmlessly idle.
2097      */
2098     create_virtio_devices(vms);
2099 
2100     vms->fw_cfg = create_fw_cfg(vms, &address_space_memory);
2101     rom_set_fw(vms->fw_cfg);
2102 
2103     create_platform_bus(vms);
2104 
2105     if (machine->nvdimms_state->is_enabled) {
2106         const struct AcpiGenericAddress arm_virt_nvdimm_acpi_dsmio = {
2107             .space_id = AML_AS_SYSTEM_MEMORY,
2108             .address = vms->memmap[VIRT_NVDIMM_ACPI].base,
2109             .bit_width = NVDIMM_ACPI_IO_LEN << 3
2110         };
2111 
2112         nvdimm_init_acpi_state(machine->nvdimms_state, sysmem,
2113                                arm_virt_nvdimm_acpi_dsmio,
2114                                vms->fw_cfg, OBJECT(vms));
2115     }
2116 
2117     vms->bootinfo.ram_size = machine->ram_size;
2118     vms->bootinfo.nb_cpus = smp_cpus;
2119     vms->bootinfo.board_id = -1;
2120     vms->bootinfo.loader_start = vms->memmap[VIRT_MEM].base;
2121     vms->bootinfo.get_dtb = machvirt_dtb;
2122     vms->bootinfo.skip_dtb_autoload = true;
2123     vms->bootinfo.firmware_loaded = firmware_loaded;
2124     arm_load_kernel(ARM_CPU(first_cpu), machine, &vms->bootinfo);
2125 
2126     vms->machine_done.notify = virt_machine_done;
2127     qemu_add_machine_init_done_notifier(&vms->machine_done);
2128 }
2129 
2130 static bool virt_get_secure(Object *obj, Error **errp)
2131 {
2132     VirtMachineState *vms = VIRT_MACHINE(obj);
2133 
2134     return vms->secure;
2135 }
2136 
2137 static void virt_set_secure(Object *obj, bool value, Error **errp)
2138 {
2139     VirtMachineState *vms = VIRT_MACHINE(obj);
2140 
2141     vms->secure = value;
2142 }
2143 
2144 static bool virt_get_virt(Object *obj, Error **errp)
2145 {
2146     VirtMachineState *vms = VIRT_MACHINE(obj);
2147 
2148     return vms->virt;
2149 }
2150 
2151 static void virt_set_virt(Object *obj, bool value, Error **errp)
2152 {
2153     VirtMachineState *vms = VIRT_MACHINE(obj);
2154 
2155     vms->virt = value;
2156 }
2157 
2158 static bool virt_get_highmem(Object *obj, Error **errp)
2159 {
2160     VirtMachineState *vms = VIRT_MACHINE(obj);
2161 
2162     return vms->highmem;
2163 }
2164 
2165 static void virt_set_highmem(Object *obj, bool value, Error **errp)
2166 {
2167     VirtMachineState *vms = VIRT_MACHINE(obj);
2168 
2169     vms->highmem = value;
2170 }
2171 
2172 static bool virt_get_its(Object *obj, Error **errp)
2173 {
2174     VirtMachineState *vms = VIRT_MACHINE(obj);
2175 
2176     return vms->its;
2177 }
2178 
2179 static void virt_set_its(Object *obj, bool value, Error **errp)
2180 {
2181     VirtMachineState *vms = VIRT_MACHINE(obj);
2182 
2183     vms->its = value;
2184 }
2185 
2186 static char *virt_get_oem_id(Object *obj, Error **errp)
2187 {
2188     VirtMachineState *vms = VIRT_MACHINE(obj);
2189 
2190     return g_strdup(vms->oem_id);
2191 }
2192 
2193 static void virt_set_oem_id(Object *obj, const char *value, Error **errp)
2194 {
2195     VirtMachineState *vms = VIRT_MACHINE(obj);
2196     size_t len = strlen(value);
2197 
2198     if (len > 6) {
2199         error_setg(errp,
2200                    "User specified oem-id value is bigger than 6 bytes in size");
2201         return;
2202     }
2203 
2204     strncpy(vms->oem_id, value, 6);
2205 }
2206 
2207 static char *virt_get_oem_table_id(Object *obj, Error **errp)
2208 {
2209     VirtMachineState *vms = VIRT_MACHINE(obj);
2210 
2211     return g_strdup(vms->oem_table_id);
2212 }
2213 
2214 static void virt_set_oem_table_id(Object *obj, const char *value,
2215                                   Error **errp)
2216 {
2217     VirtMachineState *vms = VIRT_MACHINE(obj);
2218     size_t len = strlen(value);
2219 
2220     if (len > 8) {
2221         error_setg(errp,
2222                    "User specified oem-table-id value is bigger than 8 bytes in size");
2223         return;
2224     }
2225     strncpy(vms->oem_table_id, value, 8);
2226 }
2227 
2228 
2229 bool virt_is_acpi_enabled(VirtMachineState *vms)
2230 {
2231     if (vms->acpi == ON_OFF_AUTO_OFF) {
2232         return false;
2233     }
2234     return true;
2235 }
2236 
2237 static void virt_get_acpi(Object *obj, Visitor *v, const char *name,
2238                           void *opaque, Error **errp)
2239 {
2240     VirtMachineState *vms = VIRT_MACHINE(obj);
2241     OnOffAuto acpi = vms->acpi;
2242 
2243     visit_type_OnOffAuto(v, name, &acpi, errp);
2244 }
2245 
2246 static void virt_set_acpi(Object *obj, Visitor *v, const char *name,
2247                           void *opaque, Error **errp)
2248 {
2249     VirtMachineState *vms = VIRT_MACHINE(obj);
2250 
2251     visit_type_OnOffAuto(v, name, &vms->acpi, errp);
2252 }
2253 
2254 static bool virt_get_ras(Object *obj, Error **errp)
2255 {
2256     VirtMachineState *vms = VIRT_MACHINE(obj);
2257 
2258     return vms->ras;
2259 }
2260 
2261 static void virt_set_ras(Object *obj, bool value, Error **errp)
2262 {
2263     VirtMachineState *vms = VIRT_MACHINE(obj);
2264 
2265     vms->ras = value;
2266 }
2267 
2268 static bool virt_get_mte(Object *obj, Error **errp)
2269 {
2270     VirtMachineState *vms = VIRT_MACHINE(obj);
2271 
2272     return vms->mte;
2273 }
2274 
2275 static void virt_set_mte(Object *obj, bool value, Error **errp)
2276 {
2277     VirtMachineState *vms = VIRT_MACHINE(obj);
2278 
2279     vms->mte = value;
2280 }
2281 
2282 static char *virt_get_gic_version(Object *obj, Error **errp)
2283 {
2284     VirtMachineState *vms = VIRT_MACHINE(obj);
2285     const char *val = vms->gic_version == VIRT_GIC_VERSION_3 ? "3" : "2";
2286 
2287     return g_strdup(val);
2288 }
2289 
2290 static void virt_set_gic_version(Object *obj, const char *value, Error **errp)
2291 {
2292     VirtMachineState *vms = VIRT_MACHINE(obj);
2293 
2294     if (!strcmp(value, "3")) {
2295         vms->gic_version = VIRT_GIC_VERSION_3;
2296     } else if (!strcmp(value, "2")) {
2297         vms->gic_version = VIRT_GIC_VERSION_2;
2298     } else if (!strcmp(value, "host")) {
2299         vms->gic_version = VIRT_GIC_VERSION_HOST; /* Will probe later */
2300     } else if (!strcmp(value, "max")) {
2301         vms->gic_version = VIRT_GIC_VERSION_MAX; /* Will probe later */
2302     } else {
2303         error_setg(errp, "Invalid gic-version value");
2304         error_append_hint(errp, "Valid values are 3, 2, host, max.\n");
2305     }
2306 }
2307 
2308 static char *virt_get_iommu(Object *obj, Error **errp)
2309 {
2310     VirtMachineState *vms = VIRT_MACHINE(obj);
2311 
2312     switch (vms->iommu) {
2313     case VIRT_IOMMU_NONE:
2314         return g_strdup("none");
2315     case VIRT_IOMMU_SMMUV3:
2316         return g_strdup("smmuv3");
2317     default:
2318         g_assert_not_reached();
2319     }
2320 }
2321 
2322 static void virt_set_iommu(Object *obj, const char *value, Error **errp)
2323 {
2324     VirtMachineState *vms = VIRT_MACHINE(obj);
2325 
2326     if (!strcmp(value, "smmuv3")) {
2327         vms->iommu = VIRT_IOMMU_SMMUV3;
2328     } else if (!strcmp(value, "none")) {
2329         vms->iommu = VIRT_IOMMU_NONE;
2330     } else {
2331         error_setg(errp, "Invalid iommu value");
2332         error_append_hint(errp, "Valid values are none, smmuv3.\n");
2333     }
2334 }
2335 
2336 static bool virt_get_default_bus_bypass_iommu(Object *obj, Error **errp)
2337 {
2338     VirtMachineState *vms = VIRT_MACHINE(obj);
2339 
2340     return vms->default_bus_bypass_iommu;
2341 }
2342 
2343 static void virt_set_default_bus_bypass_iommu(Object *obj, bool value,
2344                                               Error **errp)
2345 {
2346     VirtMachineState *vms = VIRT_MACHINE(obj);
2347 
2348     vms->default_bus_bypass_iommu = value;
2349 }
2350 
2351 static CpuInstanceProperties
2352 virt_cpu_index_to_props(MachineState *ms, unsigned cpu_index)
2353 {
2354     MachineClass *mc = MACHINE_GET_CLASS(ms);
2355     const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(ms);
2356 
2357     assert(cpu_index < possible_cpus->len);
2358     return possible_cpus->cpus[cpu_index].props;
2359 }
2360 
2361 static int64_t virt_get_default_cpu_node_id(const MachineState *ms, int idx)
2362 {
2363     return idx % ms->numa_state->num_nodes;
2364 }
2365 
2366 static const CPUArchIdList *virt_possible_cpu_arch_ids(MachineState *ms)
2367 {
2368     int n;
2369     unsigned int max_cpus = ms->smp.max_cpus;
2370     VirtMachineState *vms = VIRT_MACHINE(ms);
2371 
2372     if (ms->possible_cpus) {
2373         assert(ms->possible_cpus->len == max_cpus);
2374         return ms->possible_cpus;
2375     }
2376 
2377     ms->possible_cpus = g_malloc0(sizeof(CPUArchIdList) +
2378                                   sizeof(CPUArchId) * max_cpus);
2379     ms->possible_cpus->len = max_cpus;
2380     for (n = 0; n < ms->possible_cpus->len; n++) {
2381         ms->possible_cpus->cpus[n].type = ms->cpu_type;
2382         ms->possible_cpus->cpus[n].arch_id =
2383             virt_cpu_mp_affinity(vms, n);
2384         ms->possible_cpus->cpus[n].props.has_thread_id = true;
2385         ms->possible_cpus->cpus[n].props.thread_id = n;
2386     }
2387     return ms->possible_cpus;
2388 }
2389 
2390 static void virt_memory_pre_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
2391                                  Error **errp)
2392 {
2393     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2394     const MachineState *ms = MACHINE(hotplug_dev);
2395     const bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2396 
2397     if (!vms->acpi_dev) {
2398         error_setg(errp,
2399                    "memory hotplug is not enabled: missing acpi-ged device");
2400         return;
2401     }
2402 
2403     if (vms->mte) {
2404         error_setg(errp, "memory hotplug is not enabled: MTE is enabled");
2405         return;
2406     }
2407 
2408     if (is_nvdimm && !ms->nvdimms_state->is_enabled) {
2409         error_setg(errp, "nvdimm is not enabled: add 'nvdimm=on' to '-M'");
2410         return;
2411     }
2412 
2413     pc_dimm_pre_plug(PC_DIMM(dev), MACHINE(hotplug_dev), NULL, errp);
2414 }
2415 
2416 static void virt_memory_plug(HotplugHandler *hotplug_dev,
2417                              DeviceState *dev, Error **errp)
2418 {
2419     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2420     MachineState *ms = MACHINE(hotplug_dev);
2421     bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2422 
2423     pc_dimm_plug(PC_DIMM(dev), MACHINE(vms));
2424 
2425     if (is_nvdimm) {
2426         nvdimm_plug(ms->nvdimms_state);
2427     }
2428 
2429     hotplug_handler_plug(HOTPLUG_HANDLER(vms->acpi_dev),
2430                          dev, &error_abort);
2431 }
2432 
2433 static void virt_machine_device_pre_plug_cb(HotplugHandler *hotplug_dev,
2434                                             DeviceState *dev, Error **errp)
2435 {
2436     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2437 
2438     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2439         virt_memory_pre_plug(hotplug_dev, dev, errp);
2440     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2441         hwaddr db_start = 0, db_end = 0;
2442         char *resv_prop_str;
2443 
2444         switch (vms->msi_controller) {
2445         case VIRT_MSI_CTRL_NONE:
2446             return;
2447         case VIRT_MSI_CTRL_ITS:
2448             /* GITS_TRANSLATER page */
2449             db_start = base_memmap[VIRT_GIC_ITS].base + 0x10000;
2450             db_end = base_memmap[VIRT_GIC_ITS].base +
2451                      base_memmap[VIRT_GIC_ITS].size - 1;
2452             break;
2453         case VIRT_MSI_CTRL_GICV2M:
2454             /* MSI_SETSPI_NS page */
2455             db_start = base_memmap[VIRT_GIC_V2M].base;
2456             db_end = db_start + base_memmap[VIRT_GIC_V2M].size - 1;
2457             break;
2458         }
2459         resv_prop_str = g_strdup_printf("0x%"PRIx64":0x%"PRIx64":%u",
2460                                         db_start, db_end,
2461                                         VIRTIO_IOMMU_RESV_MEM_T_MSI);
2462 
2463         qdev_prop_set_uint32(dev, "len-reserved-regions", 1);
2464         qdev_prop_set_string(dev, "reserved-regions[0]", resv_prop_str);
2465         g_free(resv_prop_str);
2466     }
2467 }
2468 
2469 static void virt_machine_device_plug_cb(HotplugHandler *hotplug_dev,
2470                                         DeviceState *dev, Error **errp)
2471 {
2472     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2473 
2474     if (vms->platform_bus_dev) {
2475         MachineClass *mc = MACHINE_GET_CLASS(vms);
2476 
2477         if (device_is_dynamic_sysbus(mc, dev)) {
2478             platform_bus_link_device(PLATFORM_BUS_DEVICE(vms->platform_bus_dev),
2479                                      SYS_BUS_DEVICE(dev));
2480         }
2481     }
2482     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2483         virt_memory_plug(hotplug_dev, dev, errp);
2484     }
2485     if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2486         PCIDevice *pdev = PCI_DEVICE(dev);
2487 
2488         vms->iommu = VIRT_IOMMU_VIRTIO;
2489         vms->virtio_iommu_bdf = pci_get_bdf(pdev);
2490         create_virtio_iommu_dt_bindings(vms);
2491     }
2492 }
2493 
2494 static void virt_dimm_unplug_request(HotplugHandler *hotplug_dev,
2495                                      DeviceState *dev, Error **errp)
2496 {
2497     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2498     Error *local_err = NULL;
2499 
2500     if (!vms->acpi_dev) {
2501         error_setg(&local_err,
2502                    "memory hotplug is not enabled: missing acpi-ged device");
2503         goto out;
2504     }
2505 
2506     if (object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM)) {
2507         error_setg(&local_err,
2508                    "nvdimm device hot unplug is not supported yet.");
2509         goto out;
2510     }
2511 
2512     hotplug_handler_unplug_request(HOTPLUG_HANDLER(vms->acpi_dev), dev,
2513                                    &local_err);
2514 out:
2515     error_propagate(errp, local_err);
2516 }
2517 
2518 static void virt_dimm_unplug(HotplugHandler *hotplug_dev,
2519                              DeviceState *dev, Error **errp)
2520 {
2521     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2522     Error *local_err = NULL;
2523 
2524     hotplug_handler_unplug(HOTPLUG_HANDLER(vms->acpi_dev), dev, &local_err);
2525     if (local_err) {
2526         goto out;
2527     }
2528 
2529     pc_dimm_unplug(PC_DIMM(dev), MACHINE(vms));
2530     qdev_unrealize(dev);
2531 
2532 out:
2533     error_propagate(errp, local_err);
2534 }
2535 
2536 static void virt_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
2537                                           DeviceState *dev, Error **errp)
2538 {
2539     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2540         virt_dimm_unplug_request(hotplug_dev, dev, errp);
2541     } else {
2542         error_setg(errp, "device unplug request for unsupported device"
2543                    " type: %s", object_get_typename(OBJECT(dev)));
2544     }
2545 }
2546 
2547 static void virt_machine_device_unplug_cb(HotplugHandler *hotplug_dev,
2548                                           DeviceState *dev, Error **errp)
2549 {
2550     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2551         virt_dimm_unplug(hotplug_dev, dev, errp);
2552     } else {
2553         error_setg(errp, "virt: device unplug for unsupported device"
2554                    " type: %s", object_get_typename(OBJECT(dev)));
2555     }
2556 }
2557 
2558 static HotplugHandler *virt_machine_get_hotplug_handler(MachineState *machine,
2559                                                         DeviceState *dev)
2560 {
2561     MachineClass *mc = MACHINE_GET_CLASS(machine);
2562 
2563     if (device_is_dynamic_sysbus(mc, dev) ||
2564        (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM))) {
2565         return HOTPLUG_HANDLER(machine);
2566     }
2567     if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2568         VirtMachineState *vms = VIRT_MACHINE(machine);
2569 
2570         if (!vms->bootinfo.firmware_loaded || !virt_is_acpi_enabled(vms)) {
2571             return HOTPLUG_HANDLER(machine);
2572         }
2573     }
2574     return NULL;
2575 }
2576 
2577 /*
2578  * for arm64 kvm_type [7-0] encodes the requested number of bits
2579  * in the IPA address space
2580  */
2581 static int virt_kvm_type(MachineState *ms, const char *type_str)
2582 {
2583     VirtMachineState *vms = VIRT_MACHINE(ms);
2584     int max_vm_pa_size, requested_pa_size;
2585     bool fixed_ipa;
2586 
2587     max_vm_pa_size = kvm_arm_get_max_vm_ipa_size(ms, &fixed_ipa);
2588 
2589     /* we freeze the memory map to compute the highest gpa */
2590     virt_set_memmap(vms);
2591 
2592     requested_pa_size = 64 - clz64(vms->highest_gpa);
2593 
2594     /*
2595      * KVM requires the IPA size to be at least 32 bits.
2596      */
2597     if (requested_pa_size < 32) {
2598         requested_pa_size = 32;
2599     }
2600 
2601     if (requested_pa_size > max_vm_pa_size) {
2602         error_report("-m and ,maxmem option values "
2603                      "require an IPA range (%d bits) larger than "
2604                      "the one supported by the host (%d bits)",
2605                      requested_pa_size, max_vm_pa_size);
2606         exit(1);
2607     }
2608     /*
2609      * We return the requested PA log size, unless KVM only supports
2610      * the implicit legacy 40b IPA setting, in which case the kvm_type
2611      * must be 0.
2612      */
2613     return fixed_ipa ? 0 : requested_pa_size;
2614 }
2615 
2616 static void virt_machine_class_init(ObjectClass *oc, void *data)
2617 {
2618     MachineClass *mc = MACHINE_CLASS(oc);
2619     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
2620 
2621     mc->init = machvirt_init;
2622     /* Start with max_cpus set to 512, which is the maximum supported by KVM.
2623      * The value may be reduced later when we have more information about the
2624      * configuration of the particular instance.
2625      */
2626     mc->max_cpus = 512;
2627     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_CALXEDA_XGMAC);
2628     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_AMD_XGBE);
2629     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_RAMFB_DEVICE);
2630     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_PLATFORM);
2631 #ifdef CONFIG_TPM
2632     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_TPM_TIS_SYSBUS);
2633 #endif
2634     mc->block_default_type = IF_VIRTIO;
2635     mc->no_cdrom = 1;
2636     mc->pci_allow_0_address = true;
2637     /* We know we will never create a pre-ARMv7 CPU which needs 1K pages */
2638     mc->minimum_page_bits = 12;
2639     mc->possible_cpu_arch_ids = virt_possible_cpu_arch_ids;
2640     mc->cpu_index_to_instance_props = virt_cpu_index_to_props;
2641     mc->default_cpu_type = ARM_CPU_TYPE_NAME("cortex-a15");
2642     mc->get_default_cpu_node_id = virt_get_default_cpu_node_id;
2643     mc->kvm_type = virt_kvm_type;
2644     assert(!mc->get_hotplug_handler);
2645     mc->get_hotplug_handler = virt_machine_get_hotplug_handler;
2646     hc->pre_plug = virt_machine_device_pre_plug_cb;
2647     hc->plug = virt_machine_device_plug_cb;
2648     hc->unplug_request = virt_machine_device_unplug_request_cb;
2649     hc->unplug = virt_machine_device_unplug_cb;
2650     mc->nvdimm_supported = true;
2651     mc->auto_enable_numa_with_memhp = true;
2652     mc->auto_enable_numa_with_memdev = true;
2653     mc->default_ram_id = "mach-virt.ram";
2654 
2655     object_class_property_add(oc, "acpi", "OnOffAuto",
2656         virt_get_acpi, virt_set_acpi,
2657         NULL, NULL);
2658     object_class_property_set_description(oc, "acpi",
2659         "Enable ACPI");
2660     object_class_property_add_bool(oc, "secure", virt_get_secure,
2661                                    virt_set_secure);
2662     object_class_property_set_description(oc, "secure",
2663                                                 "Set on/off to enable/disable the ARM "
2664                                                 "Security Extensions (TrustZone)");
2665 
2666     object_class_property_add_bool(oc, "virtualization", virt_get_virt,
2667                                    virt_set_virt);
2668     object_class_property_set_description(oc, "virtualization",
2669                                           "Set on/off to enable/disable emulating a "
2670                                           "guest CPU which implements the ARM "
2671                                           "Virtualization Extensions");
2672 
2673     object_class_property_add_bool(oc, "highmem", virt_get_highmem,
2674                                    virt_set_highmem);
2675     object_class_property_set_description(oc, "highmem",
2676                                           "Set on/off to enable/disable using "
2677                                           "physical address space above 32 bits");
2678 
2679     object_class_property_add_str(oc, "gic-version", virt_get_gic_version,
2680                                   virt_set_gic_version);
2681     object_class_property_set_description(oc, "gic-version",
2682                                           "Set GIC version. "
2683                                           "Valid values are 2, 3, host and max");
2684 
2685     object_class_property_add_str(oc, "iommu", virt_get_iommu, virt_set_iommu);
2686     object_class_property_set_description(oc, "iommu",
2687                                           "Set the IOMMU type. "
2688                                           "Valid values are none and smmuv3");
2689 
2690     object_class_property_add_bool(oc, "default_bus_bypass_iommu",
2691                                    virt_get_default_bus_bypass_iommu,
2692                                    virt_set_default_bus_bypass_iommu);
2693     object_class_property_set_description(oc, "default_bus_bypass_iommu",
2694                                           "Set on/off to enable/disable "
2695                                           "bypass_iommu for default root bus");
2696 
2697     object_class_property_add_bool(oc, "ras", virt_get_ras,
2698                                    virt_set_ras);
2699     object_class_property_set_description(oc, "ras",
2700                                           "Set on/off to enable/disable reporting host memory errors "
2701                                           "to a KVM guest using ACPI and guest external abort exceptions");
2702 
2703     object_class_property_add_bool(oc, "mte", virt_get_mte, virt_set_mte);
2704     object_class_property_set_description(oc, "mte",
2705                                           "Set on/off to enable/disable emulating a "
2706                                           "guest CPU which implements the ARM "
2707                                           "Memory Tagging Extension");
2708 
2709     object_class_property_add_bool(oc, "its", virt_get_its,
2710                                    virt_set_its);
2711     object_class_property_set_description(oc, "its",
2712                                           "Set on/off to enable/disable "
2713                                           "ITS instantiation");
2714 
2715     object_class_property_add_str(oc, "x-oem-id",
2716                                   virt_get_oem_id,
2717                                   virt_set_oem_id);
2718     object_class_property_set_description(oc, "x-oem-id",
2719                                           "Override the default value of field OEMID "
2720                                           "in ACPI table header."
2721                                           "The string may be up to 6 bytes in size");
2722 
2723 
2724     object_class_property_add_str(oc, "x-oem-table-id",
2725                                   virt_get_oem_table_id,
2726                                   virt_set_oem_table_id);
2727     object_class_property_set_description(oc, "x-oem-table-id",
2728                                           "Override the default value of field OEM Table ID "
2729                                           "in ACPI table header."
2730                                           "The string may be up to 8 bytes in size");
2731 
2732 }
2733 
2734 static void virt_instance_init(Object *obj)
2735 {
2736     VirtMachineState *vms = VIRT_MACHINE(obj);
2737     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
2738 
2739     /* EL3 is disabled by default on virt: this makes us consistent
2740      * between KVM and TCG for this board, and it also allows us to
2741      * boot UEFI blobs which assume no TrustZone support.
2742      */
2743     vms->secure = false;
2744 
2745     /* EL2 is also disabled by default, for similar reasons */
2746     vms->virt = false;
2747 
2748     /* High memory is enabled by default */
2749     vms->highmem = true;
2750     vms->gic_version = VIRT_GIC_VERSION_NOSEL;
2751 
2752     vms->highmem_ecam = !vmc->no_highmem_ecam;
2753 
2754     if (vmc->no_its) {
2755         vms->its = false;
2756     } else {
2757         /* Default allows ITS instantiation */
2758         vms->its = true;
2759 
2760         if (vmc->no_tcg_its) {
2761             vms->tcg_its = false;
2762         } else {
2763             vms->tcg_its = true;
2764         }
2765     }
2766 
2767     /* Default disallows iommu instantiation */
2768     vms->iommu = VIRT_IOMMU_NONE;
2769 
2770     /* The default root bus is attached to iommu by default */
2771     vms->default_bus_bypass_iommu = false;
2772 
2773     /* Default disallows RAS instantiation */
2774     vms->ras = false;
2775 
2776     /* MTE is disabled by default.  */
2777     vms->mte = false;
2778 
2779     vms->irqmap = a15irqmap;
2780 
2781     virt_flash_create(vms);
2782 
2783     vms->oem_id = g_strndup(ACPI_BUILD_APPNAME6, 6);
2784     vms->oem_table_id = g_strndup(ACPI_BUILD_APPNAME8, 8);
2785 }
2786 
2787 static const TypeInfo virt_machine_info = {
2788     .name          = TYPE_VIRT_MACHINE,
2789     .parent        = TYPE_MACHINE,
2790     .abstract      = true,
2791     .instance_size = sizeof(VirtMachineState),
2792     .class_size    = sizeof(VirtMachineClass),
2793     .class_init    = virt_machine_class_init,
2794     .instance_init = virt_instance_init,
2795     .interfaces = (InterfaceInfo[]) {
2796          { TYPE_HOTPLUG_HANDLER },
2797          { }
2798     },
2799 };
2800 
2801 static void machvirt_machine_init(void)
2802 {
2803     type_register_static(&virt_machine_info);
2804 }
2805 type_init(machvirt_machine_init);
2806 
2807 static void virt_machine_6_2_options(MachineClass *mc)
2808 {
2809 }
2810 DEFINE_VIRT_MACHINE_AS_LATEST(6, 2)
2811 
2812 static void virt_machine_6_1_options(MachineClass *mc)
2813 {
2814     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2815 
2816     virt_machine_6_2_options(mc);
2817     compat_props_add(mc->compat_props, hw_compat_6_1, hw_compat_6_1_len);
2818 
2819     /* qemu ITS was introduced with 6.2 */
2820     vmc->no_tcg_its = true;
2821 }
2822 DEFINE_VIRT_MACHINE(6, 1)
2823 
2824 static void virt_machine_6_0_options(MachineClass *mc)
2825 {
2826     virt_machine_6_1_options(mc);
2827     compat_props_add(mc->compat_props, hw_compat_6_0, hw_compat_6_0_len);
2828 }
2829 DEFINE_VIRT_MACHINE(6, 0)
2830 
2831 static void virt_machine_5_2_options(MachineClass *mc)
2832 {
2833     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2834 
2835     virt_machine_6_0_options(mc);
2836     compat_props_add(mc->compat_props, hw_compat_5_2, hw_compat_5_2_len);
2837     vmc->no_secure_gpio = true;
2838 }
2839 DEFINE_VIRT_MACHINE(5, 2)
2840 
2841 static void virt_machine_5_1_options(MachineClass *mc)
2842 {
2843     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2844 
2845     virt_machine_5_2_options(mc);
2846     compat_props_add(mc->compat_props, hw_compat_5_1, hw_compat_5_1_len);
2847     vmc->no_kvm_steal_time = true;
2848 }
2849 DEFINE_VIRT_MACHINE(5, 1)
2850 
2851 static void virt_machine_5_0_options(MachineClass *mc)
2852 {
2853     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2854 
2855     virt_machine_5_1_options(mc);
2856     compat_props_add(mc->compat_props, hw_compat_5_0, hw_compat_5_0_len);
2857     mc->numa_mem_supported = true;
2858     vmc->acpi_expose_flash = true;
2859     mc->auto_enable_numa_with_memdev = false;
2860 }
2861 DEFINE_VIRT_MACHINE(5, 0)
2862 
2863 static void virt_machine_4_2_options(MachineClass *mc)
2864 {
2865     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2866 
2867     virt_machine_5_0_options(mc);
2868     compat_props_add(mc->compat_props, hw_compat_4_2, hw_compat_4_2_len);
2869     vmc->kvm_no_adjvtime = true;
2870 }
2871 DEFINE_VIRT_MACHINE(4, 2)
2872 
2873 static void virt_machine_4_1_options(MachineClass *mc)
2874 {
2875     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2876 
2877     virt_machine_4_2_options(mc);
2878     compat_props_add(mc->compat_props, hw_compat_4_1, hw_compat_4_1_len);
2879     vmc->no_ged = true;
2880     mc->auto_enable_numa_with_memhp = false;
2881 }
2882 DEFINE_VIRT_MACHINE(4, 1)
2883 
2884 static void virt_machine_4_0_options(MachineClass *mc)
2885 {
2886     virt_machine_4_1_options(mc);
2887     compat_props_add(mc->compat_props, hw_compat_4_0, hw_compat_4_0_len);
2888 }
2889 DEFINE_VIRT_MACHINE(4, 0)
2890 
2891 static void virt_machine_3_1_options(MachineClass *mc)
2892 {
2893     virt_machine_4_0_options(mc);
2894     compat_props_add(mc->compat_props, hw_compat_3_1, hw_compat_3_1_len);
2895 }
2896 DEFINE_VIRT_MACHINE(3, 1)
2897 
2898 static void virt_machine_3_0_options(MachineClass *mc)
2899 {
2900     virt_machine_3_1_options(mc);
2901     compat_props_add(mc->compat_props, hw_compat_3_0, hw_compat_3_0_len);
2902 }
2903 DEFINE_VIRT_MACHINE(3, 0)
2904 
2905 static void virt_machine_2_12_options(MachineClass *mc)
2906 {
2907     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2908 
2909     virt_machine_3_0_options(mc);
2910     compat_props_add(mc->compat_props, hw_compat_2_12, hw_compat_2_12_len);
2911     vmc->no_highmem_ecam = true;
2912     mc->max_cpus = 255;
2913 }
2914 DEFINE_VIRT_MACHINE(2, 12)
2915 
2916 static void virt_machine_2_11_options(MachineClass *mc)
2917 {
2918     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2919 
2920     virt_machine_2_12_options(mc);
2921     compat_props_add(mc->compat_props, hw_compat_2_11, hw_compat_2_11_len);
2922     vmc->smbios_old_sys_ver = true;
2923 }
2924 DEFINE_VIRT_MACHINE(2, 11)
2925 
2926 static void virt_machine_2_10_options(MachineClass *mc)
2927 {
2928     virt_machine_2_11_options(mc);
2929     compat_props_add(mc->compat_props, hw_compat_2_10, hw_compat_2_10_len);
2930     /* before 2.11 we never faulted accesses to bad addresses */
2931     mc->ignore_memory_transaction_failures = true;
2932 }
2933 DEFINE_VIRT_MACHINE(2, 10)
2934 
2935 static void virt_machine_2_9_options(MachineClass *mc)
2936 {
2937     virt_machine_2_10_options(mc);
2938     compat_props_add(mc->compat_props, hw_compat_2_9, hw_compat_2_9_len);
2939 }
2940 DEFINE_VIRT_MACHINE(2, 9)
2941 
2942 static void virt_machine_2_8_options(MachineClass *mc)
2943 {
2944     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2945 
2946     virt_machine_2_9_options(mc);
2947     compat_props_add(mc->compat_props, hw_compat_2_8, hw_compat_2_8_len);
2948     /* For 2.8 and earlier we falsely claimed in the DT that
2949      * our timers were edge-triggered, not level-triggered.
2950      */
2951     vmc->claim_edge_triggered_timers = true;
2952 }
2953 DEFINE_VIRT_MACHINE(2, 8)
2954 
2955 static void virt_machine_2_7_options(MachineClass *mc)
2956 {
2957     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2958 
2959     virt_machine_2_8_options(mc);
2960     compat_props_add(mc->compat_props, hw_compat_2_7, hw_compat_2_7_len);
2961     /* ITS was introduced with 2.8 */
2962     vmc->no_its = true;
2963     /* Stick with 1K pages for migration compatibility */
2964     mc->minimum_page_bits = 0;
2965 }
2966 DEFINE_VIRT_MACHINE(2, 7)
2967 
2968 static void virt_machine_2_6_options(MachineClass *mc)
2969 {
2970     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
2971 
2972     virt_machine_2_7_options(mc);
2973     compat_props_add(mc->compat_props, hw_compat_2_6, hw_compat_2_6_len);
2974     vmc->disallow_affinity_adjustment = true;
2975     /* Disable PMU for 2.6 as PMU support was first introduced in 2.7 */
2976     vmc->no_pmu = true;
2977 }
2978 DEFINE_VIRT_MACHINE(2, 6)
2979