xref: /qemu/hw/arm/virt.c (revision db65ac5e258e75e9aec45626bf1071626094e057)
1 /*
2  * ARM mach-virt emulation
3  *
4  * Copyright (c) 2013 Linaro Limited
5  *
6  * This program is free software; you can redistribute it and/or modify it
7  * under the terms and conditions of the GNU General Public License,
8  * version 2 or later, as published by the Free Software Foundation.
9  *
10  * This program is distributed in the hope it will be useful, but WITHOUT
11  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
12  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
13  * more details.
14  *
15  * You should have received a copy of the GNU General Public License along with
16  * this program.  If not, see <http://www.gnu.org/licenses/>.
17  *
18  * Emulate a virtual board which works by passing Linux all the information
19  * it needs about what devices are present via the device tree.
20  * There are some restrictions about what we can do here:
21  *  + we can only present devices whose Linux drivers will work based
22  *    purely on the device tree with no platform data at all
23  *  + we want to present a very stripped-down minimalist platform,
24  *    both because this reduces the security attack surface from the guest
25  *    and also because it reduces our exposure to being broken when
26  *    the kernel updates its device tree bindings and requires further
27  *    information in a device binding that we aren't providing.
28  * This is essentially the same approach kvmtool uses.
29  */
30 
31 #include "qemu/osdep.h"
32 #include "qemu/datadir.h"
33 #include "qemu/units.h"
34 #include "qemu/option.h"
35 #include "monitor/qdev.h"
36 #include "hw/sysbus.h"
37 #include "hw/arm/boot.h"
38 #include "hw/arm/primecell.h"
39 #include "hw/arm/virt.h"
40 #include "hw/block/flash.h"
41 #include "hw/vfio/vfio-calxeda-xgmac.h"
42 #include "hw/vfio/vfio-amd-xgbe.h"
43 #include "hw/display/ramfb.h"
44 #include "net/net.h"
45 #include "system/device_tree.h"
46 #include "system/numa.h"
47 #include "system/runstate.h"
48 #include "system/tpm.h"
49 #include "system/tcg.h"
50 #include "system/kvm.h"
51 #include "system/hvf.h"
52 #include "system/qtest.h"
53 #include "hw/loader.h"
54 #include "qapi/error.h"
55 #include "qemu/bitops.h"
56 #include "qemu/error-report.h"
57 #include "qemu/module.h"
58 #include "hw/pci-host/gpex.h"
59 #include "hw/virtio/virtio-pci.h"
60 #include "hw/core/sysbus-fdt.h"
61 #include "hw/platform-bus.h"
62 #include "hw/qdev-properties.h"
63 #include "hw/arm/fdt.h"
64 #include "hw/intc/arm_gic.h"
65 #include "hw/intc/arm_gicv3_common.h"
66 #include "hw/intc/arm_gicv3_its_common.h"
67 #include "hw/irq.h"
68 #include "kvm_arm.h"
69 #include "hvf_arm.h"
70 #include "hw/firmware/smbios.h"
71 #include "qapi/visitor.h"
72 #include "qapi/qapi-visit-common.h"
73 #include "qapi/qmp/qlist.h"
74 #include "standard-headers/linux/input.h"
75 #include "hw/arm/smmuv3.h"
76 #include "hw/acpi/acpi.h"
77 #include "target/arm/cpu-qom.h"
78 #include "target/arm/internals.h"
79 #include "target/arm/multiprocessing.h"
80 #include "target/arm/gtimer.h"
81 #include "hw/mem/pc-dimm.h"
82 #include "hw/mem/nvdimm.h"
83 #include "hw/acpi/generic_event_device.h"
84 #include "hw/virtio/virtio-md-pci.h"
85 #include "hw/virtio/virtio-iommu.h"
86 #include "hw/char/pl011.h"
87 #include "qemu/guest-random.h"
88 
89 static GlobalProperty arm_virt_compat[] = {
90     { TYPE_VIRTIO_IOMMU_PCI, "aw-bits", "48" },
91 };
92 static const size_t arm_virt_compat_len = G_N_ELEMENTS(arm_virt_compat);
93 
94 /*
95  * This cannot be called from the virt_machine_class_init() because
96  * TYPE_VIRT_MACHINE is abstract and mc->compat_props g_ptr_array_new()
97  * only is called on virt non abstract class init.
98  */
99 static void arm_virt_compat_set(MachineClass *mc)
100 {
101     compat_props_add(mc->compat_props, arm_virt_compat,
102                      arm_virt_compat_len);
103 }
104 
105 #define DEFINE_VIRT_MACHINE_IMPL(latest, ...) \
106     static void MACHINE_VER_SYM(class_init, virt, __VA_ARGS__)( \
107         ObjectClass *oc, \
108         void *data) \
109     { \
110         MachineClass *mc = MACHINE_CLASS(oc); \
111         arm_virt_compat_set(mc); \
112         MACHINE_VER_SYM(options, virt, __VA_ARGS__)(mc); \
113         mc->desc = "QEMU " MACHINE_VER_STR(__VA_ARGS__) " ARM Virtual Machine"; \
114         MACHINE_VER_DEPRECATION(__VA_ARGS__); \
115         if (latest) { \
116             mc->alias = "virt"; \
117         } \
118     } \
119     static const TypeInfo MACHINE_VER_SYM(info, virt, __VA_ARGS__) = \
120     { \
121         .name = MACHINE_VER_TYPE_NAME("virt", __VA_ARGS__), \
122         .parent = TYPE_VIRT_MACHINE, \
123         .class_init = MACHINE_VER_SYM(class_init, virt, __VA_ARGS__), \
124     }; \
125     static void MACHINE_VER_SYM(register, virt, __VA_ARGS__)(void) \
126     { \
127         MACHINE_VER_DELETION(__VA_ARGS__); \
128         type_register_static(&MACHINE_VER_SYM(info, virt, __VA_ARGS__)); \
129     } \
130     type_init(MACHINE_VER_SYM(register, virt, __VA_ARGS__));
131 
132 #define DEFINE_VIRT_MACHINE_AS_LATEST(major, minor) \
133     DEFINE_VIRT_MACHINE_IMPL(true, major, minor)
134 #define DEFINE_VIRT_MACHINE(major, minor) \
135     DEFINE_VIRT_MACHINE_IMPL(false, major, minor)
136 
137 
138 /* Number of external interrupt lines to configure the GIC with */
139 #define NUM_IRQS 256
140 
141 #define PLATFORM_BUS_NUM_IRQS 64
142 
143 /* Legacy RAM limit in GB (< version 4.0) */
144 #define LEGACY_RAMLIMIT_GB 255
145 #define LEGACY_RAMLIMIT_BYTES (LEGACY_RAMLIMIT_GB * GiB)
146 
147 /* Addresses and sizes of our components.
148  * 0..128MB is space for a flash device so we can run bootrom code such as UEFI.
149  * 128MB..256MB is used for miscellaneous device I/O.
150  * 256MB..1GB is reserved for possible future PCI support (ie where the
151  * PCI memory window will go if we add a PCI host controller).
152  * 1GB and up is RAM (which may happily spill over into the
153  * high memory region beyond 4GB).
154  * This represents a compromise between how much RAM can be given to
155  * a 32 bit VM and leaving space for expansion and in particular for PCI.
156  * Note that devices should generally be placed at multiples of 0x10000,
157  * to accommodate guests using 64K pages.
158  */
159 static const MemMapEntry base_memmap[] = {
160     /* Space up to 0x8000000 is reserved for a boot ROM */
161     [VIRT_FLASH] =              {          0, 0x08000000 },
162     [VIRT_CPUPERIPHS] =         { 0x08000000, 0x00020000 },
163     /* GIC distributor and CPU interfaces sit inside the CPU peripheral space */
164     [VIRT_GIC_DIST] =           { 0x08000000, 0x00010000 },
165     [VIRT_GIC_CPU] =            { 0x08010000, 0x00010000 },
166     [VIRT_GIC_V2M] =            { 0x08020000, 0x00001000 },
167     [VIRT_GIC_HYP] =            { 0x08030000, 0x00010000 },
168     [VIRT_GIC_VCPU] =           { 0x08040000, 0x00010000 },
169     /* The space in between here is reserved for GICv3 CPU/vCPU/HYP */
170     [VIRT_GIC_ITS] =            { 0x08080000, 0x00020000 },
171     /* This redistributor space allows up to 2*64kB*123 CPUs */
172     [VIRT_GIC_REDIST] =         { 0x080A0000, 0x00F60000 },
173     [VIRT_UART0] =              { 0x09000000, 0x00001000 },
174     [VIRT_RTC] =                { 0x09010000, 0x00001000 },
175     [VIRT_FW_CFG] =             { 0x09020000, 0x00000018 },
176     [VIRT_GPIO] =               { 0x09030000, 0x00001000 },
177     [VIRT_UART1] =              { 0x09040000, 0x00001000 },
178     [VIRT_SMMU] =               { 0x09050000, 0x00020000 },
179     [VIRT_PCDIMM_ACPI] =        { 0x09070000, MEMORY_HOTPLUG_IO_LEN },
180     [VIRT_ACPI_GED] =           { 0x09080000, ACPI_GED_EVT_SEL_LEN },
181     [VIRT_NVDIMM_ACPI] =        { 0x09090000, NVDIMM_ACPI_IO_LEN},
182     [VIRT_PVTIME] =             { 0x090a0000, 0x00010000 },
183     [VIRT_SECURE_GPIO] =        { 0x090b0000, 0x00001000 },
184     [VIRT_MMIO] =               { 0x0a000000, 0x00000200 },
185     /* ...repeating for a total of NUM_VIRTIO_TRANSPORTS, each of that size */
186     [VIRT_PLATFORM_BUS] =       { 0x0c000000, 0x02000000 },
187     [VIRT_SECURE_MEM] =         { 0x0e000000, 0x01000000 },
188     [VIRT_PCIE_MMIO] =          { 0x10000000, 0x2eff0000 },
189     [VIRT_PCIE_PIO] =           { 0x3eff0000, 0x00010000 },
190     [VIRT_PCIE_ECAM] =          { 0x3f000000, 0x01000000 },
191     /* Actual RAM size depends on initial RAM and device memory settings */
192     [VIRT_MEM] =                { GiB, LEGACY_RAMLIMIT_BYTES },
193 };
194 
195 /*
196  * Highmem IO Regions: This memory map is floating, located after the RAM.
197  * Each MemMapEntry base (GPA) will be dynamically computed, depending on the
198  * top of the RAM, so that its base get the same alignment as the size,
199  * ie. a 512GiB entry will be aligned on a 512GiB boundary. If there is
200  * less than 256GiB of RAM, the floating area starts at the 256GiB mark.
201  * Note the extended_memmap is sized so that it eventually also includes the
202  * base_memmap entries (VIRT_HIGH_GIC_REDIST2 index is greater than the last
203  * index of base_memmap).
204  *
205  * The memory map for these Highmem IO Regions can be in legacy or compact
206  * layout, depending on 'compact-highmem' property. With legacy layout, the
207  * PA space for one specific region is always reserved, even if the region
208  * has been disabled or doesn't fit into the PA space. However, the PA space
209  * for the region won't be reserved in these circumstances with compact layout.
210  */
211 static MemMapEntry extended_memmap[] = {
212     /* Additional 64 MB redist region (can contain up to 512 redistributors) */
213     [VIRT_HIGH_GIC_REDIST2] =   { 0x0, 64 * MiB },
214     [VIRT_HIGH_PCIE_ECAM] =     { 0x0, 256 * MiB },
215     /* Second PCIe window */
216     [VIRT_HIGH_PCIE_MMIO] =     { 0x0, 512 * GiB },
217 };
218 
219 static const int a15irqmap[] = {
220     [VIRT_UART0] = 1,
221     [VIRT_RTC] = 2,
222     [VIRT_PCIE] = 3, /* ... to 6 */
223     [VIRT_GPIO] = 7,
224     [VIRT_UART1] = 8,
225     [VIRT_ACPI_GED] = 9,
226     [VIRT_MMIO] = 16, /* ...to 16 + NUM_VIRTIO_TRANSPORTS - 1 */
227     [VIRT_GIC_V2M] = 48, /* ...to 48 + NUM_GICV2M_SPIS - 1 */
228     [VIRT_SMMU] = 74,    /* ...to 74 + NUM_SMMU_IRQS - 1 */
229     [VIRT_PLATFORM_BUS] = 112, /* ...to 112 + PLATFORM_BUS_NUM_IRQS -1 */
230 };
231 
232 static void create_randomness(MachineState *ms, const char *node)
233 {
234     struct {
235         uint64_t kaslr;
236         uint8_t rng[32];
237     } seed;
238 
239     if (qemu_guest_getrandom(&seed, sizeof(seed), NULL)) {
240         return;
241     }
242     qemu_fdt_setprop_u64(ms->fdt, node, "kaslr-seed", seed.kaslr);
243     qemu_fdt_setprop(ms->fdt, node, "rng-seed", seed.rng, sizeof(seed.rng));
244 }
245 
246 /*
247  * The CPU object always exposes the NS EL2 virt timer IRQ line,
248  * but we don't want to advertise it to the guest in the dtb or ACPI
249  * table unless it's really going to do something.
250  */
251 static bool ns_el2_virt_timer_present(void)
252 {
253     ARMCPU *cpu = ARM_CPU(qemu_get_cpu(0));
254     CPUARMState *env = &cpu->env;
255 
256     return arm_feature(env, ARM_FEATURE_AARCH64) &&
257         arm_feature(env, ARM_FEATURE_EL2) && cpu_isar_feature(aa64_vh, cpu);
258 }
259 
260 static void create_fdt(VirtMachineState *vms)
261 {
262     MachineState *ms = MACHINE(vms);
263     int nb_numa_nodes = ms->numa_state->num_nodes;
264     void *fdt = create_device_tree(&vms->fdt_size);
265 
266     if (!fdt) {
267         error_report("create_device_tree() failed");
268         exit(1);
269     }
270 
271     ms->fdt = fdt;
272 
273     /* Header */
274     qemu_fdt_setprop_string(fdt, "/", "compatible", "linux,dummy-virt");
275     qemu_fdt_setprop_cell(fdt, "/", "#address-cells", 0x2);
276     qemu_fdt_setprop_cell(fdt, "/", "#size-cells", 0x2);
277     qemu_fdt_setprop_string(fdt, "/", "model", "linux,dummy-virt");
278 
279     /*
280      * For QEMU, all DMA is coherent. Advertising this in the root node
281      * has two benefits:
282      *
283      * - It avoids potential bugs where we forget to mark a DMA
284      *   capable device as being dma-coherent
285      * - It avoids spurious warnings from the Linux kernel about
286      *   devices which can't do DMA at all
287      */
288     qemu_fdt_setprop(fdt, "/", "dma-coherent", NULL, 0);
289 
290     /* /chosen must exist for load_dtb to fill in necessary properties later */
291     qemu_fdt_add_subnode(fdt, "/chosen");
292     if (vms->dtb_randomness) {
293         create_randomness(ms, "/chosen");
294     }
295 
296     if (vms->secure) {
297         qemu_fdt_add_subnode(fdt, "/secure-chosen");
298         if (vms->dtb_randomness) {
299             create_randomness(ms, "/secure-chosen");
300         }
301     }
302 
303     qemu_fdt_add_subnode(fdt, "/aliases");
304 
305     /* Clock node, for the benefit of the UART. The kernel device tree
306      * binding documentation claims the PL011 node clock properties are
307      * optional but in practice if you omit them the kernel refuses to
308      * probe for the device.
309      */
310     vms->clock_phandle = qemu_fdt_alloc_phandle(fdt);
311     qemu_fdt_add_subnode(fdt, "/apb-pclk");
312     qemu_fdt_setprop_string(fdt, "/apb-pclk", "compatible", "fixed-clock");
313     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "#clock-cells", 0x0);
314     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "clock-frequency", 24000000);
315     qemu_fdt_setprop_string(fdt, "/apb-pclk", "clock-output-names",
316                                 "clk24mhz");
317     qemu_fdt_setprop_cell(fdt, "/apb-pclk", "phandle", vms->clock_phandle);
318 
319     if (nb_numa_nodes > 0 && ms->numa_state->have_numa_distance) {
320         int size = nb_numa_nodes * nb_numa_nodes * 3 * sizeof(uint32_t);
321         uint32_t *matrix = g_malloc0(size);
322         int idx, i, j;
323 
324         for (i = 0; i < nb_numa_nodes; i++) {
325             for (j = 0; j < nb_numa_nodes; j++) {
326                 idx = (i * nb_numa_nodes + j) * 3;
327                 matrix[idx + 0] = cpu_to_be32(i);
328                 matrix[idx + 1] = cpu_to_be32(j);
329                 matrix[idx + 2] =
330                     cpu_to_be32(ms->numa_state->nodes[i].distance[j]);
331             }
332         }
333 
334         qemu_fdt_add_subnode(fdt, "/distance-map");
335         qemu_fdt_setprop_string(fdt, "/distance-map", "compatible",
336                                 "numa-distance-map-v1");
337         qemu_fdt_setprop(fdt, "/distance-map", "distance-matrix",
338                          matrix, size);
339         g_free(matrix);
340     }
341 }
342 
343 static void fdt_add_timer_nodes(const VirtMachineState *vms)
344 {
345     /* On real hardware these interrupts are level-triggered.
346      * On KVM they were edge-triggered before host kernel version 4.4,
347      * and level-triggered afterwards.
348      * On emulated QEMU they are level-triggered.
349      *
350      * Getting the DTB info about them wrong is awkward for some
351      * guest kernels:
352      *  pre-4.8 ignore the DT and leave the interrupt configured
353      *   with whatever the GIC reset value (or the bootloader) left it at
354      *  4.8 before rc6 honour the incorrect data by programming it back
355      *   into the GIC, causing problems
356      *  4.8rc6 and later ignore the DT and always write "level triggered"
357      *   into the GIC
358      *
359      * For backwards-compatibility, virt-2.8 and earlier will continue
360      * to say these are edge-triggered, but later machines will report
361      * the correct information.
362      */
363     ARMCPU *armcpu;
364     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
365     uint32_t irqflags = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
366     MachineState *ms = MACHINE(vms);
367 
368     if (vmc->claim_edge_triggered_timers) {
369         irqflags = GIC_FDT_IRQ_FLAGS_EDGE_LO_HI;
370     }
371 
372     if (vms->gic_version == VIRT_GIC_VERSION_2) {
373         irqflags = deposit32(irqflags, GIC_FDT_IRQ_PPI_CPU_START,
374                              GIC_FDT_IRQ_PPI_CPU_WIDTH,
375                              (1 << MACHINE(vms)->smp.cpus) - 1);
376     }
377 
378     qemu_fdt_add_subnode(ms->fdt, "/timer");
379 
380     armcpu = ARM_CPU(qemu_get_cpu(0));
381     if (arm_feature(&armcpu->env, ARM_FEATURE_V8)) {
382         const char compat[] = "arm,armv8-timer\0arm,armv7-timer";
383         qemu_fdt_setprop(ms->fdt, "/timer", "compatible",
384                          compat, sizeof(compat));
385     } else {
386         qemu_fdt_setprop_string(ms->fdt, "/timer", "compatible",
387                                 "arm,armv7-timer");
388     }
389     qemu_fdt_setprop(ms->fdt, "/timer", "always-on", NULL, 0);
390     if (vms->ns_el2_virt_timer_irq) {
391         qemu_fdt_setprop_cells(ms->fdt, "/timer", "interrupts",
392                                GIC_FDT_IRQ_TYPE_PPI,
393                                INTID_TO_PPI(ARCH_TIMER_S_EL1_IRQ), irqflags,
394                                GIC_FDT_IRQ_TYPE_PPI,
395                                INTID_TO_PPI(ARCH_TIMER_NS_EL1_IRQ), irqflags,
396                                GIC_FDT_IRQ_TYPE_PPI,
397                                INTID_TO_PPI(ARCH_TIMER_VIRT_IRQ), irqflags,
398                                GIC_FDT_IRQ_TYPE_PPI,
399                                INTID_TO_PPI(ARCH_TIMER_NS_EL2_IRQ), irqflags,
400                                GIC_FDT_IRQ_TYPE_PPI,
401                                INTID_TO_PPI(ARCH_TIMER_NS_EL2_VIRT_IRQ), irqflags);
402     } else {
403         qemu_fdt_setprop_cells(ms->fdt, "/timer", "interrupts",
404                                GIC_FDT_IRQ_TYPE_PPI,
405                                INTID_TO_PPI(ARCH_TIMER_S_EL1_IRQ), irqflags,
406                                GIC_FDT_IRQ_TYPE_PPI,
407                                INTID_TO_PPI(ARCH_TIMER_NS_EL1_IRQ), irqflags,
408                                GIC_FDT_IRQ_TYPE_PPI,
409                                INTID_TO_PPI(ARCH_TIMER_VIRT_IRQ), irqflags,
410                                GIC_FDT_IRQ_TYPE_PPI,
411                                INTID_TO_PPI(ARCH_TIMER_NS_EL2_IRQ), irqflags);
412     }
413 }
414 
415 static void fdt_add_cpu_nodes(const VirtMachineState *vms)
416 {
417     int cpu;
418     int addr_cells = 1;
419     const MachineState *ms = MACHINE(vms);
420     const VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
421     int smp_cpus = ms->smp.cpus;
422 
423     /*
424      * See Linux Documentation/devicetree/bindings/arm/cpus.yaml
425      * On ARM v8 64-bit systems value should be set to 2,
426      * that corresponds to the MPIDR_EL1 register size.
427      * If MPIDR_EL1[63:32] value is equal to 0 on all CPUs
428      * in the system, #address-cells can be set to 1, since
429      * MPIDR_EL1[63:32] bits are not used for CPUs
430      * identification.
431      *
432      * Here we actually don't know whether our system is 32- or 64-bit one.
433      * The simplest way to go is to examine affinity IDs of all our CPUs. If
434      * at least one of them has Aff3 populated, we set #address-cells to 2.
435      */
436     for (cpu = 0; cpu < smp_cpus; cpu++) {
437         ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(cpu));
438 
439         if (arm_cpu_mp_affinity(armcpu) & ARM_AFF3_MASK) {
440             addr_cells = 2;
441             break;
442         }
443     }
444 
445     qemu_fdt_add_subnode(ms->fdt, "/cpus");
446     qemu_fdt_setprop_cell(ms->fdt, "/cpus", "#address-cells", addr_cells);
447     qemu_fdt_setprop_cell(ms->fdt, "/cpus", "#size-cells", 0x0);
448 
449     for (cpu = smp_cpus - 1; cpu >= 0; cpu--) {
450         char *nodename = g_strdup_printf("/cpus/cpu@%d", cpu);
451         ARMCPU *armcpu = ARM_CPU(qemu_get_cpu(cpu));
452         CPUState *cs = CPU(armcpu);
453 
454         qemu_fdt_add_subnode(ms->fdt, nodename);
455         qemu_fdt_setprop_string(ms->fdt, nodename, "device_type", "cpu");
456         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
457                                     armcpu->dtb_compatible);
458 
459         if (vms->psci_conduit != QEMU_PSCI_CONDUIT_DISABLED && smp_cpus > 1) {
460             qemu_fdt_setprop_string(ms->fdt, nodename,
461                                         "enable-method", "psci");
462         }
463 
464         if (addr_cells == 2) {
465             qemu_fdt_setprop_u64(ms->fdt, nodename, "reg",
466                                  arm_cpu_mp_affinity(armcpu));
467         } else {
468             qemu_fdt_setprop_cell(ms->fdt, nodename, "reg",
469                                   arm_cpu_mp_affinity(armcpu));
470         }
471 
472         if (ms->possible_cpus->cpus[cs->cpu_index].props.has_node_id) {
473             qemu_fdt_setprop_cell(ms->fdt, nodename, "numa-node-id",
474                 ms->possible_cpus->cpus[cs->cpu_index].props.node_id);
475         }
476 
477         if (!vmc->no_cpu_topology) {
478             qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle",
479                                   qemu_fdt_alloc_phandle(ms->fdt));
480         }
481 
482         g_free(nodename);
483     }
484 
485     if (!vmc->no_cpu_topology) {
486         /*
487          * Add vCPU topology description through fdt node cpu-map.
488          *
489          * See Linux Documentation/devicetree/bindings/cpu/cpu-topology.txt
490          * In a SMP system, the hierarchy of CPUs can be defined through
491          * four entities that are used to describe the layout of CPUs in
492          * the system: socket/cluster/core/thread.
493          *
494          * A socket node represents the boundary of system physical package
495          * and its child nodes must be one or more cluster nodes. A system
496          * can contain several layers of clustering within a single physical
497          * package and cluster nodes can be contained in parent cluster nodes.
498          *
499          * Note: currently we only support one layer of clustering within
500          * each physical package.
501          */
502         qemu_fdt_add_subnode(ms->fdt, "/cpus/cpu-map");
503 
504         for (cpu = smp_cpus - 1; cpu >= 0; cpu--) {
505             char *cpu_path = g_strdup_printf("/cpus/cpu@%d", cpu);
506             char *map_path;
507 
508             if (ms->smp.threads > 1) {
509                 map_path = g_strdup_printf(
510                     "/cpus/cpu-map/socket%d/cluster%d/core%d/thread%d",
511                     cpu / (ms->smp.clusters * ms->smp.cores * ms->smp.threads),
512                     (cpu / (ms->smp.cores * ms->smp.threads)) % ms->smp.clusters,
513                     (cpu / ms->smp.threads) % ms->smp.cores,
514                     cpu % ms->smp.threads);
515             } else {
516                 map_path = g_strdup_printf(
517                     "/cpus/cpu-map/socket%d/cluster%d/core%d",
518                     cpu / (ms->smp.clusters * ms->smp.cores),
519                     (cpu / ms->smp.cores) % ms->smp.clusters,
520                     cpu % ms->smp.cores);
521             }
522             qemu_fdt_add_path(ms->fdt, map_path);
523             qemu_fdt_setprop_phandle(ms->fdt, map_path, "cpu", cpu_path);
524 
525             g_free(map_path);
526             g_free(cpu_path);
527         }
528     }
529 }
530 
531 static void fdt_add_its_gic_node(VirtMachineState *vms)
532 {
533     char *nodename;
534     MachineState *ms = MACHINE(vms);
535 
536     vms->msi_phandle = qemu_fdt_alloc_phandle(ms->fdt);
537     nodename = g_strdup_printf("/intc/its@%" PRIx64,
538                                vms->memmap[VIRT_GIC_ITS].base);
539     qemu_fdt_add_subnode(ms->fdt, nodename);
540     qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
541                             "arm,gic-v3-its");
542     qemu_fdt_setprop(ms->fdt, nodename, "msi-controller", NULL, 0);
543     qemu_fdt_setprop_cell(ms->fdt, nodename, "#msi-cells", 1);
544     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
545                                  2, vms->memmap[VIRT_GIC_ITS].base,
546                                  2, vms->memmap[VIRT_GIC_ITS].size);
547     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", vms->msi_phandle);
548     g_free(nodename);
549 }
550 
551 static void fdt_add_v2m_gic_node(VirtMachineState *vms)
552 {
553     MachineState *ms = MACHINE(vms);
554     char *nodename;
555 
556     nodename = g_strdup_printf("/intc/v2m@%" PRIx64,
557                                vms->memmap[VIRT_GIC_V2M].base);
558     vms->msi_phandle = qemu_fdt_alloc_phandle(ms->fdt);
559     qemu_fdt_add_subnode(ms->fdt, nodename);
560     qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
561                             "arm,gic-v2m-frame");
562     qemu_fdt_setprop(ms->fdt, nodename, "msi-controller", NULL, 0);
563     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
564                                  2, vms->memmap[VIRT_GIC_V2M].base,
565                                  2, vms->memmap[VIRT_GIC_V2M].size);
566     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", vms->msi_phandle);
567     g_free(nodename);
568 }
569 
570 static void fdt_add_gic_node(VirtMachineState *vms)
571 {
572     MachineState *ms = MACHINE(vms);
573     char *nodename;
574 
575     vms->gic_phandle = qemu_fdt_alloc_phandle(ms->fdt);
576     qemu_fdt_setprop_cell(ms->fdt, "/", "interrupt-parent", vms->gic_phandle);
577 
578     nodename = g_strdup_printf("/intc@%" PRIx64,
579                                vms->memmap[VIRT_GIC_DIST].base);
580     qemu_fdt_add_subnode(ms->fdt, nodename);
581     qemu_fdt_setprop_cell(ms->fdt, nodename, "#interrupt-cells", 3);
582     qemu_fdt_setprop(ms->fdt, nodename, "interrupt-controller", NULL, 0);
583     qemu_fdt_setprop_cell(ms->fdt, nodename, "#address-cells", 0x2);
584     qemu_fdt_setprop_cell(ms->fdt, nodename, "#size-cells", 0x2);
585     qemu_fdt_setprop(ms->fdt, nodename, "ranges", NULL, 0);
586     if (vms->gic_version != VIRT_GIC_VERSION_2) {
587         int nb_redist_regions = virt_gicv3_redist_region_count(vms);
588 
589         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
590                                 "arm,gic-v3");
591 
592         qemu_fdt_setprop_cell(ms->fdt, nodename,
593                               "#redistributor-regions", nb_redist_regions);
594 
595         if (nb_redist_regions == 1) {
596             qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
597                                          2, vms->memmap[VIRT_GIC_DIST].base,
598                                          2, vms->memmap[VIRT_GIC_DIST].size,
599                                          2, vms->memmap[VIRT_GIC_REDIST].base,
600                                          2, vms->memmap[VIRT_GIC_REDIST].size);
601         } else {
602             qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
603                                  2, vms->memmap[VIRT_GIC_DIST].base,
604                                  2, vms->memmap[VIRT_GIC_DIST].size,
605                                  2, vms->memmap[VIRT_GIC_REDIST].base,
606                                  2, vms->memmap[VIRT_GIC_REDIST].size,
607                                  2, vms->memmap[VIRT_HIGH_GIC_REDIST2].base,
608                                  2, vms->memmap[VIRT_HIGH_GIC_REDIST2].size);
609         }
610 
611         if (vms->virt) {
612             qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
613                                    GIC_FDT_IRQ_TYPE_PPI,
614                                    INTID_TO_PPI(ARCH_GIC_MAINT_IRQ),
615                                    GIC_FDT_IRQ_FLAGS_LEVEL_HI);
616         }
617     } else {
618         /* 'cortex-a15-gic' means 'GIC v2' */
619         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible",
620                                 "arm,cortex-a15-gic");
621         if (!vms->virt) {
622             qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
623                                          2, vms->memmap[VIRT_GIC_DIST].base,
624                                          2, vms->memmap[VIRT_GIC_DIST].size,
625                                          2, vms->memmap[VIRT_GIC_CPU].base,
626                                          2, vms->memmap[VIRT_GIC_CPU].size);
627         } else {
628             qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
629                                          2, vms->memmap[VIRT_GIC_DIST].base,
630                                          2, vms->memmap[VIRT_GIC_DIST].size,
631                                          2, vms->memmap[VIRT_GIC_CPU].base,
632                                          2, vms->memmap[VIRT_GIC_CPU].size,
633                                          2, vms->memmap[VIRT_GIC_HYP].base,
634                                          2, vms->memmap[VIRT_GIC_HYP].size,
635                                          2, vms->memmap[VIRT_GIC_VCPU].base,
636                                          2, vms->memmap[VIRT_GIC_VCPU].size);
637             qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
638                                    GIC_FDT_IRQ_TYPE_PPI,
639                                    INTID_TO_PPI(ARCH_GIC_MAINT_IRQ),
640                                    GIC_FDT_IRQ_FLAGS_LEVEL_HI);
641         }
642     }
643 
644     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", vms->gic_phandle);
645     g_free(nodename);
646 }
647 
648 static void fdt_add_pmu_nodes(const VirtMachineState *vms)
649 {
650     ARMCPU *armcpu = ARM_CPU(first_cpu);
651     uint32_t irqflags = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
652     MachineState *ms = MACHINE(vms);
653 
654     if (!arm_feature(&armcpu->env, ARM_FEATURE_PMU)) {
655         assert(!object_property_get_bool(OBJECT(armcpu), "pmu", NULL));
656         return;
657     }
658 
659     if (vms->gic_version == VIRT_GIC_VERSION_2) {
660         irqflags = deposit32(irqflags, GIC_FDT_IRQ_PPI_CPU_START,
661                              GIC_FDT_IRQ_PPI_CPU_WIDTH,
662                              (1 << MACHINE(vms)->smp.cpus) - 1);
663     }
664 
665     qemu_fdt_add_subnode(ms->fdt, "/pmu");
666     if (arm_feature(&armcpu->env, ARM_FEATURE_V8)) {
667         const char compat[] = "arm,armv8-pmuv3";
668         qemu_fdt_setprop(ms->fdt, "/pmu", "compatible",
669                          compat, sizeof(compat));
670         qemu_fdt_setprop_cells(ms->fdt, "/pmu", "interrupts",
671                                GIC_FDT_IRQ_TYPE_PPI,
672                                INTID_TO_PPI(VIRTUAL_PMU_IRQ), irqflags);
673     }
674 }
675 
676 static inline DeviceState *create_acpi_ged(VirtMachineState *vms)
677 {
678     DeviceState *dev;
679     MachineState *ms = MACHINE(vms);
680     int irq = vms->irqmap[VIRT_ACPI_GED];
681     uint32_t event = ACPI_GED_PWR_DOWN_EVT;
682 
683     if (ms->ram_slots) {
684         event |= ACPI_GED_MEM_HOTPLUG_EVT;
685     }
686 
687     if (ms->nvdimms_state->is_enabled) {
688         event |= ACPI_GED_NVDIMM_HOTPLUG_EVT;
689     }
690 
691     dev = qdev_new(TYPE_ACPI_GED);
692     qdev_prop_set_uint32(dev, "ged-event", event);
693     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
694 
695     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_ACPI_GED].base);
696     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 1, vms->memmap[VIRT_PCDIMM_ACPI].base);
697     sysbus_connect_irq(SYS_BUS_DEVICE(dev), 0, qdev_get_gpio_in(vms->gic, irq));
698 
699     return dev;
700 }
701 
702 static void create_its(VirtMachineState *vms)
703 {
704     const char *itsclass = its_class_name();
705     DeviceState *dev;
706 
707     if (!strcmp(itsclass, "arm-gicv3-its")) {
708         if (!vms->tcg_its) {
709             itsclass = NULL;
710         }
711     }
712 
713     if (!itsclass) {
714         /* Do nothing if not supported */
715         return;
716     }
717 
718     dev = qdev_new(itsclass);
719 
720     object_property_set_link(OBJECT(dev), "parent-gicv3", OBJECT(vms->gic),
721                              &error_abort);
722     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
723     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_GIC_ITS].base);
724 
725     fdt_add_its_gic_node(vms);
726     vms->msi_controller = VIRT_MSI_CTRL_ITS;
727 }
728 
729 static void create_v2m(VirtMachineState *vms)
730 {
731     int i;
732     int irq = vms->irqmap[VIRT_GIC_V2M];
733     DeviceState *dev;
734 
735     dev = qdev_new("arm-gicv2m");
736     qdev_prop_set_uint32(dev, "base-spi", irq);
737     qdev_prop_set_uint32(dev, "num-spi", NUM_GICV2M_SPIS);
738     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
739     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, vms->memmap[VIRT_GIC_V2M].base);
740 
741     for (i = 0; i < NUM_GICV2M_SPIS; i++) {
742         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
743                            qdev_get_gpio_in(vms->gic, irq + i));
744     }
745 
746     fdt_add_v2m_gic_node(vms);
747     vms->msi_controller = VIRT_MSI_CTRL_GICV2M;
748 }
749 
750 /*
751  * If the CPU has FEAT_NMI, then turn on the NMI support in the GICv3 too.
752  * It's permitted to have a configuration with NMI in the CPU (and thus the
753  * GICv3 CPU interface) but not in the distributor/redistributors, but it's
754  * not very useful.
755  */
756 static bool gicv3_nmi_present(VirtMachineState *vms)
757 {
758     ARMCPU *cpu = ARM_CPU(qemu_get_cpu(0));
759 
760     return tcg_enabled() && cpu_isar_feature(aa64_nmi, cpu) &&
761            (vms->gic_version != VIRT_GIC_VERSION_2);
762 }
763 
764 static void create_gic(VirtMachineState *vms, MemoryRegion *mem)
765 {
766     MachineState *ms = MACHINE(vms);
767     /* We create a standalone GIC */
768     SysBusDevice *gicbusdev;
769     const char *gictype;
770     int i;
771     unsigned int smp_cpus = ms->smp.cpus;
772     uint32_t nb_redist_regions = 0;
773     int revision;
774 
775     if (vms->gic_version == VIRT_GIC_VERSION_2) {
776         gictype = gic_class_name();
777     } else {
778         gictype = gicv3_class_name();
779     }
780 
781     switch (vms->gic_version) {
782     case VIRT_GIC_VERSION_2:
783         revision = 2;
784         break;
785     case VIRT_GIC_VERSION_3:
786         revision = 3;
787         break;
788     case VIRT_GIC_VERSION_4:
789         revision = 4;
790         break;
791     default:
792         g_assert_not_reached();
793     }
794     vms->gic = qdev_new(gictype);
795     qdev_prop_set_uint32(vms->gic, "revision", revision);
796     qdev_prop_set_uint32(vms->gic, "num-cpu", smp_cpus);
797     /* Note that the num-irq property counts both internal and external
798      * interrupts; there are always 32 of the former (mandated by GIC spec).
799      */
800     qdev_prop_set_uint32(vms->gic, "num-irq", NUM_IRQS + 32);
801     if (!kvm_irqchip_in_kernel()) {
802         qdev_prop_set_bit(vms->gic, "has-security-extensions", vms->secure);
803     }
804 
805     if (vms->gic_version != VIRT_GIC_VERSION_2) {
806         QList *redist_region_count;
807         uint32_t redist0_capacity = virt_redist_capacity(vms, VIRT_GIC_REDIST);
808         uint32_t redist0_count = MIN(smp_cpus, redist0_capacity);
809 
810         nb_redist_regions = virt_gicv3_redist_region_count(vms);
811 
812         redist_region_count = qlist_new();
813         qlist_append_int(redist_region_count, redist0_count);
814         if (nb_redist_regions == 2) {
815             uint32_t redist1_capacity =
816                 virt_redist_capacity(vms, VIRT_HIGH_GIC_REDIST2);
817 
818             qlist_append_int(redist_region_count,
819                 MIN(smp_cpus - redist0_count, redist1_capacity));
820         }
821         qdev_prop_set_array(vms->gic, "redist-region-count",
822                             redist_region_count);
823 
824         if (!kvm_irqchip_in_kernel()) {
825             if (vms->tcg_its) {
826                 object_property_set_link(OBJECT(vms->gic), "sysmem",
827                                          OBJECT(mem), &error_fatal);
828                 qdev_prop_set_bit(vms->gic, "has-lpi", true);
829             }
830         }
831     } else {
832         if (!kvm_irqchip_in_kernel()) {
833             qdev_prop_set_bit(vms->gic, "has-virtualization-extensions",
834                               vms->virt);
835         }
836     }
837 
838     if (gicv3_nmi_present(vms)) {
839         qdev_prop_set_bit(vms->gic, "has-nmi", true);
840     }
841 
842     gicbusdev = SYS_BUS_DEVICE(vms->gic);
843     sysbus_realize_and_unref(gicbusdev, &error_fatal);
844     sysbus_mmio_map(gicbusdev, 0, vms->memmap[VIRT_GIC_DIST].base);
845     if (vms->gic_version != VIRT_GIC_VERSION_2) {
846         sysbus_mmio_map(gicbusdev, 1, vms->memmap[VIRT_GIC_REDIST].base);
847         if (nb_redist_regions == 2) {
848             sysbus_mmio_map(gicbusdev, 2,
849                             vms->memmap[VIRT_HIGH_GIC_REDIST2].base);
850         }
851     } else {
852         sysbus_mmio_map(gicbusdev, 1, vms->memmap[VIRT_GIC_CPU].base);
853         if (vms->virt) {
854             sysbus_mmio_map(gicbusdev, 2, vms->memmap[VIRT_GIC_HYP].base);
855             sysbus_mmio_map(gicbusdev, 3, vms->memmap[VIRT_GIC_VCPU].base);
856         }
857     }
858 
859     /* Wire the outputs from each CPU's generic timer and the GICv3
860      * maintenance interrupt signal to the appropriate GIC PPI inputs,
861      * and the GIC's IRQ/FIQ/VIRQ/VFIQ/NMI/VINMI interrupt outputs to the
862      * CPU's inputs.
863      */
864     for (i = 0; i < smp_cpus; i++) {
865         DeviceState *cpudev = DEVICE(qemu_get_cpu(i));
866         int intidbase = NUM_IRQS + i * GIC_INTERNAL;
867         /* Mapping from the output timer irq lines from the CPU to the
868          * GIC PPI inputs we use for the virt board.
869          */
870         const int timer_irq[] = {
871             [GTIMER_PHYS] = ARCH_TIMER_NS_EL1_IRQ,
872             [GTIMER_VIRT] = ARCH_TIMER_VIRT_IRQ,
873             [GTIMER_HYP]  = ARCH_TIMER_NS_EL2_IRQ,
874             [GTIMER_SEC]  = ARCH_TIMER_S_EL1_IRQ,
875             [GTIMER_HYPVIRT] = ARCH_TIMER_NS_EL2_VIRT_IRQ,
876         };
877 
878         for (unsigned irq = 0; irq < ARRAY_SIZE(timer_irq); irq++) {
879             qdev_connect_gpio_out(cpudev, irq,
880                                   qdev_get_gpio_in(vms->gic,
881                                                    intidbase + timer_irq[irq]));
882         }
883 
884         if (vms->gic_version != VIRT_GIC_VERSION_2) {
885             qemu_irq irq = qdev_get_gpio_in(vms->gic,
886                                             intidbase + ARCH_GIC_MAINT_IRQ);
887             qdev_connect_gpio_out_named(cpudev, "gicv3-maintenance-interrupt",
888                                         0, irq);
889         } else if (vms->virt) {
890             qemu_irq irq = qdev_get_gpio_in(vms->gic,
891                                             intidbase + ARCH_GIC_MAINT_IRQ);
892             sysbus_connect_irq(gicbusdev, i + 4 * smp_cpus, irq);
893         }
894 
895         qdev_connect_gpio_out_named(cpudev, "pmu-interrupt", 0,
896                                     qdev_get_gpio_in(vms->gic, intidbase
897                                                      + VIRTUAL_PMU_IRQ));
898 
899         sysbus_connect_irq(gicbusdev, i, qdev_get_gpio_in(cpudev, ARM_CPU_IRQ));
900         sysbus_connect_irq(gicbusdev, i + smp_cpus,
901                            qdev_get_gpio_in(cpudev, ARM_CPU_FIQ));
902         sysbus_connect_irq(gicbusdev, i + 2 * smp_cpus,
903                            qdev_get_gpio_in(cpudev, ARM_CPU_VIRQ));
904         sysbus_connect_irq(gicbusdev, i + 3 * smp_cpus,
905                            qdev_get_gpio_in(cpudev, ARM_CPU_VFIQ));
906 
907         if (vms->gic_version != VIRT_GIC_VERSION_2) {
908             sysbus_connect_irq(gicbusdev, i + 4 * smp_cpus,
909                                qdev_get_gpio_in(cpudev, ARM_CPU_NMI));
910             sysbus_connect_irq(gicbusdev, i + 5 * smp_cpus,
911                                qdev_get_gpio_in(cpudev, ARM_CPU_VINMI));
912         }
913     }
914 
915     fdt_add_gic_node(vms);
916 
917     if (vms->gic_version != VIRT_GIC_VERSION_2 && vms->its) {
918         create_its(vms);
919     } else if (vms->gic_version == VIRT_GIC_VERSION_2) {
920         create_v2m(vms);
921     }
922 }
923 
924 static void create_uart(const VirtMachineState *vms, int uart,
925                         MemoryRegion *mem, Chardev *chr, bool secure)
926 {
927     char *nodename;
928     hwaddr base = vms->memmap[uart].base;
929     hwaddr size = vms->memmap[uart].size;
930     int irq = vms->irqmap[uart];
931     const char compat[] = "arm,pl011\0arm,primecell";
932     const char clocknames[] = "uartclk\0apb_pclk";
933     DeviceState *dev = qdev_new(TYPE_PL011);
934     SysBusDevice *s = SYS_BUS_DEVICE(dev);
935     MachineState *ms = MACHINE(vms);
936 
937     qdev_prop_set_chr(dev, "chardev", chr);
938     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
939     memory_region_add_subregion(mem, base,
940                                 sysbus_mmio_get_region(s, 0));
941     sysbus_connect_irq(s, 0, qdev_get_gpio_in(vms->gic, irq));
942 
943     nodename = g_strdup_printf("/pl011@%" PRIx64, base);
944     qemu_fdt_add_subnode(ms->fdt, nodename);
945     /* Note that we can't use setprop_string because of the embedded NUL */
946     qemu_fdt_setprop(ms->fdt, nodename, "compatible",
947                          compat, sizeof(compat));
948     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
949                                      2, base, 2, size);
950     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
951                                GIC_FDT_IRQ_TYPE_SPI, irq,
952                                GIC_FDT_IRQ_FLAGS_LEVEL_HI);
953     qemu_fdt_setprop_cells(ms->fdt, nodename, "clocks",
954                                vms->clock_phandle, vms->clock_phandle);
955     qemu_fdt_setprop(ms->fdt, nodename, "clock-names",
956                          clocknames, sizeof(clocknames));
957 
958     if (uart == VIRT_UART0) {
959         qemu_fdt_setprop_string(ms->fdt, "/chosen", "stdout-path", nodename);
960         qemu_fdt_setprop_string(ms->fdt, "/aliases", "serial0", nodename);
961     } else {
962         qemu_fdt_setprop_string(ms->fdt, "/aliases", "serial1", nodename);
963     }
964     if (secure) {
965         /* Mark as not usable by the normal world */
966         qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
967         qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
968 
969         qemu_fdt_setprop_string(ms->fdt, "/secure-chosen", "stdout-path",
970                                 nodename);
971     }
972 
973     g_free(nodename);
974 }
975 
976 static void create_rtc(const VirtMachineState *vms)
977 {
978     char *nodename;
979     hwaddr base = vms->memmap[VIRT_RTC].base;
980     hwaddr size = vms->memmap[VIRT_RTC].size;
981     int irq = vms->irqmap[VIRT_RTC];
982     const char compat[] = "arm,pl031\0arm,primecell";
983     MachineState *ms = MACHINE(vms);
984 
985     sysbus_create_simple("pl031", base, qdev_get_gpio_in(vms->gic, irq));
986 
987     nodename = g_strdup_printf("/pl031@%" PRIx64, base);
988     qemu_fdt_add_subnode(ms->fdt, nodename);
989     qemu_fdt_setprop(ms->fdt, nodename, "compatible", compat, sizeof(compat));
990     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
991                                  2, base, 2, size);
992     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
993                            GIC_FDT_IRQ_TYPE_SPI, irq,
994                            GIC_FDT_IRQ_FLAGS_LEVEL_HI);
995     qemu_fdt_setprop_cell(ms->fdt, nodename, "clocks", vms->clock_phandle);
996     qemu_fdt_setprop_string(ms->fdt, nodename, "clock-names", "apb_pclk");
997     g_free(nodename);
998 }
999 
1000 static DeviceState *gpio_key_dev;
1001 static void virt_powerdown_req(Notifier *n, void *opaque)
1002 {
1003     VirtMachineState *s = container_of(n, VirtMachineState, powerdown_notifier);
1004 
1005     if (s->acpi_dev) {
1006         acpi_send_event(s->acpi_dev, ACPI_POWER_DOWN_STATUS);
1007     } else {
1008         /* use gpio Pin for power button event */
1009         qemu_set_irq(qdev_get_gpio_in(gpio_key_dev, 0), 1);
1010     }
1011 }
1012 
1013 static void create_gpio_keys(char *fdt, DeviceState *pl061_dev,
1014                              uint32_t phandle)
1015 {
1016     gpio_key_dev = sysbus_create_simple("gpio-key", -1,
1017                                         qdev_get_gpio_in(pl061_dev,
1018                                                          GPIO_PIN_POWER_BUTTON));
1019 
1020     qemu_fdt_add_subnode(fdt, "/gpio-keys");
1021     qemu_fdt_setprop_string(fdt, "/gpio-keys", "compatible", "gpio-keys");
1022 
1023     qemu_fdt_add_subnode(fdt, "/gpio-keys/poweroff");
1024     qemu_fdt_setprop_string(fdt, "/gpio-keys/poweroff",
1025                             "label", "GPIO Key Poweroff");
1026     qemu_fdt_setprop_cell(fdt, "/gpio-keys/poweroff", "linux,code",
1027                           KEY_POWER);
1028     qemu_fdt_setprop_cells(fdt, "/gpio-keys/poweroff",
1029                            "gpios", phandle, GPIO_PIN_POWER_BUTTON, 0);
1030 }
1031 
1032 #define SECURE_GPIO_POWEROFF 0
1033 #define SECURE_GPIO_RESET    1
1034 
1035 static void create_secure_gpio_pwr(char *fdt, DeviceState *pl061_dev,
1036                                    uint32_t phandle)
1037 {
1038     DeviceState *gpio_pwr_dev;
1039 
1040     /* gpio-pwr */
1041     gpio_pwr_dev = sysbus_create_simple("gpio-pwr", -1, NULL);
1042 
1043     /* connect secure pl061 to gpio-pwr */
1044     qdev_connect_gpio_out(pl061_dev, SECURE_GPIO_RESET,
1045                           qdev_get_gpio_in_named(gpio_pwr_dev, "reset", 0));
1046     qdev_connect_gpio_out(pl061_dev, SECURE_GPIO_POWEROFF,
1047                           qdev_get_gpio_in_named(gpio_pwr_dev, "shutdown", 0));
1048 
1049     qemu_fdt_add_subnode(fdt, "/gpio-poweroff");
1050     qemu_fdt_setprop_string(fdt, "/gpio-poweroff", "compatible",
1051                             "gpio-poweroff");
1052     qemu_fdt_setprop_cells(fdt, "/gpio-poweroff",
1053                            "gpios", phandle, SECURE_GPIO_POWEROFF, 0);
1054     qemu_fdt_setprop_string(fdt, "/gpio-poweroff", "status", "disabled");
1055     qemu_fdt_setprop_string(fdt, "/gpio-poweroff", "secure-status",
1056                             "okay");
1057 
1058     qemu_fdt_add_subnode(fdt, "/gpio-restart");
1059     qemu_fdt_setprop_string(fdt, "/gpio-restart", "compatible",
1060                             "gpio-restart");
1061     qemu_fdt_setprop_cells(fdt, "/gpio-restart",
1062                            "gpios", phandle, SECURE_GPIO_RESET, 0);
1063     qemu_fdt_setprop_string(fdt, "/gpio-restart", "status", "disabled");
1064     qemu_fdt_setprop_string(fdt, "/gpio-restart", "secure-status",
1065                             "okay");
1066 }
1067 
1068 static void create_gpio_devices(const VirtMachineState *vms, int gpio,
1069                                 MemoryRegion *mem)
1070 {
1071     char *nodename;
1072     DeviceState *pl061_dev;
1073     hwaddr base = vms->memmap[gpio].base;
1074     hwaddr size = vms->memmap[gpio].size;
1075     int irq = vms->irqmap[gpio];
1076     const char compat[] = "arm,pl061\0arm,primecell";
1077     SysBusDevice *s;
1078     MachineState *ms = MACHINE(vms);
1079 
1080     pl061_dev = qdev_new("pl061");
1081     /* Pull lines down to 0 if not driven by the PL061 */
1082     qdev_prop_set_uint32(pl061_dev, "pullups", 0);
1083     qdev_prop_set_uint32(pl061_dev, "pulldowns", 0xff);
1084     s = SYS_BUS_DEVICE(pl061_dev);
1085     sysbus_realize_and_unref(s, &error_fatal);
1086     memory_region_add_subregion(mem, base, sysbus_mmio_get_region(s, 0));
1087     sysbus_connect_irq(s, 0, qdev_get_gpio_in(vms->gic, irq));
1088 
1089     uint32_t phandle = qemu_fdt_alloc_phandle(ms->fdt);
1090     nodename = g_strdup_printf("/pl061@%" PRIx64, base);
1091     qemu_fdt_add_subnode(ms->fdt, nodename);
1092     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1093                                  2, base, 2, size);
1094     qemu_fdt_setprop(ms->fdt, nodename, "compatible", compat, sizeof(compat));
1095     qemu_fdt_setprop_cell(ms->fdt, nodename, "#gpio-cells", 2);
1096     qemu_fdt_setprop(ms->fdt, nodename, "gpio-controller", NULL, 0);
1097     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
1098                            GIC_FDT_IRQ_TYPE_SPI, irq,
1099                            GIC_FDT_IRQ_FLAGS_LEVEL_HI);
1100     qemu_fdt_setprop_cell(ms->fdt, nodename, "clocks", vms->clock_phandle);
1101     qemu_fdt_setprop_string(ms->fdt, nodename, "clock-names", "apb_pclk");
1102     qemu_fdt_setprop_cell(ms->fdt, nodename, "phandle", phandle);
1103 
1104     if (gpio != VIRT_GPIO) {
1105         /* Mark as not usable by the normal world */
1106         qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
1107         qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
1108     }
1109     g_free(nodename);
1110 
1111     /* Child gpio devices */
1112     if (gpio == VIRT_GPIO) {
1113         create_gpio_keys(ms->fdt, pl061_dev, phandle);
1114     } else {
1115         create_secure_gpio_pwr(ms->fdt, pl061_dev, phandle);
1116     }
1117 }
1118 
1119 static void create_virtio_devices(const VirtMachineState *vms)
1120 {
1121     int i;
1122     hwaddr size = vms->memmap[VIRT_MMIO].size;
1123     MachineState *ms = MACHINE(vms);
1124 
1125     /* We create the transports in forwards order. Since qbus_realize()
1126      * prepends (not appends) new child buses, the incrementing loop below will
1127      * create a list of virtio-mmio buses with decreasing base addresses.
1128      *
1129      * When a -device option is processed from the command line,
1130      * qbus_find_recursive() picks the next free virtio-mmio bus in forwards
1131      * order. The upshot is that -device options in increasing command line
1132      * order are mapped to virtio-mmio buses with decreasing base addresses.
1133      *
1134      * When this code was originally written, that arrangement ensured that the
1135      * guest Linux kernel would give the lowest "name" (/dev/vda, eth0, etc) to
1136      * the first -device on the command line. (The end-to-end order is a
1137      * function of this loop, qbus_realize(), qbus_find_recursive(), and the
1138      * guest kernel's name-to-address assignment strategy.)
1139      *
1140      * Meanwhile, the kernel's traversal seems to have been reversed; see eg.
1141      * the message, if not necessarily the code, of commit 70161ff336.
1142      * Therefore the loop now establishes the inverse of the original intent.
1143      *
1144      * Unfortunately, we can't counteract the kernel change by reversing the
1145      * loop; it would break existing command lines.
1146      *
1147      * In any case, the kernel makes no guarantee about the stability of
1148      * enumeration order of virtio devices (as demonstrated by it changing
1149      * between kernel versions). For reliable and stable identification
1150      * of disks users must use UUIDs or similar mechanisms.
1151      */
1152     for (i = 0; i < NUM_VIRTIO_TRANSPORTS; i++) {
1153         int irq = vms->irqmap[VIRT_MMIO] + i;
1154         hwaddr base = vms->memmap[VIRT_MMIO].base + i * size;
1155 
1156         sysbus_create_simple("virtio-mmio", base,
1157                              qdev_get_gpio_in(vms->gic, irq));
1158     }
1159 
1160     /* We add dtb nodes in reverse order so that they appear in the finished
1161      * device tree lowest address first.
1162      *
1163      * Note that this mapping is independent of the loop above. The previous
1164      * loop influences virtio device to virtio transport assignment, whereas
1165      * this loop controls how virtio transports are laid out in the dtb.
1166      */
1167     for (i = NUM_VIRTIO_TRANSPORTS - 1; i >= 0; i--) {
1168         char *nodename;
1169         int irq = vms->irqmap[VIRT_MMIO] + i;
1170         hwaddr base = vms->memmap[VIRT_MMIO].base + i * size;
1171 
1172         nodename = g_strdup_printf("/virtio_mmio@%" PRIx64, base);
1173         qemu_fdt_add_subnode(ms->fdt, nodename);
1174         qemu_fdt_setprop_string(ms->fdt, nodename,
1175                                 "compatible", "virtio,mmio");
1176         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1177                                      2, base, 2, size);
1178         qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupts",
1179                                GIC_FDT_IRQ_TYPE_SPI, irq,
1180                                GIC_FDT_IRQ_FLAGS_EDGE_LO_HI);
1181         qemu_fdt_setprop(ms->fdt, nodename, "dma-coherent", NULL, 0);
1182         g_free(nodename);
1183     }
1184 }
1185 
1186 #define VIRT_FLASH_SECTOR_SIZE (256 * KiB)
1187 
1188 static PFlashCFI01 *virt_flash_create1(VirtMachineState *vms,
1189                                         const char *name,
1190                                         const char *alias_prop_name)
1191 {
1192     /*
1193      * Create a single flash device.  We use the same parameters as
1194      * the flash devices on the Versatile Express board.
1195      */
1196     DeviceState *dev = qdev_new(TYPE_PFLASH_CFI01);
1197 
1198     qdev_prop_set_uint64(dev, "sector-length", VIRT_FLASH_SECTOR_SIZE);
1199     qdev_prop_set_uint8(dev, "width", 4);
1200     qdev_prop_set_uint8(dev, "device-width", 2);
1201     qdev_prop_set_bit(dev, "big-endian", false);
1202     qdev_prop_set_uint16(dev, "id0", 0x89);
1203     qdev_prop_set_uint16(dev, "id1", 0x18);
1204     qdev_prop_set_uint16(dev, "id2", 0x00);
1205     qdev_prop_set_uint16(dev, "id3", 0x00);
1206     qdev_prop_set_string(dev, "name", name);
1207     object_property_add_child(OBJECT(vms), name, OBJECT(dev));
1208     object_property_add_alias(OBJECT(vms), alias_prop_name,
1209                               OBJECT(dev), "drive");
1210     return PFLASH_CFI01(dev);
1211 }
1212 
1213 static void virt_flash_create(VirtMachineState *vms)
1214 {
1215     vms->flash[0] = virt_flash_create1(vms, "virt.flash0", "pflash0");
1216     vms->flash[1] = virt_flash_create1(vms, "virt.flash1", "pflash1");
1217 }
1218 
1219 static void virt_flash_map1(PFlashCFI01 *flash,
1220                             hwaddr base, hwaddr size,
1221                             MemoryRegion *sysmem)
1222 {
1223     DeviceState *dev = DEVICE(flash);
1224 
1225     assert(QEMU_IS_ALIGNED(size, VIRT_FLASH_SECTOR_SIZE));
1226     assert(size / VIRT_FLASH_SECTOR_SIZE <= UINT32_MAX);
1227     qdev_prop_set_uint32(dev, "num-blocks", size / VIRT_FLASH_SECTOR_SIZE);
1228     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1229 
1230     memory_region_add_subregion(sysmem, base,
1231                                 sysbus_mmio_get_region(SYS_BUS_DEVICE(dev),
1232                                                        0));
1233 }
1234 
1235 static void virt_flash_map(VirtMachineState *vms,
1236                            MemoryRegion *sysmem,
1237                            MemoryRegion *secure_sysmem)
1238 {
1239     /*
1240      * Map two flash devices to fill the VIRT_FLASH space in the memmap.
1241      * sysmem is the system memory space. secure_sysmem is the secure view
1242      * of the system, and the first flash device should be made visible only
1243      * there. The second flash device is visible to both secure and nonsecure.
1244      * If sysmem == secure_sysmem this means there is no separate Secure
1245      * address space and both flash devices are generally visible.
1246      */
1247     hwaddr flashsize = vms->memmap[VIRT_FLASH].size / 2;
1248     hwaddr flashbase = vms->memmap[VIRT_FLASH].base;
1249 
1250     virt_flash_map1(vms->flash[0], flashbase, flashsize,
1251                     secure_sysmem);
1252     virt_flash_map1(vms->flash[1], flashbase + flashsize, flashsize,
1253                     sysmem);
1254 }
1255 
1256 static void virt_flash_fdt(VirtMachineState *vms,
1257                            MemoryRegion *sysmem,
1258                            MemoryRegion *secure_sysmem)
1259 {
1260     hwaddr flashsize = vms->memmap[VIRT_FLASH].size / 2;
1261     hwaddr flashbase = vms->memmap[VIRT_FLASH].base;
1262     MachineState *ms = MACHINE(vms);
1263     char *nodename;
1264 
1265     if (sysmem == secure_sysmem) {
1266         /* Report both flash devices as a single node in the DT */
1267         nodename = g_strdup_printf("/flash@%" PRIx64, flashbase);
1268         qemu_fdt_add_subnode(ms->fdt, nodename);
1269         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible", "cfi-flash");
1270         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1271                                      2, flashbase, 2, flashsize,
1272                                      2, flashbase + flashsize, 2, flashsize);
1273         qemu_fdt_setprop_cell(ms->fdt, nodename, "bank-width", 4);
1274         g_free(nodename);
1275     } else {
1276         /*
1277          * Report the devices as separate nodes so we can mark one as
1278          * only visible to the secure world.
1279          */
1280         nodename = g_strdup_printf("/secflash@%" PRIx64, flashbase);
1281         qemu_fdt_add_subnode(ms->fdt, nodename);
1282         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible", "cfi-flash");
1283         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1284                                      2, flashbase, 2, flashsize);
1285         qemu_fdt_setprop_cell(ms->fdt, nodename, "bank-width", 4);
1286         qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
1287         qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
1288         g_free(nodename);
1289 
1290         nodename = g_strdup_printf("/flash@%" PRIx64, flashbase + flashsize);
1291         qemu_fdt_add_subnode(ms->fdt, nodename);
1292         qemu_fdt_setprop_string(ms->fdt, nodename, "compatible", "cfi-flash");
1293         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1294                                      2, flashbase + flashsize, 2, flashsize);
1295         qemu_fdt_setprop_cell(ms->fdt, nodename, "bank-width", 4);
1296         g_free(nodename);
1297     }
1298 }
1299 
1300 static bool virt_firmware_init(VirtMachineState *vms,
1301                                MemoryRegion *sysmem,
1302                                MemoryRegion *secure_sysmem)
1303 {
1304     int i;
1305     const char *bios_name;
1306     BlockBackend *pflash_blk0;
1307 
1308     /* Map legacy -drive if=pflash to machine properties */
1309     for (i = 0; i < ARRAY_SIZE(vms->flash); i++) {
1310         pflash_cfi01_legacy_drive(vms->flash[i],
1311                                   drive_get(IF_PFLASH, 0, i));
1312     }
1313 
1314     virt_flash_map(vms, sysmem, secure_sysmem);
1315 
1316     pflash_blk0 = pflash_cfi01_get_blk(vms->flash[0]);
1317 
1318     bios_name = MACHINE(vms)->firmware;
1319     if (bios_name) {
1320         char *fname;
1321         MemoryRegion *mr;
1322         int image_size;
1323 
1324         if (pflash_blk0) {
1325             error_report("The contents of the first flash device may be "
1326                          "specified with -bios or with -drive if=pflash... "
1327                          "but you cannot use both options at once");
1328             exit(1);
1329         }
1330 
1331         /* Fall back to -bios */
1332 
1333         fname = qemu_find_file(QEMU_FILE_TYPE_BIOS, bios_name);
1334         if (!fname) {
1335             error_report("Could not find ROM image '%s'", bios_name);
1336             exit(1);
1337         }
1338         mr = sysbus_mmio_get_region(SYS_BUS_DEVICE(vms->flash[0]), 0);
1339         image_size = load_image_mr(fname, mr);
1340         g_free(fname);
1341         if (image_size < 0) {
1342             error_report("Could not load ROM image '%s'", bios_name);
1343             exit(1);
1344         }
1345     }
1346 
1347     return pflash_blk0 || bios_name;
1348 }
1349 
1350 static FWCfgState *create_fw_cfg(const VirtMachineState *vms, AddressSpace *as)
1351 {
1352     MachineState *ms = MACHINE(vms);
1353     hwaddr base = vms->memmap[VIRT_FW_CFG].base;
1354     hwaddr size = vms->memmap[VIRT_FW_CFG].size;
1355     FWCfgState *fw_cfg;
1356     char *nodename;
1357 
1358     fw_cfg = fw_cfg_init_mem_wide(base + 8, base, 8, base + 16, as);
1359     fw_cfg_add_i16(fw_cfg, FW_CFG_NB_CPUS, (uint16_t)ms->smp.cpus);
1360 
1361     nodename = g_strdup_printf("/fw-cfg@%" PRIx64, base);
1362     qemu_fdt_add_subnode(ms->fdt, nodename);
1363     qemu_fdt_setprop_string(ms->fdt, nodename,
1364                             "compatible", "qemu,fw-cfg-mmio");
1365     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1366                                  2, base, 2, size);
1367     qemu_fdt_setprop(ms->fdt, nodename, "dma-coherent", NULL, 0);
1368     g_free(nodename);
1369     return fw_cfg;
1370 }
1371 
1372 static void create_pcie_irq_map(const MachineState *ms,
1373                                 uint32_t gic_phandle,
1374                                 int first_irq, const char *nodename)
1375 {
1376     int devfn, pin;
1377     uint32_t full_irq_map[4 * 4 * 10] = { 0 };
1378     uint32_t *irq_map = full_irq_map;
1379 
1380     for (devfn = 0; devfn <= 0x18; devfn += 0x8) {
1381         for (pin = 0; pin < 4; pin++) {
1382             int irq_type = GIC_FDT_IRQ_TYPE_SPI;
1383             int irq_nr = first_irq + ((pin + PCI_SLOT(devfn)) % PCI_NUM_PINS);
1384             int irq_level = GIC_FDT_IRQ_FLAGS_LEVEL_HI;
1385             int i;
1386 
1387             uint32_t map[] = {
1388                 devfn << 8, 0, 0,                           /* devfn */
1389                 pin + 1,                                    /* PCI pin */
1390                 gic_phandle, 0, 0, irq_type, irq_nr, irq_level }; /* GIC irq */
1391 
1392             /* Convert map to big endian */
1393             for (i = 0; i < 10; i++) {
1394                 irq_map[i] = cpu_to_be32(map[i]);
1395             }
1396             irq_map += 10;
1397         }
1398     }
1399 
1400     qemu_fdt_setprop(ms->fdt, nodename, "interrupt-map",
1401                      full_irq_map, sizeof(full_irq_map));
1402 
1403     qemu_fdt_setprop_cells(ms->fdt, nodename, "interrupt-map-mask",
1404                            cpu_to_be16(PCI_DEVFN(3, 0)), /* Slot 3 */
1405                            0, 0,
1406                            0x7           /* PCI irq */);
1407 }
1408 
1409 static void create_smmu(const VirtMachineState *vms,
1410                         PCIBus *bus)
1411 {
1412     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
1413     char *node;
1414     const char compat[] = "arm,smmu-v3";
1415     int irq =  vms->irqmap[VIRT_SMMU];
1416     int i;
1417     hwaddr base = vms->memmap[VIRT_SMMU].base;
1418     hwaddr size = vms->memmap[VIRT_SMMU].size;
1419     const char irq_names[] = "eventq\0priq\0cmdq-sync\0gerror";
1420     DeviceState *dev;
1421     MachineState *ms = MACHINE(vms);
1422 
1423     if (vms->iommu != VIRT_IOMMU_SMMUV3 || !vms->iommu_phandle) {
1424         return;
1425     }
1426 
1427     dev = qdev_new(TYPE_ARM_SMMUV3);
1428 
1429     if (!vmc->no_nested_smmu) {
1430         object_property_set_str(OBJECT(dev), "stage", "nested", &error_fatal);
1431     }
1432     object_property_set_link(OBJECT(dev), "primary-bus", OBJECT(bus),
1433                              &error_abort);
1434     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1435     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 0, base);
1436     for (i = 0; i < NUM_SMMU_IRQS; i++) {
1437         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
1438                            qdev_get_gpio_in(vms->gic, irq + i));
1439     }
1440 
1441     node = g_strdup_printf("/smmuv3@%" PRIx64, base);
1442     qemu_fdt_add_subnode(ms->fdt, node);
1443     qemu_fdt_setprop(ms->fdt, node, "compatible", compat, sizeof(compat));
1444     qemu_fdt_setprop_sized_cells(ms->fdt, node, "reg", 2, base, 2, size);
1445 
1446     qemu_fdt_setprop_cells(ms->fdt, node, "interrupts",
1447             GIC_FDT_IRQ_TYPE_SPI, irq    , GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1448             GIC_FDT_IRQ_TYPE_SPI, irq + 1, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1449             GIC_FDT_IRQ_TYPE_SPI, irq + 2, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI,
1450             GIC_FDT_IRQ_TYPE_SPI, irq + 3, GIC_FDT_IRQ_FLAGS_EDGE_LO_HI);
1451 
1452     qemu_fdt_setprop(ms->fdt, node, "interrupt-names", irq_names,
1453                      sizeof(irq_names));
1454 
1455     qemu_fdt_setprop(ms->fdt, node, "dma-coherent", NULL, 0);
1456 
1457     qemu_fdt_setprop_cell(ms->fdt, node, "#iommu-cells", 1);
1458 
1459     qemu_fdt_setprop_cell(ms->fdt, node, "phandle", vms->iommu_phandle);
1460     g_free(node);
1461 }
1462 
1463 static void create_virtio_iommu_dt_bindings(VirtMachineState *vms)
1464 {
1465     const char compat[] = "virtio,pci-iommu\0pci1af4,1057";
1466     uint16_t bdf = vms->virtio_iommu_bdf;
1467     MachineState *ms = MACHINE(vms);
1468     char *node;
1469 
1470     vms->iommu_phandle = qemu_fdt_alloc_phandle(ms->fdt);
1471 
1472     node = g_strdup_printf("%s/virtio_iommu@%x,%x", vms->pciehb_nodename,
1473                            PCI_SLOT(bdf), PCI_FUNC(bdf));
1474     qemu_fdt_add_subnode(ms->fdt, node);
1475     qemu_fdt_setprop(ms->fdt, node, "compatible", compat, sizeof(compat));
1476     qemu_fdt_setprop_sized_cells(ms->fdt, node, "reg",
1477                                  1, bdf << 8, 1, 0, 1, 0,
1478                                  1, 0, 1, 0);
1479 
1480     qemu_fdt_setprop_cell(ms->fdt, node, "#iommu-cells", 1);
1481     qemu_fdt_setprop_cell(ms->fdt, node, "phandle", vms->iommu_phandle);
1482     g_free(node);
1483 
1484     qemu_fdt_setprop_cells(ms->fdt, vms->pciehb_nodename, "iommu-map",
1485                            0x0, vms->iommu_phandle, 0x0, bdf,
1486                            bdf + 1, vms->iommu_phandle, bdf + 1, 0xffff - bdf);
1487 }
1488 
1489 static void create_pcie(VirtMachineState *vms)
1490 {
1491     hwaddr base_mmio = vms->memmap[VIRT_PCIE_MMIO].base;
1492     hwaddr size_mmio = vms->memmap[VIRT_PCIE_MMIO].size;
1493     hwaddr base_mmio_high = vms->memmap[VIRT_HIGH_PCIE_MMIO].base;
1494     hwaddr size_mmio_high = vms->memmap[VIRT_HIGH_PCIE_MMIO].size;
1495     hwaddr base_pio = vms->memmap[VIRT_PCIE_PIO].base;
1496     hwaddr size_pio = vms->memmap[VIRT_PCIE_PIO].size;
1497     hwaddr base_ecam, size_ecam;
1498     hwaddr base = base_mmio;
1499     int nr_pcie_buses;
1500     int irq = vms->irqmap[VIRT_PCIE];
1501     MemoryRegion *mmio_alias;
1502     MemoryRegion *mmio_reg;
1503     MemoryRegion *ecam_alias;
1504     MemoryRegion *ecam_reg;
1505     DeviceState *dev;
1506     char *nodename;
1507     int i, ecam_id;
1508     PCIHostState *pci;
1509     MachineState *ms = MACHINE(vms);
1510     MachineClass *mc = MACHINE_GET_CLASS(ms);
1511 
1512     dev = qdev_new(TYPE_GPEX_HOST);
1513     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1514 
1515     ecam_id = VIRT_ECAM_ID(vms->highmem_ecam);
1516     base_ecam = vms->memmap[ecam_id].base;
1517     size_ecam = vms->memmap[ecam_id].size;
1518     nr_pcie_buses = size_ecam / PCIE_MMCFG_SIZE_MIN;
1519     /* Map only the first size_ecam bytes of ECAM space */
1520     ecam_alias = g_new0(MemoryRegion, 1);
1521     ecam_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 0);
1522     memory_region_init_alias(ecam_alias, OBJECT(dev), "pcie-ecam",
1523                              ecam_reg, 0, size_ecam);
1524     memory_region_add_subregion(get_system_memory(), base_ecam, ecam_alias);
1525 
1526     /* Map the MMIO window into system address space so as to expose
1527      * the section of PCI MMIO space which starts at the same base address
1528      * (ie 1:1 mapping for that part of PCI MMIO space visible through
1529      * the window).
1530      */
1531     mmio_alias = g_new0(MemoryRegion, 1);
1532     mmio_reg = sysbus_mmio_get_region(SYS_BUS_DEVICE(dev), 1);
1533     memory_region_init_alias(mmio_alias, OBJECT(dev), "pcie-mmio",
1534                              mmio_reg, base_mmio, size_mmio);
1535     memory_region_add_subregion(get_system_memory(), base_mmio, mmio_alias);
1536 
1537     if (vms->highmem_mmio) {
1538         /* Map high MMIO space */
1539         MemoryRegion *high_mmio_alias = g_new0(MemoryRegion, 1);
1540 
1541         memory_region_init_alias(high_mmio_alias, OBJECT(dev), "pcie-mmio-high",
1542                                  mmio_reg, base_mmio_high, size_mmio_high);
1543         memory_region_add_subregion(get_system_memory(), base_mmio_high,
1544                                     high_mmio_alias);
1545     }
1546 
1547     /* Map IO port space */
1548     sysbus_mmio_map(SYS_BUS_DEVICE(dev), 2, base_pio);
1549 
1550     for (i = 0; i < PCI_NUM_PINS; i++) {
1551         sysbus_connect_irq(SYS_BUS_DEVICE(dev), i,
1552                            qdev_get_gpio_in(vms->gic, irq + i));
1553         gpex_set_irq_num(GPEX_HOST(dev), i, irq + i);
1554     }
1555 
1556     pci = PCI_HOST_BRIDGE(dev);
1557     pci->bypass_iommu = vms->default_bus_bypass_iommu;
1558     vms->bus = pci->bus;
1559     if (vms->bus) {
1560         pci_init_nic_devices(pci->bus, mc->default_nic);
1561     }
1562 
1563     nodename = vms->pciehb_nodename = g_strdup_printf("/pcie@%" PRIx64, base);
1564     qemu_fdt_add_subnode(ms->fdt, nodename);
1565     qemu_fdt_setprop_string(ms->fdt, nodename,
1566                             "compatible", "pci-host-ecam-generic");
1567     qemu_fdt_setprop_string(ms->fdt, nodename, "device_type", "pci");
1568     qemu_fdt_setprop_cell(ms->fdt, nodename, "#address-cells", 3);
1569     qemu_fdt_setprop_cell(ms->fdt, nodename, "#size-cells", 2);
1570     qemu_fdt_setprop_cell(ms->fdt, nodename, "linux,pci-domain", 0);
1571     qemu_fdt_setprop_cells(ms->fdt, nodename, "bus-range", 0,
1572                            nr_pcie_buses - 1);
1573     qemu_fdt_setprop(ms->fdt, nodename, "dma-coherent", NULL, 0);
1574 
1575     if (vms->msi_phandle) {
1576         qemu_fdt_setprop_cells(ms->fdt, nodename, "msi-map",
1577                                0, vms->msi_phandle, 0, 0x10000);
1578     }
1579 
1580     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg",
1581                                  2, base_ecam, 2, size_ecam);
1582 
1583     if (vms->highmem_mmio) {
1584         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "ranges",
1585                                      1, FDT_PCI_RANGE_IOPORT, 2, 0,
1586                                      2, base_pio, 2, size_pio,
1587                                      1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
1588                                      2, base_mmio, 2, size_mmio,
1589                                      1, FDT_PCI_RANGE_MMIO_64BIT,
1590                                      2, base_mmio_high,
1591                                      2, base_mmio_high, 2, size_mmio_high);
1592     } else {
1593         qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "ranges",
1594                                      1, FDT_PCI_RANGE_IOPORT, 2, 0,
1595                                      2, base_pio, 2, size_pio,
1596                                      1, FDT_PCI_RANGE_MMIO, 2, base_mmio,
1597                                      2, base_mmio, 2, size_mmio);
1598     }
1599 
1600     qemu_fdt_setprop_cell(ms->fdt, nodename, "#interrupt-cells", 1);
1601     create_pcie_irq_map(ms, vms->gic_phandle, irq, nodename);
1602 
1603     if (vms->iommu) {
1604         vms->iommu_phandle = qemu_fdt_alloc_phandle(ms->fdt);
1605 
1606         switch (vms->iommu) {
1607         case VIRT_IOMMU_SMMUV3:
1608             create_smmu(vms, vms->bus);
1609             qemu_fdt_setprop_cells(ms->fdt, nodename, "iommu-map",
1610                                    0x0, vms->iommu_phandle, 0x0, 0x10000);
1611             break;
1612         default:
1613             g_assert_not_reached();
1614         }
1615     }
1616 }
1617 
1618 static void create_platform_bus(VirtMachineState *vms)
1619 {
1620     DeviceState *dev;
1621     SysBusDevice *s;
1622     int i;
1623     MemoryRegion *sysmem = get_system_memory();
1624 
1625     dev = qdev_new(TYPE_PLATFORM_BUS_DEVICE);
1626     dev->id = g_strdup(TYPE_PLATFORM_BUS_DEVICE);
1627     qdev_prop_set_uint32(dev, "num_irqs", PLATFORM_BUS_NUM_IRQS);
1628     qdev_prop_set_uint32(dev, "mmio_size", vms->memmap[VIRT_PLATFORM_BUS].size);
1629     sysbus_realize_and_unref(SYS_BUS_DEVICE(dev), &error_fatal);
1630     vms->platform_bus_dev = dev;
1631 
1632     s = SYS_BUS_DEVICE(dev);
1633     for (i = 0; i < PLATFORM_BUS_NUM_IRQS; i++) {
1634         int irq = vms->irqmap[VIRT_PLATFORM_BUS] + i;
1635         sysbus_connect_irq(s, i, qdev_get_gpio_in(vms->gic, irq));
1636     }
1637 
1638     memory_region_add_subregion(sysmem,
1639                                 vms->memmap[VIRT_PLATFORM_BUS].base,
1640                                 sysbus_mmio_get_region(s, 0));
1641 }
1642 
1643 static void create_tag_ram(MemoryRegion *tag_sysmem,
1644                            hwaddr base, hwaddr size,
1645                            const char *name)
1646 {
1647     MemoryRegion *tagram = g_new(MemoryRegion, 1);
1648 
1649     memory_region_init_ram(tagram, NULL, name, size / 32, &error_fatal);
1650     memory_region_add_subregion(tag_sysmem, base / 32, tagram);
1651 }
1652 
1653 static void create_secure_ram(VirtMachineState *vms,
1654                               MemoryRegion *secure_sysmem,
1655                               MemoryRegion *secure_tag_sysmem)
1656 {
1657     MemoryRegion *secram = g_new(MemoryRegion, 1);
1658     char *nodename;
1659     hwaddr base = vms->memmap[VIRT_SECURE_MEM].base;
1660     hwaddr size = vms->memmap[VIRT_SECURE_MEM].size;
1661     MachineState *ms = MACHINE(vms);
1662 
1663     memory_region_init_ram(secram, NULL, "virt.secure-ram", size,
1664                            &error_fatal);
1665     memory_region_add_subregion(secure_sysmem, base, secram);
1666 
1667     nodename = g_strdup_printf("/secram@%" PRIx64, base);
1668     qemu_fdt_add_subnode(ms->fdt, nodename);
1669     qemu_fdt_setprop_string(ms->fdt, nodename, "device_type", "memory");
1670     qemu_fdt_setprop_sized_cells(ms->fdt, nodename, "reg", 2, base, 2, size);
1671     qemu_fdt_setprop_string(ms->fdt, nodename, "status", "disabled");
1672     qemu_fdt_setprop_string(ms->fdt, nodename, "secure-status", "okay");
1673 
1674     if (secure_tag_sysmem) {
1675         create_tag_ram(secure_tag_sysmem, base, size, "mach-virt.secure-tag");
1676     }
1677 
1678     g_free(nodename);
1679 }
1680 
1681 static void *machvirt_dtb(const struct arm_boot_info *binfo, int *fdt_size)
1682 {
1683     const VirtMachineState *board = container_of(binfo, VirtMachineState,
1684                                                  bootinfo);
1685     MachineState *ms = MACHINE(board);
1686 
1687 
1688     *fdt_size = board->fdt_size;
1689     return ms->fdt;
1690 }
1691 
1692 static void virt_build_smbios(VirtMachineState *vms)
1693 {
1694     MachineClass *mc = MACHINE_GET_CLASS(vms);
1695     MachineState *ms = MACHINE(vms);
1696     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
1697     uint8_t *smbios_tables, *smbios_anchor;
1698     size_t smbios_tables_len, smbios_anchor_len;
1699     struct smbios_phys_mem_area mem_array;
1700     const char *product = "QEMU Virtual Machine";
1701 
1702     if (kvm_enabled()) {
1703         product = "KVM Virtual Machine";
1704     }
1705 
1706     smbios_set_defaults("QEMU", product,
1707                         vmc->smbios_old_sys_ver ? "1.0" : mc->name);
1708 
1709     /* build the array of physical mem area from base_memmap */
1710     mem_array.address = vms->memmap[VIRT_MEM].base;
1711     mem_array.length = ms->ram_size;
1712 
1713     smbios_get_tables(ms, SMBIOS_ENTRY_POINT_TYPE_64, &mem_array, 1,
1714                       &smbios_tables, &smbios_tables_len,
1715                       &smbios_anchor, &smbios_anchor_len,
1716                       &error_fatal);
1717 
1718     if (smbios_anchor) {
1719         fw_cfg_add_file(vms->fw_cfg, "etc/smbios/smbios-tables",
1720                         smbios_tables, smbios_tables_len);
1721         fw_cfg_add_file(vms->fw_cfg, "etc/smbios/smbios-anchor",
1722                         smbios_anchor, smbios_anchor_len);
1723     }
1724 }
1725 
1726 static
1727 void virt_machine_done(Notifier *notifier, void *data)
1728 {
1729     VirtMachineState *vms = container_of(notifier, VirtMachineState,
1730                                          machine_done);
1731     MachineState *ms = MACHINE(vms);
1732     ARMCPU *cpu = ARM_CPU(first_cpu);
1733     struct arm_boot_info *info = &vms->bootinfo;
1734     AddressSpace *as = arm_boot_address_space(cpu, info);
1735 
1736     /*
1737      * If the user provided a dtb, we assume the dynamic sysbus nodes
1738      * already are integrated there. This corresponds to a use case where
1739      * the dynamic sysbus nodes are complex and their generation is not yet
1740      * supported. In that case the user can take charge of the guest dt
1741      * while qemu takes charge of the qom stuff.
1742      */
1743     if (info->dtb_filename == NULL) {
1744         platform_bus_add_all_fdt_nodes(ms->fdt, "/intc",
1745                                        vms->memmap[VIRT_PLATFORM_BUS].base,
1746                                        vms->memmap[VIRT_PLATFORM_BUS].size,
1747                                        vms->irqmap[VIRT_PLATFORM_BUS]);
1748     }
1749     if (arm_load_dtb(info->dtb_start, info, info->dtb_limit, as, ms) < 0) {
1750         exit(1);
1751     }
1752 
1753     pci_bus_add_fw_cfg_extra_pci_roots(vms->fw_cfg, vms->bus,
1754                                        &error_abort);
1755 
1756     virt_acpi_setup(vms);
1757     virt_build_smbios(vms);
1758 }
1759 
1760 static uint64_t virt_cpu_mp_affinity(VirtMachineState *vms, int idx)
1761 {
1762     uint8_t clustersz = ARM_DEFAULT_CPUS_PER_CLUSTER;
1763     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
1764 
1765     if (!vmc->disallow_affinity_adjustment) {
1766         /* Adjust MPIDR like 64-bit KVM hosts, which incorporate the
1767          * GIC's target-list limitations. 32-bit KVM hosts currently
1768          * always create clusters of 4 CPUs, but that is expected to
1769          * change when they gain support for gicv3. When KVM is enabled
1770          * it will override the changes we make here, therefore our
1771          * purposes are to make TCG consistent (with 64-bit KVM hosts)
1772          * and to improve SGI efficiency.
1773          */
1774         if (vms->gic_version == VIRT_GIC_VERSION_2) {
1775             clustersz = GIC_TARGETLIST_BITS;
1776         } else {
1777             clustersz = GICV3_TARGETLIST_BITS;
1778         }
1779     }
1780     return arm_build_mp_affinity(idx, clustersz);
1781 }
1782 
1783 static inline bool *virt_get_high_memmap_enabled(VirtMachineState *vms,
1784                                                  int index)
1785 {
1786     bool *enabled_array[] = {
1787         &vms->highmem_redists,
1788         &vms->highmem_ecam,
1789         &vms->highmem_mmio,
1790     };
1791 
1792     assert(ARRAY_SIZE(extended_memmap) - VIRT_LOWMEMMAP_LAST ==
1793            ARRAY_SIZE(enabled_array));
1794     assert(index - VIRT_LOWMEMMAP_LAST < ARRAY_SIZE(enabled_array));
1795 
1796     return enabled_array[index - VIRT_LOWMEMMAP_LAST];
1797 }
1798 
1799 static void virt_set_high_memmap(VirtMachineState *vms,
1800                                  hwaddr base, int pa_bits)
1801 {
1802     hwaddr region_base, region_size;
1803     bool *region_enabled, fits;
1804     int i;
1805 
1806     for (i = VIRT_LOWMEMMAP_LAST; i < ARRAY_SIZE(extended_memmap); i++) {
1807         region_enabled = virt_get_high_memmap_enabled(vms, i);
1808         region_base = ROUND_UP(base, extended_memmap[i].size);
1809         region_size = extended_memmap[i].size;
1810 
1811         vms->memmap[i].base = region_base;
1812         vms->memmap[i].size = region_size;
1813 
1814         /*
1815          * Check each device to see if it fits in the PA space,
1816          * moving highest_gpa as we go. For compatibility, move
1817          * highest_gpa for disabled fitting devices as well, if
1818          * the compact layout has been disabled.
1819          *
1820          * For each device that doesn't fit, disable it.
1821          */
1822         fits = (region_base + region_size) <= BIT_ULL(pa_bits);
1823         *region_enabled &= fits;
1824         if (vms->highmem_compact && !*region_enabled) {
1825             continue;
1826         }
1827 
1828         base = region_base + region_size;
1829         if (fits) {
1830             vms->highest_gpa = base - 1;
1831         }
1832     }
1833 }
1834 
1835 static void virt_set_memmap(VirtMachineState *vms, int pa_bits)
1836 {
1837     MachineState *ms = MACHINE(vms);
1838     hwaddr base, device_memory_base, device_memory_size, memtop;
1839     int i;
1840 
1841     vms->memmap = extended_memmap;
1842 
1843     for (i = 0; i < ARRAY_SIZE(base_memmap); i++) {
1844         vms->memmap[i] = base_memmap[i];
1845     }
1846 
1847     if (ms->ram_slots > ACPI_MAX_RAM_SLOTS) {
1848         error_report("unsupported number of memory slots: %"PRIu64,
1849                      ms->ram_slots);
1850         exit(EXIT_FAILURE);
1851     }
1852 
1853     /*
1854      * !highmem is exactly the same as limiting the PA space to 32bit,
1855      * irrespective of the underlying capabilities of the HW.
1856      */
1857     if (!vms->highmem) {
1858         pa_bits = 32;
1859     }
1860 
1861     /*
1862      * We compute the base of the high IO region depending on the
1863      * amount of initial and device memory. The device memory start/size
1864      * is aligned on 1GiB. We never put the high IO region below 256GiB
1865      * so that if maxram_size is < 255GiB we keep the legacy memory map.
1866      * The device region size assumes 1GiB page max alignment per slot.
1867      */
1868     device_memory_base =
1869         ROUND_UP(vms->memmap[VIRT_MEM].base + ms->ram_size, GiB);
1870     device_memory_size = ms->maxram_size - ms->ram_size + ms->ram_slots * GiB;
1871 
1872     /* Base address of the high IO region */
1873     memtop = base = device_memory_base + ROUND_UP(device_memory_size, GiB);
1874     if (memtop > BIT_ULL(pa_bits)) {
1875         error_report("Addressing limited to %d bits, but memory exceeds it by %llu bytes",
1876                      pa_bits, memtop - BIT_ULL(pa_bits));
1877         exit(EXIT_FAILURE);
1878     }
1879     if (base < device_memory_base) {
1880         error_report("maxmem/slots too huge");
1881         exit(EXIT_FAILURE);
1882     }
1883     if (base < vms->memmap[VIRT_MEM].base + LEGACY_RAMLIMIT_BYTES) {
1884         base = vms->memmap[VIRT_MEM].base + LEGACY_RAMLIMIT_BYTES;
1885     }
1886 
1887     /* We know for sure that at least the memory fits in the PA space */
1888     vms->highest_gpa = memtop - 1;
1889 
1890     virt_set_high_memmap(vms, base, pa_bits);
1891 
1892     if (device_memory_size > 0) {
1893         machine_memory_devices_init(ms, device_memory_base, device_memory_size);
1894     }
1895 }
1896 
1897 static VirtGICType finalize_gic_version_do(const char *accel_name,
1898                                            VirtGICType gic_version,
1899                                            int gics_supported,
1900                                            unsigned int max_cpus)
1901 {
1902     /* Convert host/max/nosel to GIC version number */
1903     switch (gic_version) {
1904     case VIRT_GIC_VERSION_HOST:
1905         if (!kvm_enabled()) {
1906             error_report("gic-version=host requires KVM");
1907             exit(1);
1908         }
1909 
1910         /* For KVM, gic-version=host means gic-version=max */
1911         return finalize_gic_version_do(accel_name, VIRT_GIC_VERSION_MAX,
1912                                        gics_supported, max_cpus);
1913     case VIRT_GIC_VERSION_MAX:
1914         if (gics_supported & VIRT_GIC_VERSION_4_MASK) {
1915             gic_version = VIRT_GIC_VERSION_4;
1916         } else if (gics_supported & VIRT_GIC_VERSION_3_MASK) {
1917             gic_version = VIRT_GIC_VERSION_3;
1918         } else {
1919             gic_version = VIRT_GIC_VERSION_2;
1920         }
1921         break;
1922     case VIRT_GIC_VERSION_NOSEL:
1923         if ((gics_supported & VIRT_GIC_VERSION_2_MASK) &&
1924             max_cpus <= GIC_NCPU) {
1925             gic_version = VIRT_GIC_VERSION_2;
1926         } else if (gics_supported & VIRT_GIC_VERSION_3_MASK) {
1927             /*
1928              * in case the host does not support v2 emulation or
1929              * the end-user requested more than 8 VCPUs we now default
1930              * to v3. In any case defaulting to v2 would be broken.
1931              */
1932             gic_version = VIRT_GIC_VERSION_3;
1933         } else if (max_cpus > GIC_NCPU) {
1934             error_report("%s only supports GICv2 emulation but more than 8 "
1935                          "vcpus are requested", accel_name);
1936             exit(1);
1937         }
1938         break;
1939     case VIRT_GIC_VERSION_2:
1940     case VIRT_GIC_VERSION_3:
1941     case VIRT_GIC_VERSION_4:
1942         break;
1943     }
1944 
1945     /* Check chosen version is effectively supported */
1946     switch (gic_version) {
1947     case VIRT_GIC_VERSION_2:
1948         if (!(gics_supported & VIRT_GIC_VERSION_2_MASK)) {
1949             error_report("%s does not support GICv2 emulation", accel_name);
1950             exit(1);
1951         }
1952         break;
1953     case VIRT_GIC_VERSION_3:
1954         if (!(gics_supported & VIRT_GIC_VERSION_3_MASK)) {
1955             error_report("%s does not support GICv3 emulation", accel_name);
1956             exit(1);
1957         }
1958         break;
1959     case VIRT_GIC_VERSION_4:
1960         if (!(gics_supported & VIRT_GIC_VERSION_4_MASK)) {
1961             error_report("%s does not support GICv4 emulation, is virtualization=on?",
1962                          accel_name);
1963             exit(1);
1964         }
1965         break;
1966     default:
1967         error_report("logic error in finalize_gic_version");
1968         exit(1);
1969         break;
1970     }
1971 
1972     return gic_version;
1973 }
1974 
1975 /*
1976  * finalize_gic_version - Determines the final gic_version
1977  * according to the gic-version property
1978  *
1979  * Default GIC type is v2
1980  */
1981 static void finalize_gic_version(VirtMachineState *vms)
1982 {
1983     const char *accel_name = current_accel_name();
1984     unsigned int max_cpus = MACHINE(vms)->smp.max_cpus;
1985     int gics_supported = 0;
1986 
1987     /* Determine which GIC versions the current environment supports */
1988     if (kvm_enabled() && kvm_irqchip_in_kernel()) {
1989         int probe_bitmap = kvm_arm_vgic_probe();
1990 
1991         if (!probe_bitmap) {
1992             error_report("Unable to determine GIC version supported by host");
1993             exit(1);
1994         }
1995 
1996         if (probe_bitmap & KVM_ARM_VGIC_V2) {
1997             gics_supported |= VIRT_GIC_VERSION_2_MASK;
1998         }
1999         if (probe_bitmap & KVM_ARM_VGIC_V3) {
2000             gics_supported |= VIRT_GIC_VERSION_3_MASK;
2001         }
2002     } else if (kvm_enabled() && !kvm_irqchip_in_kernel()) {
2003         /* KVM w/o kernel irqchip can only deal with GICv2 */
2004         gics_supported |= VIRT_GIC_VERSION_2_MASK;
2005         accel_name = "KVM with kernel-irqchip=off";
2006     } else if (tcg_enabled() || hvf_enabled() || qtest_enabled())  {
2007         gics_supported |= VIRT_GIC_VERSION_2_MASK;
2008         if (module_object_class_by_name("arm-gicv3")) {
2009             gics_supported |= VIRT_GIC_VERSION_3_MASK;
2010             if (vms->virt) {
2011                 /* GICv4 only makes sense if CPU has EL2 */
2012                 gics_supported |= VIRT_GIC_VERSION_4_MASK;
2013             }
2014         }
2015     } else {
2016         error_report("Unsupported accelerator, can not determine GIC support");
2017         exit(1);
2018     }
2019 
2020     /*
2021      * Then convert helpers like host/max to concrete GIC versions and ensure
2022      * the desired version is supported
2023      */
2024     vms->gic_version = finalize_gic_version_do(accel_name, vms->gic_version,
2025                                                gics_supported, max_cpus);
2026 }
2027 
2028 /*
2029  * virt_cpu_post_init() must be called after the CPUs have
2030  * been realized and the GIC has been created.
2031  */
2032 static void virt_cpu_post_init(VirtMachineState *vms, MemoryRegion *sysmem)
2033 {
2034     int max_cpus = MACHINE(vms)->smp.max_cpus;
2035     bool aarch64, pmu, steal_time;
2036     CPUState *cpu;
2037 
2038     aarch64 = object_property_get_bool(OBJECT(first_cpu), "aarch64", NULL);
2039     pmu = object_property_get_bool(OBJECT(first_cpu), "pmu", NULL);
2040     steal_time = object_property_get_bool(OBJECT(first_cpu),
2041                                           "kvm-steal-time", NULL);
2042 
2043     if (kvm_enabled()) {
2044         hwaddr pvtime_reg_base = vms->memmap[VIRT_PVTIME].base;
2045         hwaddr pvtime_reg_size = vms->memmap[VIRT_PVTIME].size;
2046 
2047         if (steal_time) {
2048             MemoryRegion *pvtime = g_new(MemoryRegion, 1);
2049             hwaddr pvtime_size = max_cpus * PVTIME_SIZE_PER_CPU;
2050 
2051             /* The memory region size must be a multiple of host page size. */
2052             pvtime_size = REAL_HOST_PAGE_ALIGN(pvtime_size);
2053 
2054             if (pvtime_size > pvtime_reg_size) {
2055                 error_report("pvtime requires a %" HWADDR_PRId
2056                              " byte memory region for %d CPUs,"
2057                              " but only %" HWADDR_PRId " has been reserved",
2058                              pvtime_size, max_cpus, pvtime_reg_size);
2059                 exit(1);
2060             }
2061 
2062             memory_region_init_ram(pvtime, NULL, "pvtime", pvtime_size, NULL);
2063             memory_region_add_subregion(sysmem, pvtime_reg_base, pvtime);
2064         }
2065 
2066         CPU_FOREACH(cpu) {
2067             if (pmu) {
2068                 assert(arm_feature(&ARM_CPU(cpu)->env, ARM_FEATURE_PMU));
2069                 if (kvm_irqchip_in_kernel()) {
2070                     kvm_arm_pmu_set_irq(ARM_CPU(cpu), VIRTUAL_PMU_IRQ);
2071                 }
2072                 kvm_arm_pmu_init(ARM_CPU(cpu));
2073             }
2074             if (steal_time) {
2075                 kvm_arm_pvtime_init(ARM_CPU(cpu), pvtime_reg_base
2076                                                   + cpu->cpu_index
2077                                                     * PVTIME_SIZE_PER_CPU);
2078             }
2079         }
2080     } else {
2081         if (aarch64 && vms->highmem) {
2082             int requested_pa_size = 64 - clz64(vms->highest_gpa);
2083             int pamax = arm_pamax(ARM_CPU(first_cpu));
2084 
2085             if (pamax < requested_pa_size) {
2086                 error_report("VCPU supports less PA bits (%d) than "
2087                              "requested by the memory map (%d)",
2088                              pamax, requested_pa_size);
2089                 exit(1);
2090             }
2091         }
2092     }
2093 }
2094 
2095 static void machvirt_init(MachineState *machine)
2096 {
2097     VirtMachineState *vms = VIRT_MACHINE(machine);
2098     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(machine);
2099     MachineClass *mc = MACHINE_GET_CLASS(machine);
2100     const CPUArchIdList *possible_cpus;
2101     MemoryRegion *sysmem = get_system_memory();
2102     MemoryRegion *secure_sysmem = NULL;
2103     MemoryRegion *tag_sysmem = NULL;
2104     MemoryRegion *secure_tag_sysmem = NULL;
2105     int n, virt_max_cpus;
2106     bool firmware_loaded;
2107     bool aarch64 = true;
2108     bool has_ged = !vmc->no_ged;
2109     unsigned int smp_cpus = machine->smp.cpus;
2110     unsigned int max_cpus = machine->smp.max_cpus;
2111 
2112     possible_cpus = mc->possible_cpu_arch_ids(machine);
2113 
2114     /*
2115      * In accelerated mode, the memory map is computed earlier in kvm_type()
2116      * for Linux, or hvf_get_physical_address_range() for macOS to create a
2117      * VM with the right number of IPA bits.
2118      */
2119     if (!vms->memmap) {
2120         Object *cpuobj;
2121         ARMCPU *armcpu;
2122         int pa_bits;
2123 
2124         /*
2125          * Instantiate a temporary CPU object to find out about what
2126          * we are about to deal with. Once this is done, get rid of
2127          * the object.
2128          */
2129         cpuobj = object_new(possible_cpus->cpus[0].type);
2130         armcpu = ARM_CPU(cpuobj);
2131 
2132         pa_bits = arm_pamax(armcpu);
2133 
2134         object_unref(cpuobj);
2135 
2136         virt_set_memmap(vms, pa_bits);
2137     }
2138 
2139     /* We can probe only here because during property set
2140      * KVM is not available yet
2141      */
2142     finalize_gic_version(vms);
2143 
2144     if (vms->secure) {
2145         /*
2146          * The Secure view of the world is the same as the NonSecure,
2147          * but with a few extra devices. Create it as a container region
2148          * containing the system memory at low priority; any secure-only
2149          * devices go in at higher priority and take precedence.
2150          */
2151         secure_sysmem = g_new(MemoryRegion, 1);
2152         memory_region_init(secure_sysmem, OBJECT(machine), "secure-memory",
2153                            UINT64_MAX);
2154         memory_region_add_subregion_overlap(secure_sysmem, 0, sysmem, -1);
2155     }
2156 
2157     firmware_loaded = virt_firmware_init(vms, sysmem,
2158                                          secure_sysmem ?: sysmem);
2159 
2160     /* If we have an EL3 boot ROM then the assumption is that it will
2161      * implement PSCI itself, so disable QEMU's internal implementation
2162      * so it doesn't get in the way. Instead of starting secondary
2163      * CPUs in PSCI powerdown state we will start them all running and
2164      * let the boot ROM sort them out.
2165      * The usual case is that we do use QEMU's PSCI implementation;
2166      * if the guest has EL2 then we will use SMC as the conduit,
2167      * and otherwise we will use HVC (for backwards compatibility and
2168      * because if we're using KVM then we must use HVC).
2169      */
2170     if (vms->secure && firmware_loaded) {
2171         vms->psci_conduit = QEMU_PSCI_CONDUIT_DISABLED;
2172     } else if (vms->virt) {
2173         vms->psci_conduit = QEMU_PSCI_CONDUIT_SMC;
2174     } else {
2175         vms->psci_conduit = QEMU_PSCI_CONDUIT_HVC;
2176     }
2177 
2178     /*
2179      * The maximum number of CPUs depends on the GIC version, or on how
2180      * many redistributors we can fit into the memory map (which in turn
2181      * depends on whether this is a GICv3 or v4).
2182      */
2183     if (vms->gic_version == VIRT_GIC_VERSION_2) {
2184         virt_max_cpus = GIC_NCPU;
2185     } else {
2186         virt_max_cpus = virt_redist_capacity(vms, VIRT_GIC_REDIST);
2187         if (vms->highmem_redists) {
2188             virt_max_cpus += virt_redist_capacity(vms, VIRT_HIGH_GIC_REDIST2);
2189         }
2190     }
2191 
2192     if (max_cpus > virt_max_cpus) {
2193         error_report("Number of SMP CPUs requested (%d) exceeds max CPUs "
2194                      "supported by machine 'mach-virt' (%d)",
2195                      max_cpus, virt_max_cpus);
2196         if (vms->gic_version != VIRT_GIC_VERSION_2 && !vms->highmem_redists) {
2197             error_printf("Try 'highmem-redists=on' for more CPUs\n");
2198         }
2199 
2200         exit(1);
2201     }
2202 
2203     if (vms->secure && (kvm_enabled() || hvf_enabled())) {
2204         error_report("mach-virt: %s does not support providing "
2205                      "Security extensions (TrustZone) to the guest CPU",
2206                      current_accel_name());
2207         exit(1);
2208     }
2209 
2210     if (vms->virt && (kvm_enabled() || hvf_enabled())) {
2211         error_report("mach-virt: %s does not support providing "
2212                      "Virtualization extensions to the guest CPU",
2213                      current_accel_name());
2214         exit(1);
2215     }
2216 
2217     if (vms->mte && hvf_enabled()) {
2218         error_report("mach-virt: %s does not support providing "
2219                      "MTE to the guest CPU",
2220                      current_accel_name());
2221         exit(1);
2222     }
2223 
2224     create_fdt(vms);
2225 
2226     assert(possible_cpus->len == max_cpus);
2227     for (n = 0; n < possible_cpus->len; n++) {
2228         Object *cpuobj;
2229         CPUState *cs;
2230 
2231         if (n >= smp_cpus) {
2232             break;
2233         }
2234 
2235         cpuobj = object_new(possible_cpus->cpus[n].type);
2236         object_property_set_int(cpuobj, "mp-affinity",
2237                                 possible_cpus->cpus[n].arch_id, NULL);
2238 
2239         cs = CPU(cpuobj);
2240         cs->cpu_index = n;
2241 
2242         numa_cpu_pre_plug(&possible_cpus->cpus[cs->cpu_index], DEVICE(cpuobj),
2243                           &error_fatal);
2244 
2245         aarch64 &= object_property_get_bool(cpuobj, "aarch64", NULL);
2246 
2247         if (!vms->secure) {
2248             object_property_set_bool(cpuobj, "has_el3", false, NULL);
2249         }
2250 
2251         if (!vms->virt && object_property_find(cpuobj, "has_el2")) {
2252             object_property_set_bool(cpuobj, "has_el2", false, NULL);
2253         }
2254 
2255         if (vmc->kvm_no_adjvtime &&
2256             object_property_find(cpuobj, "kvm-no-adjvtime")) {
2257             object_property_set_bool(cpuobj, "kvm-no-adjvtime", true, NULL);
2258         }
2259 
2260         if (vmc->no_kvm_steal_time &&
2261             object_property_find(cpuobj, "kvm-steal-time")) {
2262             object_property_set_bool(cpuobj, "kvm-steal-time", false, NULL);
2263         }
2264 
2265         if (vmc->no_pmu && object_property_find(cpuobj, "pmu")) {
2266             object_property_set_bool(cpuobj, "pmu", false, NULL);
2267         }
2268 
2269         if (vmc->no_tcg_lpa2 && object_property_find(cpuobj, "lpa2")) {
2270             object_property_set_bool(cpuobj, "lpa2", false, NULL);
2271         }
2272 
2273         if (object_property_find(cpuobj, "reset-cbar")) {
2274             object_property_set_int(cpuobj, "reset-cbar",
2275                                     vms->memmap[VIRT_CPUPERIPHS].base,
2276                                     &error_abort);
2277         }
2278 
2279         object_property_set_link(cpuobj, "memory", OBJECT(sysmem),
2280                                  &error_abort);
2281         if (vms->secure) {
2282             object_property_set_link(cpuobj, "secure-memory",
2283                                      OBJECT(secure_sysmem), &error_abort);
2284         }
2285 
2286         if (vms->mte) {
2287             if (tcg_enabled()) {
2288                 /* Create the memory region only once, but link to all cpus. */
2289                 if (!tag_sysmem) {
2290                     /*
2291                      * The property exists only if MemTag is supported.
2292                      * If it is, we must allocate the ram to back that up.
2293                      */
2294                     if (!object_property_find(cpuobj, "tag-memory")) {
2295                         error_report("MTE requested, but not supported "
2296                                      "by the guest CPU");
2297                         exit(1);
2298                     }
2299 
2300                     tag_sysmem = g_new(MemoryRegion, 1);
2301                     memory_region_init(tag_sysmem, OBJECT(machine),
2302                                        "tag-memory", UINT64_MAX / 32);
2303 
2304                     if (vms->secure) {
2305                         secure_tag_sysmem = g_new(MemoryRegion, 1);
2306                         memory_region_init(secure_tag_sysmem, OBJECT(machine),
2307                                            "secure-tag-memory",
2308                                            UINT64_MAX / 32);
2309 
2310                         /* As with ram, secure-tag takes precedence over tag. */
2311                         memory_region_add_subregion_overlap(secure_tag_sysmem,
2312                                                             0, tag_sysmem, -1);
2313                     }
2314                 }
2315 
2316                 object_property_set_link(cpuobj, "tag-memory",
2317                                          OBJECT(tag_sysmem), &error_abort);
2318                 if (vms->secure) {
2319                     object_property_set_link(cpuobj, "secure-tag-memory",
2320                                              OBJECT(secure_tag_sysmem),
2321                                              &error_abort);
2322                 }
2323             } else if (kvm_enabled()) {
2324                 if (!kvm_arm_mte_supported()) {
2325                     error_report("MTE requested, but not supported by KVM");
2326                     exit(1);
2327                 }
2328                 kvm_arm_enable_mte(cpuobj, &error_abort);
2329             } else {
2330                     error_report("MTE requested, but not supported ");
2331                     exit(1);
2332             }
2333         }
2334 
2335         qdev_realize(DEVICE(cpuobj), NULL, &error_fatal);
2336         object_unref(cpuobj);
2337     }
2338 
2339     /* Now we've created the CPUs we can see if they have the hypvirt timer */
2340     vms->ns_el2_virt_timer_irq = ns_el2_virt_timer_present() &&
2341         !vmc->no_ns_el2_virt_timer_irq;
2342 
2343     fdt_add_timer_nodes(vms);
2344     fdt_add_cpu_nodes(vms);
2345 
2346     memory_region_add_subregion(sysmem, vms->memmap[VIRT_MEM].base,
2347                                 machine->ram);
2348 
2349     virt_flash_fdt(vms, sysmem, secure_sysmem ?: sysmem);
2350 
2351     create_gic(vms, sysmem);
2352 
2353     virt_cpu_post_init(vms, sysmem);
2354 
2355     fdt_add_pmu_nodes(vms);
2356 
2357     /*
2358      * The first UART always exists. If the security extensions are
2359      * enabled, the second UART also always exists. Otherwise, it only exists
2360      * if a backend is configured explicitly via '-serial <backend>'.
2361      * This avoids potentially breaking existing user setups that expect
2362      * only one NonSecure UART to be present (for instance, older EDK2
2363      * binaries).
2364      *
2365      * The nodes end up in the DTB in reverse order of creation, so we must
2366      * create UART0 last to ensure it appears as the first node in the DTB,
2367      * for compatibility with guest software that just iterates through the
2368      * DTB to find the first UART, as older versions of EDK2 do.
2369      * DTB readers that follow the spec, as Linux does, should honour the
2370      * aliases node information and /chosen/stdout-path regardless of
2371      * the order that nodes appear in the DTB.
2372      *
2373      * For similar back-compatibility reasons, if UART1 is the secure UART
2374      * we create it second (and so it appears first in the DTB), because
2375      * that's what QEMU has always done.
2376      */
2377     if (!vms->secure) {
2378         Chardev *serial1 = serial_hd(1);
2379 
2380         if (serial1) {
2381             vms->second_ns_uart_present = true;
2382             create_uart(vms, VIRT_UART1, sysmem, serial1, false);
2383         }
2384     }
2385     create_uart(vms, VIRT_UART0, sysmem, serial_hd(0), false);
2386     if (vms->secure) {
2387         create_uart(vms, VIRT_UART1, secure_sysmem, serial_hd(1), true);
2388     }
2389 
2390     if (vms->secure) {
2391         create_secure_ram(vms, secure_sysmem, secure_tag_sysmem);
2392     }
2393 
2394     if (tag_sysmem) {
2395         create_tag_ram(tag_sysmem, vms->memmap[VIRT_MEM].base,
2396                        machine->ram_size, "mach-virt.tag");
2397     }
2398 
2399     vms->highmem_ecam &= (!firmware_loaded || aarch64);
2400 
2401     create_rtc(vms);
2402 
2403     create_pcie(vms);
2404 
2405     if (has_ged && aarch64 && firmware_loaded && virt_is_acpi_enabled(vms)) {
2406         vms->acpi_dev = create_acpi_ged(vms);
2407     } else {
2408         create_gpio_devices(vms, VIRT_GPIO, sysmem);
2409     }
2410 
2411     if (vms->secure && !vmc->no_secure_gpio) {
2412         create_gpio_devices(vms, VIRT_SECURE_GPIO, secure_sysmem);
2413     }
2414 
2415      /* connect powerdown request */
2416      vms->powerdown_notifier.notify = virt_powerdown_req;
2417      qemu_register_powerdown_notifier(&vms->powerdown_notifier);
2418 
2419     /* Create mmio transports, so the user can create virtio backends
2420      * (which will be automatically plugged in to the transports). If
2421      * no backend is created the transport will just sit harmlessly idle.
2422      */
2423     create_virtio_devices(vms);
2424 
2425     vms->fw_cfg = create_fw_cfg(vms, &address_space_memory);
2426     rom_set_fw(vms->fw_cfg);
2427 
2428     create_platform_bus(vms);
2429 
2430     if (machine->nvdimms_state->is_enabled) {
2431         const struct AcpiGenericAddress arm_virt_nvdimm_acpi_dsmio = {
2432             .space_id = AML_AS_SYSTEM_MEMORY,
2433             .address = vms->memmap[VIRT_NVDIMM_ACPI].base,
2434             .bit_width = NVDIMM_ACPI_IO_LEN << 3
2435         };
2436 
2437         nvdimm_init_acpi_state(machine->nvdimms_state, sysmem,
2438                                arm_virt_nvdimm_acpi_dsmio,
2439                                vms->fw_cfg, OBJECT(vms));
2440     }
2441 
2442     vms->bootinfo.ram_size = machine->ram_size;
2443     vms->bootinfo.board_id = -1;
2444     vms->bootinfo.loader_start = vms->memmap[VIRT_MEM].base;
2445     vms->bootinfo.get_dtb = machvirt_dtb;
2446     vms->bootinfo.skip_dtb_autoload = true;
2447     vms->bootinfo.firmware_loaded = firmware_loaded;
2448     vms->bootinfo.psci_conduit = vms->psci_conduit;
2449     arm_load_kernel(ARM_CPU(first_cpu), machine, &vms->bootinfo);
2450 
2451     vms->machine_done.notify = virt_machine_done;
2452     qemu_add_machine_init_done_notifier(&vms->machine_done);
2453 }
2454 
2455 static bool virt_get_secure(Object *obj, Error **errp)
2456 {
2457     VirtMachineState *vms = VIRT_MACHINE(obj);
2458 
2459     return vms->secure;
2460 }
2461 
2462 static void virt_set_secure(Object *obj, bool value, Error **errp)
2463 {
2464     VirtMachineState *vms = VIRT_MACHINE(obj);
2465 
2466     vms->secure = value;
2467 }
2468 
2469 static bool virt_get_virt(Object *obj, Error **errp)
2470 {
2471     VirtMachineState *vms = VIRT_MACHINE(obj);
2472 
2473     return vms->virt;
2474 }
2475 
2476 static void virt_set_virt(Object *obj, bool value, Error **errp)
2477 {
2478     VirtMachineState *vms = VIRT_MACHINE(obj);
2479 
2480     vms->virt = value;
2481 }
2482 
2483 static bool virt_get_highmem(Object *obj, Error **errp)
2484 {
2485     VirtMachineState *vms = VIRT_MACHINE(obj);
2486 
2487     return vms->highmem;
2488 }
2489 
2490 static void virt_set_highmem(Object *obj, bool value, Error **errp)
2491 {
2492     VirtMachineState *vms = VIRT_MACHINE(obj);
2493 
2494     vms->highmem = value;
2495 }
2496 
2497 static bool virt_get_compact_highmem(Object *obj, Error **errp)
2498 {
2499     VirtMachineState *vms = VIRT_MACHINE(obj);
2500 
2501     return vms->highmem_compact;
2502 }
2503 
2504 static void virt_set_compact_highmem(Object *obj, bool value, Error **errp)
2505 {
2506     VirtMachineState *vms = VIRT_MACHINE(obj);
2507 
2508     vms->highmem_compact = value;
2509 }
2510 
2511 static bool virt_get_highmem_redists(Object *obj, Error **errp)
2512 {
2513     VirtMachineState *vms = VIRT_MACHINE(obj);
2514 
2515     return vms->highmem_redists;
2516 }
2517 
2518 static void virt_set_highmem_redists(Object *obj, bool value, Error **errp)
2519 {
2520     VirtMachineState *vms = VIRT_MACHINE(obj);
2521 
2522     vms->highmem_redists = value;
2523 }
2524 
2525 static bool virt_get_highmem_ecam(Object *obj, Error **errp)
2526 {
2527     VirtMachineState *vms = VIRT_MACHINE(obj);
2528 
2529     return vms->highmem_ecam;
2530 }
2531 
2532 static void virt_set_highmem_ecam(Object *obj, bool value, Error **errp)
2533 {
2534     VirtMachineState *vms = VIRT_MACHINE(obj);
2535 
2536     vms->highmem_ecam = value;
2537 }
2538 
2539 static bool virt_get_highmem_mmio(Object *obj, Error **errp)
2540 {
2541     VirtMachineState *vms = VIRT_MACHINE(obj);
2542 
2543     return vms->highmem_mmio;
2544 }
2545 
2546 static void virt_set_highmem_mmio(Object *obj, bool value, Error **errp)
2547 {
2548     VirtMachineState *vms = VIRT_MACHINE(obj);
2549 
2550     vms->highmem_mmio = value;
2551 }
2552 
2553 
2554 static bool virt_get_its(Object *obj, Error **errp)
2555 {
2556     VirtMachineState *vms = VIRT_MACHINE(obj);
2557 
2558     return vms->its;
2559 }
2560 
2561 static void virt_set_its(Object *obj, bool value, Error **errp)
2562 {
2563     VirtMachineState *vms = VIRT_MACHINE(obj);
2564 
2565     vms->its = value;
2566 }
2567 
2568 static bool virt_get_dtb_randomness(Object *obj, Error **errp)
2569 {
2570     VirtMachineState *vms = VIRT_MACHINE(obj);
2571 
2572     return vms->dtb_randomness;
2573 }
2574 
2575 static void virt_set_dtb_randomness(Object *obj, bool value, Error **errp)
2576 {
2577     VirtMachineState *vms = VIRT_MACHINE(obj);
2578 
2579     vms->dtb_randomness = value;
2580 }
2581 
2582 static char *virt_get_oem_id(Object *obj, Error **errp)
2583 {
2584     VirtMachineState *vms = VIRT_MACHINE(obj);
2585 
2586     return g_strdup(vms->oem_id);
2587 }
2588 
2589 static void virt_set_oem_id(Object *obj, const char *value, Error **errp)
2590 {
2591     VirtMachineState *vms = VIRT_MACHINE(obj);
2592     size_t len = strlen(value);
2593 
2594     if (len > 6) {
2595         error_setg(errp,
2596                    "User specified oem-id value is bigger than 6 bytes in size");
2597         return;
2598     }
2599 
2600     strncpy(vms->oem_id, value, 6);
2601 }
2602 
2603 static char *virt_get_oem_table_id(Object *obj, Error **errp)
2604 {
2605     VirtMachineState *vms = VIRT_MACHINE(obj);
2606 
2607     return g_strdup(vms->oem_table_id);
2608 }
2609 
2610 static void virt_set_oem_table_id(Object *obj, const char *value,
2611                                   Error **errp)
2612 {
2613     VirtMachineState *vms = VIRT_MACHINE(obj);
2614     size_t len = strlen(value);
2615 
2616     if (len > 8) {
2617         error_setg(errp,
2618                    "User specified oem-table-id value is bigger than 8 bytes in size");
2619         return;
2620     }
2621     strncpy(vms->oem_table_id, value, 8);
2622 }
2623 
2624 
2625 bool virt_is_acpi_enabled(VirtMachineState *vms)
2626 {
2627     if (vms->acpi == ON_OFF_AUTO_OFF) {
2628         return false;
2629     }
2630     return true;
2631 }
2632 
2633 static void virt_get_acpi(Object *obj, Visitor *v, const char *name,
2634                           void *opaque, Error **errp)
2635 {
2636     VirtMachineState *vms = VIRT_MACHINE(obj);
2637     OnOffAuto acpi = vms->acpi;
2638 
2639     visit_type_OnOffAuto(v, name, &acpi, errp);
2640 }
2641 
2642 static void virt_set_acpi(Object *obj, Visitor *v, const char *name,
2643                           void *opaque, Error **errp)
2644 {
2645     VirtMachineState *vms = VIRT_MACHINE(obj);
2646 
2647     visit_type_OnOffAuto(v, name, &vms->acpi, errp);
2648 }
2649 
2650 static bool virt_get_ras(Object *obj, Error **errp)
2651 {
2652     VirtMachineState *vms = VIRT_MACHINE(obj);
2653 
2654     return vms->ras;
2655 }
2656 
2657 static void virt_set_ras(Object *obj, bool value, Error **errp)
2658 {
2659     VirtMachineState *vms = VIRT_MACHINE(obj);
2660 
2661     vms->ras = value;
2662 }
2663 
2664 static bool virt_get_mte(Object *obj, Error **errp)
2665 {
2666     VirtMachineState *vms = VIRT_MACHINE(obj);
2667 
2668     return vms->mte;
2669 }
2670 
2671 static void virt_set_mte(Object *obj, bool value, Error **errp)
2672 {
2673     VirtMachineState *vms = VIRT_MACHINE(obj);
2674 
2675     vms->mte = value;
2676 }
2677 
2678 static char *virt_get_gic_version(Object *obj, Error **errp)
2679 {
2680     VirtMachineState *vms = VIRT_MACHINE(obj);
2681     const char *val;
2682 
2683     switch (vms->gic_version) {
2684     case VIRT_GIC_VERSION_4:
2685         val = "4";
2686         break;
2687     case VIRT_GIC_VERSION_3:
2688         val = "3";
2689         break;
2690     default:
2691         val = "2";
2692         break;
2693     }
2694     return g_strdup(val);
2695 }
2696 
2697 static void virt_set_gic_version(Object *obj, const char *value, Error **errp)
2698 {
2699     VirtMachineState *vms = VIRT_MACHINE(obj);
2700 
2701     if (!strcmp(value, "4")) {
2702         vms->gic_version = VIRT_GIC_VERSION_4;
2703     } else if (!strcmp(value, "3")) {
2704         vms->gic_version = VIRT_GIC_VERSION_3;
2705     } else if (!strcmp(value, "2")) {
2706         vms->gic_version = VIRT_GIC_VERSION_2;
2707     } else if (!strcmp(value, "host")) {
2708         vms->gic_version = VIRT_GIC_VERSION_HOST; /* Will probe later */
2709     } else if (!strcmp(value, "max")) {
2710         vms->gic_version = VIRT_GIC_VERSION_MAX; /* Will probe later */
2711     } else {
2712         error_setg(errp, "Invalid gic-version value");
2713         error_append_hint(errp, "Valid values are 3, 2, host, max.\n");
2714     }
2715 }
2716 
2717 static char *virt_get_iommu(Object *obj, Error **errp)
2718 {
2719     VirtMachineState *vms = VIRT_MACHINE(obj);
2720 
2721     switch (vms->iommu) {
2722     case VIRT_IOMMU_NONE:
2723         return g_strdup("none");
2724     case VIRT_IOMMU_SMMUV3:
2725         return g_strdup("smmuv3");
2726     default:
2727         g_assert_not_reached();
2728     }
2729 }
2730 
2731 static void virt_set_iommu(Object *obj, const char *value, Error **errp)
2732 {
2733     VirtMachineState *vms = VIRT_MACHINE(obj);
2734 
2735     if (!strcmp(value, "smmuv3")) {
2736         vms->iommu = VIRT_IOMMU_SMMUV3;
2737     } else if (!strcmp(value, "none")) {
2738         vms->iommu = VIRT_IOMMU_NONE;
2739     } else {
2740         error_setg(errp, "Invalid iommu value");
2741         error_append_hint(errp, "Valid values are none, smmuv3.\n");
2742     }
2743 }
2744 
2745 static bool virt_get_default_bus_bypass_iommu(Object *obj, Error **errp)
2746 {
2747     VirtMachineState *vms = VIRT_MACHINE(obj);
2748 
2749     return vms->default_bus_bypass_iommu;
2750 }
2751 
2752 static void virt_set_default_bus_bypass_iommu(Object *obj, bool value,
2753                                               Error **errp)
2754 {
2755     VirtMachineState *vms = VIRT_MACHINE(obj);
2756 
2757     vms->default_bus_bypass_iommu = value;
2758 }
2759 
2760 static CpuInstanceProperties
2761 virt_cpu_index_to_props(MachineState *ms, unsigned cpu_index)
2762 {
2763     MachineClass *mc = MACHINE_GET_CLASS(ms);
2764     const CPUArchIdList *possible_cpus = mc->possible_cpu_arch_ids(ms);
2765 
2766     assert(cpu_index < possible_cpus->len);
2767     return possible_cpus->cpus[cpu_index].props;
2768 }
2769 
2770 static int64_t virt_get_default_cpu_node_id(const MachineState *ms, int idx)
2771 {
2772     int64_t socket_id = ms->possible_cpus->cpus[idx].props.socket_id;
2773 
2774     return socket_id % ms->numa_state->num_nodes;
2775 }
2776 
2777 static const CPUArchIdList *virt_possible_cpu_arch_ids(MachineState *ms)
2778 {
2779     int n;
2780     unsigned int max_cpus = ms->smp.max_cpus;
2781     VirtMachineState *vms = VIRT_MACHINE(ms);
2782     MachineClass *mc = MACHINE_GET_CLASS(vms);
2783 
2784     if (ms->possible_cpus) {
2785         assert(ms->possible_cpus->len == max_cpus);
2786         return ms->possible_cpus;
2787     }
2788 
2789     ms->possible_cpus = g_malloc0(sizeof(CPUArchIdList) +
2790                                   sizeof(CPUArchId) * max_cpus);
2791     ms->possible_cpus->len = max_cpus;
2792     for (n = 0; n < ms->possible_cpus->len; n++) {
2793         ms->possible_cpus->cpus[n].type = ms->cpu_type;
2794         ms->possible_cpus->cpus[n].arch_id =
2795             virt_cpu_mp_affinity(vms, n);
2796 
2797         assert(!mc->smp_props.dies_supported);
2798         ms->possible_cpus->cpus[n].props.has_socket_id = true;
2799         ms->possible_cpus->cpus[n].props.socket_id =
2800             n / (ms->smp.clusters * ms->smp.cores * ms->smp.threads);
2801         ms->possible_cpus->cpus[n].props.has_cluster_id = true;
2802         ms->possible_cpus->cpus[n].props.cluster_id =
2803             (n / (ms->smp.cores * ms->smp.threads)) % ms->smp.clusters;
2804         ms->possible_cpus->cpus[n].props.has_core_id = true;
2805         ms->possible_cpus->cpus[n].props.core_id =
2806             (n / ms->smp.threads) % ms->smp.cores;
2807         ms->possible_cpus->cpus[n].props.has_thread_id = true;
2808         ms->possible_cpus->cpus[n].props.thread_id =
2809             n % ms->smp.threads;
2810     }
2811     return ms->possible_cpus;
2812 }
2813 
2814 static void virt_memory_pre_plug(HotplugHandler *hotplug_dev, DeviceState *dev,
2815                                  Error **errp)
2816 {
2817     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2818     const MachineState *ms = MACHINE(hotplug_dev);
2819     const bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2820 
2821     if (!vms->acpi_dev) {
2822         error_setg(errp,
2823                    "memory hotplug is not enabled: missing acpi-ged device");
2824         return;
2825     }
2826 
2827     if (vms->mte) {
2828         error_setg(errp, "memory hotplug is not enabled: MTE is enabled");
2829         return;
2830     }
2831 
2832     if (is_nvdimm && !ms->nvdimms_state->is_enabled) {
2833         error_setg(errp, "nvdimm is not enabled: add 'nvdimm=on' to '-M'");
2834         return;
2835     }
2836 
2837     pc_dimm_pre_plug(PC_DIMM(dev), MACHINE(hotplug_dev), errp);
2838 }
2839 
2840 static void virt_memory_plug(HotplugHandler *hotplug_dev,
2841                              DeviceState *dev, Error **errp)
2842 {
2843     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2844     MachineState *ms = MACHINE(hotplug_dev);
2845     bool is_nvdimm = object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM);
2846 
2847     pc_dimm_plug(PC_DIMM(dev), MACHINE(vms));
2848 
2849     if (is_nvdimm) {
2850         nvdimm_plug(ms->nvdimms_state);
2851     }
2852 
2853     hotplug_handler_plug(HOTPLUG_HANDLER(vms->acpi_dev),
2854                          dev, &error_abort);
2855 }
2856 
2857 static void virt_machine_device_pre_plug_cb(HotplugHandler *hotplug_dev,
2858                                             DeviceState *dev, Error **errp)
2859 {
2860     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2861 
2862     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2863         virt_memory_pre_plug(hotplug_dev, dev, errp);
2864     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI)) {
2865         virtio_md_pci_pre_plug(VIRTIO_MD_PCI(dev), MACHINE(hotplug_dev), errp);
2866     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2867         hwaddr db_start = 0, db_end = 0;
2868         QList *reserved_regions;
2869         char *resv_prop_str;
2870 
2871         if (vms->iommu != VIRT_IOMMU_NONE) {
2872             error_setg(errp, "virt machine does not support multiple IOMMUs");
2873             return;
2874         }
2875 
2876         switch (vms->msi_controller) {
2877         case VIRT_MSI_CTRL_NONE:
2878             return;
2879         case VIRT_MSI_CTRL_ITS:
2880             /* GITS_TRANSLATER page */
2881             db_start = base_memmap[VIRT_GIC_ITS].base + 0x10000;
2882             db_end = base_memmap[VIRT_GIC_ITS].base +
2883                      base_memmap[VIRT_GIC_ITS].size - 1;
2884             break;
2885         case VIRT_MSI_CTRL_GICV2M:
2886             /* MSI_SETSPI_NS page */
2887             db_start = base_memmap[VIRT_GIC_V2M].base;
2888             db_end = db_start + base_memmap[VIRT_GIC_V2M].size - 1;
2889             break;
2890         }
2891         resv_prop_str = g_strdup_printf("0x%"PRIx64":0x%"PRIx64":%u",
2892                                         db_start, db_end,
2893                                         VIRTIO_IOMMU_RESV_MEM_T_MSI);
2894 
2895         reserved_regions = qlist_new();
2896         qlist_append_str(reserved_regions, resv_prop_str);
2897         qdev_prop_set_array(dev, "reserved-regions", reserved_regions);
2898         g_free(resv_prop_str);
2899     }
2900 }
2901 
2902 static void virt_machine_device_plug_cb(HotplugHandler *hotplug_dev,
2903                                         DeviceState *dev, Error **errp)
2904 {
2905     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2906 
2907     if (vms->platform_bus_dev) {
2908         MachineClass *mc = MACHINE_GET_CLASS(vms);
2909 
2910         if (device_is_dynamic_sysbus(mc, dev)) {
2911             platform_bus_link_device(PLATFORM_BUS_DEVICE(vms->platform_bus_dev),
2912                                      SYS_BUS_DEVICE(dev));
2913         }
2914     }
2915 
2916     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2917         virt_memory_plug(hotplug_dev, dev, errp);
2918     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI)) {
2919         virtio_md_pci_plug(VIRTIO_MD_PCI(dev), MACHINE(hotplug_dev), errp);
2920     }
2921 
2922     if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
2923         PCIDevice *pdev = PCI_DEVICE(dev);
2924 
2925         vms->iommu = VIRT_IOMMU_VIRTIO;
2926         vms->virtio_iommu_bdf = pci_get_bdf(pdev);
2927         create_virtio_iommu_dt_bindings(vms);
2928     }
2929 }
2930 
2931 static void virt_dimm_unplug_request(HotplugHandler *hotplug_dev,
2932                                      DeviceState *dev, Error **errp)
2933 {
2934     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2935 
2936     if (!vms->acpi_dev) {
2937         error_setg(errp,
2938                    "memory hotplug is not enabled: missing acpi-ged device");
2939         return;
2940     }
2941 
2942     if (object_dynamic_cast(OBJECT(dev), TYPE_NVDIMM)) {
2943         error_setg(errp, "nvdimm device hot unplug is not supported yet.");
2944         return;
2945     }
2946 
2947     hotplug_handler_unplug_request(HOTPLUG_HANDLER(vms->acpi_dev), dev,
2948                                    errp);
2949 }
2950 
2951 static void virt_dimm_unplug(HotplugHandler *hotplug_dev,
2952                              DeviceState *dev, Error **errp)
2953 {
2954     VirtMachineState *vms = VIRT_MACHINE(hotplug_dev);
2955     Error *local_err = NULL;
2956 
2957     hotplug_handler_unplug(HOTPLUG_HANDLER(vms->acpi_dev), dev, &local_err);
2958     if (local_err) {
2959         goto out;
2960     }
2961 
2962     pc_dimm_unplug(PC_DIMM(dev), MACHINE(vms));
2963     qdev_unrealize(dev);
2964 
2965 out:
2966     error_propagate(errp, local_err);
2967 }
2968 
2969 static void virt_machine_device_unplug_request_cb(HotplugHandler *hotplug_dev,
2970                                           DeviceState *dev, Error **errp)
2971 {
2972     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2973         virt_dimm_unplug_request(hotplug_dev, dev, errp);
2974     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI)) {
2975         virtio_md_pci_unplug_request(VIRTIO_MD_PCI(dev), MACHINE(hotplug_dev),
2976                                      errp);
2977     } else {
2978         error_setg(errp, "device unplug request for unsupported device"
2979                    " type: %s", object_get_typename(OBJECT(dev)));
2980     }
2981 }
2982 
2983 static void virt_machine_device_unplug_cb(HotplugHandler *hotplug_dev,
2984                                           DeviceState *dev, Error **errp)
2985 {
2986     if (object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM)) {
2987         virt_dimm_unplug(hotplug_dev, dev, errp);
2988     } else if (object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI)) {
2989         virtio_md_pci_unplug(VIRTIO_MD_PCI(dev), MACHINE(hotplug_dev), errp);
2990     } else {
2991         error_setg(errp, "virt: device unplug for unsupported device"
2992                    " type: %s", object_get_typename(OBJECT(dev)));
2993     }
2994 }
2995 
2996 static HotplugHandler *virt_machine_get_hotplug_handler(MachineState *machine,
2997                                                         DeviceState *dev)
2998 {
2999     MachineClass *mc = MACHINE_GET_CLASS(machine);
3000 
3001     if (device_is_dynamic_sysbus(mc, dev) ||
3002         object_dynamic_cast(OBJECT(dev), TYPE_PC_DIMM) ||
3003         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_MD_PCI) ||
3004         object_dynamic_cast(OBJECT(dev), TYPE_VIRTIO_IOMMU_PCI)) {
3005         return HOTPLUG_HANDLER(machine);
3006     }
3007     return NULL;
3008 }
3009 
3010 /*
3011  * for arm64 kvm_type [7-0] encodes the requested number of bits
3012  * in the IPA address space
3013  */
3014 static int virt_kvm_type(MachineState *ms, const char *type_str)
3015 {
3016     VirtMachineState *vms = VIRT_MACHINE(ms);
3017     int max_vm_pa_size, requested_pa_size;
3018     bool fixed_ipa;
3019 
3020     max_vm_pa_size = kvm_arm_get_max_vm_ipa_size(ms, &fixed_ipa);
3021 
3022     /* we freeze the memory map to compute the highest gpa */
3023     virt_set_memmap(vms, max_vm_pa_size);
3024 
3025     requested_pa_size = 64 - clz64(vms->highest_gpa);
3026 
3027     /*
3028      * KVM requires the IPA size to be at least 32 bits.
3029      */
3030     if (requested_pa_size < 32) {
3031         requested_pa_size = 32;
3032     }
3033 
3034     if (requested_pa_size > max_vm_pa_size) {
3035         error_report("-m and ,maxmem option values "
3036                      "require an IPA range (%d bits) larger than "
3037                      "the one supported by the host (%d bits)",
3038                      requested_pa_size, max_vm_pa_size);
3039         return -1;
3040     }
3041     /*
3042      * We return the requested PA log size, unless KVM only supports
3043      * the implicit legacy 40b IPA setting, in which case the kvm_type
3044      * must be 0.
3045      */
3046     return fixed_ipa ? 0 : requested_pa_size;
3047 }
3048 
3049 static int virt_hvf_get_physical_address_range(MachineState *ms)
3050 {
3051     VirtMachineState *vms = VIRT_MACHINE(ms);
3052 
3053     int default_ipa_size = hvf_arm_get_default_ipa_bit_size();
3054     int max_ipa_size = hvf_arm_get_max_ipa_bit_size();
3055 
3056     /* We freeze the memory map to compute the highest gpa */
3057     virt_set_memmap(vms, max_ipa_size);
3058 
3059     int requested_ipa_size = 64 - clz64(vms->highest_gpa);
3060 
3061     /*
3062      * If we're <= the default IPA size just use the default.
3063      * If we're above the default but below the maximum, round up to
3064      * the maximum. hvf_arm_get_max_ipa_bit_size() conveniently only
3065      * returns values that are valid ARM PARange values.
3066      */
3067     if (requested_ipa_size <= default_ipa_size) {
3068         requested_ipa_size = default_ipa_size;
3069     } else if (requested_ipa_size <= max_ipa_size) {
3070         requested_ipa_size = max_ipa_size;
3071     } else {
3072         error_report("-m and ,maxmem option values "
3073                      "require an IPA range (%d bits) larger than "
3074                      "the one supported by the host (%d bits)",
3075                      requested_ipa_size, max_ipa_size);
3076         return -1;
3077     }
3078 
3079     return requested_ipa_size;
3080 }
3081 
3082 static void virt_machine_class_init(ObjectClass *oc, void *data)
3083 {
3084     MachineClass *mc = MACHINE_CLASS(oc);
3085     HotplugHandlerClass *hc = HOTPLUG_HANDLER_CLASS(oc);
3086     static const char * const valid_cpu_types[] = {
3087 #ifdef CONFIG_TCG
3088         ARM_CPU_TYPE_NAME("cortex-a7"),
3089         ARM_CPU_TYPE_NAME("cortex-a15"),
3090 #ifdef TARGET_AARCH64
3091         ARM_CPU_TYPE_NAME("cortex-a35"),
3092         ARM_CPU_TYPE_NAME("cortex-a55"),
3093         ARM_CPU_TYPE_NAME("cortex-a72"),
3094         ARM_CPU_TYPE_NAME("cortex-a76"),
3095         ARM_CPU_TYPE_NAME("cortex-a710"),
3096         ARM_CPU_TYPE_NAME("a64fx"),
3097         ARM_CPU_TYPE_NAME("neoverse-n1"),
3098         ARM_CPU_TYPE_NAME("neoverse-v1"),
3099         ARM_CPU_TYPE_NAME("neoverse-n2"),
3100 #endif /* TARGET_AARCH64 */
3101 #endif /* CONFIG_TCG */
3102 #ifdef TARGET_AARCH64
3103         ARM_CPU_TYPE_NAME("cortex-a53"),
3104         ARM_CPU_TYPE_NAME("cortex-a57"),
3105 #if defined(CONFIG_KVM) || defined(CONFIG_HVF)
3106         ARM_CPU_TYPE_NAME("host"),
3107 #endif /* CONFIG_KVM || CONFIG_HVF */
3108 #endif /* TARGET_AARCH64 */
3109         ARM_CPU_TYPE_NAME("max"),
3110         NULL
3111     };
3112 
3113     mc->init = machvirt_init;
3114     /* Start with max_cpus set to 512, which is the maximum supported by KVM.
3115      * The value may be reduced later when we have more information about the
3116      * configuration of the particular instance.
3117      */
3118     mc->max_cpus = 512;
3119     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_CALXEDA_XGMAC);
3120     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_AMD_XGBE);
3121     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_RAMFB_DEVICE);
3122     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_VFIO_PLATFORM);
3123 #ifdef CONFIG_TPM
3124     machine_class_allow_dynamic_sysbus_dev(mc, TYPE_TPM_TIS_SYSBUS);
3125 #endif
3126     mc->block_default_type = IF_VIRTIO;
3127     mc->no_cdrom = 1;
3128     mc->pci_allow_0_address = true;
3129     /* We know we will never create a pre-ARMv7 CPU which needs 1K pages */
3130     mc->minimum_page_bits = 12;
3131     mc->possible_cpu_arch_ids = virt_possible_cpu_arch_ids;
3132     mc->cpu_index_to_instance_props = virt_cpu_index_to_props;
3133 #ifdef CONFIG_TCG
3134     mc->default_cpu_type = ARM_CPU_TYPE_NAME("cortex-a15");
3135 #else
3136     mc->default_cpu_type = ARM_CPU_TYPE_NAME("max");
3137 #endif
3138     mc->valid_cpu_types = valid_cpu_types;
3139     mc->get_default_cpu_node_id = virt_get_default_cpu_node_id;
3140     mc->kvm_type = virt_kvm_type;
3141     mc->hvf_get_physical_address_range = virt_hvf_get_physical_address_range;
3142     assert(!mc->get_hotplug_handler);
3143     mc->get_hotplug_handler = virt_machine_get_hotplug_handler;
3144     hc->pre_plug = virt_machine_device_pre_plug_cb;
3145     hc->plug = virt_machine_device_plug_cb;
3146     hc->unplug_request = virt_machine_device_unplug_request_cb;
3147     hc->unplug = virt_machine_device_unplug_cb;
3148     mc->nvdimm_supported = true;
3149     mc->smp_props.clusters_supported = true;
3150     mc->auto_enable_numa_with_memhp = true;
3151     mc->auto_enable_numa_with_memdev = true;
3152     /* platform instead of architectural choice */
3153     mc->cpu_cluster_has_numa_boundary = true;
3154     mc->default_ram_id = "mach-virt.ram";
3155     mc->default_nic = "virtio-net-pci";
3156 
3157     object_class_property_add(oc, "acpi", "OnOffAuto",
3158         virt_get_acpi, virt_set_acpi,
3159         NULL, NULL);
3160     object_class_property_set_description(oc, "acpi",
3161         "Enable ACPI");
3162     object_class_property_add_bool(oc, "secure", virt_get_secure,
3163                                    virt_set_secure);
3164     object_class_property_set_description(oc, "secure",
3165                                                 "Set on/off to enable/disable the ARM "
3166                                                 "Security Extensions (TrustZone)");
3167 
3168     object_class_property_add_bool(oc, "virtualization", virt_get_virt,
3169                                    virt_set_virt);
3170     object_class_property_set_description(oc, "virtualization",
3171                                           "Set on/off to enable/disable emulating a "
3172                                           "guest CPU which implements the ARM "
3173                                           "Virtualization Extensions");
3174 
3175     object_class_property_add_bool(oc, "highmem", virt_get_highmem,
3176                                    virt_set_highmem);
3177     object_class_property_set_description(oc, "highmem",
3178                                           "Set on/off to enable/disable using "
3179                                           "physical address space above 32 bits");
3180 
3181     object_class_property_add_bool(oc, "compact-highmem",
3182                                    virt_get_compact_highmem,
3183                                    virt_set_compact_highmem);
3184     object_class_property_set_description(oc, "compact-highmem",
3185                                           "Set on/off to enable/disable compact "
3186                                           "layout for high memory regions");
3187 
3188     object_class_property_add_bool(oc, "highmem-redists",
3189                                    virt_get_highmem_redists,
3190                                    virt_set_highmem_redists);
3191     object_class_property_set_description(oc, "highmem-redists",
3192                                           "Set on/off to enable/disable high "
3193                                           "memory region for GICv3 or GICv4 "
3194                                           "redistributor");
3195 
3196     object_class_property_add_bool(oc, "highmem-ecam",
3197                                    virt_get_highmem_ecam,
3198                                    virt_set_highmem_ecam);
3199     object_class_property_set_description(oc, "highmem-ecam",
3200                                           "Set on/off to enable/disable high "
3201                                           "memory region for PCI ECAM");
3202 
3203     object_class_property_add_bool(oc, "highmem-mmio",
3204                                    virt_get_highmem_mmio,
3205                                    virt_set_highmem_mmio);
3206     object_class_property_set_description(oc, "highmem-mmio",
3207                                           "Set on/off to enable/disable high "
3208                                           "memory region for PCI MMIO");
3209 
3210     object_class_property_add_str(oc, "gic-version", virt_get_gic_version,
3211                                   virt_set_gic_version);
3212     object_class_property_set_description(oc, "gic-version",
3213                                           "Set GIC version. "
3214                                           "Valid values are 2, 3, 4, host and max");
3215 
3216     object_class_property_add_str(oc, "iommu", virt_get_iommu, virt_set_iommu);
3217     object_class_property_set_description(oc, "iommu",
3218                                           "Set the IOMMU type. "
3219                                           "Valid values are none and smmuv3");
3220 
3221     object_class_property_add_bool(oc, "default-bus-bypass-iommu",
3222                                    virt_get_default_bus_bypass_iommu,
3223                                    virt_set_default_bus_bypass_iommu);
3224     object_class_property_set_description(oc, "default-bus-bypass-iommu",
3225                                           "Set on/off to enable/disable "
3226                                           "bypass_iommu for default root bus");
3227 
3228     object_class_property_add_bool(oc, "ras", virt_get_ras,
3229                                    virt_set_ras);
3230     object_class_property_set_description(oc, "ras",
3231                                           "Set on/off to enable/disable reporting host memory errors "
3232                                           "to a KVM guest using ACPI and guest external abort exceptions");
3233 
3234     object_class_property_add_bool(oc, "mte", virt_get_mte, virt_set_mte);
3235     object_class_property_set_description(oc, "mte",
3236                                           "Set on/off to enable/disable emulating a "
3237                                           "guest CPU which implements the ARM "
3238                                           "Memory Tagging Extension");
3239 
3240     object_class_property_add_bool(oc, "its", virt_get_its,
3241                                    virt_set_its);
3242     object_class_property_set_description(oc, "its",
3243                                           "Set on/off to enable/disable "
3244                                           "ITS instantiation");
3245 
3246     object_class_property_add_bool(oc, "dtb-randomness",
3247                                    virt_get_dtb_randomness,
3248                                    virt_set_dtb_randomness);
3249     object_class_property_set_description(oc, "dtb-randomness",
3250                                           "Set off to disable passing random or "
3251                                           "non-deterministic dtb nodes to guest");
3252 
3253     object_class_property_add_bool(oc, "dtb-kaslr-seed",
3254                                    virt_get_dtb_randomness,
3255                                    virt_set_dtb_randomness);
3256     object_class_property_set_description(oc, "dtb-kaslr-seed",
3257                                           "Deprecated synonym of dtb-randomness");
3258 
3259     object_class_property_add_str(oc, "x-oem-id",
3260                                   virt_get_oem_id,
3261                                   virt_set_oem_id);
3262     object_class_property_set_description(oc, "x-oem-id",
3263                                           "Override the default value of field OEMID "
3264                                           "in ACPI table header."
3265                                           "The string may be up to 6 bytes in size");
3266 
3267 
3268     object_class_property_add_str(oc, "x-oem-table-id",
3269                                   virt_get_oem_table_id,
3270                                   virt_set_oem_table_id);
3271     object_class_property_set_description(oc, "x-oem-table-id",
3272                                           "Override the default value of field OEM Table ID "
3273                                           "in ACPI table header."
3274                                           "The string may be up to 8 bytes in size");
3275 
3276 }
3277 
3278 static void virt_instance_init(Object *obj)
3279 {
3280     VirtMachineState *vms = VIRT_MACHINE(obj);
3281     VirtMachineClass *vmc = VIRT_MACHINE_GET_CLASS(vms);
3282 
3283     /* EL3 is disabled by default on virt: this makes us consistent
3284      * between KVM and TCG for this board, and it also allows us to
3285      * boot UEFI blobs which assume no TrustZone support.
3286      */
3287     vms->secure = false;
3288 
3289     /* EL2 is also disabled by default, for similar reasons */
3290     vms->virt = false;
3291 
3292     /* High memory is enabled by default */
3293     vms->highmem = true;
3294     vms->highmem_compact = !vmc->no_highmem_compact;
3295     vms->gic_version = VIRT_GIC_VERSION_NOSEL;
3296 
3297     vms->highmem_ecam = !vmc->no_highmem_ecam;
3298     vms->highmem_mmio = true;
3299     vms->highmem_redists = true;
3300 
3301     if (vmc->no_its) {
3302         vms->its = false;
3303     } else {
3304         /* Default allows ITS instantiation */
3305         vms->its = true;
3306 
3307         if (vmc->no_tcg_its) {
3308             vms->tcg_its = false;
3309         } else {
3310             vms->tcg_its = true;
3311         }
3312     }
3313 
3314     /* Default disallows iommu instantiation */
3315     vms->iommu = VIRT_IOMMU_NONE;
3316 
3317     /* The default root bus is attached to iommu by default */
3318     vms->default_bus_bypass_iommu = false;
3319 
3320     /* Default disallows RAS instantiation */
3321     vms->ras = false;
3322 
3323     /* MTE is disabled by default.  */
3324     vms->mte = false;
3325 
3326     /* Supply kaslr-seed and rng-seed by default */
3327     vms->dtb_randomness = true;
3328 
3329     vms->irqmap = a15irqmap;
3330 
3331     virt_flash_create(vms);
3332 
3333     vms->oem_id = g_strndup(ACPI_BUILD_APPNAME6, 6);
3334     vms->oem_table_id = g_strndup(ACPI_BUILD_APPNAME8, 8);
3335 }
3336 
3337 static const TypeInfo virt_machine_info = {
3338     .name          = TYPE_VIRT_MACHINE,
3339     .parent        = TYPE_MACHINE,
3340     .abstract      = true,
3341     .instance_size = sizeof(VirtMachineState),
3342     .class_size    = sizeof(VirtMachineClass),
3343     .class_init    = virt_machine_class_init,
3344     .instance_init = virt_instance_init,
3345     .interfaces = (InterfaceInfo[]) {
3346          { TYPE_HOTPLUG_HANDLER },
3347          { }
3348     },
3349 };
3350 
3351 static void machvirt_machine_init(void)
3352 {
3353     type_register_static(&virt_machine_info);
3354 }
3355 type_init(machvirt_machine_init);
3356 
3357 static void virt_machine_10_0_options(MachineClass *mc)
3358 {
3359 }
3360 DEFINE_VIRT_MACHINE_AS_LATEST(10, 0)
3361 
3362 static void virt_machine_9_2_options(MachineClass *mc)
3363 {
3364     virt_machine_10_0_options(mc);
3365     compat_props_add(mc->compat_props, hw_compat_9_2, hw_compat_9_2_len);
3366 }
3367 DEFINE_VIRT_MACHINE(9, 2)
3368 
3369 static void virt_machine_9_1_options(MachineClass *mc)
3370 {
3371     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3372 
3373     virt_machine_9_2_options(mc);
3374     compat_props_add(mc->compat_props, hw_compat_9_1, hw_compat_9_1_len);
3375     /* 9.1 and earlier have only a stage-1 SMMU, not a nested s1+2 one */
3376     vmc->no_nested_smmu = true;
3377 }
3378 DEFINE_VIRT_MACHINE(9, 1)
3379 
3380 static void virt_machine_9_0_options(MachineClass *mc)
3381 {
3382     virt_machine_9_1_options(mc);
3383     mc->smbios_memory_device_size = 16 * GiB;
3384     compat_props_add(mc->compat_props, hw_compat_9_0, hw_compat_9_0_len);
3385 }
3386 DEFINE_VIRT_MACHINE(9, 0)
3387 
3388 static void virt_machine_8_2_options(MachineClass *mc)
3389 {
3390     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3391 
3392     virt_machine_9_0_options(mc);
3393     compat_props_add(mc->compat_props, hw_compat_8_2, hw_compat_8_2_len);
3394     /*
3395      * Don't expose NS_EL2_VIRT timer IRQ in DTB on ACPI on 8.2 and
3396      * earlier machines. (Exposing it tickles a bug in older EDK2
3397      * guest BIOS binaries.)
3398      */
3399     vmc->no_ns_el2_virt_timer_irq = true;
3400 }
3401 DEFINE_VIRT_MACHINE(8, 2)
3402 
3403 static void virt_machine_8_1_options(MachineClass *mc)
3404 {
3405     virt_machine_8_2_options(mc);
3406     compat_props_add(mc->compat_props, hw_compat_8_1, hw_compat_8_1_len);
3407 }
3408 DEFINE_VIRT_MACHINE(8, 1)
3409 
3410 static void virt_machine_8_0_options(MachineClass *mc)
3411 {
3412     virt_machine_8_1_options(mc);
3413     compat_props_add(mc->compat_props, hw_compat_8_0, hw_compat_8_0_len);
3414 }
3415 DEFINE_VIRT_MACHINE(8, 0)
3416 
3417 static void virt_machine_7_2_options(MachineClass *mc)
3418 {
3419     virt_machine_8_0_options(mc);
3420     compat_props_add(mc->compat_props, hw_compat_7_2, hw_compat_7_2_len);
3421 }
3422 DEFINE_VIRT_MACHINE(7, 2)
3423 
3424 static void virt_machine_7_1_options(MachineClass *mc)
3425 {
3426     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3427 
3428     virt_machine_7_2_options(mc);
3429     compat_props_add(mc->compat_props, hw_compat_7_1, hw_compat_7_1_len);
3430     /* Compact layout for high memory regions was introduced with 7.2 */
3431     vmc->no_highmem_compact = true;
3432 }
3433 DEFINE_VIRT_MACHINE(7, 1)
3434 
3435 static void virt_machine_7_0_options(MachineClass *mc)
3436 {
3437     virt_machine_7_1_options(mc);
3438     compat_props_add(mc->compat_props, hw_compat_7_0, hw_compat_7_0_len);
3439 }
3440 DEFINE_VIRT_MACHINE(7, 0)
3441 
3442 static void virt_machine_6_2_options(MachineClass *mc)
3443 {
3444     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3445 
3446     virt_machine_7_0_options(mc);
3447     compat_props_add(mc->compat_props, hw_compat_6_2, hw_compat_6_2_len);
3448     vmc->no_tcg_lpa2 = true;
3449 }
3450 DEFINE_VIRT_MACHINE(6, 2)
3451 
3452 static void virt_machine_6_1_options(MachineClass *mc)
3453 {
3454     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3455 
3456     virt_machine_6_2_options(mc);
3457     compat_props_add(mc->compat_props, hw_compat_6_1, hw_compat_6_1_len);
3458     mc->smp_props.prefer_sockets = true;
3459     vmc->no_cpu_topology = true;
3460 
3461     /* qemu ITS was introduced with 6.2 */
3462     vmc->no_tcg_its = true;
3463 }
3464 DEFINE_VIRT_MACHINE(6, 1)
3465 
3466 static void virt_machine_6_0_options(MachineClass *mc)
3467 {
3468     virt_machine_6_1_options(mc);
3469     compat_props_add(mc->compat_props, hw_compat_6_0, hw_compat_6_0_len);
3470 }
3471 DEFINE_VIRT_MACHINE(6, 0)
3472 
3473 static void virt_machine_5_2_options(MachineClass *mc)
3474 {
3475     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3476 
3477     virt_machine_6_0_options(mc);
3478     compat_props_add(mc->compat_props, hw_compat_5_2, hw_compat_5_2_len);
3479     vmc->no_secure_gpio = true;
3480 }
3481 DEFINE_VIRT_MACHINE(5, 2)
3482 
3483 static void virt_machine_5_1_options(MachineClass *mc)
3484 {
3485     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3486 
3487     virt_machine_5_2_options(mc);
3488     compat_props_add(mc->compat_props, hw_compat_5_1, hw_compat_5_1_len);
3489     vmc->no_kvm_steal_time = true;
3490 }
3491 DEFINE_VIRT_MACHINE(5, 1)
3492 
3493 static void virt_machine_5_0_options(MachineClass *mc)
3494 {
3495     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3496 
3497     virt_machine_5_1_options(mc);
3498     compat_props_add(mc->compat_props, hw_compat_5_0, hw_compat_5_0_len);
3499     mc->numa_mem_supported = true;
3500     vmc->acpi_expose_flash = true;
3501     mc->auto_enable_numa_with_memdev = false;
3502 }
3503 DEFINE_VIRT_MACHINE(5, 0)
3504 
3505 static void virt_machine_4_2_options(MachineClass *mc)
3506 {
3507     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3508 
3509     virt_machine_5_0_options(mc);
3510     compat_props_add(mc->compat_props, hw_compat_4_2, hw_compat_4_2_len);
3511     vmc->kvm_no_adjvtime = true;
3512 }
3513 DEFINE_VIRT_MACHINE(4, 2)
3514 
3515 static void virt_machine_4_1_options(MachineClass *mc)
3516 {
3517     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3518 
3519     virt_machine_4_2_options(mc);
3520     compat_props_add(mc->compat_props, hw_compat_4_1, hw_compat_4_1_len);
3521     vmc->no_ged = true;
3522     mc->auto_enable_numa_with_memhp = false;
3523 }
3524 DEFINE_VIRT_MACHINE(4, 1)
3525 
3526 static void virt_machine_4_0_options(MachineClass *mc)
3527 {
3528     virt_machine_4_1_options(mc);
3529     compat_props_add(mc->compat_props, hw_compat_4_0, hw_compat_4_0_len);
3530 }
3531 DEFINE_VIRT_MACHINE(4, 0)
3532 
3533 static void virt_machine_3_1_options(MachineClass *mc)
3534 {
3535     virt_machine_4_0_options(mc);
3536     compat_props_add(mc->compat_props, hw_compat_3_1, hw_compat_3_1_len);
3537 }
3538 DEFINE_VIRT_MACHINE(3, 1)
3539 
3540 static void virt_machine_3_0_options(MachineClass *mc)
3541 {
3542     virt_machine_3_1_options(mc);
3543     compat_props_add(mc->compat_props, hw_compat_3_0, hw_compat_3_0_len);
3544 }
3545 DEFINE_VIRT_MACHINE(3, 0)
3546 
3547 static void virt_machine_2_12_options(MachineClass *mc)
3548 {
3549     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3550 
3551     virt_machine_3_0_options(mc);
3552     compat_props_add(mc->compat_props, hw_compat_2_12, hw_compat_2_12_len);
3553     vmc->no_highmem_ecam = true;
3554     mc->max_cpus = 255;
3555 }
3556 DEFINE_VIRT_MACHINE(2, 12)
3557 
3558 static void virt_machine_2_11_options(MachineClass *mc)
3559 {
3560     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3561 
3562     virt_machine_2_12_options(mc);
3563     compat_props_add(mc->compat_props, hw_compat_2_11, hw_compat_2_11_len);
3564     vmc->smbios_old_sys_ver = true;
3565 }
3566 DEFINE_VIRT_MACHINE(2, 11)
3567 
3568 static void virt_machine_2_10_options(MachineClass *mc)
3569 {
3570     virt_machine_2_11_options(mc);
3571     compat_props_add(mc->compat_props, hw_compat_2_10, hw_compat_2_10_len);
3572     /* before 2.11 we never faulted accesses to bad addresses */
3573     mc->ignore_memory_transaction_failures = true;
3574 }
3575 DEFINE_VIRT_MACHINE(2, 10)
3576 
3577 static void virt_machine_2_9_options(MachineClass *mc)
3578 {
3579     virt_machine_2_10_options(mc);
3580     compat_props_add(mc->compat_props, hw_compat_2_9, hw_compat_2_9_len);
3581 }
3582 DEFINE_VIRT_MACHINE(2, 9)
3583 
3584 static void virt_machine_2_8_options(MachineClass *mc)
3585 {
3586     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3587 
3588     virt_machine_2_9_options(mc);
3589     compat_props_add(mc->compat_props, hw_compat_2_8, hw_compat_2_8_len);
3590     /* For 2.8 and earlier we falsely claimed in the DT that
3591      * our timers were edge-triggered, not level-triggered.
3592      */
3593     vmc->claim_edge_triggered_timers = true;
3594 }
3595 DEFINE_VIRT_MACHINE(2, 8)
3596 
3597 static void virt_machine_2_7_options(MachineClass *mc)
3598 {
3599     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3600 
3601     virt_machine_2_8_options(mc);
3602     compat_props_add(mc->compat_props, hw_compat_2_7, hw_compat_2_7_len);
3603     /* ITS was introduced with 2.8 */
3604     vmc->no_its = true;
3605     /* Stick with 1K pages for migration compatibility */
3606     mc->minimum_page_bits = 0;
3607 }
3608 DEFINE_VIRT_MACHINE(2, 7)
3609 
3610 static void virt_machine_2_6_options(MachineClass *mc)
3611 {
3612     VirtMachineClass *vmc = VIRT_MACHINE_CLASS(OBJECT_CLASS(mc));
3613 
3614     virt_machine_2_7_options(mc);
3615     compat_props_add(mc->compat_props, hw_compat_2_6, hw_compat_2_6_len);
3616     vmc->disallow_affinity_adjustment = true;
3617     /* Disable PMU for 2.6 as PMU support was first introduced in 2.7 */
3618     vmc->no_pmu = true;
3619 }
3620 DEFINE_VIRT_MACHINE(2, 6)
3621